diff options
Diffstat (limited to 'vp9')
56 files changed, 325 insertions, 6310 deletions
diff --git a/vp9/common/arm/neon/vp9_save_reg_neon.asm b/vp9/common/arm/neon/vp9_save_reg_neon.asm deleted file mode 100644 index 71c3e7077..000000000 --- a/vp9/common/arm/neon/vp9_save_reg_neon.asm +++ /dev/null @@ -1,36 +0,0 @@ -; -; Copyright (c) 2010 The WebM project authors. All Rights Reserved. -; -; Use of this source code is governed by a BSD-style license -; that can be found in the LICENSE file in the root of the source -; tree. An additional intellectual property rights grant can be found -; in the file PATENTS. All contributing project authors may -; be found in the AUTHORS file in the root of the source tree. -; - - - EXPORT |vp9_push_neon| - EXPORT |vp9_pop_neon| - - ARM - REQUIRE8 - PRESERVE8 - - AREA ||.text||, CODE, READONLY, ALIGN=2 - -|vp9_push_neon| PROC - vst1.i64 {d8, d9, d10, d11}, [r0]! - vst1.i64 {d12, d13, d14, d15}, [r0]! - bx lr - - ENDP - -|vp9_pop_neon| PROC - vld1.i64 {d8, d9, d10, d11}, [r0]! - vld1.i64 {d12, d13, d14, d15}, [r0]! - bx lr - - ENDP - - END - diff --git a/vp9/common/mips/dspr2/vp9_common_dspr2.c b/vp9/common/mips/dspr2/vp9_common_dspr2.c deleted file mode 100644 index 6498a7e9e..000000000 --- a/vp9/common/mips/dspr2/vp9_common_dspr2.c +++ /dev/null @@ -1,32 +0,0 @@ -/* - * Copyright (c) 2013 The WebM project authors. All Rights Reserved. - * - * Use of this source code is governed by a BSD-style license - * that can be found in the LICENSE file in the root of the source - * tree. An additional intellectual property rights grant can be found - * in the file PATENTS. All contributing project authors may - * be found in the AUTHORS file in the root of the source tree. - */ - -#include "vp9/common/mips/dspr2/vp9_common_dspr2.h" -#include "vpx_dsp/vpx_dsp_common.h" -#include "vpx_ports/mem.h" - -#if HAVE_DSPR2 -uint8_t vp9_ff_cropTbl_a[256 + 2 * CROP_WIDTH]; -uint8_t *vp9_ff_cropTbl; - -void vp9_dsputil_static_init(void) { - int i; - - for (i = 0; i < 256; i++) vp9_ff_cropTbl_a[i + CROP_WIDTH] = i; - - for (i = 0; i < CROP_WIDTH; i++) { - vp9_ff_cropTbl_a[i] = 0; - vp9_ff_cropTbl_a[i + CROP_WIDTH + 256] = 255; - } - - vp9_ff_cropTbl = &vp9_ff_cropTbl_a[CROP_WIDTH]; -} - -#endif diff --git a/vp9/common/mips/dspr2/vp9_common_dspr2.h b/vp9/common/mips/dspr2/vp9_common_dspr2.h deleted file mode 100644 index 4e6b3552e..000000000 --- a/vp9/common/mips/dspr2/vp9_common_dspr2.h +++ /dev/null @@ -1,62 +0,0 @@ -/* - * Copyright (c) 2013 The WebM project authors. All Rights Reserved. - * - * Use of this source code is governed by a BSD-style license - * that can be found in the LICENSE file in the root of the source - * tree. An additional intellectual property rights grant can be found - * in the file PATENTS. All contributing project authors may - * be found in the AUTHORS file in the root of the source tree. - */ - -#ifndef VP9_COMMON_MIPS_DSPR2_VP9_COMMON_DSPR2_H_ -#define VP9_COMMON_MIPS_DSPR2_VP9_COMMON_DSPR2_H_ - -#include <assert.h> - -#include "./vpx_config.h" -#include "vpx/vpx_integer.h" -#include "vpx_dsp/mips/common_dspr2.h" - -#ifdef __cplusplus -extern "C" { -#endif - -#if HAVE_DSPR2 - -extern uint8_t *vpx_ff_cropTbl; - -#define DCT_CONST_ROUND_SHIFT_TWICE_COSPI_16_64(input) ({ \ - \ - int32_t tmp, out; \ - int dct_cost_rounding = DCT_CONST_ROUNDING; \ - int in = input; \ - \ - __asm__ __volatile__ ( \ - /* out = dct_const_round_shift(input_dc * cospi_16_64); */ \ - "mtlo %[dct_cost_rounding], $ac1 \n\t"\ - "mthi $zero, $ac1 \n\t"\ - "madd $ac1, %[in], %[cospi_16_64] \n\t"\ - "extp %[tmp], $ac1, 31 \n\t"\ - \ - /* out = dct_const_round_shift(out * cospi_16_64); */ \ - "mtlo %[dct_cost_rounding], $ac2 \n\t"\ - "mthi $zero, $ac2 \n\t"\ - "madd $ac2, %[tmp], %[cospi_16_64] \n\t"\ - "extp %[out], $ac2, 31 \n\t"\ - \ - : [tmp] "=&r" (tmp), [out] "=r" (out) \ - : [in] "r" (in), \ - [dct_cost_rounding] "r" (dct_cost_rounding), \ - [cospi_16_64] "r" (cospi_16_64) \ - ); \ - out; }) - -void vp9_idct32_cols_add_blk_dspr2(int16_t *input, uint8_t *dest, - int dest_stride); - -#endif // #if HAVE_DSPR2 -#ifdef __cplusplus -} // extern "C" -#endif - -#endif // VP9_COMMON_MIPS_DSPR2_VP9_COMMON_DSPR2_H_ diff --git a/vp9/common/mips/dspr2/vp9_itrans16_dspr2.c b/vp9/common/mips/dspr2/vp9_itrans16_dspr2.c index aca6550c8..6ca83a00c 100644 --- a/vp9/common/mips/dspr2/vp9_itrans16_dspr2.c +++ b/vp9/common/mips/dspr2/vp9_itrans16_dspr2.c @@ -16,1074 +16,11 @@ #include "vp9/common/vp9_common.h" #include "vp9/common/vp9_blockd.h" #include "vp9/common/vp9_idct.h" -#include "vp9/common/mips/dspr2/vp9_common_dspr2.h" +#include "vpx_dsp/mips/inv_txfm_dspr2.h" #include "vpx_dsp/txfm_common.h" #include "vpx_ports/mem.h" #if HAVE_DSPR2 -static void idct16_rows_dspr2(const int16_t *input, int16_t *output, - uint32_t no_rows) { - int i; - int step1_0, step1_1, step1_2, step1_3, step1_4, step1_5, step1_6, step1_7; - int step1_10, step1_11, step1_12, step1_13; - int step2_0, step2_1, step2_2, step2_3; - int step2_8, step2_9, step2_10, step2_11; - int step2_12, step2_13, step2_14, step2_15; - int load1, load2, load3, load4, load5, load6, load7, load8; - int result1, result2, result3, result4; - const int const_2_power_13 = 8192; - - for (i = no_rows; i--; ) { - /* prefetch row */ - prefetch_load((const uint8_t *)(input + 16)); - - __asm__ __volatile__ ( - "lh %[load1], 0(%[input]) \n\t" - "lh %[load2], 16(%[input]) \n\t" - "lh %[load3], 8(%[input]) \n\t" - "lh %[load4], 24(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - "add %[result1], %[load1], %[load2] \n\t" - "sub %[result2], %[load1], %[load2] \n\t" - "madd $ac1, %[result1], %[cospi_16_64] \n\t" - "madd $ac2, %[result2], %[cospi_16_64] \n\t" - "extp %[step2_0], $ac1, 31 \n\t" - "extp %[step2_1], $ac2, 31 \n\t" - - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - "madd $ac3, %[load3], %[cospi_24_64] \n\t" - "msub $ac3, %[load4], %[cospi_8_64] \n\t" - "extp %[step2_2], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "madd $ac1, %[load3], %[cospi_8_64] \n\t" - "madd $ac1, %[load4], %[cospi_24_64] \n\t" - "extp %[step2_3], $ac1, 31 \n\t" - - "add %[step1_0], %[step2_0], %[step2_3] \n\t" - "add %[step1_1], %[step2_1], %[step2_2] \n\t" - "sub %[step1_2], %[step2_1], %[step2_2] \n\t" - "sub %[step1_3], %[step2_0], %[step2_3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), - [load3] "=&r" (load3), [load4] "=&r" (load4), - [result1] "=&r" (result1), [result2] "=&r" (result2), - [step2_0] "=&r" (step2_0), [step2_1] "=&r" (step2_1), - [step2_2] "=&r" (step2_2), [step2_3] "=&r" (step2_3), - [step1_0] "=r" (step1_0), [step1_1] "=r" (step1_1), - [step1_2] "=r" (step1_2), [step1_3] "=r" (step1_3) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64), - [cospi_16_64] "r" (cospi_16_64) - ); - - __asm__ __volatile__ ( - "lh %[load5], 2(%[input]) \n\t" - "lh %[load6], 30(%[input]) \n\t" - "lh %[load7], 18(%[input]) \n\t" - "lh %[load8], 14(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load5], %[cospi_30_64] \n\t" - "msub $ac1, %[load6], %[cospi_2_64] \n\t" - "extp %[result1], $ac1, 31 \n\t" - - "madd $ac3, %[load7], %[cospi_14_64] \n\t" - "msub $ac3, %[load8], %[cospi_18_64] \n\t" - "extp %[result2], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac1, %[load7], %[cospi_18_64] \n\t" - "madd $ac1, %[load8], %[cospi_14_64] \n\t" - "extp %[result3], $ac1, 31 \n\t" - - "madd $ac2, %[load5], %[cospi_2_64] \n\t" - "madd $ac2, %[load6], %[cospi_30_64] \n\t" - "extp %[result4], $ac2, 31 \n\t" - - "sub %[load5], %[result1], %[result2] \n\t" - "sub %[load6], %[result4], %[result3] \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load6], %[cospi_24_64] \n\t" - "msub $ac1, %[load5], %[cospi_8_64] \n\t" - "madd $ac3, %[load5], %[cospi_24_64] \n\t" - "madd $ac3, %[load6], %[cospi_8_64] \n\t" - - "extp %[step2_9], $ac1, 31 \n\t" - "extp %[step2_14], $ac3, 31 \n\t" - "add %[step2_8], %[result1], %[result2] \n\t" - "add %[step2_15], %[result4], %[result3] \n\t" - - : [load5] "=&r" (load5), [load6] "=&r" (load6), - [load7] "=&r" (load7), [load8] "=&r" (load8), - [result1] "=&r" (result1), [result2] "=&r" (result2), - [result3] "=&r" (result3), [result4] "=&r" (result4), - [step2_8] "=r" (step2_8), [step2_15] "=r" (step2_15), - [step2_9] "=r" (step2_9), [step2_14] "=r" (step2_14) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_30_64] "r" (cospi_30_64), [cospi_2_64] "r" (cospi_2_64), - [cospi_14_64] "r" (cospi_14_64), [cospi_18_64] "r" (cospi_18_64), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) - ); - - __asm__ __volatile__ ( - "lh %[load1], 10(%[input]) \n\t" - "lh %[load2], 22(%[input]) \n\t" - "lh %[load3], 26(%[input]) \n\t" - "lh %[load4], 6(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_22_64] \n\t" - "msub $ac1, %[load2], %[cospi_10_64] \n\t" - "extp %[result1], $ac1, 31 \n\t" - - "madd $ac3, %[load3], %[cospi_6_64] \n\t" - "msub $ac3, %[load4], %[cospi_26_64] \n\t" - "extp %[result2], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac1, %[load1], %[cospi_10_64] \n\t" - "madd $ac1, %[load2], %[cospi_22_64] \n\t" - "extp %[result3], $ac1, 31 \n\t" - - "madd $ac2, %[load3], %[cospi_26_64] \n\t" - "madd $ac2, %[load4], %[cospi_6_64] \n\t" - "extp %[result4], $ac2, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[result2], %[result1] \n\t" - "sub %[load2], %[result4], %[result3] \n\t" - - "msub $ac1, %[load1], %[cospi_24_64] \n\t" - "msub $ac1, %[load2], %[cospi_8_64] \n\t" - "madd $ac3, %[load2], %[cospi_24_64] \n\t" - "msub $ac3, %[load1], %[cospi_8_64] \n\t" - - "extp %[step2_10], $ac1, 31 \n\t" - "extp %[step2_13], $ac3, 31 \n\t" - "add %[step2_11], %[result1], %[result2] \n\t" - "add %[step2_12], %[result4], %[result3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), - [load3] "=&r" (load3), [load4] "=&r" (load4), - [result1] "=&r" (result1), [result2] "=&r" (result2), - [result3] "=&r" (result3), [result4] "=&r" (result4), - [step2_10] "=r" (step2_10), [step2_11] "=r" (step2_11), - [step2_12] "=r" (step2_12), [step2_13] "=r" (step2_13) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_22_64] "r" (cospi_22_64), [cospi_10_64] "r" (cospi_10_64), - [cospi_6_64] "r" (cospi_6_64), [cospi_26_64] "r" (cospi_26_64), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) - ); - - __asm__ __volatile__ ( - "lh %[load5], 4(%[input]) \n\t" - "lh %[load6], 28(%[input]) \n\t" - "lh %[load7], 20(%[input]) \n\t" - "lh %[load8], 12(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load5], %[cospi_28_64] \n\t" - "msub $ac1, %[load6], %[cospi_4_64] \n\t" - "extp %[result1], $ac1, 31 \n\t" - - "madd $ac3, %[load7], %[cospi_12_64] \n\t" - "msub $ac3, %[load8], %[cospi_20_64] \n\t" - "extp %[result2], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac1, %[load7], %[cospi_20_64] \n\t" - "madd $ac1, %[load8], %[cospi_12_64] \n\t" - "extp %[result3], $ac1, 31 \n\t" - - "madd $ac2, %[load5], %[cospi_4_64] \n\t" - "madd $ac2, %[load6], %[cospi_28_64] \n\t" - "extp %[result4], $ac2, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load5], %[result4], %[result3] \n\t" - "sub %[load5], %[load5], %[result1] \n\t" - "add %[load5], %[load5], %[result2] \n\t" - - "sub %[load6], %[result1], %[result2] \n\t" - "sub %[load6], %[load6], %[result3] \n\t" - "add %[load6], %[load6], %[result4] \n\t" - - "madd $ac1, %[load5], %[cospi_16_64] \n\t" - "madd $ac3, %[load6], %[cospi_16_64] \n\t" - - "extp %[step1_5], $ac1, 31 \n\t" - "extp %[step1_6], $ac3, 31 \n\t" - "add %[step1_4], %[result1], %[result2] \n\t" - "add %[step1_7], %[result4], %[result3] \n\t" - - : [load5] "=&r" (load5), [load6] "=&r" (load6), - [load7] "=&r" (load7), [load8] "=&r" (load8), - [result1] "=&r" (result1), [result2] "=&r" (result2), - [result3] "=&r" (result3), [result4] "=&r" (result4), - [step1_4] "=r" (step1_4), [step1_5] "=r" (step1_5), - [step1_6] "=r" (step1_6), [step1_7] "=r" (step1_7) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_20_64] "r" (cospi_20_64), [cospi_12_64] "r" (cospi_12_64), - [cospi_4_64] "r" (cospi_4_64), [cospi_28_64] "r" (cospi_28_64), - [cospi_16_64] "r" (cospi_16_64) - ); - - __asm__ __volatile__ ( - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - - "sub %[load5], %[step2_14], %[step2_13] \n\t" - "sub %[load5], %[load5], %[step2_9] \n\t" - "add %[load5], %[load5], %[step2_10] \n\t" - - "madd $ac0, %[load5], %[cospi_16_64] \n\t" - - "sub %[load6], %[step2_14], %[step2_13] \n\t" - "sub %[load6], %[load6], %[step2_10] \n\t" - "add %[load6], %[load6], %[step2_9] \n\t" - - "madd $ac1, %[load6], %[cospi_16_64] \n\t" - - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load5], %[step2_15], %[step2_12] \n\t" - "sub %[load5], %[load5], %[step2_8] \n\t" - "add %[load5], %[load5], %[step2_11] \n\t" - - "madd $ac2, %[load5], %[cospi_16_64] \n\t" - - "sub %[load6], %[step2_15], %[step2_12] \n\t" - "sub %[load6], %[load6], %[step2_11] \n\t" - "add %[load6], %[load6], %[step2_8] \n\t" - - "madd $ac3, %[load6], %[cospi_16_64] \n\t" - - "extp %[step1_10], $ac0, 31 \n\t" - "extp %[step1_13], $ac1, 31 \n\t" - "extp %[step1_11], $ac2, 31 \n\t" - "extp %[step1_12], $ac3, 31 \n\t" - - : [load5] "=&r" (load5), [load6] "=&r" (load6), - [step1_10] "=r" (step1_10), [step1_11] "=r" (step1_11), - [step1_12] "=r" (step1_12), [step1_13] "=r" (step1_13) - : [const_2_power_13] "r" (const_2_power_13), - [step2_14] "r" (step2_14), [step2_13] "r" (step2_13), - [step2_9] "r" (step2_9), [step2_10] "r" (step2_10), - [step2_15] "r" (step2_15), [step2_12] "r" (step2_12), - [step2_8] "r" (step2_8), [step2_11] "r" (step2_11), - [cospi_16_64] "r" (cospi_16_64) - ); - - __asm__ __volatile__ ( - "add %[load5], %[step1_0], %[step1_7] \n\t" - "add %[load5], %[load5], %[step2_12] \n\t" - "add %[load5], %[load5], %[step2_15] \n\t" - "add %[load6], %[step1_1], %[step1_6] \n\t" - "add %[load6], %[load6], %[step2_13] \n\t" - "add %[load6], %[load6], %[step2_14] \n\t" - "sh %[load5], 0(%[output]) \n\t" - "sh %[load6], 32(%[output]) \n\t" - "sub %[load5], %[step1_1], %[step1_6] \n\t" - "add %[load5], %[load5], %[step2_9] \n\t" - "add %[load5], %[load5], %[step2_10] \n\t" - "sub %[load6], %[step1_0], %[step1_7] \n\t" - "add %[load6], %[load6], %[step2_8] \n\t" - "add %[load6], %[load6], %[step2_11] \n\t" - "sh %[load5], 192(%[output]) \n\t" - "sh %[load6], 224(%[output]) \n\t" - "sub %[load5], %[step1_0], %[step1_7] \n\t" - "sub %[load5], %[load5], %[step2_8] \n\t" - "sub %[load5], %[load5], %[step2_11] \n\t" - "sub %[load6], %[step1_1], %[step1_6] \n\t" - "sub %[load6], %[load6], %[step2_9] \n\t" - "sub %[load6], %[load6], %[step2_10] \n\t" - "sh %[load5], 256(%[output]) \n\t" - "sh %[load6], 288(%[output]) \n\t" - "add %[load5], %[step1_1], %[step1_6] \n\t" - "sub %[load5], %[load5], %[step2_13] \n\t" - "sub %[load5], %[load5], %[step2_14] \n\t" - "add %[load6], %[step1_0], %[step1_7] \n\t" - "sub %[load6], %[load6], %[step2_12] \n\t" - "sub %[load6], %[load6], %[step2_15] \n\t" - "sh %[load5], 448(%[output]) \n\t" - "sh %[load6], 480(%[output]) \n\t" - - : [load5] "=&r" (load5), [load6] "=&r" (load6) - : [output] "r" (output), - [step1_0] "r" (step1_0), [step1_1] "r" (step1_1), - [step1_6] "r" (step1_6), [step1_7] "r" (step1_7), - [step2_8] "r" (step2_8), [step2_9] "r" (step2_9), - [step2_10] "r" (step2_10), [step2_11] "r" (step2_11), - [step2_12] "r" (step2_12), [step2_13] "r" (step2_13), - [step2_14] "r" (step2_14), [step2_15] "r" (step2_15) - ); - - __asm__ __volatile__ ( - "add %[load5], %[step1_2], %[step1_5] \n\t" - "add %[load5], %[load5], %[step1_13] \n\t" - "add %[load6], %[step1_3], %[step1_4] \n\t" - "add %[load6], %[load6], %[step1_12] \n\t" - "sh %[load5], 64(%[output]) \n\t" - "sh %[load6], 96(%[output]) \n\t" - "sub %[load5], %[step1_3], %[step1_4] \n\t" - "add %[load5], %[load5], %[step1_11] \n\t" - "sub %[load6], %[step1_2], %[step1_5] \n\t" - "add %[load6], %[load6], %[step1_10] \n\t" - "sh %[load5], 128(%[output]) \n\t" - "sh %[load6], 160(%[output]) \n\t" - "sub %[load5], %[step1_2], %[step1_5] \n\t" - "sub %[load5], %[load5], %[step1_10] \n\t" - "sub %[load6], %[step1_3], %[step1_4] \n\t" - "sub %[load6], %[load6], %[step1_11] \n\t" - "sh %[load5], 320(%[output]) \n\t" - "sh %[load6], 352(%[output]) \n\t" - "add %[load5], %[step1_3], %[step1_4] \n\t" - "sub %[load5], %[load5], %[step1_12] \n\t" - "add %[load6], %[step1_2], %[step1_5] \n\t" - "sub %[load6], %[load6], %[step1_13] \n\t" - "sh %[load5], 384(%[output]) \n\t" - "sh %[load6], 416(%[output]) \n\t" - - : [load5] "=&r" (load5), [load6] "=&r" (load6) - : [output] "r" (output), - [step1_2] "r" (step1_2), [step1_3] "r" (step1_3), - [step1_4] "r" (step1_4), [step1_5] "r" (step1_5), - [step1_10] "r" (step1_10), [step1_11] "r" (step1_11), - [step1_12] "r" (step1_12), [step1_13] "r" (step1_13) - ); - - input += 16; - output += 1; - } -} - -static void idct16_cols_add_blk_dspr2(int16_t *input, uint8_t *dest, - int dest_stride) { - int i; - int step1_0, step1_1, step1_2, step1_3, step1_4, step1_5, step1_6, step1_7; - int step1_8, step1_9, step1_10, step1_11; - int step1_12, step1_13, step1_14, step1_15; - int step2_0, step2_1, step2_2, step2_3; - int step2_8, step2_9, step2_10, step2_11; - int step2_12, step2_13, step2_14, step2_15; - int load1, load2, load3, load4, load5, load6, load7, load8; - int result1, result2, result3, result4; - const int const_2_power_13 = 8192; - uint8_t *dest_pix; - uint8_t *cm = vpx_ff_cropTbl; - - /* prefetch vpx_ff_cropTbl */ - prefetch_load(vpx_ff_cropTbl); - prefetch_load(vpx_ff_cropTbl + 32); - prefetch_load(vpx_ff_cropTbl + 64); - prefetch_load(vpx_ff_cropTbl + 96); - prefetch_load(vpx_ff_cropTbl + 128); - prefetch_load(vpx_ff_cropTbl + 160); - prefetch_load(vpx_ff_cropTbl + 192); - prefetch_load(vpx_ff_cropTbl + 224); - - for (i = 0; i < 16; ++i) { - dest_pix = (dest + i); - __asm__ __volatile__ ( - "lh %[load1], 0(%[input]) \n\t" - "lh %[load2], 16(%[input]) \n\t" - "lh %[load3], 8(%[input]) \n\t" - "lh %[load4], 24(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - "add %[result1], %[load1], %[load2] \n\t" - "sub %[result2], %[load1], %[load2] \n\t" - "madd $ac1, %[result1], %[cospi_16_64] \n\t" - "madd $ac2, %[result2], %[cospi_16_64] \n\t" - "extp %[step2_0], $ac1, 31 \n\t" - "extp %[step2_1], $ac2, 31 \n\t" - - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - "madd $ac3, %[load3], %[cospi_24_64] \n\t" - "msub $ac3, %[load4], %[cospi_8_64] \n\t" - "extp %[step2_2], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "madd $ac1, %[load3], %[cospi_8_64] \n\t" - "madd $ac1, %[load4], %[cospi_24_64] \n\t" - "extp %[step2_3], $ac1, 31 \n\t" - - "add %[step1_0], %[step2_0], %[step2_3] \n\t" - "add %[step1_1], %[step2_1], %[step2_2] \n\t" - "sub %[step1_2], %[step2_1], %[step2_2] \n\t" - "sub %[step1_3], %[step2_0], %[step2_3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), - [load3] "=&r" (load3), [load4] "=&r" (load4), - [result1] "=&r" (result1), [result2] "=&r" (result2), - [step2_0] "=&r" (step2_0), [step2_1] "=&r" (step2_1), - [step2_2] "=&r" (step2_2), [step2_3] "=&r" (step2_3), - [step1_0] "=r" (step1_0), [step1_1] "=r" (step1_1), - [step1_2] "=r" (step1_2), [step1_3] "=r" (step1_3) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64), - [cospi_16_64] "r" (cospi_16_64) - ); - - __asm__ __volatile__ ( - "lh %[load5], 2(%[input]) \n\t" - "lh %[load6], 30(%[input]) \n\t" - "lh %[load7], 18(%[input]) \n\t" - "lh %[load8], 14(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load5], %[cospi_30_64] \n\t" - "msub $ac1, %[load6], %[cospi_2_64] \n\t" - "extp %[result1], $ac1, 31 \n\t" - - "madd $ac3, %[load7], %[cospi_14_64] \n\t" - "msub $ac3, %[load8], %[cospi_18_64] \n\t" - "extp %[result2], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac1, %[load7], %[cospi_18_64] \n\t" - "madd $ac1, %[load8], %[cospi_14_64] \n\t" - "extp %[result3], $ac1, 31 \n\t" - - "madd $ac2, %[load5], %[cospi_2_64] \n\t" - "madd $ac2, %[load6], %[cospi_30_64] \n\t" - "extp %[result4], $ac2, 31 \n\t" - - "sub %[load5], %[result1], %[result2] \n\t" - "sub %[load6], %[result4], %[result3] \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load6], %[cospi_24_64] \n\t" - "msub $ac1, %[load5], %[cospi_8_64] \n\t" - "madd $ac3, %[load5], %[cospi_24_64] \n\t" - "madd $ac3, %[load6], %[cospi_8_64] \n\t" - - "extp %[step2_9], $ac1, 31 \n\t" - "extp %[step2_14], $ac3, 31 \n\t" - "add %[step2_8], %[result1], %[result2] \n\t" - "add %[step2_15], %[result4], %[result3] \n\t" - - : [load5] "=&r" (load5), [load6] "=&r" (load6), - [load7] "=&r" (load7), [load8] "=&r" (load8), - [result1] "=&r" (result1), [result2] "=&r" (result2), - [result3] "=&r" (result3), [result4] "=&r" (result4), - [step2_8] "=r" (step2_8), [step2_15] "=r" (step2_15), - [step2_9] "=r" (step2_9), [step2_14] "=r" (step2_14) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_30_64] "r" (cospi_30_64), [cospi_2_64] "r" (cospi_2_64), - [cospi_14_64] "r" (cospi_14_64), [cospi_18_64] "r" (cospi_18_64), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) - ); - - __asm__ __volatile__ ( - "lh %[load1], 10(%[input]) \n\t" - "lh %[load2], 22(%[input]) \n\t" - "lh %[load3], 26(%[input]) \n\t" - "lh %[load4], 6(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_22_64] \n\t" - "msub $ac1, %[load2], %[cospi_10_64] \n\t" - "extp %[result1], $ac1, 31 \n\t" - - "madd $ac3, %[load3], %[cospi_6_64] \n\t" - "msub $ac3, %[load4], %[cospi_26_64] \n\t" - "extp %[result2], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac1, %[load1], %[cospi_10_64] \n\t" - "madd $ac1, %[load2], %[cospi_22_64] \n\t" - "extp %[result3], $ac1, 31 \n\t" - - "madd $ac2, %[load3], %[cospi_26_64] \n\t" - "madd $ac2, %[load4], %[cospi_6_64] \n\t" - "extp %[result4], $ac2, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[result2], %[result1] \n\t" - "sub %[load2], %[result4], %[result3] \n\t" - - "msub $ac1, %[load1], %[cospi_24_64] \n\t" - "msub $ac1, %[load2], %[cospi_8_64] \n\t" - "madd $ac3, %[load2], %[cospi_24_64] \n\t" - "msub $ac3, %[load1], %[cospi_8_64] \n\t" - - "extp %[step2_10], $ac1, 31 \n\t" - "extp %[step2_13], $ac3, 31 \n\t" - "add %[step2_11], %[result1], %[result2] \n\t" - "add %[step2_12], %[result4], %[result3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), - [load3] "=&r" (load3), [load4] "=&r" (load4), - [result1] "=&r" (result1), [result2] "=&r" (result2), - [result3] "=&r" (result3), [result4] "=&r" (result4), - [step2_10] "=r" (step2_10), [step2_11] "=r" (step2_11), - [step2_12] "=r" (step2_12), [step2_13] "=r" (step2_13) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_22_64] "r" (cospi_22_64), [cospi_10_64] "r" (cospi_10_64), - [cospi_6_64] "r" (cospi_6_64), [cospi_26_64] "r" (cospi_26_64), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) - ); - - __asm__ __volatile__ ( - "lh %[load5], 4(%[input]) \n\t" - "lh %[load6], 28(%[input]) \n\t" - "lh %[load7], 20(%[input]) \n\t" - "lh %[load8], 12(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load5], %[cospi_28_64] \n\t" - "msub $ac1, %[load6], %[cospi_4_64] \n\t" - "extp %[result1], $ac1, 31 \n\t" - - "madd $ac3, %[load7], %[cospi_12_64] \n\t" - "msub $ac3, %[load8], %[cospi_20_64] \n\t" - "extp %[result2], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac1, %[load7], %[cospi_20_64] \n\t" - "madd $ac1, %[load8], %[cospi_12_64] \n\t" - "extp %[result3], $ac1, 31 \n\t" - - "madd $ac2, %[load5], %[cospi_4_64] \n\t" - "madd $ac2, %[load6], %[cospi_28_64] \n\t" - "extp %[result4], $ac2, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load5], %[result4], %[result3] \n\t" - "sub %[load5], %[load5], %[result1] \n\t" - "add %[load5], %[load5], %[result2] \n\t" - - "sub %[load6], %[result1], %[result2] \n\t" - "sub %[load6], %[load6], %[result3] \n\t" - "add %[load6], %[load6], %[result4] \n\t" - - "madd $ac1, %[load5], %[cospi_16_64] \n\t" - "madd $ac3, %[load6], %[cospi_16_64] \n\t" - - "extp %[step1_5], $ac1, 31 \n\t" - "extp %[step1_6], $ac3, 31 \n\t" - - "add %[step1_4], %[result1], %[result2] \n\t" - "add %[step1_7], %[result4], %[result3] \n\t" - - : [load5] "=&r" (load5), [load6] "=&r" (load6), - [load7] "=&r" (load7), [load8] "=&r" (load8), - [result1] "=&r" (result1), [result2] "=&r" (result2), - [result3] "=&r" (result3), [result4] "=&r" (result4), - [step1_4] "=r" (step1_4), [step1_5] "=r" (step1_5), - [step1_6] "=r" (step1_6), [step1_7] "=r" (step1_7) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_20_64] "r" (cospi_20_64), [cospi_12_64] "r" (cospi_12_64), - [cospi_4_64] "r" (cospi_4_64), [cospi_28_64] "r" (cospi_28_64), - [cospi_16_64] "r" (cospi_16_64) - ); - - __asm__ __volatile__ ( - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - - "sub %[load5], %[step2_14], %[step2_13] \n\t" - "sub %[load5], %[load5], %[step2_9] \n\t" - "add %[load5], %[load5], %[step2_10] \n\t" - - "madd $ac0, %[load5], %[cospi_16_64] \n\t" - - "sub %[load6], %[step2_14], %[step2_13] \n\t" - "sub %[load6], %[load6], %[step2_10] \n\t" - "add %[load6], %[load6], %[step2_9] \n\t" - - "madd $ac1, %[load6], %[cospi_16_64] \n\t" - - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load5], %[step2_15], %[step2_12] \n\t" - "sub %[load5], %[load5], %[step2_8] \n\t" - "add %[load5], %[load5], %[step2_11] \n\t" - - "madd $ac2, %[load5], %[cospi_16_64] \n\t" - - "sub %[load6], %[step2_15], %[step2_12] \n\t" - "sub %[load6], %[load6], %[step2_11] \n\t" - "add %[load6], %[load6], %[step2_8] \n\t" - - "madd $ac3, %[load6], %[cospi_16_64] \n\t" - - "extp %[step1_10], $ac0, 31 \n\t" - "extp %[step1_13], $ac1, 31 \n\t" - "extp %[step1_11], $ac2, 31 \n\t" - "extp %[step1_12], $ac3, 31 \n\t" - - : [load5] "=&r" (load5), [load6] "=&r" (load6), - [step1_10] "=r" (step1_10), [step1_11] "=r" (step1_11), - [step1_12] "=r" (step1_12), [step1_13] "=r" (step1_13) - : [const_2_power_13] "r" (const_2_power_13), - [step2_14] "r" (step2_14), [step2_13] "r" (step2_13), - [step2_9] "r" (step2_9), [step2_10] "r" (step2_10), - [step2_15] "r" (step2_15), [step2_12] "r" (step2_12), - [step2_8] "r" (step2_8), [step2_11] "r" (step2_11), - [cospi_16_64] "r" (cospi_16_64) - ); - - step1_8 = step2_8 + step2_11; - step1_9 = step2_9 + step2_10; - step1_14 = step2_13 + step2_14; - step1_15 = step2_12 + step2_15; - - __asm__ __volatile__ ( - "lbu %[load7], 0(%[dest_pix]) \n\t" - "add %[load5], %[step1_0], %[step1_7] \n\t" - "add %[load5], %[load5], %[step1_15] \n\t" - "addi %[load5], %[load5], 32 \n\t" - "sra %[load5], %[load5], 6 \n\t" - "add %[load7], %[load7], %[load5] \n\t" - "lbux %[load5], %[load7](%[cm]) \n\t" - "add %[load6], %[step1_1], %[step1_6] \n\t" - "add %[load6], %[load6], %[step1_14] \n\t" - "sb %[load5], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[load8], 0(%[dest_pix]) \n\t" - "addi %[load6], %[load6], 32 \n\t" - "sra %[load6], %[load6], 6 \n\t" - "add %[load8], %[load8], %[load6] \n\t" - "lbux %[load6], %[load8](%[cm]) \n\t" - "sb %[load6], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[load7], 0(%[dest_pix]) \n\t" - "add %[load5], %[step1_2], %[step1_5] \n\t" - "add %[load5], %[load5], %[step1_13] \n\t" - "addi %[load5], %[load5], 32 \n\t" - "sra %[load5], %[load5], 6 \n\t" - "add %[load7], %[load7], %[load5] \n\t" - "lbux %[load5], %[load7](%[cm]) \n\t" - "add %[load6], %[step1_3], %[step1_4] \n\t" - "add %[load6], %[load6], %[step1_12] \n\t" - "sb %[load5], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[load8], 0(%[dest_pix]) \n\t" - "addi %[load6], %[load6], 32 \n\t" - "sra %[load6], %[load6], 6 \n\t" - "add %[load8], %[load8], %[load6] \n\t" - "lbux %[load6], %[load8](%[cm]) \n\t" - "sb %[load6], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[load7], 0(%[dest_pix]) \n\t" - "sub %[load5], %[step1_3], %[step1_4] \n\t" - "add %[load5], %[load5], %[step1_11] \n\t" - "addi %[load5], %[load5], 32 \n\t" - "sra %[load5], %[load5], 6 \n\t" - "add %[load7], %[load7], %[load5] \n\t" - "lbux %[load5], %[load7](%[cm]) \n\t" - "sub %[load6], %[step1_2], %[step1_5] \n\t" - "add %[load6], %[load6], %[step1_10] \n\t" - "sb %[load5], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[load8], 0(%[dest_pix]) \n\t" - "addi %[load6], %[load6], 32 \n\t" - "sra %[load6], %[load6], 6 \n\t" - "add %[load8], %[load8], %[load6] \n\t" - "lbux %[load6], %[load8](%[cm]) \n\t" - "sb %[load6], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "sub %[load5], %[step1_1], %[step1_6] \n\t" - "lbu %[load7], 0(%[dest_pix]) \n\t" - "add %[load5], %[load5], %[step1_9] \n\t" - "addi %[load5], %[load5], 32 \n\t" - "sra %[load5], %[load5], 6 \n\t" - "add %[load7], %[load7], %[load5] \n\t" - "lbux %[load5], %[load7](%[cm]) \n\t" - "sub %[load6], %[step1_0], %[step1_7] \n\t" - "add %[load6], %[load6], %[step1_8] \n\t" - "sb %[load5], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[load8], 0(%[dest_pix]) \n\t" - "addi %[load6], %[load6], 32 \n\t" - "sra %[load6], %[load6], 6 \n\t" - "add %[load8], %[load8], %[load6] \n\t" - "lbux %[load6], %[load8](%[cm]) \n\t" - "sb %[load6], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[load7], 0(%[dest_pix]) \n\t" - "sub %[load5], %[step1_0], %[step1_7] \n\t" - "sub %[load5], %[load5], %[step1_8] \n\t" - "addi %[load5], %[load5], 32 \n\t" - "sra %[load5], %[load5], 6 \n\t" - "add %[load7], %[load7], %[load5] \n\t" - "lbux %[load5], %[load7](%[cm]) \n\t" - "sub %[load6], %[step1_1], %[step1_6] \n\t" - "sub %[load6], %[load6], %[step1_9] \n\t" - "sb %[load5], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[load8], 0(%[dest_pix]) \n\t" - "addi %[load6], %[load6], 32 \n\t" - "sra %[load6], %[load6], 6 \n\t" - "add %[load8], %[load8], %[load6] \n\t" - "lbux %[load6], %[load8](%[cm]) \n\t" - "sb %[load6], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[load7], 0(%[dest_pix]) \n\t" - "sub %[load5], %[step1_2], %[step1_5] \n\t" - "sub %[load5], %[load5], %[step1_10] \n\t" - "addi %[load5], %[load5], 32 \n\t" - "sra %[load5], %[load5], 6 \n\t" - "add %[load7], %[load7], %[load5] \n\t" - "lbux %[load5], %[load7](%[cm]) \n\t" - "sub %[load6], %[step1_3], %[step1_4] \n\t" - "sub %[load6], %[load6], %[step1_11] \n\t" - "sb %[load5], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[load8], 0(%[dest_pix]) \n\t" - "addi %[load6], %[load6], 32 \n\t" - "sra %[load6], %[load6], 6 \n\t" - "add %[load8], %[load8], %[load6] \n\t" - "lbux %[load6], %[load8](%[cm]) \n\t" - "sb %[load6], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[load7], 0(%[dest_pix]) \n\t" - "add %[load5], %[step1_3], %[step1_4] \n\t" - "sub %[load5], %[load5], %[step1_12] \n\t" - "addi %[load5], %[load5], 32 \n\t" - "sra %[load5], %[load5], 6 \n\t" - "add %[load7], %[load7], %[load5] \n\t" - "lbux %[load5], %[load7](%[cm]) \n\t" - "add %[load6], %[step1_2], %[step1_5] \n\t" - "sub %[load6], %[load6], %[step1_13] \n\t" - "sb %[load5], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[load8], 0(%[dest_pix]) \n\t" - "addi %[load6], %[load6], 32 \n\t" - "sra %[load6], %[load6], 6 \n\t" - "add %[load8], %[load8], %[load6] \n\t" - "lbux %[load6], %[load8](%[cm]) \n\t" - "sb %[load6], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[load7], 0(%[dest_pix]) \n\t" - "add %[load5], %[step1_1], %[step1_6] \n\t" - "sub %[load5], %[load5], %[step1_14] \n\t" - "addi %[load5], %[load5], 32 \n\t" - "sra %[load5], %[load5], 6 \n\t" - "add %[load7], %[load7], %[load5] \n\t" - "lbux %[load5], %[load7](%[cm]) \n\t" - "add %[load6], %[step1_0], %[step1_7] \n\t" - "sub %[load6], %[load6], %[step1_15] \n\t" - "sb %[load5], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[load8], 0(%[dest_pix]) \n\t" - "addi %[load6], %[load6], 32 \n\t" - "sra %[load6], %[load6], 6 \n\t" - "add %[load8], %[load8], %[load6] \n\t" - "lbux %[load6], %[load8](%[cm]) \n\t" - "sb %[load6], 0(%[dest_pix]) \n\t" - - : [load5] "=&r" (load5), [load6] "=&r" (load6), [load7] "=&r" (load7), - [load8] "=&r" (load8), [dest_pix] "+r" (dest_pix) - : [cm] "r" (cm), [dest_stride] "r" (dest_stride), - [step1_0] "r" (step1_0), [step1_1] "r" (step1_1), - [step1_2] "r" (step1_2), [step1_3] "r" (step1_3), - [step1_4] "r" (step1_4), [step1_5] "r" (step1_5), - [step1_6] "r" (step1_6), [step1_7] "r" (step1_7), - [step1_8] "r" (step1_8), [step1_9] "r" (step1_9), - [step1_10] "r" (step1_10), [step1_11] "r" (step1_11), - [step1_12] "r" (step1_12), [step1_13] "r" (step1_13), - [step1_14] "r" (step1_14), [step1_15] "r" (step1_15) - ); - - input += 16; - } -} - -void vp9_idct16x16_256_add_dspr2(const int16_t *input, uint8_t *dest, - int dest_stride) { - DECLARE_ALIGNED(32, int16_t, out[16 * 16]); - uint32_t pos = 45; - - /* bit positon for extract from acc */ - __asm__ __volatile__ ( - "wrdsp %[pos], 1 \n\t" - : - : [pos] "r" (pos) - ); - - // First transform rows - idct16_rows_dspr2(input, out, 16); - - // Then transform columns and add to dest - idct16_cols_add_blk_dspr2(out, dest, dest_stride); -} - -static void iadst16_dspr2(const int16_t *input, int16_t *output) { - int s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, s13, s14, s15; - - int x0 = input[15]; - int x1 = input[0]; - int x2 = input[13]; - int x3 = input[2]; - int x4 = input[11]; - int x5 = input[4]; - int x6 = input[9]; - int x7 = input[6]; - int x8 = input[7]; - int x9 = input[8]; - int x10 = input[5]; - int x11 = input[10]; - int x12 = input[3]; - int x13 = input[12]; - int x14 = input[1]; - int x15 = input[14]; - - if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7 | x8 - | x9 | x10 | x11 | x12 | x13 | x14 | x15)) { - output[0] = output[1] = output[2] = output[3] = output[4] - = output[5] = output[6] = output[7] = output[8] - = output[9] = output[10] = output[11] = output[12] - = output[13] = output[14] = output[15] = 0; - return; - } - - // stage 1 - s0 = x0 * cospi_1_64 + x1 * cospi_31_64; - s1 = x0 * cospi_31_64 - x1 * cospi_1_64; - s2 = x2 * cospi_5_64 + x3 * cospi_27_64; - s3 = x2 * cospi_27_64 - x3 * cospi_5_64; - s4 = x4 * cospi_9_64 + x5 * cospi_23_64; - s5 = x4 * cospi_23_64 - x5 * cospi_9_64; - s6 = x6 * cospi_13_64 + x7 * cospi_19_64; - s7 = x6 * cospi_19_64 - x7 * cospi_13_64; - s8 = x8 * cospi_17_64 + x9 * cospi_15_64; - s9 = x8 * cospi_15_64 - x9 * cospi_17_64; - s10 = x10 * cospi_21_64 + x11 * cospi_11_64; - s11 = x10 * cospi_11_64 - x11 * cospi_21_64; - s12 = x12 * cospi_25_64 + x13 * cospi_7_64; - s13 = x12 * cospi_7_64 - x13 * cospi_25_64; - s14 = x14 * cospi_29_64 + x15 * cospi_3_64; - s15 = x14 * cospi_3_64 - x15 * cospi_29_64; - - x0 = dct_const_round_shift(s0 + s8); - x1 = dct_const_round_shift(s1 + s9); - x2 = dct_const_round_shift(s2 + s10); - x3 = dct_const_round_shift(s3 + s11); - x4 = dct_const_round_shift(s4 + s12); - x5 = dct_const_round_shift(s5 + s13); - x6 = dct_const_round_shift(s6 + s14); - x7 = dct_const_round_shift(s7 + s15); - x8 = dct_const_round_shift(s0 - s8); - x9 = dct_const_round_shift(s1 - s9); - x10 = dct_const_round_shift(s2 - s10); - x11 = dct_const_round_shift(s3 - s11); - x12 = dct_const_round_shift(s4 - s12); - x13 = dct_const_round_shift(s5 - s13); - x14 = dct_const_round_shift(s6 - s14); - x15 = dct_const_round_shift(s7 - s15); - - // stage 2 - s0 = x0; - s1 = x1; - s2 = x2; - s3 = x3; - s4 = x4; - s5 = x5; - s6 = x6; - s7 = x7; - s8 = x8 * cospi_4_64 + x9 * cospi_28_64; - s9 = x8 * cospi_28_64 - x9 * cospi_4_64; - s10 = x10 * cospi_20_64 + x11 * cospi_12_64; - s11 = x10 * cospi_12_64 - x11 * cospi_20_64; - s12 = - x12 * cospi_28_64 + x13 * cospi_4_64; - s13 = x12 * cospi_4_64 + x13 * cospi_28_64; - s14 = - x14 * cospi_12_64 + x15 * cospi_20_64; - s15 = x14 * cospi_20_64 + x15 * cospi_12_64; - - x0 = s0 + s4; - x1 = s1 + s5; - x2 = s2 + s6; - x3 = s3 + s7; - x4 = s0 - s4; - x5 = s1 - s5; - x6 = s2 - s6; - x7 = s3 - s7; - x8 = dct_const_round_shift(s8 + s12); - x9 = dct_const_round_shift(s9 + s13); - x10 = dct_const_round_shift(s10 + s14); - x11 = dct_const_round_shift(s11 + s15); - x12 = dct_const_round_shift(s8 - s12); - x13 = dct_const_round_shift(s9 - s13); - x14 = dct_const_round_shift(s10 - s14); - x15 = dct_const_round_shift(s11 - s15); - - // stage 3 - s0 = x0; - s1 = x1; - s2 = x2; - s3 = x3; - s4 = x4 * cospi_8_64 + x5 * cospi_24_64; - s5 = x4 * cospi_24_64 - x5 * cospi_8_64; - s6 = - x6 * cospi_24_64 + x7 * cospi_8_64; - s7 = x6 * cospi_8_64 + x7 * cospi_24_64; - s8 = x8; - s9 = x9; - s10 = x10; - s11 = x11; - s12 = x12 * cospi_8_64 + x13 * cospi_24_64; - s13 = x12 * cospi_24_64 - x13 * cospi_8_64; - s14 = - x14 * cospi_24_64 + x15 * cospi_8_64; - s15 = x14 * cospi_8_64 + x15 * cospi_24_64; - - x0 = s0 + s2; - x1 = s1 + s3; - x2 = s0 - s2; - x3 = s1 - s3; - x4 = dct_const_round_shift(s4 + s6); - x5 = dct_const_round_shift(s5 + s7); - x6 = dct_const_round_shift(s4 - s6); - x7 = dct_const_round_shift(s5 - s7); - x8 = s8 + s10; - x9 = s9 + s11; - x10 = s8 - s10; - x11 = s9 - s11; - x12 = dct_const_round_shift(s12 + s14); - x13 = dct_const_round_shift(s13 + s15); - x14 = dct_const_round_shift(s12 - s14); - x15 = dct_const_round_shift(s13 - s15); - - // stage 4 - s2 = (- cospi_16_64) * (x2 + x3); - s3 = cospi_16_64 * (x2 - x3); - s6 = cospi_16_64 * (x6 + x7); - s7 = cospi_16_64 * (- x6 + x7); - s10 = cospi_16_64 * (x10 + x11); - s11 = cospi_16_64 * (- x10 + x11); - s14 = (- cospi_16_64) * (x14 + x15); - s15 = cospi_16_64 * (x14 - x15); - - x2 = dct_const_round_shift(s2); - x3 = dct_const_round_shift(s3); - x6 = dct_const_round_shift(s6); - x7 = dct_const_round_shift(s7); - x10 = dct_const_round_shift(s10); - x11 = dct_const_round_shift(s11); - x14 = dct_const_round_shift(s14); - x15 = dct_const_round_shift(s15); - - output[0] = x0; - output[1] = -x8; - output[2] = x12; - output[3] = -x4; - output[4] = x6; - output[5] = x14; - output[6] = x10; - output[7] = x2; - output[8] = x3; - output[9] = x11; - output[10] = x15; - output[11] = x7; - output[12] = x5; - output[13] = -x13; - output[14] = x9; - output[15] = -x1; -} - void vp9_iht16x16_256_add_dspr2(const int16_t *input, uint8_t *dest, int pitch, int tx_type) { int i, j; @@ -1168,150 +105,4 @@ void vp9_iht16x16_256_add_dspr2(const int16_t *input, uint8_t *dest, break; } } - -void vp9_idct16x16_10_add_dspr2(const int16_t *input, uint8_t *dest, - int dest_stride) { - DECLARE_ALIGNED(32, int16_t, out[16 * 16]); - int16_t *outptr = out; - uint32_t i; - uint32_t pos = 45; - - /* bit positon for extract from acc */ - __asm__ __volatile__ ( - "wrdsp %[pos], 1 \n\t" - : - : [pos] "r" (pos) - ); - - // First transform rows. Since all non-zero dct coefficients are in - // upper-left 4x4 area, we only need to calculate first 4 rows here. - idct16_rows_dspr2(input, outptr, 4); - - outptr += 4; - for (i = 0; i < 6; ++i) { - __asm__ __volatile__ ( - "sw $zero, 0(%[outptr]) \n\t" - "sw $zero, 32(%[outptr]) \n\t" - "sw $zero, 64(%[outptr]) \n\t" - "sw $zero, 96(%[outptr]) \n\t" - "sw $zero, 128(%[outptr]) \n\t" - "sw $zero, 160(%[outptr]) \n\t" - "sw $zero, 192(%[outptr]) \n\t" - "sw $zero, 224(%[outptr]) \n\t" - "sw $zero, 256(%[outptr]) \n\t" - "sw $zero, 288(%[outptr]) \n\t" - "sw $zero, 320(%[outptr]) \n\t" - "sw $zero, 352(%[outptr]) \n\t" - "sw $zero, 384(%[outptr]) \n\t" - "sw $zero, 416(%[outptr]) \n\t" - "sw $zero, 448(%[outptr]) \n\t" - "sw $zero, 480(%[outptr]) \n\t" - - : - : [outptr] "r" (outptr) - ); - - outptr += 2; - } - - // Then transform columns - idct16_cols_add_blk_dspr2(out, dest, dest_stride); -} - -void vp9_idct16x16_1_add_dspr2(const int16_t *input, uint8_t *dest, - int dest_stride) { - uint32_t pos = 45; - int32_t out; - int32_t r; - int32_t a1, absa1; - int32_t vector_a1; - int32_t t1, t2, t3, t4; - int32_t vector_1, vector_2, vector_3, vector_4; - - /* bit positon for extract from acc */ - __asm__ __volatile__ ( - "wrdsp %[pos], 1 \n\t" - - : - : [pos] "r" (pos) - ); - - out = DCT_CONST_ROUND_SHIFT_TWICE_COSPI_16_64(input[0]); - __asm__ __volatile__ ( - "addi %[out], %[out], 32 \n\t" - "sra %[a1], %[out], 6 \n\t" - - : [out] "+r" (out), [a1] "=r" (a1) - : - ); - - if (a1 < 0) { - /* use quad-byte - * input and output memory are four byte aligned */ - __asm__ __volatile__ ( - "abs %[absa1], %[a1] \n\t" - "replv.qb %[vector_a1], %[absa1] \n\t" - - : [absa1] "=r" (absa1), [vector_a1] "=r" (vector_a1) - : [a1] "r" (a1) - ); - - for (r = 16; r--;) { - __asm__ __volatile__ ( - "lw %[t1], 0(%[dest]) \n\t" - "lw %[t2], 4(%[dest]) \n\t" - "lw %[t3], 8(%[dest]) \n\t" - "lw %[t4], 12(%[dest]) \n\t" - "subu_s.qb %[vector_1], %[t1], %[vector_a1] \n\t" - "subu_s.qb %[vector_2], %[t2], %[vector_a1] \n\t" - "subu_s.qb %[vector_3], %[t3], %[vector_a1] \n\t" - "subu_s.qb %[vector_4], %[t4], %[vector_a1] \n\t" - "sw %[vector_1], 0(%[dest]) \n\t" - "sw %[vector_2], 4(%[dest]) \n\t" - "sw %[vector_3], 8(%[dest]) \n\t" - "sw %[vector_4], 12(%[dest]) \n\t" - "add %[dest], %[dest], %[dest_stride] \n\t" - - : [t1] "=&r" (t1), [t2] "=&r" (t2), [t3] "=&r" (t3), [t4] "=&r" (t4), - [vector_1] "=&r" (vector_1), [vector_2] "=&r" (vector_2), - [vector_3] "=&r" (vector_3), [vector_4] "=&r" (vector_4), - [dest] "+&r" (dest) - : [dest_stride] "r" (dest_stride), [vector_a1] "r" (vector_a1) - ); - } - } else { - /* use quad-byte - * input and output memory are four byte aligned */ - __asm__ __volatile__ ( - "replv.qb %[vector_a1], %[a1] \n\t" - - : [vector_a1] "=r" (vector_a1) - : [a1] "r" (a1) - ); - - for (r = 16; r--;) { - __asm__ __volatile__ ( - "lw %[t1], 0(%[dest]) \n\t" - "lw %[t2], 4(%[dest]) \n\t" - "lw %[t3], 8(%[dest]) \n\t" - "lw %[t4], 12(%[dest]) \n\t" - "addu_s.qb %[vector_1], %[t1], %[vector_a1] \n\t" - "addu_s.qb %[vector_2], %[t2], %[vector_a1] \n\t" - "addu_s.qb %[vector_3], %[t3], %[vector_a1] \n\t" - "addu_s.qb %[vector_4], %[t4], %[vector_a1] \n\t" - "sw %[vector_1], 0(%[dest]) \n\t" - "sw %[vector_2], 4(%[dest]) \n\t" - "sw %[vector_3], 8(%[dest]) \n\t" - "sw %[vector_4], 12(%[dest]) \n\t" - "add %[dest], %[dest], %[dest_stride] \n\t" - - : [t1] "=&r" (t1), [t2] "=&r" (t2), [t3] "=&r" (t3), [t4] "=&r" (t4), - [vector_1] "=&r" (vector_1), [vector_2] "=&r" (vector_2), - [vector_3] "=&r" (vector_3), [vector_4] "=&r" (vector_4), - [dest] "+&r" (dest) - : [dest_stride] "r" (dest_stride), [vector_a1] "r" (vector_a1) - ); - } - } -} #endif // #if HAVE_DSPR2 diff --git a/vp9/common/mips/dspr2/vp9_itrans32_cols_dspr2.c b/vp9/common/mips/dspr2/vp9_itrans32_cols_dspr2.c deleted file mode 100644 index 48da85cbc..000000000 --- a/vp9/common/mips/dspr2/vp9_itrans32_cols_dspr2.c +++ /dev/null @@ -1,1074 +0,0 @@ -/* - * Copyright (c) 2013 The WebM project authors. All Rights Reserved. - * - * Use of this source code is governed by a BSD-style license - * that can be found in the LICENSE file in the root of the source - * tree. An additional intellectual property rights grant can be found - * in the file PATENTS. All contributing project authors may - * be found in the AUTHORS file in the root of the source tree. - */ - -#include <assert.h> - -#include "./vpx_config.h" -#include "./vp9_rtcd.h" -#include "vp9/common/vp9_common.h" -#include "vp9/common/vp9_blockd.h" -#include "vp9/common/mips/dspr2/vp9_common_dspr2.h" -#include "vpx_dsp/txfm_common.h" -#include "vpx_ports/mem.h" - -#if HAVE_DSPR2 -void vp9_idct32_cols_add_blk_dspr2(int16_t *input, uint8_t *dest, - int dest_stride) { - int16_t step1_0, step1_1, step1_2, step1_3, step1_4, step1_5, step1_6; - int16_t step1_7, step1_8, step1_9, step1_10, step1_11, step1_12, step1_13; - int16_t step1_14, step1_15, step1_16, step1_17, step1_18, step1_19; - int16_t step1_20, step1_21, step1_22, step1_23, step1_24, step1_25, step1_26; - int16_t step1_27, step1_28, step1_29, step1_30, step1_31; - int16_t step2_0, step2_1, step2_2, step2_3, step2_4, step2_5, step2_6; - int16_t step2_7, step2_8, step2_9, step2_10, step2_11, step2_12, step2_13; - int16_t step2_14, step2_15, step2_16, step2_17, step2_18, step2_19, step2_20; - int16_t step2_21, step2_22, step2_23, step2_24, step2_25, step2_26, step2_27; - int16_t step2_28, step2_29, step2_30, step2_31; - int16_t step3_8, step3_9, step3_10, step3_11, step3_12, step3_13, step3_14; - int16_t step3_15, step3_16, step3_17, step3_18, step3_19, step3_20, step3_21; - int16_t step3_22, step3_23, step3_24, step3_25, step3_26, step3_27; - int16_t step3_28, step3_29, step3_30, step3_31; - int temp0, temp1, temp2, temp3; - int load1, load2, load3, load4; - int result1, result2; - int i, temp21; - uint8_t *dest_pix, *dest_pix1; - const int const_2_power_13 = 8192; - uint8_t *cm = vpx_ff_cropTbl; - - /* prefetch vpx_ff_cropTbl */ - prefetch_load(vpx_ff_cropTbl); - prefetch_load(vpx_ff_cropTbl + 32); - prefetch_load(vpx_ff_cropTbl + 64); - prefetch_load(vpx_ff_cropTbl + 96); - prefetch_load(vpx_ff_cropTbl + 128); - prefetch_load(vpx_ff_cropTbl + 160); - prefetch_load(vpx_ff_cropTbl + 192); - prefetch_load(vpx_ff_cropTbl + 224); - - for (i = 0; i < 32; ++i) { - dest_pix = dest + i; - dest_pix1 = dest + i + 31 * dest_stride; - - __asm__ __volatile__ ( - "lh %[load1], 2(%[input]) \n\t" - "lh %[load2], 62(%[input]) \n\t" - "lh %[load3], 34(%[input]) \n\t" - "lh %[load4], 30(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_31_64] \n\t" - "msub $ac1, %[load2], %[cospi_1_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - - "madd $ac3, %[load1], %[cospi_1_64] \n\t" - "madd $ac3, %[load2], %[cospi_31_64] \n\t" - "extp %[temp3], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac2, %[load3], %[cospi_15_64] \n\t" - "msub $ac2, %[load4], %[cospi_17_64] \n\t" - "extp %[temp1], $ac2, 31 \n\t" - - "madd $ac1, %[load3], %[cospi_17_64] \n\t" - "madd $ac1, %[load4], %[cospi_15_64] \n\t" - "extp %[temp2], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[temp3], %[temp2] \n\t" - "sub %[load2], %[temp0], %[temp1] \n\t" - - "madd $ac1, %[load1], %[cospi_28_64] \n\t" - "msub $ac1, %[load2], %[cospi_4_64] \n\t" - "madd $ac3, %[load1], %[cospi_4_64] \n\t" - "madd $ac3, %[load2], %[cospi_28_64] \n\t" - - "extp %[step1_17], $ac1, 31 \n\t" - "extp %[step1_30], $ac3, 31 \n\t" - "add %[step1_16], %[temp0], %[temp1] \n\t" - "add %[step1_31], %[temp2], %[temp3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3), - [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step1_16] "=r" (step1_16), [step1_17] "=r" (step1_17), - [step1_30] "=r" (step1_30), [step1_31] "=r" (step1_31) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_31_64] "r" (cospi_31_64), [cospi_1_64] "r" (cospi_1_64), - [cospi_4_64] "r" (cospi_4_64), [cospi_17_64] "r" (cospi_17_64), - [cospi_15_64] "r" (cospi_15_64), [cospi_28_64] "r" (cospi_28_64) - ); - - __asm__ __volatile__ ( - "lh %[load1], 18(%[input]) \n\t" - "lh %[load2], 46(%[input]) \n\t" - "lh %[load3], 50(%[input]) \n\t" - "lh %[load4], 14(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_23_64] \n\t" - "msub $ac1, %[load2], %[cospi_9_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - - "madd $ac3, %[load1], %[cospi_9_64] \n\t" - "madd $ac3, %[load2], %[cospi_23_64] \n\t" - "extp %[temp3], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac2, %[load3], %[cospi_7_64] \n\t" - "msub $ac2, %[load4], %[cospi_25_64] \n\t" - "extp %[temp1], $ac2, 31 \n\t" - - "madd $ac1, %[load3], %[cospi_25_64] \n\t" - "madd $ac1, %[load4], %[cospi_7_64] \n\t" - "extp %[temp2], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[temp1], %[temp0] \n\t" - "sub %[load2], %[temp2], %[temp3] \n\t" - - "msub $ac1, %[load1], %[cospi_28_64] \n\t" - "msub $ac1, %[load2], %[cospi_4_64] \n\t" - "msub $ac3, %[load1], %[cospi_4_64] \n\t" - "madd $ac3, %[load2], %[cospi_28_64] \n\t" - - "extp %[step1_18], $ac1, 31 \n\t" - "extp %[step1_29], $ac3, 31 \n\t" - "add %[step1_19], %[temp0], %[temp1] \n\t" - "add %[step1_28], %[temp2], %[temp3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3), - [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step1_18] "=r" (step1_18), [step1_19] "=r" (step1_19), - [step1_28] "=r" (step1_28), [step1_29] "=r" (step1_29) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_23_64] "r" (cospi_23_64), [cospi_9_64] "r" (cospi_9_64), - [cospi_4_64] "r" (cospi_4_64), [cospi_7_64] "r" (cospi_7_64), - [cospi_25_64] "r" (cospi_25_64), [cospi_28_64] "r" (cospi_28_64) - ); - - __asm__ __volatile__ ( - "lh %[load1], 10(%[input]) \n\t" - "lh %[load2], 54(%[input]) \n\t" - "lh %[load3], 42(%[input]) \n\t" - "lh %[load4], 22(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_27_64] \n\t" - "msub $ac1, %[load2], %[cospi_5_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - - "madd $ac3, %[load1], %[cospi_5_64] \n\t" - "madd $ac3, %[load2], %[cospi_27_64] \n\t" - "extp %[temp3], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac2, %[load3], %[cospi_11_64] \n\t" - "msub $ac2, %[load4], %[cospi_21_64] \n\t" - "extp %[temp1], $ac2, 31 \n\t" - - "madd $ac1, %[load3], %[cospi_21_64] \n\t" - "madd $ac1, %[load4], %[cospi_11_64] \n\t" - "extp %[temp2], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[temp0], %[temp1] \n\t" - "sub %[load2], %[temp3], %[temp2] \n\t" - - "madd $ac1, %[load2], %[cospi_12_64] \n\t" - "msub $ac1, %[load1], %[cospi_20_64] \n\t" - "madd $ac3, %[load1], %[cospi_12_64] \n\t" - "madd $ac3, %[load2], %[cospi_20_64] \n\t" - - "extp %[step1_21], $ac1, 31 \n\t" - "extp %[step1_26], $ac3, 31 \n\t" - "add %[step1_20], %[temp0], %[temp1] \n\t" - "add %[step1_27], %[temp2], %[temp3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3), - [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step1_20] "=r" (step1_20), [step1_21] "=r" (step1_21), - [step1_26] "=r" (step1_26), [step1_27] "=r" (step1_27) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_27_64] "r" (cospi_27_64), [cospi_5_64] "r" (cospi_5_64), - [cospi_11_64] "r" (cospi_11_64), [cospi_21_64] "r" (cospi_21_64), - [cospi_12_64] "r" (cospi_12_64), [cospi_20_64] "r" (cospi_20_64) - ); - - __asm__ __volatile__ ( - "lh %[load1], 26(%[input]) \n\t" - "lh %[load2], 38(%[input]) \n\t" - "lh %[load3], 58(%[input]) \n\t" - "lh %[load4], 6(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_19_64] \n\t" - "msub $ac1, %[load2], %[cospi_13_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - "madd $ac3, %[load1], %[cospi_13_64] \n\t" - "madd $ac3, %[load2], %[cospi_19_64] \n\t" - "extp %[temp3], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac2, %[load3], %[cospi_3_64] \n\t" - "msub $ac2, %[load4], %[cospi_29_64] \n\t" - "extp %[temp1], $ac2, 31 \n\t" - "madd $ac1, %[load3], %[cospi_29_64] \n\t" - "madd $ac1, %[load4], %[cospi_3_64] \n\t" - "extp %[temp2], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[temp1], %[temp0] \n\t" - "sub %[load2], %[temp2], %[temp3] \n\t" - "msub $ac1, %[load1], %[cospi_12_64] \n\t" - "msub $ac1, %[load2], %[cospi_20_64] \n\t" - "msub $ac3, %[load1], %[cospi_20_64] \n\t" - "madd $ac3, %[load2], %[cospi_12_64] \n\t" - "extp %[step1_22], $ac1, 31 \n\t" - "extp %[step1_25], $ac3, 31 \n\t" - "add %[step1_23], %[temp0], %[temp1] \n\t" - "add %[step1_24], %[temp2], %[temp3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3), - [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step1_22] "=r" (step1_22), [step1_23] "=r" (step1_23), - [step1_24] "=r" (step1_24), [step1_25] "=r" (step1_25) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_19_64] "r" (cospi_19_64), [cospi_13_64] "r" (cospi_13_64), - [cospi_3_64] "r" (cospi_3_64), [cospi_29_64] "r" (cospi_29_64), - [cospi_12_64] "r" (cospi_12_64), [cospi_20_64] "r" (cospi_20_64) - ); - - __asm__ __volatile__ ( - "lh %[load1], 4(%[input]) \n\t" - "lh %[load2], 60(%[input]) \n\t" - "lh %[load3], 36(%[input]) \n\t" - "lh %[load4], 28(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_30_64] \n\t" - "msub $ac1, %[load2], %[cospi_2_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - "madd $ac3, %[load1], %[cospi_2_64] \n\t" - "madd $ac3, %[load2], %[cospi_30_64] \n\t" - "extp %[temp3], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac2, %[load3], %[cospi_14_64] \n\t" - "msub $ac2, %[load4], %[cospi_18_64] \n\t" - "extp %[temp1], $ac2, 31 \n\t" - "madd $ac1, %[load3], %[cospi_18_64] \n\t" - "madd $ac1, %[load4], %[cospi_14_64] \n\t" - "extp %[temp2], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[temp0], %[temp1] \n\t" - "sub %[load2], %[temp3], %[temp2] \n\t" - "msub $ac1, %[load1], %[cospi_8_64] \n\t" - "madd $ac1, %[load2], %[cospi_24_64] \n\t" - "madd $ac3, %[load1], %[cospi_24_64] \n\t" - "madd $ac3, %[load2], %[cospi_8_64] \n\t" - "extp %[step2_9], $ac1, 31 \n\t" - "extp %[step2_14], $ac3, 31 \n\t" - "add %[step2_8], %[temp0], %[temp1] \n\t" - "add %[step2_15], %[temp2], %[temp3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3), - [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step2_8] "=r" (step2_8), [step2_9] "=r" (step2_9), - [step2_14] "=r" (step2_14), [step2_15] "=r" (step2_15) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_30_64] "r" (cospi_30_64), [cospi_2_64] "r" (cospi_2_64), - [cospi_14_64] "r" (cospi_14_64), [cospi_18_64] "r" (cospi_18_64), - [cospi_8_64] "r" (cospi_8_64), [cospi_24_64] "r" (cospi_24_64) - ); - - __asm__ __volatile__ ( - "lh %[load1], 20(%[input]) \n\t" - "lh %[load2], 44(%[input]) \n\t" - "lh %[load3], 52(%[input]) \n\t" - "lh %[load4], 12(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_22_64] \n\t" - "msub $ac1, %[load2], %[cospi_10_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - "madd $ac3, %[load1], %[cospi_10_64] \n\t" - "madd $ac3, %[load2], %[cospi_22_64] \n\t" - "extp %[temp3], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac2, %[load3], %[cospi_6_64] \n\t" - "msub $ac2, %[load4], %[cospi_26_64] \n\t" - "extp %[temp1], $ac2, 31 \n\t" - "madd $ac1, %[load3], %[cospi_26_64] \n\t" - "madd $ac1, %[load4], %[cospi_6_64] \n\t" - "extp %[temp2], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[temp1], %[temp0] \n\t" - "sub %[load2], %[temp2], %[temp3] \n\t" - "msub $ac1, %[load1], %[cospi_24_64] \n\t" - "msub $ac1, %[load2], %[cospi_8_64] \n\t" - "madd $ac3, %[load2], %[cospi_24_64] \n\t" - "msub $ac3, %[load1], %[cospi_8_64] \n\t" - "extp %[step2_10], $ac1, 31 \n\t" - "extp %[step2_13], $ac3, 31 \n\t" - "add %[step2_11], %[temp0], %[temp1] \n\t" - "add %[step2_12], %[temp2], %[temp3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), [load3] "=&r" (load3), - [load4] "=&r" (load4), [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step2_10] "=r" (step2_10), [step2_11] "=r" (step2_11), - [step2_12] "=r" (step2_12), [step2_13] "=r" (step2_13) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_22_64] "r" (cospi_22_64), [cospi_10_64] "r" (cospi_10_64), - [cospi_6_64] "r" (cospi_6_64), [cospi_26_64] "r" (cospi_26_64), - [cospi_8_64] "r" (cospi_8_64), [cospi_24_64] "r" (cospi_24_64) - ); - - __asm__ __volatile__ ( - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "sub %[temp0], %[step2_14], %[step2_13] \n\t" - "sub %[temp0], %[temp0], %[step2_9] \n\t" - "add %[temp0], %[temp0], %[step2_10] \n\t" - "madd $ac0, %[temp0], %[cospi_16_64] \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "sub %[temp1], %[step2_14], %[step2_13] \n\t" - "add %[temp1], %[temp1], %[step2_9] \n\t" - "sub %[temp1], %[temp1], %[step2_10] \n\t" - "madd $ac1, %[temp1], %[cospi_16_64] \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - "sub %[temp0], %[step2_15], %[step2_12] \n\t" - "sub %[temp0], %[temp0], %[step2_8] \n\t" - "add %[temp0], %[temp0], %[step2_11] \n\t" - "madd $ac2, %[temp0], %[cospi_16_64] \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - "sub %[temp1], %[step2_15], %[step2_12] \n\t" - "add %[temp1], %[temp1], %[step2_8] \n\t" - "sub %[temp1], %[temp1], %[step2_11] \n\t" - "madd $ac3, %[temp1], %[cospi_16_64] \n\t" - - "add %[step3_8], %[step2_8], %[step2_11] \n\t" - "add %[step3_9], %[step2_9], %[step2_10] \n\t" - "add %[step3_14], %[step2_13], %[step2_14] \n\t" - "add %[step3_15], %[step2_12], %[step2_15] \n\t" - "extp %[step3_10], $ac0, 31 \n\t" - "extp %[step3_13], $ac1, 31 \n\t" - "extp %[step3_11], $ac2, 31 \n\t" - "extp %[step3_12], $ac3, 31 \n\t" - - : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [step3_8] "=r" (step3_8), [step3_9] "=r" (step3_9), - [step3_10] "=r" (step3_10), [step3_11] "=r" (step3_11), - [step3_12] "=r" (step3_12), [step3_13] "=r" (step3_13), - [step3_14] "=r" (step3_14), [step3_15] "=r" (step3_15) - : [const_2_power_13] "r" (const_2_power_13), [step2_8] "r" (step2_8), - [step2_9] "r" (step2_9), [step2_10] "r" (step2_10), - [step2_11] "r" (step2_11), [step2_12] "r" (step2_12), - [step2_13] "r" (step2_13), [step2_14] "r" (step2_14), - [step2_15] "r" (step2_15), [cospi_16_64] "r" (cospi_16_64) - ); - - step2_18 = step1_17 - step1_18; - step2_29 = step1_30 - step1_29; - - __asm__ __volatile__ ( - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "msub $ac0, %[step2_18], %[cospi_8_64] \n\t" - "madd $ac0, %[step2_29], %[cospi_24_64] \n\t" - "extp %[step3_18], $ac0, 31 \n\t" - - : [step3_18] "=r" (step3_18) - : [const_2_power_13] "r" (const_2_power_13), - [step2_18] "r" (step2_18), [step2_29] "r" (step2_29), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) - ); - - temp21 = step2_18 * cospi_24_64 + step2_29 * cospi_8_64; - step3_29 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - step2_19 = step1_16 - step1_19; - step2_28 = step1_31 - step1_28; - - __asm__ __volatile__ ( - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "msub $ac0, %[step2_19], %[cospi_8_64] \n\t" - "madd $ac0, %[step2_28], %[cospi_24_64] \n\t" - "extp %[step3_19], $ac0, 31 \n\t" - - : [step3_19] "=r" (step3_19) - : [const_2_power_13] "r" (const_2_power_13), - [step2_19] "r" (step2_19), [step2_28] "r" (step2_28), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) - ); - - temp21 = step2_19 * cospi_24_64 + step2_28 * cospi_8_64; - step3_28 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - step3_16 = step1_16 + step1_19; - step3_17 = step1_17 + step1_18; - step3_30 = step1_29 + step1_30; - step3_31 = step1_28 + step1_31; - - step2_20 = step1_23 - step1_20; - step2_27 = step1_24 - step1_27; - - __asm__ __volatile__ ( - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "msub $ac0, %[step2_20], %[cospi_24_64] \n\t" - "msub $ac0, %[step2_27], %[cospi_8_64] \n\t" - "extp %[step3_20], $ac0, 31 \n\t" - - : [step3_20] "=r" (step3_20) - : [const_2_power_13] "r" (const_2_power_13), - [step2_20] "r" (step2_20), [step2_27] "r" (step2_27), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) - ); - - temp21 = -step2_20 * cospi_8_64 + step2_27 * cospi_24_64; - step3_27 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - step2_21 = step1_22 - step1_21; - step2_26 = step1_25 - step1_26; - - __asm__ __volatile__ ( - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "msub $ac1, %[step2_21], %[cospi_24_64] \n\t" - "msub $ac1, %[step2_26], %[cospi_8_64] \n\t" - "extp %[step3_21], $ac1, 31 \n\t" - - : [step3_21] "=r" (step3_21) - : [const_2_power_13] "r" (const_2_power_13), - [step2_21] "r" (step2_21), [step2_26] "r" (step2_26), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) - ); - - temp21 = -step2_21 * cospi_8_64 + step2_26 * cospi_24_64; - step3_26 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - step3_22 = step1_21 + step1_22; - step3_23 = step1_20 + step1_23; - step3_24 = step1_24 + step1_27; - step3_25 = step1_25 + step1_26; - - step2_16 = step3_16 + step3_23; - step2_17 = step3_17 + step3_22; - step2_18 = step3_18 + step3_21; - step2_19 = step3_19 + step3_20; - step2_20 = step3_19 - step3_20; - step2_21 = step3_18 - step3_21; - step2_22 = step3_17 - step3_22; - step2_23 = step3_16 - step3_23; - - step2_24 = step3_31 - step3_24; - step2_25 = step3_30 - step3_25; - step2_26 = step3_29 - step3_26; - step2_27 = step3_28 - step3_27; - step2_28 = step3_28 + step3_27; - step2_29 = step3_29 + step3_26; - step2_30 = step3_30 + step3_25; - step2_31 = step3_31 + step3_24; - - __asm__ __volatile__ ( - "lh %[load1], 0(%[input]) \n\t" - "lh %[load2], 32(%[input]) \n\t" - "lh %[load3], 16(%[input]) \n\t" - "lh %[load4], 48(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - "add %[result1], %[load1], %[load2] \n\t" - "sub %[result2], %[load1], %[load2] \n\t" - "madd $ac1, %[result1], %[cospi_16_64] \n\t" - "madd $ac2, %[result2], %[cospi_16_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - "extp %[temp1], $ac2, 31 \n\t" - - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - "madd $ac3, %[load3], %[cospi_24_64] \n\t" - "msub $ac3, %[load4], %[cospi_8_64] \n\t" - "extp %[temp2], $ac3, 31 \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "madd $ac1, %[load3], %[cospi_8_64] \n\t" - "madd $ac1, %[load4], %[cospi_24_64] \n\t" - "extp %[temp3], $ac1, 31 \n\t" - "add %[step1_0], %[temp0], %[temp3] \n\t" - "add %[step1_1], %[temp1], %[temp2] \n\t" - "sub %[step1_2], %[temp1], %[temp2] \n\t" - "sub %[step1_3], %[temp0], %[temp3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), - [load3] "=&r" (load3), [load4] "=&r" (load4), - [result1] "=&r" (result1), [result2] "=&r" (result2), - [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step1_0] "=r" (step1_0), [step1_1] "=r" (step1_1), - [step1_2] "=r" (step1_2), [step1_3] "=r" (step1_3) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64), - [cospi_16_64] "r" (cospi_16_64) - ); - - __asm__ __volatile__ ( - "lh %[load1], 8(%[input]) \n\t" - "lh %[load2], 56(%[input]) \n\t" - "lh %[load3], 40(%[input]) \n\t" - "lh %[load4], 24(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_28_64] \n\t" - "msub $ac1, %[load2], %[cospi_4_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - "madd $ac3, %[load1], %[cospi_4_64] \n\t" - "madd $ac3, %[load2], %[cospi_28_64] \n\t" - "extp %[temp3], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac2, %[load3], %[cospi_12_64] \n\t" - "msub $ac2, %[load4], %[cospi_20_64] \n\t" - "extp %[temp1], $ac2, 31 \n\t" - "madd $ac1, %[load3], %[cospi_20_64] \n\t" - "madd $ac1, %[load4], %[cospi_12_64] \n\t" - "extp %[temp2], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[temp3], %[temp2] \n\t" - "sub %[load1], %[load1], %[temp0] \n\t" - "add %[load1], %[load1], %[temp1] \n\t" - "sub %[load2], %[temp0], %[temp1] \n\t" - "sub %[load2], %[load2], %[temp2] \n\t" - "add %[load2], %[load2], %[temp3] \n\t" - "madd $ac1, %[load1], %[cospi_16_64] \n\t" - "madd $ac3, %[load2], %[cospi_16_64] \n\t" - - "extp %[step1_5], $ac1, 31 \n\t" - "extp %[step1_6], $ac3, 31 \n\t" - "add %[step1_4], %[temp0], %[temp1] \n\t" - "add %[step1_7], %[temp3], %[temp2] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), - [load3] "=&r" (load3), [load4] "=&r" (load4), - [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step1_4] "=r" (step1_4), [step1_5] "=r" (step1_5), - [step1_6] "=r" (step1_6), [step1_7] "=r" (step1_7) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_20_64] "r" (cospi_20_64), [cospi_12_64] "r" (cospi_12_64), - [cospi_4_64] "r" (cospi_4_64), [cospi_28_64] "r" (cospi_28_64), - [cospi_16_64] "r" (cospi_16_64) - ); - - step2_0 = step1_0 + step1_7; - step2_1 = step1_1 + step1_6; - step2_2 = step1_2 + step1_5; - step2_3 = step1_3 + step1_4; - step2_4 = step1_3 - step1_4; - step2_5 = step1_2 - step1_5; - step2_6 = step1_1 - step1_6; - step2_7 = step1_0 - step1_7; - - // stage 7 - step1_0 = step2_0 + step3_15; - step1_1 = step2_1 + step3_14; - step1_2 = step2_2 + step3_13; - step1_3 = step2_3 + step3_12; - step1_4 = step2_4 + step3_11; - step1_5 = step2_5 + step3_10; - step1_6 = step2_6 + step3_9; - step1_7 = step2_7 + step3_8; - step1_8 = step2_7 - step3_8; - step1_9 = step2_6 - step3_9; - step1_10 = step2_5 - step3_10; - step1_11 = step2_4 - step3_11; - step1_12 = step2_3 - step3_12; - step1_13 = step2_2 - step3_13; - step1_14 = step2_1 - step3_14; - step1_15 = step2_0 - step3_15; - - __asm__ __volatile__ ( - "sub %[temp0], %[step2_27], %[step2_20] \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "madd $ac0, %[temp0], %[cospi_16_64] \n\t" - "extp %[step1_20], $ac0, 31 \n\t" - - : [temp0] "=&r" (temp0), [step1_20] "=r" (step1_20) - : [const_2_power_13] "r" (const_2_power_13), [step2_20] "r" (step2_20), - [step2_27] "r" (step2_27), [cospi_16_64] "r" (cospi_16_64) - ); - - temp21 = (step2_20 + step2_27) * cospi_16_64; - step1_27 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - __asm__ __volatile__ ( - "sub %[temp0], %[step2_26], %[step2_21] \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "madd $ac0, %[temp0], %[cospi_16_64] \n\t" - "extp %[step1_21], $ac0, 31 \n\t" - - : [temp0] "=&r" (temp0), [step1_21] "=r" (step1_21) - : [const_2_power_13] "r" (const_2_power_13), [step2_26] "r" (step2_26), - [step2_21] "r" (step2_21), [cospi_16_64] "r" (cospi_16_64) - ); - - temp21 = (step2_21 + step2_26) * cospi_16_64; - step1_26 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - __asm__ __volatile__ ( - "sub %[temp0], %[step2_25], %[step2_22] \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "madd $ac0, %[temp0], %[cospi_16_64] \n\t" - "extp %[step1_22], $ac0, 31 \n\t" - - : [temp0] "=&r" (temp0), [step1_22] "=r" (step1_22) - : [const_2_power_13] "r" (const_2_power_13), [step2_25] "r" (step2_25), - [step2_22] "r" (step2_22), [cospi_16_64] "r" (cospi_16_64) - ); - - temp21 = (step2_22 + step2_25) * cospi_16_64; - step1_25 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - __asm__ __volatile__ ( - "sub %[temp0], %[step2_24], %[step2_23] \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "madd $ac0, %[temp0], %[cospi_16_64] \n\t" - "extp %[step1_23], $ac0, 31 \n\t" - - : [temp0] "=&r" (temp0), [step1_23] "=r" (step1_23) - : [const_2_power_13] "r" (const_2_power_13), [step2_24] "r" (step2_24), - [step2_23] "r" (step2_23), [cospi_16_64] "r" (cospi_16_64) - ); - - temp21 = (step2_23 + step2_24) * cospi_16_64; - step1_24 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - __asm__ __volatile__ ( - "lbu %[temp2], 0(%[dest_pix]) \n\t" - "add %[temp0], %[step1_0], %[step2_31] \n\t" - "addi %[temp0], %[temp0], 32 \n\t" - "sra %[temp0], %[temp0], 6 \n\t" - "add %[temp2], %[temp2], %[temp0] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "add %[temp1], %[step1_1], %[step2_30] \n\t" - "sb %[temp0], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix]) \n\t" - "addi %[temp1], %[temp1], 32 \n\t" - "sra %[temp1], %[temp1], 6 \n\t" - "add %[temp3], %[temp3], %[temp1] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[temp2], 0(%[dest_pix]) \n\t" - "add %[temp0], %[step1_2], %[step2_29] \n\t" - "addi %[temp0], %[temp0], 32 \n\t" - "sra %[temp0], %[temp0], 6 \n\t" - "add %[temp2], %[temp2], %[temp0] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "add %[temp1], %[step1_3], %[step2_28] \n\t" - "sb %[temp0], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix]) \n\t" - "addi %[temp1], %[temp1], 32 \n\t" - "sra %[temp1], %[temp1], 6 \n\t" - "add %[temp3], %[temp3], %[temp1] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), - [temp3] "=&r" (temp3), [dest_pix] "+r" (dest_pix) - : [cm] "r" (cm), [dest_stride] "r" (dest_stride), - [step1_0] "r" (step1_0), [step1_1] "r" (step1_1), - [step1_2] "r" (step1_2), [step1_3] "r" (step1_3), - [step2_28] "r" (step2_28), [step2_29] "r" (step2_29), - [step2_30] "r" (step2_30), [step2_31] "r" (step2_31) - ); - - step3_12 = ROUND_POWER_OF_TWO((step1_3 - step2_28), 6); - step3_13 = ROUND_POWER_OF_TWO((step1_2 - step2_29), 6); - step3_14 = ROUND_POWER_OF_TWO((step1_1 - step2_30), 6); - step3_15 = ROUND_POWER_OF_TWO((step1_0 - step2_31), 6); - - __asm__ __volatile__ ( - "lbu %[temp2], 0(%[dest_pix1]) \n\t" - "add %[temp2], %[temp2], %[step3_15] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "sb %[temp0], 0(%[dest_pix1]) \n\t" - "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix1]) \n\t" - "add %[temp3], %[temp3], %[step3_14] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix1]) \n\t" - "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" - - "lbu %[temp2], 0(%[dest_pix1]) \n\t" - "add %[temp2], %[temp2], %[step3_13] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "sb %[temp0], 0(%[dest_pix1]) \n\t" - "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix1]) \n\t" - "add %[temp3], %[temp3], %[step3_12] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix1]) \n\t" - "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" - - : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), - [temp3] "=&r" (temp3), [dest_pix1] "+r" (dest_pix1) - : [cm] "r" (cm), [dest_stride] "r" (dest_stride), - [step3_12] "r" (step3_12), [step3_13] "r" (step3_13), - [step3_14] "r" (step3_14), [step3_15] "r" (step3_15) - ); - - __asm__ __volatile__ ( - "lbu %[temp2], 0(%[dest_pix]) \n\t" - "add %[temp0], %[step1_4], %[step1_27] \n\t" - "addi %[temp0], %[temp0], 32 \n\t" - "sra %[temp0], %[temp0], 6 \n\t" - "add %[temp2], %[temp2], %[temp0] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "add %[temp1], %[step1_5], %[step1_26] \n\t" - "sb %[temp0], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix]) \n\t" - "addi %[temp1], %[temp1], 32 \n\t" - "sra %[temp1], %[temp1], 6 \n\t" - "add %[temp3], %[temp3], %[temp1] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[temp2], 0(%[dest_pix]) \n\t" - "add %[temp0], %[step1_6], %[step1_25] \n\t" - "addi %[temp0], %[temp0], 32 \n\t" - "sra %[temp0], %[temp0], 6 \n\t" - "add %[temp2], %[temp2], %[temp0] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "add %[temp1], %[step1_7], %[step1_24] \n\t" - "sb %[temp0], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix]) \n\t" - "addi %[temp1], %[temp1], 32 \n\t" - "sra %[temp1], %[temp1], 6 \n\t" - "add %[temp3], %[temp3], %[temp1] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), - [temp3] "=&r" (temp3), [dest_pix] "+r" (dest_pix) - : [cm] "r" (cm), [dest_stride] "r" (dest_stride), - [step1_4] "r" (step1_4), [step1_5] "r" (step1_5), - [step1_6] "r" (step1_6), [step1_7] "r" (step1_7), - [step1_24] "r" (step1_24), [step1_25] "r" (step1_25), - [step1_26] "r" (step1_26), [step1_27] "r" (step1_27) - ); - - step3_12 = ROUND_POWER_OF_TWO((step1_7 - step1_24), 6); - step3_13 = ROUND_POWER_OF_TWO((step1_6 - step1_25), 6); - step3_14 = ROUND_POWER_OF_TWO((step1_5 - step1_26), 6); - step3_15 = ROUND_POWER_OF_TWO((step1_4 - step1_27), 6); - - __asm__ __volatile__ ( - "lbu %[temp2], 0(%[dest_pix1]) \n\t" - "add %[temp2], %[temp2], %[step3_15] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "sb %[temp0], 0(%[dest_pix1]) \n\t" - "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix1]) \n\t" - "add %[temp3], %[temp3], %[step3_14] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix1]) \n\t" - "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" - - "lbu %[temp2], 0(%[dest_pix1]) \n\t" - "add %[temp2], %[temp2], %[step3_13] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "sb %[temp0], 0(%[dest_pix1]) \n\t" - "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix1]) \n\t" - "add %[temp3], %[temp3], %[step3_12] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix1]) \n\t" - "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" - - : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), - [temp3] "=&r" (temp3), [dest_pix1] "+r" (dest_pix1) - : [cm] "r" (cm), [dest_stride] "r" (dest_stride), - [step3_12] "r" (step3_12), [step3_13] "r" (step3_13), - [step3_14] "r" (step3_14), [step3_15] "r" (step3_15) - ); - - __asm__ __volatile__ ( - "lbu %[temp2], 0(%[dest_pix]) \n\t" - "add %[temp0], %[step1_8], %[step1_23] \n\t" - "addi %[temp0], %[temp0], 32 \n\t" - "sra %[temp0], %[temp0], 6 \n\t" - "add %[temp2], %[temp2], %[temp0] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "add %[temp1], %[step1_9], %[step1_22] \n\t" - "sb %[temp0], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix]) \n\t" - "addi %[temp1], %[temp1], 32 \n\t" - "sra %[temp1], %[temp1], 6 \n\t" - "add %[temp3], %[temp3], %[temp1] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[temp2], 0(%[dest_pix]) \n\t" - "add %[temp0], %[step1_10], %[step1_21] \n\t" - "addi %[temp0], %[temp0], 32 \n\t" - "sra %[temp0], %[temp0], 6 \n\t" - "add %[temp2], %[temp2], %[temp0] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "add %[temp1], %[step1_11], %[step1_20] \n\t" - "sb %[temp0], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix]) \n\t" - "addi %[temp1], %[temp1], 32 \n\t" - "sra %[temp1], %[temp1], 6 \n\t" - "add %[temp3], %[temp3], %[temp1] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), - [temp3] "=&r" (temp3), [dest_pix] "+r" (dest_pix) - : [cm] "r" (cm), [dest_stride] "r" (dest_stride), - [step1_8] "r" (step1_8), [step1_9] "r" (step1_9), - [step1_10] "r" (step1_10), [step1_11] "r" (step1_11), - [step1_20] "r" (step1_20), [step1_21] "r" (step1_21), - [step1_22] "r" (step1_22), [step1_23] "r" (step1_23) - ); - - step3_12 = ROUND_POWER_OF_TWO((step1_11 - step1_20), 6); - step3_13 = ROUND_POWER_OF_TWO((step1_10 - step1_21), 6); - step3_14 = ROUND_POWER_OF_TWO((step1_9 - step1_22), 6); - step3_15 = ROUND_POWER_OF_TWO((step1_8 - step1_23), 6); - - __asm__ __volatile__ ( - "lbu %[temp2], 0(%[dest_pix1]) \n\t" - "add %[temp2], %[temp2], %[step3_15] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "sb %[temp0], 0(%[dest_pix1]) \n\t" - "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix1]) \n\t" - "add %[temp3], %[temp3], %[step3_14] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix1]) \n\t" - "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" - - "lbu %[temp2], 0(%[dest_pix1]) \n\t" - "add %[temp2], %[temp2], %[step3_13] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "sb %[temp0], 0(%[dest_pix1]) \n\t" - "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix1]) \n\t" - "add %[temp3], %[temp3], %[step3_12] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix1]) \n\t" - "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" - - : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), - [temp3] "=&r" (temp3), [dest_pix1] "+r" (dest_pix1) - : [cm] "r" (cm), [dest_stride] "r" (dest_stride), - [step3_12] "r" (step3_12), [step3_13] "r" (step3_13), - [step3_14] "r" (step3_14), [step3_15] "r" (step3_15) - ); - - __asm__ __volatile__ ( - "lbu %[temp2], 0(%[dest_pix]) \n\t" - "add %[temp0], %[step1_12], %[step2_19] \n\t" - "addi %[temp0], %[temp0], 32 \n\t" - "sra %[temp0], %[temp0], 6 \n\t" - "add %[temp2], %[temp2], %[temp0] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "add %[temp1], %[step1_13], %[step2_18] \n\t" - "sb %[temp0], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix]) \n\t" - "addi %[temp1], %[temp1], 32 \n\t" - "sra %[temp1], %[temp1], 6 \n\t" - "add %[temp3], %[temp3], %[temp1] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[temp2], 0(%[dest_pix]) \n\t" - "add %[temp0], %[step1_14], %[step2_17] \n\t" - "addi %[temp0], %[temp0], 32 \n\t" - "sra %[temp0], %[temp0], 6 \n\t" - "add %[temp2], %[temp2], %[temp0] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "add %[temp1], %[step1_15], %[step2_16] \n\t" - "sb %[temp0], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix]) \n\t" - "addi %[temp1], %[temp1], 32 \n\t" - "sra %[temp1], %[temp1], 6 \n\t" - "add %[temp3], %[temp3], %[temp1] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix]) \n\t" - - : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), - [temp3] "=&r" (temp3), [dest_pix] "+r" (dest_pix) - : [cm] "r" (cm), [dest_stride] "r" (dest_stride), - [step1_12] "r" (step1_12), [step1_13] "r" (step1_13), - [step1_14] "r" (step1_14), [step1_15] "r" (step1_15), - [step2_16] "r" (step2_16), [step2_17] "r" (step2_17), - [step2_18] "r" (step2_18), [step2_19] "r" (step2_19) - ); - - step3_12 = ROUND_POWER_OF_TWO((step1_15 - step2_16), 6); - step3_13 = ROUND_POWER_OF_TWO((step1_14 - step2_17), 6); - step3_14 = ROUND_POWER_OF_TWO((step1_13 - step2_18), 6); - step3_15 = ROUND_POWER_OF_TWO((step1_12 - step2_19), 6); - - __asm__ __volatile__ ( - "lbu %[temp2], 0(%[dest_pix1]) \n\t" - "add %[temp2], %[temp2], %[step3_15] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "sb %[temp0], 0(%[dest_pix1]) \n\t" - "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix1]) \n\t" - "add %[temp3], %[temp3], %[step3_14] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix1]) \n\t" - "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" - - "lbu %[temp2], 0(%[dest_pix1]) \n\t" - "add %[temp2], %[temp2], %[step3_13] \n\t" - "lbux %[temp0], %[temp2](%[cm]) \n\t" - "sb %[temp0], 0(%[dest_pix1]) \n\t" - "subu %[dest_pix1], %[dest_pix1], %[dest_stride] \n\t" - "lbu %[temp3], 0(%[dest_pix1]) \n\t" - "add %[temp3], %[temp3], %[step3_12] \n\t" - "lbux %[temp1], %[temp3](%[cm]) \n\t" - "sb %[temp1], 0(%[dest_pix1]) \n\t" - - : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), [temp2] "=&r" (temp2), - [temp3] "=&r" (temp3), [dest_pix1] "+r" (dest_pix1) - : [cm] "r" (cm), [dest_stride] "r" (dest_stride), - [step3_12] "r" (step3_12), [step3_13] "r" (step3_13), - [step3_14] "r" (step3_14), [step3_15] "r" (step3_15) - ); - - input += 32; - } -} -#endif // #if HAVE_DSPR2 diff --git a/vp9/common/mips/dspr2/vp9_itrans32_dspr2.c b/vp9/common/mips/dspr2/vp9_itrans32_dspr2.c deleted file mode 100644 index b4b0d248c..000000000 --- a/vp9/common/mips/dspr2/vp9_itrans32_dspr2.c +++ /dev/null @@ -1,1076 +0,0 @@ -/* - * Copyright (c) 2013 The WebM project authors. All Rights Reserved. - * - * Use of this source code is governed by a BSD-style license - * that can be found in the LICENSE file in the root of the source - * tree. An additional intellectual property rights grant can be found - * in the file PATENTS. All contributing project authors may - * be found in the AUTHORS file in the root of the source tree. - */ - -#include <assert.h> -#include <stdio.h> - -#include "./vpx_config.h" -#include "./vp9_rtcd.h" -#include "vp9/common/vp9_common.h" -#include "vp9/common/vp9_blockd.h" -#include "vp9/common/mips/dspr2/vp9_common_dspr2.h" -#include "vpx_dsp/txfm_common.h" - -#if HAVE_DSPR2 -static void idct32_rows_dspr2(const int16_t *input, int16_t *output, - uint32_t no_rows) { - int16_t step1_0, step1_1, step1_2, step1_3, step1_4, step1_5, step1_6; - int16_t step1_7, step1_8, step1_9, step1_10, step1_11, step1_12, step1_13; - int16_t step1_14, step1_15, step1_16, step1_17, step1_18, step1_19, step1_20; - int16_t step1_21, step1_22, step1_23, step1_24, step1_25, step1_26, step1_27; - int16_t step1_28, step1_29, step1_30, step1_31; - int16_t step2_0, step2_1, step2_2, step2_3, step2_4, step2_5, step2_6; - int16_t step2_7, step2_8, step2_9, step2_10, step2_11, step2_12, step2_13; - int16_t step2_14, step2_15, step2_16, step2_17, step2_18, step2_19, step2_20; - int16_t step2_21, step2_22, step2_23, step2_24, step2_25, step2_26, step2_27; - int16_t step2_28, step2_29, step2_30, step2_31; - int16_t step3_8, step3_9, step3_10, step3_11, step3_12, step3_13, step3_14; - int16_t step3_15, step3_16, step3_17, step3_18, step3_19, step3_20, step3_21; - int16_t step3_22, step3_23, step3_24, step3_25, step3_26, step3_27, step3_28; - int16_t step3_29, step3_30, step3_31; - int temp0, temp1, temp2, temp3; - int load1, load2, load3, load4; - int result1, result2; - int temp21; - int i; - const int const_2_power_13 = 8192; - const int32_t *input_int; - - for (i = no_rows; i--; ) { - input_int = (const int32_t *)input; - - if (!(input_int[0] | input_int[1] | input_int[2] | input_int[3] | - input_int[4] | input_int[5] | input_int[6] | input_int[7] | - input_int[8] | input_int[9] | input_int[10] | input_int[11] | - input_int[12] | input_int[13] | input_int[14] | input_int[15])) { - input += 32; - - __asm__ __volatile__ ( - "sh $zero, 0(%[output]) \n\t" - "sh $zero, 64(%[output]) \n\t" - "sh $zero, 128(%[output]) \n\t" - "sh $zero, 192(%[output]) \n\t" - "sh $zero, 256(%[output]) \n\t" - "sh $zero, 320(%[output]) \n\t" - "sh $zero, 384(%[output]) \n\t" - "sh $zero, 448(%[output]) \n\t" - "sh $zero, 512(%[output]) \n\t" - "sh $zero, 576(%[output]) \n\t" - "sh $zero, 640(%[output]) \n\t" - "sh $zero, 704(%[output]) \n\t" - "sh $zero, 768(%[output]) \n\t" - "sh $zero, 832(%[output]) \n\t" - "sh $zero, 896(%[output]) \n\t" - "sh $zero, 960(%[output]) \n\t" - "sh $zero, 1024(%[output]) \n\t" - "sh $zero, 1088(%[output]) \n\t" - "sh $zero, 1152(%[output]) \n\t" - "sh $zero, 1216(%[output]) \n\t" - "sh $zero, 1280(%[output]) \n\t" - "sh $zero, 1344(%[output]) \n\t" - "sh $zero, 1408(%[output]) \n\t" - "sh $zero, 1472(%[output]) \n\t" - "sh $zero, 1536(%[output]) \n\t" - "sh $zero, 1600(%[output]) \n\t" - "sh $zero, 1664(%[output]) \n\t" - "sh $zero, 1728(%[output]) \n\t" - "sh $zero, 1792(%[output]) \n\t" - "sh $zero, 1856(%[output]) \n\t" - "sh $zero, 1920(%[output]) \n\t" - "sh $zero, 1984(%[output]) \n\t" - - : - : [output] "r" (output) - ); - - output += 1; - - continue; - } - - /* prefetch row */ - prefetch_load((const uint8_t *)(input + 32)); - prefetch_load((const uint8_t *)(input + 48)); - - __asm__ __volatile__ ( - "lh %[load1], 2(%[input]) \n\t" - "lh %[load2], 62(%[input]) \n\t" - "lh %[load3], 34(%[input]) \n\t" - "lh %[load4], 30(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_31_64] \n\t" - "msub $ac1, %[load2], %[cospi_1_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - - "madd $ac3, %[load1], %[cospi_1_64] \n\t" - "madd $ac3, %[load2], %[cospi_31_64] \n\t" - "extp %[temp3], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac2, %[load3], %[cospi_15_64] \n\t" - "msub $ac2, %[load4], %[cospi_17_64] \n\t" - "extp %[temp1], $ac2, 31 \n\t" - - "madd $ac1, %[load3], %[cospi_17_64] \n\t" - "madd $ac1, %[load4], %[cospi_15_64] \n\t" - "extp %[temp2], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[temp3], %[temp2] \n\t" - "sub %[load2], %[temp0], %[temp1] \n\t" - - "madd $ac1, %[load1], %[cospi_28_64] \n\t" - "msub $ac1, %[load2], %[cospi_4_64] \n\t" - "madd $ac3, %[load1], %[cospi_4_64] \n\t" - "madd $ac3, %[load2], %[cospi_28_64] \n\t" - - "extp %[step1_17], $ac1, 31 \n\t" - "extp %[step1_30], $ac3, 31 \n\t" - "add %[step1_16], %[temp0], %[temp1] \n\t" - "add %[step1_31], %[temp2], %[temp3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), - [load3] "=&r" (load3), [load4] "=&r" (load4), - [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step1_16] "=r" (step1_16), [step1_17] "=r" (step1_17), - [step1_30] "=r" (step1_30), [step1_31] "=r" (step1_31) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_31_64] "r" (cospi_31_64), [cospi_1_64] "r" (cospi_1_64), - [cospi_4_64] "r" (cospi_4_64), [cospi_17_64] "r" (cospi_17_64), - [cospi_15_64] "r" (cospi_15_64), [cospi_28_64] "r" (cospi_28_64) - ); - - __asm__ __volatile__ ( - "lh %[load1], 18(%[input]) \n\t" - "lh %[load2], 46(%[input]) \n\t" - "lh %[load3], 50(%[input]) \n\t" - "lh %[load4], 14(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_23_64] \n\t" - "msub $ac1, %[load2], %[cospi_9_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - - "madd $ac3, %[load1], %[cospi_9_64] \n\t" - "madd $ac3, %[load2], %[cospi_23_64] \n\t" - "extp %[temp3], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac2, %[load3], %[cospi_7_64] \n\t" - "msub $ac2, %[load4], %[cospi_25_64] \n\t" - "extp %[temp1], $ac2, 31 \n\t" - - "madd $ac1, %[load3], %[cospi_25_64] \n\t" - "madd $ac1, %[load4], %[cospi_7_64] \n\t" - "extp %[temp2], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[temp1], %[temp0] \n\t" - "sub %[load2], %[temp2], %[temp3] \n\t" - - "msub $ac1, %[load1], %[cospi_28_64] \n\t" - "msub $ac1, %[load2], %[cospi_4_64] \n\t" - "msub $ac3, %[load1], %[cospi_4_64] \n\t" - "madd $ac3, %[load2], %[cospi_28_64] \n\t" - - "extp %[step1_18], $ac1, 31 \n\t" - "extp %[step1_29], $ac3, 31 \n\t" - "add %[step1_19], %[temp0], %[temp1] \n\t" - "add %[step1_28], %[temp2], %[temp3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), - [load3] "=&r" (load3), [load4] "=&r" (load4), - [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step1_18] "=r" (step1_18), [step1_19] "=r" (step1_19), - [step1_28] "=r" (step1_28), [step1_29] "=r" (step1_29) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_23_64] "r" (cospi_23_64), [cospi_9_64] "r" (cospi_9_64), - [cospi_4_64] "r" (cospi_4_64), [cospi_7_64] "r" (cospi_7_64), - [cospi_25_64] "r" (cospi_25_64), [cospi_28_64] "r" (cospi_28_64) - ); - - __asm__ __volatile__ ( - "lh %[load1], 10(%[input]) \n\t" - "lh %[load2], 54(%[input]) \n\t" - "lh %[load3], 42(%[input]) \n\t" - "lh %[load4], 22(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_27_64] \n\t" - "msub $ac1, %[load2], %[cospi_5_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - - "madd $ac3, %[load1], %[cospi_5_64] \n\t" - "madd $ac3, %[load2], %[cospi_27_64] \n\t" - "extp %[temp3], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac2, %[load3], %[cospi_11_64] \n\t" - "msub $ac2, %[load4], %[cospi_21_64] \n\t" - "extp %[temp1], $ac2, 31 \n\t" - - "madd $ac1, %[load3], %[cospi_21_64] \n\t" - "madd $ac1, %[load4], %[cospi_11_64] \n\t" - "extp %[temp2], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[temp0], %[temp1] \n\t" - "sub %[load2], %[temp3], %[temp2] \n\t" - - "madd $ac1, %[load2], %[cospi_12_64] \n\t" - "msub $ac1, %[load1], %[cospi_20_64] \n\t" - "madd $ac3, %[load1], %[cospi_12_64] \n\t" - "madd $ac3, %[load2], %[cospi_20_64] \n\t" - - "extp %[step1_21], $ac1, 31 \n\t" - "extp %[step1_26], $ac3, 31 \n\t" - "add %[step1_20], %[temp0], %[temp1] \n\t" - "add %[step1_27], %[temp2], %[temp3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), - [load3] "=&r" (load3), [load4] "=&r" (load4), - [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step1_20] "=r" (step1_20), [step1_21] "=r" (step1_21), - [step1_26] "=r" (step1_26), [step1_27] "=r" (step1_27) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_27_64] "r" (cospi_27_64), [cospi_5_64] "r" (cospi_5_64), - [cospi_11_64] "r" (cospi_11_64), [cospi_21_64] "r" (cospi_21_64), - [cospi_12_64] "r" (cospi_12_64), [cospi_20_64] "r" (cospi_20_64) - ); - - __asm__ __volatile__ ( - "lh %[load1], 26(%[input]) \n\t" - "lh %[load2], 38(%[input]) \n\t" - "lh %[load3], 58(%[input]) \n\t" - "lh %[load4], 6(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_19_64] \n\t" - "msub $ac1, %[load2], %[cospi_13_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - - "madd $ac3, %[load1], %[cospi_13_64] \n\t" - "madd $ac3, %[load2], %[cospi_19_64] \n\t" - "extp %[temp3], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac2, %[load3], %[cospi_3_64] \n\t" - "msub $ac2, %[load4], %[cospi_29_64] \n\t" - "extp %[temp1], $ac2, 31 \n\t" - - "madd $ac1, %[load3], %[cospi_29_64] \n\t" - "madd $ac1, %[load4], %[cospi_3_64] \n\t" - "extp %[temp2], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[temp1], %[temp0] \n\t" - "sub %[load2], %[temp2], %[temp3] \n\t" - - "msub $ac1, %[load1], %[cospi_12_64] \n\t" - "msub $ac1, %[load2], %[cospi_20_64] \n\t" - "msub $ac3, %[load1], %[cospi_20_64] \n\t" - "madd $ac3, %[load2], %[cospi_12_64] \n\t" - - "extp %[step1_22], $ac1, 31 \n\t" - "extp %[step1_25], $ac3, 31 \n\t" - "add %[step1_23], %[temp0], %[temp1] \n\t" - "add %[step1_24], %[temp2], %[temp3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), - [load3] "=&r" (load3), [load4] "=&r" (load4), - [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step1_22] "=r" (step1_22), [step1_23] "=r" (step1_23), - [step1_24] "=r" (step1_24), [step1_25] "=r" (step1_25) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_19_64] "r" (cospi_19_64), [cospi_13_64] "r" (cospi_13_64), - [cospi_3_64] "r" (cospi_3_64), [cospi_29_64] "r" (cospi_29_64), - [cospi_12_64] "r" (cospi_12_64), [cospi_20_64] "r" (cospi_20_64) - ); - - __asm__ __volatile__ ( - "lh %[load1], 4(%[input]) \n\t" - "lh %[load2], 60(%[input]) \n\t" - "lh %[load3], 36(%[input]) \n\t" - "lh %[load4], 28(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_30_64] \n\t" - "msub $ac1, %[load2], %[cospi_2_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - - "madd $ac3, %[load1], %[cospi_2_64] \n\t" - "madd $ac3, %[load2], %[cospi_30_64] \n\t" - "extp %[temp3], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac2, %[load3], %[cospi_14_64] \n\t" - "msub $ac2, %[load4], %[cospi_18_64] \n\t" - "extp %[temp1], $ac2, 31 \n\t" - - "madd $ac1, %[load3], %[cospi_18_64] \n\t" - "madd $ac1, %[load4], %[cospi_14_64] \n\t" - "extp %[temp2], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[temp0], %[temp1] \n\t" - "sub %[load2], %[temp3], %[temp2] \n\t" - - "msub $ac1, %[load1], %[cospi_8_64] \n\t" - "madd $ac1, %[load2], %[cospi_24_64] \n\t" - "madd $ac3, %[load1], %[cospi_24_64] \n\t" - "madd $ac3, %[load2], %[cospi_8_64] \n\t" - - "extp %[step2_9], $ac1, 31 \n\t" - "extp %[step2_14], $ac3, 31 \n\t" - "add %[step2_8], %[temp0], %[temp1] \n\t" - "add %[step2_15], %[temp2], %[temp3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), - [load3] "=&r" (load3), [load4] "=&r" (load4), - [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step2_8] "=r" (step2_8), [step2_9] "=r" (step2_9), - [step2_14] "=r" (step2_14), [step2_15] "=r" (step2_15) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_30_64] "r" (cospi_30_64), [cospi_2_64] "r" (cospi_2_64), - [cospi_14_64] "r" (cospi_14_64), [cospi_18_64] "r" (cospi_18_64), - [cospi_8_64] "r" (cospi_8_64), [cospi_24_64] "r" (cospi_24_64) - ); - - __asm__ __volatile__ ( - "lh %[load1], 20(%[input]) \n\t" - "lh %[load2], 44(%[input]) \n\t" - "lh %[load3], 52(%[input]) \n\t" - "lh %[load4], 12(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_22_64] \n\t" - "msub $ac1, %[load2], %[cospi_10_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - - "madd $ac3, %[load1], %[cospi_10_64] \n\t" - "madd $ac3, %[load2], %[cospi_22_64] \n\t" - "extp %[temp3], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac2, %[load3], %[cospi_6_64] \n\t" - "msub $ac2, %[load4], %[cospi_26_64] \n\t" - "extp %[temp1], $ac2, 31 \n\t" - - "madd $ac1, %[load3], %[cospi_26_64] \n\t" - "madd $ac1, %[load4], %[cospi_6_64] \n\t" - "extp %[temp2], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[temp1], %[temp0] \n\t" - "sub %[load2], %[temp2], %[temp3] \n\t" - - "msub $ac1, %[load1], %[cospi_24_64] \n\t" - "msub $ac1, %[load2], %[cospi_8_64] \n\t" - "madd $ac3, %[load2], %[cospi_24_64] \n\t" - "msub $ac3, %[load1], %[cospi_8_64] \n\t" - - "extp %[step2_10], $ac1, 31 \n\t" - "extp %[step2_13], $ac3, 31 \n\t" - "add %[step2_11], %[temp0], %[temp1] \n\t" - "add %[step2_12], %[temp2], %[temp3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), - [load3] "=&r" (load3), [load4] "=&r" (load4), - [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step2_10] "=r" (step2_10), [step2_11] "=r" (step2_11), - [step2_12] "=r" (step2_12), [step2_13] "=r" (step2_13) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_22_64] "r" (cospi_22_64), [cospi_10_64] "r" (cospi_10_64), - [cospi_6_64] "r" (cospi_6_64), [cospi_26_64] "r" (cospi_26_64), - [cospi_8_64] "r" (cospi_8_64), [cospi_24_64] "r" (cospi_24_64) - ); - - __asm__ __volatile__ ( - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "sub %[temp0], %[step2_14], %[step2_13] \n\t" - "sub %[temp0], %[temp0], %[step2_9] \n\t" - "add %[temp0], %[temp0], %[step2_10] \n\t" - "madd $ac0, %[temp0], %[cospi_16_64] \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "sub %[temp1], %[step2_14], %[step2_13] \n\t" - "add %[temp1], %[temp1], %[step2_9] \n\t" - "sub %[temp1], %[temp1], %[step2_10] \n\t" - "madd $ac1, %[temp1], %[cospi_16_64] \n\t" - - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - "sub %[temp0], %[step2_15], %[step2_12] \n\t" - "sub %[temp0], %[temp0], %[step2_8] \n\t" - "add %[temp0], %[temp0], %[step2_11] \n\t" - "madd $ac2, %[temp0], %[cospi_16_64] \n\t" - - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - "sub %[temp1], %[step2_15], %[step2_12] \n\t" - "add %[temp1], %[temp1], %[step2_8] \n\t" - "sub %[temp1], %[temp1], %[step2_11] \n\t" - "madd $ac3, %[temp1], %[cospi_16_64] \n\t" - - "add %[step3_8], %[step2_8], %[step2_11] \n\t" - "add %[step3_9], %[step2_9], %[step2_10] \n\t" - "add %[step3_14], %[step2_13], %[step2_14] \n\t" - "add %[step3_15], %[step2_12], %[step2_15] \n\t" - - "extp %[step3_10], $ac0, 31 \n\t" - "extp %[step3_13], $ac1, 31 \n\t" - "extp %[step3_11], $ac2, 31 \n\t" - "extp %[step3_12], $ac3, 31 \n\t" - - : [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [step3_8] "=r" (step3_8), [step3_9] "=r" (step3_9), - [step3_10] "=r" (step3_10), [step3_11] "=r" (step3_11), - [step3_12] "=r" (step3_12), [step3_13] "=r" (step3_13), - [step3_14] "=r" (step3_14), [step3_15] "=r" (step3_15) - : [const_2_power_13] "r" (const_2_power_13), - [step2_8] "r" (step2_8), [step2_9] "r" (step2_9), - [step2_10] "r" (step2_10), [step2_11] "r" (step2_11), - [step2_12] "r" (step2_12), [step2_13] "r" (step2_13), - [step2_14] "r" (step2_14), [step2_15] "r" (step2_15), - [cospi_16_64] "r" (cospi_16_64) - ); - - step2_18 = step1_17 - step1_18; - step2_29 = step1_30 - step1_29; - - __asm__ __volatile__ ( - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "msub $ac0, %[step2_18], %[cospi_8_64] \n\t" - "madd $ac0, %[step2_29], %[cospi_24_64] \n\t" - "extp %[step3_18], $ac0, 31 \n\t" - - : [step3_18] "=r" (step3_18) - : [const_2_power_13] "r" (const_2_power_13), - [step2_18] "r" (step2_18), [step2_29] "r" (step2_29), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) - ); - - temp21 = step2_18 * cospi_24_64 + step2_29 * cospi_8_64; - step3_29 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - step2_19 = step1_16 - step1_19; - step2_28 = step1_31 - step1_28; - - __asm__ __volatile__ ( - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "msub $ac0, %[step2_19], %[cospi_8_64] \n\t" - "madd $ac0, %[step2_28], %[cospi_24_64] \n\t" - "extp %[step3_19], $ac0, 31 \n\t" - - : [step3_19] "=r" (step3_19) - : [const_2_power_13] "r" (const_2_power_13), - [step2_19] "r" (step2_19), [step2_28] "r" (step2_28), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) - ); - - temp21 = step2_19 * cospi_24_64 + step2_28 * cospi_8_64; - step3_28 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - step3_16 = step1_16 + step1_19; - step3_17 = step1_17 + step1_18; - step3_30 = step1_29 + step1_30; - step3_31 = step1_28 + step1_31; - - step2_20 = step1_23 - step1_20; - step2_27 = step1_24 - step1_27; - - __asm__ __volatile__ ( - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "msub $ac0, %[step2_20], %[cospi_24_64] \n\t" - "msub $ac0, %[step2_27], %[cospi_8_64] \n\t" - "extp %[step3_20], $ac0, 31 \n\t" - - : [step3_20] "=r" (step3_20) - : [const_2_power_13] "r" (const_2_power_13), - [step2_20] "r" (step2_20), [step2_27] "r" (step2_27), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) - ); - - temp21 = -step2_20 * cospi_8_64 + step2_27 * cospi_24_64; - step3_27 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - step2_21 = step1_22 - step1_21; - step2_26 = step1_25 - step1_26; - - __asm__ __volatile__ ( - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "msub $ac1, %[step2_21], %[cospi_24_64] \n\t" - "msub $ac1, %[step2_26], %[cospi_8_64] \n\t" - "extp %[step3_21], $ac1, 31 \n\t" - - : [step3_21] "=r" (step3_21) - : [const_2_power_13] "r" (const_2_power_13), - [step2_21] "r" (step2_21), [step2_26] "r" (step2_26), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) - ); - - temp21 = -step2_21 * cospi_8_64 + step2_26 * cospi_24_64; - step3_26 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - step3_22 = step1_21 + step1_22; - step3_23 = step1_20 + step1_23; - step3_24 = step1_24 + step1_27; - step3_25 = step1_25 + step1_26; - - step2_16 = step3_16 + step3_23; - step2_17 = step3_17 + step3_22; - step2_18 = step3_18 + step3_21; - step2_19 = step3_19 + step3_20; - step2_20 = step3_19 - step3_20; - step2_21 = step3_18 - step3_21; - step2_22 = step3_17 - step3_22; - step2_23 = step3_16 - step3_23; - - step2_24 = step3_31 - step3_24; - step2_25 = step3_30 - step3_25; - step2_26 = step3_29 - step3_26; - step2_27 = step3_28 - step3_27; - step2_28 = step3_28 + step3_27; - step2_29 = step3_29 + step3_26; - step2_30 = step3_30 + step3_25; - step2_31 = step3_31 + step3_24; - - __asm__ __volatile__ ( - "lh %[load1], 0(%[input]) \n\t" - "lh %[load2], 32(%[input]) \n\t" - "lh %[load3], 16(%[input]) \n\t" - "lh %[load4], 48(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - "add %[result1], %[load1], %[load2] \n\t" - "sub %[result2], %[load1], %[load2] \n\t" - "madd $ac1, %[result1], %[cospi_16_64] \n\t" - "madd $ac2, %[result2], %[cospi_16_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - "extp %[temp1], $ac2, 31 \n\t" - - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - "madd $ac3, %[load3], %[cospi_24_64] \n\t" - "msub $ac3, %[load4], %[cospi_8_64] \n\t" - "extp %[temp2], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "madd $ac1, %[load3], %[cospi_8_64] \n\t" - "madd $ac1, %[load4], %[cospi_24_64] \n\t" - "extp %[temp3], $ac1, 31 \n\t" - - "add %[step1_0], %[temp0], %[temp3] \n\t" - "add %[step1_1], %[temp1], %[temp2] \n\t" - "sub %[step1_2], %[temp1], %[temp2] \n\t" - "sub %[step1_3], %[temp0], %[temp3] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), - [load3] "=&r" (load3), [load4] "=&r" (load4), - [result1] "=&r" (result1), [result2] "=&r" (result2), - [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step1_0] "=r" (step1_0), [step1_1] "=r" (step1_1), - [step1_2] "=r" (step1_2), [step1_3] "=r" (step1_3) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_16_64] "r" (cospi_16_64), - [cospi_24_64] "r" (cospi_24_64), [cospi_8_64] "r" (cospi_8_64) - - ); - - __asm__ __volatile__ ( - "lh %[load1], 8(%[input]) \n\t" - "lh %[load2], 56(%[input]) \n\t" - "lh %[load3], 40(%[input]) \n\t" - "lh %[load4], 24(%[input]) \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "madd $ac1, %[load1], %[cospi_28_64] \n\t" - "msub $ac1, %[load2], %[cospi_4_64] \n\t" - "extp %[temp0], $ac1, 31 \n\t" - - "madd $ac3, %[load1], %[cospi_4_64] \n\t" - "madd $ac3, %[load2], %[cospi_28_64] \n\t" - "extp %[temp3], $ac3, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac2 \n\t" - "mthi $zero, $ac2 \n\t" - - "madd $ac2, %[load3], %[cospi_12_64] \n\t" - "msub $ac2, %[load4], %[cospi_20_64] \n\t" - "extp %[temp1], $ac2, 31 \n\t" - - "madd $ac1, %[load3], %[cospi_20_64] \n\t" - "madd $ac1, %[load4], %[cospi_12_64] \n\t" - "extp %[temp2], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "mtlo %[const_2_power_13], $ac3 \n\t" - "mthi $zero, $ac3 \n\t" - - "sub %[load1], %[temp3], %[temp2] \n\t" - "sub %[load1], %[load1], %[temp0] \n\t" - "add %[load1], %[load1], %[temp1] \n\t" - - "sub %[load2], %[temp0], %[temp1] \n\t" - "sub %[load2], %[load2], %[temp2] \n\t" - "add %[load2], %[load2], %[temp3] \n\t" - - "madd $ac1, %[load1], %[cospi_16_64] \n\t" - "madd $ac3, %[load2], %[cospi_16_64] \n\t" - - "extp %[step1_5], $ac1, 31 \n\t" - "extp %[step1_6], $ac3, 31 \n\t" - "add %[step1_4], %[temp0], %[temp1] \n\t" - "add %[step1_7], %[temp3], %[temp2] \n\t" - - : [load1] "=&r" (load1), [load2] "=&r" (load2), - [load3] "=&r" (load3), [load4] "=&r" (load4), - [temp0] "=&r" (temp0), [temp1] "=&r" (temp1), - [temp2] "=&r" (temp2), [temp3] "=&r" (temp3), - [step1_4] "=r" (step1_4), [step1_5] "=r" (step1_5), - [step1_6] "=r" (step1_6), [step1_7] "=r" (step1_7) - : [const_2_power_13] "r" (const_2_power_13), [input] "r" (input), - [cospi_20_64] "r" (cospi_20_64), [cospi_12_64] "r" (cospi_12_64), - [cospi_4_64] "r" (cospi_4_64), [cospi_28_64] "r" (cospi_28_64), - [cospi_16_64] "r" (cospi_16_64) - ); - - step2_0 = step1_0 + step1_7; - step2_1 = step1_1 + step1_6; - step2_2 = step1_2 + step1_5; - step2_3 = step1_3 + step1_4; - step2_4 = step1_3 - step1_4; - step2_5 = step1_2 - step1_5; - step2_6 = step1_1 - step1_6; - step2_7 = step1_0 - step1_7; - - step1_0 = step2_0 + step3_15; - step1_1 = step2_1 + step3_14; - step1_2 = step2_2 + step3_13; - step1_3 = step2_3 + step3_12; - step1_4 = step2_4 + step3_11; - step1_5 = step2_5 + step3_10; - step1_6 = step2_6 + step3_9; - step1_7 = step2_7 + step3_8; - step1_8 = step2_7 - step3_8; - step1_9 = step2_6 - step3_9; - step1_10 = step2_5 - step3_10; - step1_11 = step2_4 - step3_11; - step1_12 = step2_3 - step3_12; - step1_13 = step2_2 - step3_13; - step1_14 = step2_1 - step3_14; - step1_15 = step2_0 - step3_15; - - __asm__ __volatile__ ( - "sub %[temp0], %[step2_27], %[step2_20] \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "madd $ac0, %[temp0], %[cospi_16_64] \n\t" - "extp %[step1_20], $ac0, 31 \n\t" - - : [temp0] "=&r" (temp0), [step1_20] "=r" (step1_20) - : [const_2_power_13] "r" (const_2_power_13), - [step2_20] "r" (step2_20), [step2_27] "r" (step2_27), - [cospi_16_64] "r" (cospi_16_64) - ); - - temp21 = (step2_20 + step2_27) * cospi_16_64; - step1_27 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - __asm__ __volatile__ ( - "sub %[temp0], %[step2_26], %[step2_21] \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "madd $ac0, %[temp0], %[cospi_16_64] \n\t" - "extp %[step1_21], $ac0, 31 \n\t" - - : [temp0] "=&r" (temp0), [step1_21] "=r" (step1_21) - : [const_2_power_13] "r" (const_2_power_13), - [step2_26] "r" (step2_26), [step2_21] "r" (step2_21), - [cospi_16_64] "r" (cospi_16_64) - ); - - temp21 = (step2_21 + step2_26) * cospi_16_64; - step1_26 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - __asm__ __volatile__ ( - "sub %[temp0], %[step2_25], %[step2_22] \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "madd $ac0, %[temp0], %[cospi_16_64] \n\t" - "extp %[step1_22], $ac0, 31 \n\t" - - : [temp0] "=&r" (temp0), [step1_22] "=r" (step1_22) - : [const_2_power_13] "r" (const_2_power_13), - [step2_25] "r" (step2_25), [step2_22] "r" (step2_22), - [cospi_16_64] "r" (cospi_16_64) - ); - - temp21 = (step2_22 + step2_25) * cospi_16_64; - step1_25 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - __asm__ __volatile__ ( - "sub %[temp0], %[step2_24], %[step2_23] \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "madd $ac0, %[temp0], %[cospi_16_64] \n\t" - "extp %[step1_23], $ac0, 31 \n\t" - - : [temp0] "=&r" (temp0), [step1_23] "=r" (step1_23) - : [const_2_power_13] "r" (const_2_power_13), - [step2_24] "r" (step2_24), [step2_23] "r" (step2_23), - [cospi_16_64] "r" (cospi_16_64) - ); - - temp21 = (step2_23 + step2_24) * cospi_16_64; - step1_24 = (temp21 + DCT_CONST_ROUNDING) >> DCT_CONST_BITS; - - // final stage - output[0 * 32] = step1_0 + step2_31; - output[1 * 32] = step1_1 + step2_30; - output[2 * 32] = step1_2 + step2_29; - output[3 * 32] = step1_3 + step2_28; - output[4 * 32] = step1_4 + step1_27; - output[5 * 32] = step1_5 + step1_26; - output[6 * 32] = step1_6 + step1_25; - output[7 * 32] = step1_7 + step1_24; - output[8 * 32] = step1_8 + step1_23; - output[9 * 32] = step1_9 + step1_22; - output[10 * 32] = step1_10 + step1_21; - output[11 * 32] = step1_11 + step1_20; - output[12 * 32] = step1_12 + step2_19; - output[13 * 32] = step1_13 + step2_18; - output[14 * 32] = step1_14 + step2_17; - output[15 * 32] = step1_15 + step2_16; - output[16 * 32] = step1_15 - step2_16; - output[17 * 32] = step1_14 - step2_17; - output[18 * 32] = step1_13 - step2_18; - output[19 * 32] = step1_12 - step2_19; - output[20 * 32] = step1_11 - step1_20; - output[21 * 32] = step1_10 - step1_21; - output[22 * 32] = step1_9 - step1_22; - output[23 * 32] = step1_8 - step1_23; - output[24 * 32] = step1_7 - step1_24; - output[25 * 32] = step1_6 - step1_25; - output[26 * 32] = step1_5 - step1_26; - output[27 * 32] = step1_4 - step1_27; - output[28 * 32] = step1_3 - step2_28; - output[29 * 32] = step1_2 - step2_29; - output[30 * 32] = step1_1 - step2_30; - output[31 * 32] = step1_0 - step2_31; - - input += 32; - output += 1; - } -} - -void vp9_idct32x32_1024_add_dspr2(const int16_t *input, uint8_t *dest, - int dest_stride) { - DECLARE_ALIGNED(32, int16_t, out[32 * 32]); - int16_t *outptr = out; - uint32_t pos = 45; - - /* bit positon for extract from acc */ - __asm__ __volatile__ ( - "wrdsp %[pos], 1 \n\t" - : - : [pos] "r" (pos) - ); - - // Rows - idct32_rows_dspr2(input, outptr, 32); - - // Columns - vp9_idct32_cols_add_blk_dspr2(out, dest, dest_stride); -} - -void vp9_idct32x32_34_add_dspr2(const int16_t *input, uint8_t *dest, - int stride) { - DECLARE_ALIGNED(32, int16_t, out[32 * 32]); - int16_t *outptr = out; - uint32_t i; - uint32_t pos = 45; - - /* bit positon for extract from acc */ - __asm__ __volatile__ ( - "wrdsp %[pos], 1 \n\t" - : - : [pos] "r" (pos) - ); - - // Rows - idct32_rows_dspr2(input, outptr, 8); - - outptr += 8; - __asm__ __volatile__ ( - "sw $zero, 0(%[outptr]) \n\t" - "sw $zero, 4(%[outptr]) \n\t" - "sw $zero, 8(%[outptr]) \n\t" - "sw $zero, 12(%[outptr]) \n\t" - "sw $zero, 16(%[outptr]) \n\t" - "sw $zero, 20(%[outptr]) \n\t" - "sw $zero, 24(%[outptr]) \n\t" - "sw $zero, 28(%[outptr]) \n\t" - "sw $zero, 32(%[outptr]) \n\t" - "sw $zero, 36(%[outptr]) \n\t" - "sw $zero, 40(%[outptr]) \n\t" - "sw $zero, 44(%[outptr]) \n\t" - - : - : [outptr] "r" (outptr) - ); - - for (i = 0; i < 31; ++i) { - outptr += 32; - - __asm__ __volatile__ ( - "sw $zero, 0(%[outptr]) \n\t" - "sw $zero, 4(%[outptr]) \n\t" - "sw $zero, 8(%[outptr]) \n\t" - "sw $zero, 12(%[outptr]) \n\t" - "sw $zero, 16(%[outptr]) \n\t" - "sw $zero, 20(%[outptr]) \n\t" - "sw $zero, 24(%[outptr]) \n\t" - "sw $zero, 28(%[outptr]) \n\t" - "sw $zero, 32(%[outptr]) \n\t" - "sw $zero, 36(%[outptr]) \n\t" - "sw $zero, 40(%[outptr]) \n\t" - "sw $zero, 44(%[outptr]) \n\t" - - : - : [outptr] "r" (outptr) - ); - } - - // Columns - vp9_idct32_cols_add_blk_dspr2(out, dest, stride); -} - -void vp9_idct32x32_1_add_dspr2(const int16_t *input, uint8_t *dest, - int stride) { - int r, out; - int32_t a1, absa1; - int32_t vector_a1; - int32_t t1, t2, t3, t4; - int32_t vector_1, vector_2, vector_3, vector_4; - uint32_t pos = 45; - - /* bit positon for extract from acc */ - __asm__ __volatile__ ( - "wrdsp %[pos], 1 \n\t" - - : - : [pos] "r" (pos) - ); - - out = DCT_CONST_ROUND_SHIFT_TWICE_COSPI_16_64(input[0]); - __asm__ __volatile__ ( - "addi %[out], %[out], 32 \n\t" - "sra %[a1], %[out], 6 \n\t" - - : [out] "+r" (out), [a1] "=r" (a1) - : - ); - - if (a1 < 0) { - /* use quad-byte - * input and output memory are four byte aligned */ - __asm__ __volatile__ ( - "abs %[absa1], %[a1] \n\t" - "replv.qb %[vector_a1], %[absa1] \n\t" - - : [absa1] "=r" (absa1), [vector_a1] "=r" (vector_a1) - : [a1] "r" (a1) - ); - - for (r = 32; r--;) { - __asm__ __volatile__ ( - "lw %[t1], 0(%[dest]) \n\t" - "lw %[t2], 4(%[dest]) \n\t" - "lw %[t3], 8(%[dest]) \n\t" - "lw %[t4], 12(%[dest]) \n\t" - "subu_s.qb %[vector_1], %[t1], %[vector_a1] \n\t" - "subu_s.qb %[vector_2], %[t2], %[vector_a1] \n\t" - "subu_s.qb %[vector_3], %[t3], %[vector_a1] \n\t" - "subu_s.qb %[vector_4], %[t4], %[vector_a1] \n\t" - "sw %[vector_1], 0(%[dest]) \n\t" - "sw %[vector_2], 4(%[dest]) \n\t" - "sw %[vector_3], 8(%[dest]) \n\t" - "sw %[vector_4], 12(%[dest]) \n\t" - - "lw %[t1], 16(%[dest]) \n\t" - "lw %[t2], 20(%[dest]) \n\t" - "lw %[t3], 24(%[dest]) \n\t" - "lw %[t4], 28(%[dest]) \n\t" - "subu_s.qb %[vector_1], %[t1], %[vector_a1] \n\t" - "subu_s.qb %[vector_2], %[t2], %[vector_a1] \n\t" - "subu_s.qb %[vector_3], %[t3], %[vector_a1] \n\t" - "subu_s.qb %[vector_4], %[t4], %[vector_a1] \n\t" - "sw %[vector_1], 16(%[dest]) \n\t" - "sw %[vector_2], 20(%[dest]) \n\t" - "sw %[vector_3], 24(%[dest]) \n\t" - "sw %[vector_4], 28(%[dest]) \n\t" - - "add %[dest], %[dest], %[stride] \n\t" - - : [t1] "=&r" (t1), [t2] "=&r" (t2), [t3] "=&r" (t3), [t4] "=&r" (t4), - [vector_1] "=&r" (vector_1), [vector_2] "=&r" (vector_2), - [vector_3] "=&r" (vector_3), [vector_4] "=&r" (vector_4), - [dest] "+&r" (dest) - : [stride] "r" (stride), [vector_a1] "r" (vector_a1) - ); - } - } else { - /* use quad-byte - * input and output memory are four byte aligned */ - __asm__ __volatile__ ( - "replv.qb %[vector_a1], %[a1] \n\t" - - : [vector_a1] "=r" (vector_a1) - : [a1] "r" (a1) - ); - - for (r = 32; r--;) { - __asm__ __volatile__ ( - "lw %[t1], 0(%[dest]) \n\t" - "lw %[t2], 4(%[dest]) \n\t" - "lw %[t3], 8(%[dest]) \n\t" - "lw %[t4], 12(%[dest]) \n\t" - "addu_s.qb %[vector_1], %[t1], %[vector_a1] \n\t" - "addu_s.qb %[vector_2], %[t2], %[vector_a1] \n\t" - "addu_s.qb %[vector_3], %[t3], %[vector_a1] \n\t" - "addu_s.qb %[vector_4], %[t4], %[vector_a1] \n\t" - "sw %[vector_1], 0(%[dest]) \n\t" - "sw %[vector_2], 4(%[dest]) \n\t" - "sw %[vector_3], 8(%[dest]) \n\t" - "sw %[vector_4], 12(%[dest]) \n\t" - - "lw %[t1], 16(%[dest]) \n\t" - "lw %[t2], 20(%[dest]) \n\t" - "lw %[t3], 24(%[dest]) \n\t" - "lw %[t4], 28(%[dest]) \n\t" - "addu_s.qb %[vector_1], %[t1], %[vector_a1] \n\t" - "addu_s.qb %[vector_2], %[t2], %[vector_a1] \n\t" - "addu_s.qb %[vector_3], %[t3], %[vector_a1] \n\t" - "addu_s.qb %[vector_4], %[t4], %[vector_a1] \n\t" - "sw %[vector_1], 16(%[dest]) \n\t" - "sw %[vector_2], 20(%[dest]) \n\t" - "sw %[vector_3], 24(%[dest]) \n\t" - "sw %[vector_4], 28(%[dest]) \n\t" - - "add %[dest], %[dest], %[stride] \n\t" - - : [t1] "=&r" (t1), [t2] "=&r" (t2), [t3] "=&r" (t3), [t4] "=&r" (t4), - [vector_1] "=&r" (vector_1), [vector_2] "=&r" (vector_2), - [vector_3] "=&r" (vector_3), [vector_4] "=&r" (vector_4), - [dest] "+&r" (dest) - : [stride] "r" (stride), [vector_a1] "r" (vector_a1) - ); - } - } -} -#endif // #if HAVE_DSPR2 diff --git a/vp9/common/mips/dspr2/vp9_itrans4_dspr2.c b/vp9/common/mips/dspr2/vp9_itrans4_dspr2.c index aa801ecb6..c10979b64 100644 --- a/vp9/common/mips/dspr2/vp9_itrans4_dspr2.c +++ b/vp9/common/mips/dspr2/vp9_itrans4_dspr2.c @@ -16,354 +16,11 @@ #include "vp9/common/vp9_common.h" #include "vp9/common/vp9_blockd.h" #include "vp9/common/vp9_idct.h" -#include "vp9/common/mips/dspr2/vp9_common_dspr2.h" +#include "vpx_dsp/mips/inv_txfm_dspr2.h" #include "vpx_dsp/txfm_common.h" #include "vpx_ports/mem.h" #if HAVE_DSPR2 -static void vp9_idct4_rows_dspr2(const int16_t *input, int16_t *output) { - int16_t step_0, step_1, step_2, step_3; - int Temp0, Temp1, Temp2, Temp3; - const int const_2_power_13 = 8192; - int i; - - for (i = 4; i--; ) { - __asm__ __volatile__ ( - /* - temp_1 = (input[0] + input[2]) * cospi_16_64; - step_0 = dct_const_round_shift(temp_1); - - temp_2 = (input[0] - input[2]) * cospi_16_64; - step_1 = dct_const_round_shift(temp_2); - */ - "lh %[Temp0], 0(%[input]) \n\t" - "lh %[Temp1], 4(%[input]) \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "add %[Temp2], %[Temp0], %[Temp1] \n\t" - "sub %[Temp3], %[Temp0], %[Temp1] \n\t" - "madd $ac0, %[Temp2], %[cospi_16_64] \n\t" - "lh %[Temp0], 2(%[input]) \n\t" - "lh %[Temp1], 6(%[input]) \n\t" - "extp %[step_0], $ac0, 31 \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - - "madd $ac1, %[Temp3], %[cospi_16_64] \n\t" - "extp %[step_1], $ac1, 31 \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - - /* - temp1 = input[1] * cospi_24_64 - input[3] * cospi_8_64; - step_2 = dct_const_round_shift(temp1); - */ - "madd $ac0, %[Temp0], %[cospi_24_64] \n\t" - "msub $ac0, %[Temp1], %[cospi_8_64] \n\t" - "extp %[step_2], $ac0, 31 \n\t" - - /* - temp2 = input[1] * cospi_8_64 + input[3] * cospi_24_64; - step_3 = dct_const_round_shift(temp2); - */ - "madd $ac1, %[Temp0], %[cospi_8_64] \n\t" - "madd $ac1, %[Temp1], %[cospi_24_64] \n\t" - "extp %[step_3], $ac1, 31 \n\t" - - /* - output[0] = step_0 + step_3; - output[4] = step_1 + step_2; - output[8] = step_1 - step_2; - output[12] = step_0 - step_3; - */ - "add %[Temp0], %[step_0], %[step_3] \n\t" - "sh %[Temp0], 0(%[output]) \n\t" - - "add %[Temp1], %[step_1], %[step_2] \n\t" - "sh %[Temp1], 8(%[output]) \n\t" - - "sub %[Temp2], %[step_1], %[step_2] \n\t" - "sh %[Temp2], 16(%[output]) \n\t" - - "sub %[Temp3], %[step_0], %[step_3] \n\t" - "sh %[Temp3], 24(%[output]) \n\t" - - : [Temp0] "=&r" (Temp0), [Temp1] "=&r" (Temp1), - [Temp2] "=&r" (Temp2), [Temp3] "=&r" (Temp3), - [step_0] "=&r" (step_0), [step_1] "=&r" (step_1), - [step_2] "=&r" (step_2), [step_3] "=&r" (step_3), - [output] "+r" (output) - : [const_2_power_13] "r" (const_2_power_13), - [cospi_8_64] "r" (cospi_8_64), [cospi_16_64] "r" (cospi_16_64), - [cospi_24_64] "r" (cospi_24_64), - [input] "r" (input) - ); - - input += 4; - output += 1; - } -} - -static void vp9_idct4_columns_add_blk_dspr2(int16_t *input, uint8_t *dest, - int dest_stride) { - int16_t step_0, step_1, step_2, step_3; - int Temp0, Temp1, Temp2, Temp3; - const int const_2_power_13 = 8192; - int i; - uint8_t *dest_pix; - uint8_t *cm = vpx_ff_cropTbl; - - /* prefetch vpx_ff_cropTbl */ - prefetch_load(vpx_ff_cropTbl); - prefetch_load(vpx_ff_cropTbl + 32); - prefetch_load(vpx_ff_cropTbl + 64); - prefetch_load(vpx_ff_cropTbl + 96); - prefetch_load(vpx_ff_cropTbl + 128); - prefetch_load(vpx_ff_cropTbl + 160); - prefetch_load(vpx_ff_cropTbl + 192); - prefetch_load(vpx_ff_cropTbl + 224); - - for (i = 0; i < 4; ++i) { - dest_pix = (dest + i); - - __asm__ __volatile__ ( - /* - temp_1 = (input[0] + input[2]) * cospi_16_64; - step_0 = dct_const_round_shift(temp_1); - - temp_2 = (input[0] - input[2]) * cospi_16_64; - step_1 = dct_const_round_shift(temp_2); - */ - "lh %[Temp0], 0(%[input]) \n\t" - "lh %[Temp1], 4(%[input]) \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "add %[Temp2], %[Temp0], %[Temp1] \n\t" - "sub %[Temp3], %[Temp0], %[Temp1] \n\t" - "madd $ac0, %[Temp2], %[cospi_16_64] \n\t" - "lh %[Temp0], 2(%[input]) \n\t" - "lh %[Temp1], 6(%[input]) \n\t" - "extp %[step_0], $ac0, 31 \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - - "madd $ac1, %[Temp3], %[cospi_16_64] \n\t" - "extp %[step_1], $ac1, 31 \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - - /* - temp1 = input[1] * cospi_24_64 - input[3] * cospi_8_64; - step_2 = dct_const_round_shift(temp1); - */ - "madd $ac0, %[Temp0], %[cospi_24_64] \n\t" - "msub $ac0, %[Temp1], %[cospi_8_64] \n\t" - "extp %[step_2], $ac0, 31 \n\t" - - /* - temp2 = input[1] * cospi_8_64 + input[3] * cospi_24_64; - step_3 = dct_const_round_shift(temp2); - */ - "madd $ac1, %[Temp0], %[cospi_8_64] \n\t" - "madd $ac1, %[Temp1], %[cospi_24_64] \n\t" - "extp %[step_3], $ac1, 31 \n\t" - - /* - output[0] = step_0 + step_3; - output[4] = step_1 + step_2; - output[8] = step_1 - step_2; - output[12] = step_0 - step_3; - */ - "add %[Temp0], %[step_0], %[step_3] \n\t" - "addi %[Temp0], %[Temp0], 8 \n\t" - "sra %[Temp0], %[Temp0], 4 \n\t" - "lbu %[Temp1], 0(%[dest_pix]) \n\t" - "add %[Temp1], %[Temp1], %[Temp0] \n\t" - "add %[Temp0], %[step_1], %[step_2] \n\t" - "lbux %[Temp2], %[Temp1](%[cm]) \n\t" - "sb %[Temp2], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "addi %[Temp0], %[Temp0], 8 \n\t" - "sra %[Temp0], %[Temp0], 4 \n\t" - "lbu %[Temp1], 0(%[dest_pix]) \n\t" - "add %[Temp1], %[Temp1], %[Temp0] \n\t" - "sub %[Temp0], %[step_1], %[step_2] \n\t" - "lbux %[Temp2], %[Temp1](%[cm]) \n\t" - "sb %[Temp2], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "addi %[Temp0], %[Temp0], 8 \n\t" - "sra %[Temp0], %[Temp0], 4 \n\t" - "lbu %[Temp1], 0(%[dest_pix]) \n\t" - "add %[Temp1], %[Temp1], %[Temp0] \n\t" - "sub %[Temp0], %[step_0], %[step_3] \n\t" - "lbux %[Temp2], %[Temp1](%[cm]) \n\t" - "sb %[Temp2], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "addi %[Temp0], %[Temp0], 8 \n\t" - "sra %[Temp0], %[Temp0], 4 \n\t" - "lbu %[Temp1], 0(%[dest_pix]) \n\t" - "add %[Temp1], %[Temp1], %[Temp0] \n\t" - "lbux %[Temp2], %[Temp1](%[cm]) \n\t" - "sb %[Temp2], 0(%[dest_pix]) \n\t" - - : [Temp0] "=&r" (Temp0), [Temp1] "=&r" (Temp1), - [Temp2] "=&r" (Temp2), [Temp3] "=&r" (Temp3), - [step_0] "=&r" (step_0), [step_1] "=&r" (step_1), - [step_2] "=&r" (step_2), [step_3] "=&r" (step_3), - [dest_pix] "+r" (dest_pix) - : [const_2_power_13] "r" (const_2_power_13), - [cospi_8_64] "r" (cospi_8_64), [cospi_16_64] "r" (cospi_16_64), - [cospi_24_64] "r" (cospi_24_64), - [input] "r" (input), [cm] "r" (cm), [dest_stride] "r" (dest_stride) - ); - - input += 4; - } -} - -void vp9_idct4x4_16_add_dspr2(const int16_t *input, uint8_t *dest, - int dest_stride) { - DECLARE_ALIGNED(32, int16_t, out[4 * 4]); - int16_t *outptr = out; - uint32_t pos = 45; - - /* bit positon for extract from acc */ - __asm__ __volatile__ ( - "wrdsp %[pos], 1 \n\t" - : - : [pos] "r" (pos) - ); - - // Rows - vp9_idct4_rows_dspr2(input, outptr); - - // Columns - vp9_idct4_columns_add_blk_dspr2(&out[0], dest, dest_stride); -} - -void vp9_idct4x4_1_add_dspr2(const int16_t *input, uint8_t *dest, - int dest_stride) { - int a1, absa1; - int r; - int32_t out; - int t2, vector_a1, vector_a; - uint32_t pos = 45; - int16_t input_dc = input[0]; - - /* bit positon for extract from acc */ - __asm__ __volatile__ ( - "wrdsp %[pos], 1 \n\t" - - : - : [pos] "r" (pos) - ); - - out = DCT_CONST_ROUND_SHIFT_TWICE_COSPI_16_64(input_dc); - __asm__ __volatile__ ( - "addi %[out], %[out], 8 \n\t" - "sra %[a1], %[out], 4 \n\t" - - : [out] "+r" (out), [a1] "=r" (a1) - : - ); - - if (a1 < 0) { - /* use quad-byte - * input and output memory are four byte aligned */ - __asm__ __volatile__ ( - "abs %[absa1], %[a1] \n\t" - "replv.qb %[vector_a1], %[absa1] \n\t" - - : [absa1] "=r" (absa1), [vector_a1] "=r" (vector_a1) - : [a1] "r" (a1) - ); - - for (r = 4; r--;) { - __asm__ __volatile__ ( - "lw %[t2], 0(%[dest]) \n\t" - "subu_s.qb %[vector_a], %[t2], %[vector_a1] \n\t" - "sw %[vector_a], 0(%[dest]) \n\t" - "add %[dest], %[dest], %[dest_stride] \n\t" - - : [t2] "=&r" (t2), [vector_a] "=&r" (vector_a), - [dest] "+&r" (dest) - : [dest_stride] "r" (dest_stride), [vector_a1] "r" (vector_a1) - ); - } - } else { - /* use quad-byte - * input and output memory are four byte aligned */ - __asm__ __volatile__ ( - "replv.qb %[vector_a1], %[a1] \n\t" - : [vector_a1] "=r" (vector_a1) - : [a1] "r" (a1) - ); - - for (r = 4; r--;) { - __asm__ __volatile__ ( - "lw %[t2], 0(%[dest]) \n\t" - "addu_s.qb %[vector_a], %[t2], %[vector_a1] \n\t" - "sw %[vector_a], 0(%[dest]) \n\t" - "add %[dest], %[dest], %[dest_stride] \n\t" - - : [t2] "=&r" (t2), [vector_a] "=&r" (vector_a), - [dest] "+&r" (dest) - : [dest_stride] "r" (dest_stride), [vector_a1] "r" (vector_a1) - ); - } - } -} - -static void iadst4_dspr2(const int16_t *input, int16_t *output) { - int s0, s1, s2, s3, s4, s5, s6, s7; - int x0, x1, x2, x3; - - x0 = input[0]; - x1 = input[1]; - x2 = input[2]; - x3 = input[3]; - - if (!(x0 | x1 | x2 | x3)) { - output[0] = output[1] = output[2] = output[3] = 0; - return; - } - - s0 = sinpi_1_9 * x0; - s1 = sinpi_2_9 * x0; - s2 = sinpi_3_9 * x1; - s3 = sinpi_4_9 * x2; - s4 = sinpi_1_9 * x2; - s5 = sinpi_2_9 * x3; - s6 = sinpi_4_9 * x3; - s7 = x0 - x2 + x3; - - x0 = s0 + s3 + s5; - x1 = s1 - s4 - s6; - x2 = sinpi_3_9 * s7; - x3 = s2; - - s0 = x0 + x3; - s1 = x1 + x3; - s2 = x2; - s3 = x0 + x1 - x3; - - // 1-D transform scaling factor is sqrt(2). - // The overall dynamic range is 14b (input) + 14b (multiplication scaling) - // + 1b (addition) = 29b. - // Hence the output bit depth is 15b. - output[0] = dct_const_round_shift(s0); - output[1] = dct_const_round_shift(s1); - output[2] = dct_const_round_shift(s2); - output[3] = dct_const_round_shift(s3); -} - void vp9_iht4x4_16_add_dspr2(const int16_t *input, uint8_t *dest, int dest_stride, int tx_type) { int i, j; @@ -381,11 +38,11 @@ void vp9_iht4x4_16_add_dspr2(const int16_t *input, uint8_t *dest, switch (tx_type) { case DCT_DCT: // DCT in both horizontal and vertical - vp9_idct4_rows_dspr2(input, outptr); - vp9_idct4_columns_add_blk_dspr2(&out[0], dest, dest_stride); + vpx_idct4_rows_dspr2(input, outptr); + vpx_idct4_columns_add_blk_dspr2(&out[0], dest, dest_stride); break; case ADST_DCT: // ADST in vertical, DCT in horizontal - vp9_idct4_rows_dspr2(input, outptr); + vpx_idct4_rows_dspr2(input, outptr); outptr = out; @@ -412,7 +69,7 @@ void vp9_iht4x4_16_add_dspr2(const int16_t *input, uint8_t *dest, temp_in[i * 4 + j] = out[j * 4 + i]; } } - vp9_idct4_columns_add_blk_dspr2(&temp_in[0], dest, dest_stride); + vpx_idct4_columns_add_blk_dspr2(&temp_in[0], dest, dest_stride); break; case ADST_ADST: // ADST in both directions for (i = 0; i < 4; ++i) { diff --git a/vp9/common/mips/dspr2/vp9_itrans8_dspr2.c b/vp9/common/mips/dspr2/vp9_itrans8_dspr2.c index 5270fa17f..37f3ca9fc 100644 --- a/vp9/common/mips/dspr2/vp9_itrans8_dspr2.c +++ b/vp9/common/mips/dspr2/vp9_itrans8_dspr2.c @@ -15,538 +15,11 @@ #include "./vp9_rtcd.h" #include "vp9/common/vp9_common.h" #include "vp9/common/vp9_blockd.h" -#include "vp9/common/mips/dspr2/vp9_common_dspr2.h" +#include "vpx_dsp/mips/inv_txfm_dspr2.h" #include "vpx_dsp/txfm_common.h" #include "vpx_ports/mem.h" #if HAVE_DSPR2 -static void idct8_rows_dspr2(const int16_t *input, int16_t *output, - uint32_t no_rows) { - int step1_0, step1_1, step1_2, step1_3, step1_4, step1_5, step1_6, step1_7; - const int const_2_power_13 = 8192; - int Temp0, Temp1, Temp2, Temp3, Temp4; - int i; - - for (i = no_rows; i--; ) { - __asm__ __volatile__ ( - /* - temp_1 = (input[0] + input[4]) * cospi_16_64; - step2_0 = dct_const_round_shift(temp_1); - - temp_2 = (input[0] - input[4]) * cospi_16_64; - step2_1 = dct_const_round_shift(temp_2); - */ - "lh %[Temp0], 0(%[input]) \n\t" - "lh %[Temp1], 8(%[input]) \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "add %[Temp2], %[Temp0], %[Temp1] \n\t" - "madd $ac0, %[Temp2], %[cospi_16_64] \n\t" - "extp %[Temp4], $ac0, 31 \n\t" - - "sub %[Temp3], %[Temp0], %[Temp1] \n\t" - "madd $ac1, %[Temp3], %[cospi_16_64] \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "extp %[Temp2], $ac1, 31 \n\t" - - /* - temp_1 = input[2] * cospi_24_64 - input[6] * cospi_8_64; - step2_2 = dct_const_round_shift(temp_1); - */ - "lh %[Temp0], 4(%[input]) \n\t" - "lh %[Temp1], 12(%[input]) \n\t" - "madd $ac0, %[Temp0], %[cospi_24_64] \n\t" - "msub $ac0, %[Temp1], %[cospi_8_64] \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "extp %[Temp3], $ac0, 31 \n\t" - - /* - step1_1 = step2_1 + step2_2; - step1_2 = step2_1 - step2_2; - */ - "add %[step1_1], %[Temp2], %[Temp3] \n\t" - "sub %[step1_2], %[Temp2], %[Temp3] \n\t" - - /* - temp_2 = input[2] * cospi_8_64 + input[6] * cospi_24_64; - step2_3 = dct_const_round_shift(temp_2); - */ - "madd $ac1, %[Temp0], %[cospi_8_64] \n\t" - "madd $ac1, %[Temp1], %[cospi_24_64] \n\t" - "extp %[Temp1], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - - /* - step1_0 = step2_0 + step2_3; - step1_3 = step2_0 - step2_3; - */ - "add %[step1_0], %[Temp4], %[Temp1] \n\t" - "sub %[step1_3], %[Temp4], %[Temp1] \n\t" - - /* - temp_1 = input[1] * cospi_28_64 - input[7] * cospi_4_64; - step1_4 = dct_const_round_shift(temp_1); - */ - "lh %[Temp0], 2(%[input]) \n\t" - "madd $ac0, %[Temp0], %[cospi_28_64] \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "lh %[Temp1], 14(%[input]) \n\t" - "lh %[Temp0], 2(%[input]) \n\t" - "msub $ac0, %[Temp1], %[cospi_4_64] \n\t" - "extp %[step1_4], $ac0, 31 \n\t" - - /* - temp_2 = input[1] * cospi_4_64 + input[7] * cospi_28_64; - step1_7 = dct_const_round_shift(temp_2); - */ - "madd $ac1, %[Temp0], %[cospi_4_64] \n\t" - "madd $ac1, %[Temp1], %[cospi_28_64] \n\t" - "extp %[step1_7], $ac1, 31 \n\t" - - /* - temp_1 = input[5] * cospi_12_64 - input[3] * cospi_20_64; - step1_5 = dct_const_round_shift(temp_1); - */ - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "lh %[Temp0], 10(%[input]) \n\t" - "madd $ac0, %[Temp0], %[cospi_12_64] \n\t" - "lh %[Temp1], 6(%[input]) \n\t" - "msub $ac0, %[Temp1], %[cospi_20_64] \n\t" - "extp %[step1_5], $ac0, 31 \n\t" - - /* - temp_2 = input[5] * cospi_20_64 + input[3] * cospi_12_64; - step1_6 = dct_const_round_shift(temp_2); - */ - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "lh %[Temp0], 10(%[input]) \n\t" - "madd $ac1, %[Temp0], %[cospi_20_64] \n\t" - "lh %[Temp1], 6(%[input]) \n\t" - "madd $ac1, %[Temp1], %[cospi_12_64] \n\t" - "extp %[step1_6], $ac1, 31 \n\t" - - /* - temp_1 = (step1_7 - step1_6 - step1_4 + step1_5) * cospi_16_64; - temp_2 = (step1_4 - step1_5 - step1_6 + step1_7) * cospi_16_64; - */ - "sub %[Temp0], %[step1_7], %[step1_6] \n\t" - "sub %[Temp0], %[Temp0], %[step1_4] \n\t" - "add %[Temp0], %[Temp0], %[step1_5] \n\t" - "sub %[Temp1], %[step1_4], %[step1_5] \n\t" - "sub %[Temp1], %[Temp1], %[step1_6] \n\t" - "add %[Temp1], %[Temp1], %[step1_7] \n\t" - - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - - "madd $ac0, %[Temp0], %[cospi_16_64] \n\t" - "madd $ac1, %[Temp1], %[cospi_16_64] \n\t" - - /* - step1_4 = step1_4 + step1_5; - step1_7 = step1_6 + step1_7; - */ - "add %[step1_4], %[step1_4], %[step1_5] \n\t" - "add %[step1_7], %[step1_7], %[step1_6] \n\t" - - "extp %[step1_5], $ac0, 31 \n\t" - "extp %[step1_6], $ac1, 31 \n\t" - - "add %[Temp0], %[step1_0], %[step1_7] \n\t" - "sh %[Temp0], 0(%[output]) \n\t" - "add %[Temp1], %[step1_1], %[step1_6] \n\t" - "sh %[Temp1], 16(%[output]) \n\t" - "add %[Temp0], %[step1_2], %[step1_5] \n\t" - "sh %[Temp0], 32(%[output]) \n\t" - "add %[Temp1], %[step1_3], %[step1_4] \n\t" - "sh %[Temp1], 48(%[output]) \n\t" - - "sub %[Temp0], %[step1_3], %[step1_4] \n\t" - "sh %[Temp0], 64(%[output]) \n\t" - "sub %[Temp1], %[step1_2], %[step1_5] \n\t" - "sh %[Temp1], 80(%[output]) \n\t" - "sub %[Temp0], %[step1_1], %[step1_6] \n\t" - "sh %[Temp0], 96(%[output]) \n\t" - "sub %[Temp1], %[step1_0], %[step1_7] \n\t" - "sh %[Temp1], 112(%[output]) \n\t" - - : [step1_0] "=&r" (step1_0), [step1_1] "=&r" (step1_1), - [step1_2] "=&r" (step1_2), [step1_3] "=&r" (step1_3), - [step1_4] "=&r" (step1_4), [step1_5] "=&r" (step1_5), - [step1_6] "=&r" (step1_6), [step1_7] "=&r" (step1_7), - [Temp0] "=&r" (Temp0), [Temp1] "=&r" (Temp1), - [Temp2] "=&r" (Temp2), [Temp3] "=&r" (Temp3), - [Temp4] "=&r" (Temp4) - : [const_2_power_13] "r" (const_2_power_13), - [cospi_16_64] "r" (cospi_16_64), [cospi_28_64] "r" (cospi_28_64), - [cospi_4_64] "r" (cospi_4_64), [cospi_12_64] "r" (cospi_12_64), - [cospi_20_64] "r" (cospi_20_64), [cospi_8_64] "r" (cospi_8_64), - [cospi_24_64] "r" (cospi_24_64), - [output] "r" (output), [input] "r" (input) - ); - - input += 8; - output += 1; - } -} - -static void idct8_columns_add_blk_dspr2(int16_t *input, uint8_t *dest, - int dest_stride) { - int step1_0, step1_1, step1_2, step1_3, step1_4, step1_5, step1_6, step1_7; - int Temp0, Temp1, Temp2, Temp3; - int i; - const int const_2_power_13 = 8192; - uint8_t *dest_pix; - uint8_t *cm = vpx_ff_cropTbl; - - /* prefetch vpx_ff_cropTbl */ - prefetch_load(vpx_ff_cropTbl); - prefetch_load(vpx_ff_cropTbl + 32); - prefetch_load(vpx_ff_cropTbl + 64); - prefetch_load(vpx_ff_cropTbl + 96); - prefetch_load(vpx_ff_cropTbl + 128); - prefetch_load(vpx_ff_cropTbl + 160); - prefetch_load(vpx_ff_cropTbl + 192); - prefetch_load(vpx_ff_cropTbl + 224); - - for (i = 0; i < 8; ++i) { - dest_pix = (dest + i); - - __asm__ __volatile__ ( - /* - temp_1 = (input[0] + input[4]) * cospi_16_64; - step2_0 = dct_const_round_shift(temp_1); - - temp_2 = (input[0] - input[4]) * cospi_16_64; - step2_1 = dct_const_round_shift(temp_2); - */ - "lh %[Temp0], 0(%[input]) \n\t" - "lh %[Temp1], 8(%[input]) \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "add %[Temp2], %[Temp0], %[Temp1] \n\t" - "madd $ac0, %[Temp2], %[cospi_16_64] \n\t" - "extp %[step1_6], $ac0, 31 \n\t" - - "sub %[Temp3], %[Temp0], %[Temp1] \n\t" - "madd $ac1, %[Temp3], %[cospi_16_64] \n\t" - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "extp %[Temp2], $ac1, 31 \n\t" - - /* - temp_1 = input[2] * cospi_24_64 - input[6] * cospi_8_64; - step2_2 = dct_const_round_shift(temp_1); - */ - "lh %[Temp0], 4(%[input]) \n\t" - "lh %[Temp1], 12(%[input]) \n\t" - "madd $ac0, %[Temp0], %[cospi_24_64] \n\t" - "msub $ac0, %[Temp1], %[cospi_8_64] \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "extp %[Temp3], $ac0, 31 \n\t" - - /* - step1_1 = step2_1 + step2_2; - step1_2 = step2_1 - step2_2; - */ - "add %[step1_1], %[Temp2], %[Temp3] \n\t" - "sub %[step1_2], %[Temp2], %[Temp3] \n\t" - - /* - temp_2 = input[2] * cospi_8_64 + input[6] * cospi_24_64; - step2_3 = dct_const_round_shift(temp_2); - */ - "madd $ac1, %[Temp0], %[cospi_8_64] \n\t" - "madd $ac1, %[Temp1], %[cospi_24_64] \n\t" - "extp %[Temp1], $ac1, 31 \n\t" - - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - - /* - step1_0 = step2_0 + step2_3; - step1_3 = step2_0 - step2_3; - */ - "add %[step1_0], %[step1_6], %[Temp1] \n\t" - "sub %[step1_3], %[step1_6], %[Temp1] \n\t" - - /* - temp_1 = input[1] * cospi_28_64 - input[7] * cospi_4_64; - step1_4 = dct_const_round_shift(temp_1); - */ - "lh %[Temp0], 2(%[input]) \n\t" - "madd $ac0, %[Temp0], %[cospi_28_64] \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "lh %[Temp1], 14(%[input]) \n\t" - "lh %[Temp0], 2(%[input]) \n\t" - "msub $ac0, %[Temp1], %[cospi_4_64] \n\t" - "extp %[step1_4], $ac0, 31 \n\t" - - /* - temp_2 = input[1] * cospi_4_64 + input[7] * cospi_28_64; - step1_7 = dct_const_round_shift(temp_2); - */ - "madd $ac1, %[Temp0], %[cospi_4_64] \n\t" - "madd $ac1, %[Temp1], %[cospi_28_64] \n\t" - "extp %[step1_7], $ac1, 31 \n\t" - - /* - temp_1 = input[5] * cospi_12_64 - input[3] * cospi_20_64; - step1_5 = dct_const_round_shift(temp_1); - */ - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "lh %[Temp0], 10(%[input]) \n\t" - "madd $ac0, %[Temp0], %[cospi_12_64] \n\t" - "lh %[Temp1], 6(%[input]) \n\t" - "msub $ac0, %[Temp1], %[cospi_20_64] \n\t" - "extp %[step1_5], $ac0, 31 \n\t" - - /* - temp_2 = input[5] * cospi_20_64 + input[3] * cospi_12_64; - step1_6 = dct_const_round_shift(temp_2); - */ - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - "lh %[Temp0], 10(%[input]) \n\t" - "madd $ac1, %[Temp0], %[cospi_20_64] \n\t" - "lh %[Temp1], 6(%[input]) \n\t" - "madd $ac1, %[Temp1], %[cospi_12_64] \n\t" - "extp %[step1_6], $ac1, 31 \n\t" - - /* - temp_1 = (step1_7 - step1_6 - step1_4 + step1_5) * cospi_16_64; - temp_2 = (step1_4 - step1_5 - step1_6 + step1_7) * cospi_16_64; - */ - "sub %[Temp0], %[step1_7], %[step1_6] \n\t" - "sub %[Temp0], %[Temp0], %[step1_4] \n\t" - "add %[Temp0], %[Temp0], %[step1_5] \n\t" - "sub %[Temp1], %[step1_4], %[step1_5] \n\t" - "sub %[Temp1], %[Temp1], %[step1_6] \n\t" - "add %[Temp1], %[Temp1], %[step1_7] \n\t" - - "mtlo %[const_2_power_13], $ac0 \n\t" - "mthi $zero, $ac0 \n\t" - "mtlo %[const_2_power_13], $ac1 \n\t" - "mthi $zero, $ac1 \n\t" - - "madd $ac0, %[Temp0], %[cospi_16_64] \n\t" - "madd $ac1, %[Temp1], %[cospi_16_64] \n\t" - - /* - step1_4 = step1_4 + step1_5; - step1_7 = step1_6 + step1_7; - */ - "add %[step1_4], %[step1_4], %[step1_5] \n\t" - "add %[step1_7], %[step1_7], %[step1_6] \n\t" - - "extp %[step1_5], $ac0, 31 \n\t" - "extp %[step1_6], $ac1, 31 \n\t" - - /* add block */ - "lbu %[Temp1], 0(%[dest_pix]) \n\t" - "add %[Temp0], %[step1_0], %[step1_7] \n\t" - "addi %[Temp0], %[Temp0], 16 \n\t" - "sra %[Temp0], %[Temp0], 5 \n\t" - "add %[Temp1], %[Temp1], %[Temp0] \n\t" - "add %[Temp0], %[step1_1], %[step1_6] \n\t" - "lbux %[Temp2], %[Temp1](%[cm]) \n\t" - "sb %[Temp2], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[Temp1], 0(%[dest_pix]) \n\t" - "addi %[Temp0], %[Temp0], 16 \n\t" - "sra %[Temp0], %[Temp0], 5 \n\t" - "add %[Temp1], %[Temp1], %[Temp0] \n\t" - "add %[Temp0], %[step1_2], %[step1_5] \n\t" - "lbux %[Temp2], %[Temp1](%[cm]) \n\t" - "sb %[Temp2], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[Temp1], 0(%[dest_pix]) \n\t" - "addi %[Temp0], %[Temp0], 16 \n\t" - "sra %[Temp0], %[Temp0], 5 \n\t" - "add %[Temp1], %[Temp1], %[Temp0] \n\t" - "add %[Temp0], %[step1_3], %[step1_4] \n\t" - "lbux %[Temp2], %[Temp1](%[cm]) \n\t" - "sb %[Temp2], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[Temp1], 0(%[dest_pix]) \n\t" - "addi %[Temp0], %[Temp0], 16 \n\t" - "sra %[Temp0], %[Temp0], 5 \n\t" - "add %[Temp1], %[Temp1], %[Temp0] \n\t" - "sub %[Temp0], %[step1_3], %[step1_4] \n\t" - "lbux %[Temp2], %[Temp1](%[cm]) \n\t" - "sb %[Temp2], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[Temp1], 0(%[dest_pix]) \n\t" - "addi %[Temp0], %[Temp0], 16 \n\t" - "sra %[Temp0], %[Temp0], 5 \n\t" - "add %[Temp1], %[Temp1], %[Temp0] \n\t" - "sub %[Temp0], %[step1_2], %[step1_5] \n\t" - "lbux %[Temp2], %[Temp1](%[cm]) \n\t" - "sb %[Temp2], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[Temp1], 0(%[dest_pix]) \n\t" - "addi %[Temp0], %[Temp0], 16 \n\t" - "sra %[Temp0], %[Temp0], 5 \n\t" - "add %[Temp1], %[Temp1], %[Temp0] \n\t" - "sub %[Temp0], %[step1_1], %[step1_6] \n\t" - "lbux %[Temp2], %[Temp1](%[cm]) \n\t" - "sb %[Temp2], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[Temp1], 0(%[dest_pix]) \n\t" - "addi %[Temp0], %[Temp0], 16 \n\t" - "sra %[Temp0], %[Temp0], 5 \n\t" - "add %[Temp1], %[Temp1], %[Temp0] \n\t" - "sub %[Temp0], %[step1_0], %[step1_7] \n\t" - "lbux %[Temp2], %[Temp1](%[cm]) \n\t" - "sb %[Temp2], 0(%[dest_pix]) \n\t" - "addu %[dest_pix], %[dest_pix], %[dest_stride] \n\t" - - "lbu %[Temp1], 0(%[dest_pix]) \n\t" - "addi %[Temp0], %[Temp0], 16 \n\t" - "sra %[Temp0], %[Temp0], 5 \n\t" - "add %[Temp1], %[Temp1], %[Temp0] \n\t" - "lbux %[Temp2], %[Temp1](%[cm]) \n\t" - "sb %[Temp2], 0(%[dest_pix]) \n\t" - - : [step1_0] "=&r" (step1_0), [step1_1] "=&r" (step1_1), - [step1_2] "=&r" (step1_2), [step1_3] "=&r" (step1_3), - [step1_4] "=&r" (step1_4), [step1_5] "=&r" (step1_5), - [step1_6] "=&r" (step1_6), [step1_7] "=&r" (step1_7), - [Temp0] "=&r" (Temp0), [Temp1] "=&r" (Temp1), - [Temp2] "=&r" (Temp2), [Temp3] "=&r" (Temp3), - [dest_pix] "+r" (dest_pix) - : [const_2_power_13] "r" (const_2_power_13), - [cospi_16_64] "r" (cospi_16_64), [cospi_28_64] "r" (cospi_28_64), - [cospi_4_64] "r" (cospi_4_64), [cospi_12_64] "r" (cospi_12_64), - [cospi_20_64] "r" (cospi_20_64), [cospi_8_64] "r" (cospi_8_64), - [cospi_24_64] "r" (cospi_24_64), - [input] "r" (input), [cm] "r" (cm), [dest_stride] "r" (dest_stride) - ); - - input += 8; - } -} - -void vp9_idct8x8_64_add_dspr2(const int16_t *input, uint8_t *dest, - int dest_stride) { - DECLARE_ALIGNED(32, int16_t, out[8 * 8]); - int16_t *outptr = out; - uint32_t pos = 45; - - /* bit positon for extract from acc */ - __asm__ __volatile__ ( - "wrdsp %[pos], 1 \n\t" - : - : [pos] "r" (pos) - ); - - // First transform rows - idct8_rows_dspr2(input, outptr, 8); - - // Then transform columns and add to dest - idct8_columns_add_blk_dspr2(&out[0], dest, dest_stride); -} - -static void iadst8_dspr2(const int16_t *input, int16_t *output) { - int s0, s1, s2, s3, s4, s5, s6, s7; - int x0, x1, x2, x3, x4, x5, x6, x7; - - x0 = input[7]; - x1 = input[0]; - x2 = input[5]; - x3 = input[2]; - x4 = input[3]; - x5 = input[4]; - x6 = input[1]; - x7 = input[6]; - - if (!(x0 | x1 | x2 | x3 | x4 | x5 | x6 | x7)) { - output[0] = output[1] = output[2] = output[3] = output[4] - = output[5] = output[6] = output[7] = 0; - return; - } - - // stage 1 - s0 = cospi_2_64 * x0 + cospi_30_64 * x1; - s1 = cospi_30_64 * x0 - cospi_2_64 * x1; - s2 = cospi_10_64 * x2 + cospi_22_64 * x3; - s3 = cospi_22_64 * x2 - cospi_10_64 * x3; - s4 = cospi_18_64 * x4 + cospi_14_64 * x5; - s5 = cospi_14_64 * x4 - cospi_18_64 * x5; - s6 = cospi_26_64 * x6 + cospi_6_64 * x7; - s7 = cospi_6_64 * x6 - cospi_26_64 * x7; - - x0 = ROUND_POWER_OF_TWO((s0 + s4), DCT_CONST_BITS); - x1 = ROUND_POWER_OF_TWO((s1 + s5), DCT_CONST_BITS); - x2 = ROUND_POWER_OF_TWO((s2 + s6), DCT_CONST_BITS); - x3 = ROUND_POWER_OF_TWO((s3 + s7), DCT_CONST_BITS); - x4 = ROUND_POWER_OF_TWO((s0 - s4), DCT_CONST_BITS); - x5 = ROUND_POWER_OF_TWO((s1 - s5), DCT_CONST_BITS); - x6 = ROUND_POWER_OF_TWO((s2 - s6), DCT_CONST_BITS); - x7 = ROUND_POWER_OF_TWO((s3 - s7), DCT_CONST_BITS); - - // stage 2 - s0 = x0; - s1 = x1; - s2 = x2; - s3 = x3; - s4 = cospi_8_64 * x4 + cospi_24_64 * x5; - s5 = cospi_24_64 * x4 - cospi_8_64 * x5; - s6 = -cospi_24_64 * x6 + cospi_8_64 * x7; - s7 = cospi_8_64 * x6 + cospi_24_64 * x7; - - x0 = s0 + s2; - x1 = s1 + s3; - x2 = s0 - s2; - x3 = s1 - s3; - x4 = ROUND_POWER_OF_TWO((s4 + s6), DCT_CONST_BITS); - x5 = ROUND_POWER_OF_TWO((s5 + s7), DCT_CONST_BITS); - x6 = ROUND_POWER_OF_TWO((s4 - s6), DCT_CONST_BITS); - x7 = ROUND_POWER_OF_TWO((s5 - s7), DCT_CONST_BITS); - - // stage 3 - s2 = cospi_16_64 * (x2 + x3); - s3 = cospi_16_64 * (x2 - x3); - s6 = cospi_16_64 * (x6 + x7); - s7 = cospi_16_64 * (x6 - x7); - - x2 = ROUND_POWER_OF_TWO((s2), DCT_CONST_BITS); - x3 = ROUND_POWER_OF_TWO((s3), DCT_CONST_BITS); - x6 = ROUND_POWER_OF_TWO((s6), DCT_CONST_BITS); - x7 = ROUND_POWER_OF_TWO((s7), DCT_CONST_BITS); - - output[0] = x0; - output[1] = -x4; - output[2] = x6; - output[3] = -x2; - output[4] = x3; - output[5] = -x7; - output[6] = x5; - output[7] = -x1; -} - void vp9_iht8x8_64_add_dspr2(const int16_t *input, uint8_t *dest, int dest_stride, int tx_type) { int i, j; @@ -617,130 +90,4 @@ void vp9_iht8x8_64_add_dspr2(const int16_t *input, uint8_t *dest, break; } } - -void vp9_idct8x8_12_add_dspr2(const int16_t *input, uint8_t *dest, - int dest_stride) { - DECLARE_ALIGNED(32, int16_t, out[8 * 8]); - int16_t *outptr = out; - uint32_t pos = 45; - - /* bit positon for extract from acc */ - __asm__ __volatile__ ( - "wrdsp %[pos], 1 \n\t" - : - : [pos] "r" (pos) - ); - - // First transform rows - idct8_rows_dspr2(input, outptr, 4); - - outptr += 4; - - __asm__ __volatile__ ( - "sw $zero, 0(%[outptr]) \n\t" - "sw $zero, 4(%[outptr]) \n\t" - "sw $zero, 16(%[outptr]) \n\t" - "sw $zero, 20(%[outptr]) \n\t" - "sw $zero, 32(%[outptr]) \n\t" - "sw $zero, 36(%[outptr]) \n\t" - "sw $zero, 48(%[outptr]) \n\t" - "sw $zero, 52(%[outptr]) \n\t" - "sw $zero, 64(%[outptr]) \n\t" - "sw $zero, 68(%[outptr]) \n\t" - "sw $zero, 80(%[outptr]) \n\t" - "sw $zero, 84(%[outptr]) \n\t" - "sw $zero, 96(%[outptr]) \n\t" - "sw $zero, 100(%[outptr]) \n\t" - "sw $zero, 112(%[outptr]) \n\t" - "sw $zero, 116(%[outptr]) \n\t" - - : - : [outptr] "r" (outptr) - ); - - - // Then transform columns and add to dest - idct8_columns_add_blk_dspr2(&out[0], dest, dest_stride); -} - -void vp9_idct8x8_1_add_dspr2(const int16_t *input, uint8_t *dest, - int dest_stride) { - uint32_t pos = 45; - int32_t out; - int32_t r; - int32_t a1, absa1; - int32_t t1, t2, vector_a1, vector_1, vector_2; - - /* bit positon for extract from acc */ - __asm__ __volatile__ ( - "wrdsp %[pos], 1 \n\t" - - : - : [pos] "r" (pos) - ); - - out = DCT_CONST_ROUND_SHIFT_TWICE_COSPI_16_64(input[0]); - __asm__ __volatile__ ( - "addi %[out], %[out], 16 \n\t" - "sra %[a1], %[out], 5 \n\t" - - : [out] "+r" (out), [a1] "=r" (a1) - : - ); - - if (a1 < 0) { - /* use quad-byte - * input and output memory are four byte aligned */ - __asm__ __volatile__ ( - "abs %[absa1], %[a1] \n\t" - "replv.qb %[vector_a1], %[absa1] \n\t" - - : [absa1] "=r" (absa1), [vector_a1] "=r" (vector_a1) - : [a1] "r" (a1) - ); - - for (r = 8; r--;) { - __asm__ __volatile__ ( - "lw %[t1], 0(%[dest]) \n\t" - "lw %[t2], 4(%[dest]) \n\t" - "subu_s.qb %[vector_1], %[t1], %[vector_a1] \n\t" - "subu_s.qb %[vector_2], %[t2], %[vector_a1] \n\t" - "sw %[vector_1], 0(%[dest]) \n\t" - "sw %[vector_2], 4(%[dest]) \n\t" - "add %[dest], %[dest], %[dest_stride] \n\t" - - : [t1] "=&r" (t1), [t2] "=&r" (t2), - [vector_1] "=&r" (vector_1), [vector_2] "=&r" (vector_2), - [dest] "+&r" (dest) - : [dest_stride] "r" (dest_stride), [vector_a1] "r" (vector_a1) - ); - } - } else { - /* use quad-byte - * input and output memory are four byte aligned */ - __asm__ __volatile__ ( - "replv.qb %[vector_a1], %[a1] \n\t" - - : [vector_a1] "=r" (vector_a1) - : [a1] "r" (a1) - ); - - for (r = 8; r--;) { - __asm__ __volatile__ ( - "lw %[t1], 0(%[dest]) \n\t" - "lw %[t2], 4(%[dest]) \n\t" - "addu_s.qb %[vector_1], %[t1], %[vector_a1] \n\t" - "addu_s.qb %[vector_2], %[t2], %[vector_a1] \n\t" - "sw %[vector_1], 0(%[dest]) \n\t" - "sw %[vector_2], 4(%[dest]) \n\t" - "add %[dest], %[dest], %[dest_stride] \n\t" - - : [t1] "=&r" (t1), [t2] "=&r" (t2), - [vector_1] "=&r" (vector_1), [vector_2] "=&r" (vector_2), - [dest] "+r" (dest) - : [dest_stride] "r" (dest_stride), [vector_a1] "r" (vector_a1) - ); - } - } -} #endif // #if HAVE_DSPR2 diff --git a/vp9/common/mips/msa/vp9_idct16x16_msa.c b/vp9/common/mips/msa/vp9_idct16x16_msa.c index 59b8930b5..5adf0aaac 100644 --- a/vp9/common/mips/msa/vp9_idct16x16_msa.c +++ b/vp9/common/mips/msa/vp9_idct16x16_msa.c @@ -24,13 +24,13 @@ void vp9_iht16x16_256_add_msa(const int16_t *input, uint8_t *dst, /* transform rows */ for (i = 0; i < 2; ++i) { /* process 16 * 8 block */ - vp9_idct16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7))); + vpx_idct16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7))); } /* transform columns */ for (i = 0; i < 2; ++i) { /* process 8 * 16 block */ - vp9_idct16_1d_columns_addblk_msa((out_ptr + (i << 3)), (dst + (i << 3)), + vpx_idct16_1d_columns_addblk_msa((out_ptr + (i << 3)), (dst + (i << 3)), dst_stride); } break; @@ -38,12 +38,12 @@ void vp9_iht16x16_256_add_msa(const int16_t *input, uint8_t *dst, /* transform rows */ for (i = 0; i < 2; ++i) { /* process 16 * 8 block */ - vp9_idct16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7))); + vpx_idct16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7))); } /* transform columns */ for (i = 0; i < 2; ++i) { - vp9_iadst16_1d_columns_addblk_msa((out_ptr + (i << 3)), + vpx_iadst16_1d_columns_addblk_msa((out_ptr + (i << 3)), (dst + (i << 3)), dst_stride); } break; @@ -51,13 +51,13 @@ void vp9_iht16x16_256_add_msa(const int16_t *input, uint8_t *dst, /* transform rows */ for (i = 0; i < 2; ++i) { /* process 16 * 8 block */ - vp9_iadst16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7))); + vpx_iadst16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7))); } /* transform columns */ for (i = 0; i < 2; ++i) { /* process 8 * 16 block */ - vp9_idct16_1d_columns_addblk_msa((out_ptr + (i << 3)), (dst + (i << 3)), + vpx_idct16_1d_columns_addblk_msa((out_ptr + (i << 3)), (dst + (i << 3)), dst_stride); } break; @@ -65,12 +65,12 @@ void vp9_iht16x16_256_add_msa(const int16_t *input, uint8_t *dst, /* transform rows */ for (i = 0; i < 2; ++i) { /* process 16 * 8 block */ - vp9_iadst16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7))); + vpx_iadst16_1d_rows_msa((input + (i << 7)), (out_ptr + (i << 7))); } /* transform columns */ for (i = 0; i < 2; ++i) { - vp9_iadst16_1d_columns_addblk_msa((out_ptr + (i << 3)), + vpx_iadst16_1d_columns_addblk_msa((out_ptr + (i << 3)), (dst + (i << 3)), dst_stride); } break; diff --git a/vp9/common/vp9_alloccommon.c b/vp9/common/vp9_alloccommon.c index 8eda491de..a3e01bf38 100644 --- a/vp9/common/vp9_alloccommon.c +++ b/vp9/common/vp9_alloccommon.c @@ -16,7 +16,6 @@ #include "vp9/common/vp9_entropymode.h" #include "vp9/common/vp9_entropymv.h" #include "vp9/common/vp9_onyxc_int.h" -#include "vp9/common/vp9_systemdependent.h" // TODO(hkuang): Don't need to lock the whole pool after implementing atomic // frame reference count. diff --git a/vp9/common/vp9_common.h b/vp9/common/vp9_common.h index c249ad4d7..76e7cd440 100644 --- a/vp9/common/vp9_common.h +++ b/vp9/common/vp9_common.h @@ -19,7 +19,7 @@ #include "vpx_dsp/vpx_dsp_common.h" #include "vpx_mem/vpx_mem.h" #include "vpx/vpx_integer.h" -#include "vp9/common/vp9_systemdependent.h" +#include "vpx_ports/bitops.h" #ifdef __cplusplus extern "C" { diff --git a/vp9/common/vp9_idct.c b/vp9/common/vp9_idct.c index b15f7f370..d12cd76db 100644 --- a/vp9/common/vp9_idct.c +++ b/vp9/common/vp9_idct.c @@ -14,7 +14,6 @@ #include "./vpx_dsp_rtcd.h" #include "vp9/common/vp9_blockd.h" #include "vp9/common/vp9_idct.h" -#include "vp9/common/vp9_systemdependent.h" #include "vpx_dsp/inv_txfm.h" #include "vpx_ports/mem.h" @@ -123,18 +122,18 @@ void vp9_iht16x16_256_add_c(const tran_low_t *input, uint8_t *dest, int stride, void vp9_idct4x4_add(const tran_low_t *input, uint8_t *dest, int stride, int eob) { if (eob > 1) - vp9_idct4x4_16_add(input, dest, stride); + vpx_idct4x4_16_add(input, dest, stride); else - vp9_idct4x4_1_add(input, dest, stride); + vpx_idct4x4_1_add(input, dest, stride); } void vp9_iwht4x4_add(const tran_low_t *input, uint8_t *dest, int stride, int eob) { if (eob > 1) - vp9_iwht4x4_16_add(input, dest, stride); + vpx_iwht4x4_16_add(input, dest, stride); else - vp9_iwht4x4_1_add(input, dest, stride); + vpx_iwht4x4_1_add(input, dest, stride); } void vp9_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride, @@ -148,11 +147,11 @@ void vp9_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride, // Combine that with code here. if (eob == 1) // DC only DCT coefficient - vp9_idct8x8_1_add(input, dest, stride); + vpx_idct8x8_1_add(input, dest, stride); else if (eob <= 12) - vp9_idct8x8_12_add(input, dest, stride); + vpx_idct8x8_12_add(input, dest, stride); else - vp9_idct8x8_64_add(input, dest, stride); + vpx_idct8x8_64_add(input, dest, stride); } void vp9_idct16x16_add(const tran_low_t *input, uint8_t *dest, int stride, @@ -161,22 +160,22 @@ void vp9_idct16x16_add(const tran_low_t *input, uint8_t *dest, int stride, * coefficients. Use eobs to separate different cases. */ if (eob == 1) /* DC only DCT coefficient. */ - vp9_idct16x16_1_add(input, dest, stride); + vpx_idct16x16_1_add(input, dest, stride); else if (eob <= 10) - vp9_idct16x16_10_add(input, dest, stride); + vpx_idct16x16_10_add(input, dest, stride); else - vp9_idct16x16_256_add(input, dest, stride); + vpx_idct16x16_256_add(input, dest, stride); } void vp9_idct32x32_add(const tran_low_t *input, uint8_t *dest, int stride, int eob) { if (eob == 1) - vp9_idct32x32_1_add(input, dest, stride); + vpx_idct32x32_1_add(input, dest, stride); else if (eob <= 34) // non-zero coeff only in upper-left 8x8 - vp9_idct32x32_34_add(input, dest, stride); + vpx_idct32x32_34_add(input, dest, stride); else - vp9_idct32x32_1024_add(input, dest, stride); + vpx_idct32x32_1024_add(input, dest, stride); } // iht @@ -210,10 +209,10 @@ void vp9_iht16x16_add(TX_TYPE tx_type, const tran_low_t *input, uint8_t *dest, void vp9_highbd_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8, int stride, int tx_type, int bd) { const highbd_transform_2d IHT_4[] = { - { vp9_highbd_idct4_c, vp9_highbd_idct4_c }, // DCT_DCT = 0 - { highbd_iadst4_c, vp9_highbd_idct4_c }, // ADST_DCT = 1 - { vp9_highbd_idct4_c, highbd_iadst4_c }, // DCT_ADST = 2 - { highbd_iadst4_c, highbd_iadst4_c } // ADST_ADST = 3 + { vpx_highbd_idct4_c, vpx_highbd_idct4_c }, // DCT_DCT = 0 + { vpx_highbd_iadst4_c, vpx_highbd_idct4_c }, // ADST_DCT = 1 + { vpx_highbd_idct4_c, vpx_highbd_iadst4_c }, // DCT_ADST = 2 + { vpx_highbd_iadst4_c, vpx_highbd_iadst4_c } // ADST_ADST = 3 }; uint16_t *dest = CONVERT_TO_SHORTPTR(dest8); @@ -242,10 +241,10 @@ void vp9_highbd_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest8, } static const highbd_transform_2d HIGH_IHT_8[] = { - { vp9_highbd_idct8_c, vp9_highbd_idct8_c }, // DCT_DCT = 0 - { highbd_iadst8_c, vp9_highbd_idct8_c }, // ADST_DCT = 1 - { vp9_highbd_idct8_c, highbd_iadst8_c }, // DCT_ADST = 2 - { highbd_iadst8_c, highbd_iadst8_c } // ADST_ADST = 3 + { vpx_highbd_idct8_c, vpx_highbd_idct8_c }, // DCT_DCT = 0 + { vpx_highbd_iadst8_c, vpx_highbd_idct8_c }, // ADST_DCT = 1 + { vpx_highbd_idct8_c, vpx_highbd_iadst8_c }, // DCT_ADST = 2 + { vpx_highbd_iadst8_c, vpx_highbd_iadst8_c } // ADST_ADST = 3 }; void vp9_highbd_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest8, @@ -277,10 +276,10 @@ void vp9_highbd_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest8, } static const highbd_transform_2d HIGH_IHT_16[] = { - { vp9_highbd_idct16_c, vp9_highbd_idct16_c }, // DCT_DCT = 0 - { highbd_iadst16_c, vp9_highbd_idct16_c }, // ADST_DCT = 1 - { vp9_highbd_idct16_c, highbd_iadst16_c }, // DCT_ADST = 2 - { highbd_iadst16_c, highbd_iadst16_c } // ADST_ADST = 3 + { vpx_highbd_idct16_c, vpx_highbd_idct16_c }, // DCT_DCT = 0 + { vpx_highbd_iadst16_c, vpx_highbd_idct16_c }, // ADST_DCT = 1 + { vpx_highbd_idct16_c, vpx_highbd_iadst16_c }, // DCT_ADST = 2 + { vpx_highbd_iadst16_c, vpx_highbd_iadst16_c } // ADST_ADST = 3 }; void vp9_highbd_iht16x16_256_add_c(const tran_low_t *input, uint8_t *dest8, @@ -315,18 +314,18 @@ void vp9_highbd_iht16x16_256_add_c(const tran_low_t *input, uint8_t *dest8, void vp9_highbd_idct4x4_add(const tran_low_t *input, uint8_t *dest, int stride, int eob, int bd) { if (eob > 1) - vp9_highbd_idct4x4_16_add(input, dest, stride, bd); + vpx_highbd_idct4x4_16_add(input, dest, stride, bd); else - vp9_highbd_idct4x4_1_add(input, dest, stride, bd); + vpx_highbd_idct4x4_1_add(input, dest, stride, bd); } void vp9_highbd_iwht4x4_add(const tran_low_t *input, uint8_t *dest, int stride, int eob, int bd) { if (eob > 1) - vp9_highbd_iwht4x4_16_add(input, dest, stride, bd); + vpx_highbd_iwht4x4_16_add(input, dest, stride, bd); else - vp9_highbd_iwht4x4_1_add(input, dest, stride, bd); + vpx_highbd_iwht4x4_1_add(input, dest, stride, bd); } void vp9_highbd_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride, @@ -340,11 +339,11 @@ void vp9_highbd_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride, // Combine that with code here. // DC only DCT coefficient if (eob == 1) { - vp9_highbd_idct8x8_1_add(input, dest, stride, bd); + vpx_highbd_idct8x8_1_add(input, dest, stride, bd); } else if (eob <= 10) { - vp9_highbd_idct8x8_10_add(input, dest, stride, bd); + vpx_highbd_idct8x8_10_add(input, dest, stride, bd); } else { - vp9_highbd_idct8x8_64_add(input, dest, stride, bd); + vpx_highbd_idct8x8_64_add(input, dest, stride, bd); } } @@ -354,11 +353,11 @@ void vp9_highbd_idct16x16_add(const tran_low_t *input, uint8_t *dest, // coefficients. Use eobs to separate different cases. // DC only DCT coefficient. if (eob == 1) { - vp9_highbd_idct16x16_1_add(input, dest, stride, bd); + vpx_highbd_idct16x16_1_add(input, dest, stride, bd); } else if (eob <= 10) { - vp9_highbd_idct16x16_10_add(input, dest, stride, bd); + vpx_highbd_idct16x16_10_add(input, dest, stride, bd); } else { - vp9_highbd_idct16x16_256_add(input, dest, stride, bd); + vpx_highbd_idct16x16_256_add(input, dest, stride, bd); } } @@ -366,11 +365,11 @@ void vp9_highbd_idct32x32_add(const tran_low_t *input, uint8_t *dest, int stride, int eob, int bd) { // Non-zero coeff only in upper-left 8x8 if (eob == 1) { - vp9_highbd_idct32x32_1_add(input, dest, stride, bd); + vpx_highbd_idct32x32_1_add(input, dest, stride, bd); } else if (eob <= 34) { - vp9_highbd_idct32x32_34_add(input, dest, stride, bd); + vpx_highbd_idct32x32_34_add(input, dest, stride, bd); } else { - vp9_highbd_idct32x32_1024_add(input, dest, stride, bd); + vpx_highbd_idct32x32_1024_add(input, dest, stride, bd); } } diff --git a/vp9/common/vp9_idct.h b/vp9/common/vp9_idct.h index 7a7dc1d64..b5a3fbf36 100644 --- a/vp9/common/vp9_idct.h +++ b/vp9/common/vp9_idct.h @@ -44,8 +44,8 @@ void vp9_idct4x4_add(const tran_low_t *input, uint8_t *dest, int stride, int eob); void vp9_idct8x8_add(const tran_low_t *input, uint8_t *dest, int stride, int eob); -void vp9_idct16x16_add(const tran_low_t *input, uint8_t *dest, int stride, int - eob); +void vp9_idct16x16_add(const tran_low_t *input, uint8_t *dest, int stride, + int eob); void vp9_idct32x32_add(const tran_low_t *input, uint8_t *dest, int stride, int eob); diff --git a/vp9/common/vp9_mfqe.c b/vp9/common/vp9_mfqe.c index bebb37eda..6d560f438 100644 --- a/vp9/common/vp9_mfqe.c +++ b/vp9/common/vp9_mfqe.c @@ -120,8 +120,8 @@ static void copy_mem32x32(const uint8_t *src, int src_stride, dst + dst_stride * 16 + 16, dst_stride); } -void copy_mem64x64(const uint8_t *src, int src_stride, - uint8_t *dst, int dst_stride) { +static void copy_mem64x64(const uint8_t *src, int src_stride, + uint8_t *dst, int dst_stride) { copy_mem32x32(src, src_stride, dst, dst_stride); copy_mem32x32(src + 32, src_stride, dst + 32, dst_stride); copy_mem32x32(src + src_stride * 32, src_stride, diff --git a/vp9/common/vp9_postproc.c b/vp9/common/vp9_postproc.c index d26a6eb5c..1843bb95c 100644 --- a/vp9/common/vp9_postproc.c +++ b/vp9/common/vp9_postproc.c @@ -17,20 +17,20 @@ #include "./vp9_rtcd.h" #include "vpx_ports/mem.h" +#include "vpx_ports/system_state.h" #include "vpx_scale/vpx_scale.h" #include "vpx_scale/yv12config.h" #include "vp9/common/vp9_onyxc_int.h" #include "vp9/common/vp9_postproc.h" -#include "vp9/common/vp9_systemdependent.h" #include "vp9/common/vp9_textblit.h" #if CONFIG_VP9_POSTPROC -static const short kernel5[] = { +static const int16_t kernel5[] = { 1, 1, 4, 1, 1 }; -const short vp9_rv[] = { +const int16_t vp9_rv[] = { 8, 5, 2, 2, 8, 12, 4, 9, 8, 3, 0, 3, 9, 0, 0, 0, 8, 3, 14, 4, 10, 1, 11, 14, 1, 14, 9, 6, 12, 11, @@ -320,7 +320,7 @@ void vp9_mbpost_proc_down_c(uint8_t *dst, int pitch, int sumsq = 0; int sum = 0; uint8_t d[16]; - const short *rv2 = rv3 + ((c * 17) & 127); + const int16_t *rv2 = rv3 + ((c * 17) & 127); for (i = -8; i <= 6; i++) { sumsq += s[i * pitch] * s[i * pitch]; @@ -544,7 +544,7 @@ static void fillrd(struct postproc_state *state, int q, int a) { double sigma; int ai = a, qi = q, i; - vp9_clear_system_state(); + vpx_clear_system_state(); sigma = ai + .5 + .6 * (63 - qi) / 63.0; @@ -638,7 +638,7 @@ int vp9_post_proc_frame(struct VP9Common *cm, return 0; } - vp9_clear_system_state(); + vpx_clear_system_state(); // Alloc memory for prev_mip in the first frame. if (cm->current_video_frame == 1) { diff --git a/vp9/common/vp9_reconinter.c b/vp9/common/vp9_reconinter.c index db9971da7..f83f8257a 100644 --- a/vp9/common/vp9_reconinter.c +++ b/vp9/common/vp9_reconinter.c @@ -153,7 +153,7 @@ MV average_split_mvs(const struct macroblockd_plane *pd, return res; } -void build_inter_predictors(MACROBLOCKD *xd, int plane, int block, +static void build_inter_predictors(MACROBLOCKD *xd, int plane, int block, int bw, int bh, int x, int y, int w, int h, int mi_x, int mi_y) { diff --git a/vp9/common/vp9_reconinter.h b/vp9/common/vp9_reconinter.h index 9bc62900a..7d907748e 100644 --- a/vp9/common/vp9_reconinter.h +++ b/vp9/common/vp9_reconinter.h @@ -50,11 +50,6 @@ MV average_split_mvs(const struct macroblockd_plane *pd, const MODE_INFO *mi, MV clamp_mv_to_umv_border_sb(const MACROBLOCKD *xd, const MV *src_mv, int bw, int bh, int ss_x, int ss_y); -void build_inter_predictors(MACROBLOCKD *xd, int plane, int block, - int bw, int bh, - int x, int y, int w, int h, - int mi_x, int mi_y); - void vp9_build_inter_predictors_sby(MACROBLOCKD *xd, int mi_row, int mi_col, BLOCK_SIZE bsize); diff --git a/vp9/common/vp9_rtcd_defs.pl b/vp9/common/vp9_rtcd_defs.pl index c1731012f..737fc56dc 100644 --- a/vp9/common/vp9_rtcd_defs.pl +++ b/vp9/common/vp9_rtcd_defs.pl @@ -209,7 +209,7 @@ add_proto qw/int16_t vp9_int_pro_col/, "uint8_t const *ref, const int width"; specialize qw/vp9_int_pro_col sse2 neon/; add_proto qw/int vp9_vector_var/, "int16_t const *ref, int16_t const *src, const int bwl"; -specialize qw/vp9_vector_var sse2/; +specialize qw/vp9_vector_var neon sse2/; if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") { add_proto qw/unsigned int vp9_highbd_avg_8x8/, "const uint8_t *, int p"; @@ -261,17 +261,6 @@ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") { specialize qw/vp9_fdct8x8_quant sse2 ssse3 neon/; } -# -# Structured Similarity (SSIM) -# -if (vpx_config("CONFIG_INTERNAL_STATS") eq "yes") { - add_proto qw/void vp9_ssim_parms_8x8/, "uint8_t *s, int sp, uint8_t *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr"; - specialize qw/vp9_ssim_parms_8x8/, "$sse2_x86_64"; - - add_proto qw/void vp9_ssim_parms_16x16/, "uint8_t *s, int sp, uint8_t *r, int rp, unsigned long *sum_s, unsigned long *sum_r, unsigned long *sum_sq_s, unsigned long *sum_sq_r, unsigned long *sum_sxr"; - specialize qw/vp9_ssim_parms_16x16/, "$sse2_x86_64"; -} - # fdct functions if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") { @@ -330,14 +319,6 @@ if (vpx_config("CONFIG_VP9_HIGHBITDEPTH") eq "yes") { add_proto qw/void vp9_highbd_quantize_fp_32x32/, "const tran_low_t *coeff_ptr, intptr_t n_coeffs, int skip_block, const int16_t *zbin_ptr, const int16_t *round_ptr, const int16_t *quant_ptr, const int16_t *quant_shift_ptr, tran_low_t *qcoeff_ptr, tran_low_t *dqcoeff_ptr, const int16_t *dequant_ptr, uint16_t *eob_ptr, const int16_t *scan, const int16_t *iscan"; specialize qw/vp9_highbd_quantize_fp_32x32/; - # - # Structured Similarity (SSIM) - # - if (vpx_config("CONFIG_INTERNAL_STATS") eq "yes") { - add_proto qw/void vp9_highbd_ssim_parms_8x8/, "uint16_t *s, int sp, uint16_t *r, int rp, uint32_t *sum_s, uint32_t *sum_r, uint32_t *sum_sq_s, uint32_t *sum_sq_r, uint32_t *sum_sxr"; - specialize qw/vp9_highbd_ssim_parms_8x8/; - } - # fdct functions add_proto qw/void vp9_highbd_fht4x4/, "const int16_t *input, tran_low_t *output, int stride, int tx_type"; specialize qw/vp9_highbd_fht4x4/; diff --git a/vp9/common/vp9_systemdependent.h b/vp9/common/vp9_systemdependent.h deleted file mode 100644 index fc77762de..000000000 --- a/vp9/common/vp9_systemdependent.h +++ /dev/null @@ -1,84 +0,0 @@ -/* - * Copyright (c) 2010 The WebM project authors. All Rights Reserved. - * - * Use of this source code is governed by a BSD-style license - * that can be found in the LICENSE file in the root of the source - * tree. An additional intellectual property rights grant can be found - * in the file PATENTS. All contributing project authors may - * be found in the AUTHORS file in the root of the source tree. - */ - -#ifndef VP9_COMMON_VP9_SYSTEMDEPENDENT_H_ -#define VP9_COMMON_VP9_SYSTEMDEPENDENT_H_ - -#include "vpx_ports/msvc.h" - -#ifdef _MSC_VER -# include <math.h> // the ceil() definition must precede intrin.h -# if _MSC_VER > 1310 && (defined(_M_X64) || defined(_M_IX86)) -# include <intrin.h> -# define USE_MSC_INTRINSICS -# endif -#endif - -#ifdef __cplusplus -extern "C" { -#endif - -#include "./vpx_config.h" -#if ARCH_X86 || ARCH_X86_64 -void vpx_reset_mmx_state(void); -#define vp9_clear_system_state() vpx_reset_mmx_state() -#else -#define vp9_clear_system_state() -#endif - -#if defined(_MSC_VER) && _MSC_VER < 1800 -// round is not defined in MSVC before VS2013. -static INLINE int round(double x) { - if (x < 0) - return (int)ceil(x - 0.5); - else - return (int)floor(x + 0.5); -} -#endif - -// use GNU builtins where available. -#if defined(__GNUC__) && \ - ((__GNUC__ == 3 && __GNUC_MINOR__ >= 4) || __GNUC__ >= 4) -static INLINE int get_msb(unsigned int n) { - return 31 ^ __builtin_clz(n); -} -#elif defined(USE_MSC_INTRINSICS) -#pragma intrinsic(_BitScanReverse) - -static INLINE int get_msb(unsigned int n) { - unsigned long first_set_bit; - _BitScanReverse(&first_set_bit, n); - return first_set_bit; -} -#undef USE_MSC_INTRINSICS -#else -// Returns (int)floor(log2(n)). n must be > 0. -static INLINE int get_msb(unsigned int n) { - int log = 0; - unsigned int value = n; - int i; - - for (i = 4; i >= 0; --i) { - const int shift = (1 << i); - const unsigned int x = value >> shift; - if (x != 0) { - value = x; - log += shift; - } - } - return log; -} -#endif - -#ifdef __cplusplus -} // extern "C" -#endif - -#endif // VP9_COMMON_VP9_SYSTEMDEPENDENT_H_ diff --git a/vp9/decoder/vp9_decodeframe.c b/vp9/decoder/vp9_decodeframe.c index ecebe1efb..58b79ea27 100644 --- a/vp9/decoder/vp9_decodeframe.c +++ b/vp9/decoder/vp9_decodeframe.c @@ -2219,7 +2219,6 @@ void vp9_decode_frame(VP9Decoder *pbi, } else { vpx_internal_error(&cm->error, VPX_CODEC_CORRUPT_FRAME, "Decode failed. Frame data is corrupted."); - } } else { *p_data_end = decode_tiles(pbi, data + first_partition_size, data_end); diff --git a/vp9/decoder/vp9_decodemv.c b/vp9/decoder/vp9_decodemv.c index 341e6d792..33818a993 100644 --- a/vp9/decoder/vp9_decodemv.c +++ b/vp9/decoder/vp9_decodemv.c @@ -560,7 +560,7 @@ static void read_inter_block_mode_info(VP9Decoder *const pbi, is_compound, allow_hp, r)) { xd->corrupted |= 1; break; - }; + } mi->bmi[j].as_mv[0].as_int = block[0].as_int; if (is_compound) diff --git a/vp9/decoder/vp9_decoder.c b/vp9/decoder/vp9_decoder.c index 801145654..6734d0029 100644 --- a/vp9/decoder/vp9_decoder.c +++ b/vp9/decoder/vp9_decoder.c @@ -17,6 +17,7 @@ #include "./vpx_scale_rtcd.h" #include "vpx_mem/vpx_mem.h" +#include "vpx_ports/system_state.h" #include "vpx_ports/vpx_once.h" #include "vpx_ports/vpx_timer.h" #include "vpx_scale/vpx_scale.h" @@ -30,7 +31,6 @@ #endif #include "vp9/common/vp9_quant_common.h" #include "vp9/common/vp9_reconintra.h" -#include "vp9/common/vp9_systemdependent.h" #include "vp9/decoder/vp9_decodeframe.h" #include "vp9/decoder/vp9_decoder.h" @@ -365,7 +365,7 @@ int vp9_receive_compressed_data(VP9Decoder *pbi, decrease_ref_count(cm->new_fb_idx, frame_bufs, pool); unlock_buffer_pool(pool); - vp9_clear_system_state(); + vpx_clear_system_state(); return -1; } @@ -374,7 +374,7 @@ int vp9_receive_compressed_data(VP9Decoder *pbi, swap_frame_buffers(pbi); - vp9_clear_system_state(); + vpx_clear_system_state(); if (!cm->show_existing_frame) { cm->last_show_frame = cm->show_frame; @@ -440,7 +440,7 @@ int vp9_get_raw_frame(VP9Decoder *pbi, YV12_BUFFER_CONFIG *sd, *sd = *cm->frame_to_show; ret = 0; #endif /*!CONFIG_POSTPROC*/ - vp9_clear_system_state(); + vpx_clear_system_state(); return ret; } diff --git a/vp9/encoder/arm/neon/vp9_avg_neon.c b/vp9/encoder/arm/neon/vp9_avg_neon.c index 40d7e8779..d569ec95d 100644 --- a/vp9/encoder/arm/neon/vp9_avg_neon.c +++ b/vp9/encoder/arm/neon/vp9_avg_neon.c @@ -9,6 +9,8 @@ */ #include <arm_neon.h> +#include <assert.h> + #include "./vp9_rtcd.h" #include "./vpx_config.h" @@ -114,3 +116,45 @@ int16_t vp9_int_pro_col_neon(uint8_t const *ref, const int width) { return horizontal_add_u16x8(vec_sum); } + +// ref, src = [0, 510] - max diff = 16-bits +// bwl = {2, 3, 4}, width = {16, 32, 64} +int vp9_vector_var_neon(int16_t const *ref, int16_t const *src, const int bwl) { + int width = 4 << bwl; + int32x4_t sse = vdupq_n_s32(0); + int16x8_t total = vdupq_n_s16(0); + + assert(width >= 8); + assert((width % 8) == 0); + + do { + const int16x8_t r = vld1q_s16(ref); + const int16x8_t s = vld1q_s16(src); + const int16x8_t diff = vsubq_s16(r, s); // [-510, 510], 10 bits. + const int16x4_t diff_lo = vget_low_s16(diff); + const int16x4_t diff_hi = vget_high_s16(diff); + sse = vmlal_s16(sse, diff_lo, diff_lo); // dynamic range 26 bits. + sse = vmlal_s16(sse, diff_hi, diff_hi); + total = vaddq_s16(total, diff); // dynamic range 16 bits. + + ref += 8; + src += 8; + width -= 8; + } while (width != 0); + + { + // Note: 'total''s pairwise addition could be implemented similarly to + // horizontal_add_u16x8(), but one less vpaddl with 'total' when paired + // with the summation of 'sse' performed better on a Cortex-A15. + const int32x4_t t0 = vpaddlq_s16(total); // cascading summation of 'total' + const int32x2_t t1 = vadd_s32(vget_low_s32(t0), vget_high_s32(t0)); + const int32x2_t t2 = vpadd_s32(t1, t1); + const int t = vget_lane_s32(t2, 0); + const int64x2_t s0 = vpaddlq_s32(sse); // cascading summation of 'sse'. + const int32x2_t s1 = vadd_s32(vreinterpret_s32_s64(vget_low_s64(s0)), + vreinterpret_s32_s64(vget_high_s64(s0))); + const int s = vget_lane_s32(s1, 0); + const int shift_factor = bwl + 2; + return s - ((t * t) >> shift_factor); + } +} diff --git a/vp9/encoder/vp9_aq_complexity.c b/vp9/encoder/vp9_aq_complexity.c index bea7653d2..15f227fb8 100644 --- a/vp9/encoder/vp9_aq_complexity.c +++ b/vp9/encoder/vp9_aq_complexity.c @@ -10,6 +10,7 @@ #include <limits.h> #include <math.h> +#include "vpx_ports/system_state.h" #include "vp9/encoder/vp9_aq_complexity.h" #include "vp9/encoder/vp9_aq_variance.h" @@ -47,7 +48,7 @@ void vp9_setup_in_frame_q_adj(VP9_COMP *cpi) { struct segmentation *const seg = &cm->seg; // Make SURE use of floating point in this function is safe. - vp9_clear_system_state(); + vpx_clear_system_state(); if (cm->frame_type == KEY_FRAME || cpi->refresh_alt_ref_frame || @@ -133,7 +134,7 @@ void vp9_caq_select_segment(VP9_COMP *cpi, MACROBLOCK *mb, BLOCK_SIZE bs, double low_var_thresh; const int aq_strength = get_aq_c_strength(cm->base_qindex, cm->bit_depth); - vp9_clear_system_state(); + vpx_clear_system_state(); low_var_thresh = (cpi->oxcf.pass == 2) ? MAX(cpi->twopass.mb_av_energy, MIN_DEFAULT_LV_THRESH) : DEFAULT_LV_THRESH; diff --git a/vp9/encoder/vp9_aq_cyclicrefresh.c b/vp9/encoder/vp9_aq_cyclicrefresh.c index b61906388..e6b368617 100644 --- a/vp9/encoder/vp9_aq_cyclicrefresh.c +++ b/vp9/encoder/vp9_aq_cyclicrefresh.c @@ -11,6 +11,8 @@ #include <limits.h> #include <math.h> +#include "vpx_ports/system_state.h" + #include "vp9/encoder/vp9_aq_cyclicrefresh.h" #include "vp9/common/vp9_seg_common.h" @@ -459,7 +461,10 @@ void vp9_cyclic_refresh_update_parameters(VP9_COMP *const cpi) { cr->time_for_refresh = 0; // Use larger delta-qp (increase rate_ratio_qdelta) for first few (~4) // periods of the refresh cycle, after a key frame. - if (rc->frames_since_key < 4 * cr->percent_refresh) + // Account for larger interval on base layer for temporal layers. + if (cr->percent_refresh > 0 && + rc->frames_since_key < (4 * cpi->svc.number_temporal_layers) * + (100 / cr->percent_refresh)) cr->rate_ratio_qdelta = 3.0; else cr->rate_ratio_qdelta = 2.0; @@ -503,7 +508,7 @@ void vp9_cyclic_refresh_setup(VP9_COMP *const cpi) { int qindex_delta = 0; int qindex2; const double q = vp9_convert_qindex_to_q(cm->base_qindex, cm->bit_depth); - vp9_clear_system_state(); + vpx_clear_system_state(); // Set rate threshold to some multiple (set to 2 for now) of the target // rate (target is given by sb64_target_rate and scaled by 256). cr->thresh_rate_sb = ((int64_t)(rc->sb64_target_rate) << 8) << 2; diff --git a/vp9/encoder/vp9_aq_variance.c b/vp9/encoder/vp9_aq_variance.c index f072717f1..1c99105d1 100644 --- a/vp9/encoder/vp9_aq_variance.c +++ b/vp9/encoder/vp9_aq_variance.c @@ -11,6 +11,7 @@ #include <math.h> #include "vpx_ports/mem.h" +#include "vpx_ports/system_state.h" #include "vp9/encoder/vp9_aq_variance.h" @@ -19,7 +20,6 @@ #include "vp9/encoder/vp9_ratectrl.h" #include "vp9/encoder/vp9_rd.h" #include "vp9/encoder/vp9_segmentation.h" -#include "vp9/common/vp9_systemdependent.h" #define ENERGY_MIN (-4) #define ENERGY_MAX (1) @@ -56,7 +56,7 @@ void vp9_vaq_frame_setup(VP9_COMP *cpi) { seg->abs_delta = SEGMENT_DELTADATA; - vp9_clear_system_state(); + vpx_clear_system_state(); for (i = 0; i < MAX_SEGMENTS; ++i) { int qindex_delta = @@ -191,7 +191,7 @@ static unsigned int block_variance(VP9_COMP *cpi, MACROBLOCK *x, double vp9_log_block_var(VP9_COMP *cpi, MACROBLOCK *x, BLOCK_SIZE bs) { unsigned int var = block_variance(cpi, x, bs); - vp9_clear_system_state(); + vpx_clear_system_state(); return log(var + 1.0); } @@ -199,7 +199,7 @@ double vp9_log_block_var(VP9_COMP *cpi, MACROBLOCK *x, BLOCK_SIZE bs) { int vp9_block_energy(VP9_COMP *cpi, MACROBLOCK *x, BLOCK_SIZE bs) { double energy; double energy_midpoint; - vp9_clear_system_state(); + vpx_clear_system_state(); energy_midpoint = (cpi->oxcf.pass == 2) ? cpi->twopass.mb_av_energy : DEFAULT_E_MIDPOINT; energy = vp9_log_block_var(cpi, x, bs) - energy_midpoint; diff --git a/vp9/encoder/vp9_bitstream.c b/vp9/encoder/vp9_bitstream.c index f06bd56ea..d0de095ac 100644 --- a/vp9/encoder/vp9_bitstream.c +++ b/vp9/encoder/vp9_bitstream.c @@ -16,6 +16,7 @@ #include "vpx_dsp/bitwriter_buffer.h" #include "vpx_mem/vpx_mem.h" #include "vpx_ports/mem_ops.h" +#include "vpx_ports/system_state.h" #include "vp9/common/vp9_entropy.h" #include "vp9/common/vp9_entropymode.h" @@ -23,7 +24,6 @@ #include "vp9/common/vp9_mvref_common.h" #include "vp9/common/vp9_pred_common.h" #include "vp9/common/vp9_seg_common.h" -#include "vp9/common/vp9_systemdependent.h" #include "vp9/common/vp9_tile_common.h" #include "vp9/encoder/vp9_cost.h" @@ -1240,7 +1240,7 @@ void vp9_pack_bitstream(VP9_COMP *cpi, uint8_t *dest, size_t *size) { uncompressed_hdr_size = vpx_wb_bytes_written(&wb); data += uncompressed_hdr_size; - vp9_clear_system_state(); + vpx_clear_system_state(); first_part_size = write_compressed_header(cpi, data); data += first_part_size; diff --git a/vp9/encoder/vp9_blockiness.c b/vp9/encoder/vp9_blockiness.c index fc3eac6c7..661496ee0 100644 --- a/vp9/encoder/vp9_blockiness.c +++ b/vp9/encoder/vp9_blockiness.c @@ -17,6 +17,7 @@ #include "vpx_dsp/vpx_convolve.h" #include "vpx_dsp/vpx_filter.h" #include "vpx_ports/mem.h" +#include "vpx_ports/system_state.h" static int horizontal_filter(const uint8_t *s) { return (s[1] - s[-2]) * 2 + (s[-1] - s[0]) * 6; @@ -123,7 +124,7 @@ double vp9_get_blockiness(const unsigned char *img1, int img1_pitch, int width, int height ) { double blockiness = 0; int i, j; - vp9_clear_system_state(); + vpx_clear_system_state(); for (i = 0; i < height; i += 4, img1 += img1_pitch * 4, img2 += img2_pitch * 4) { for (j = 0; j < width; j += 4) { diff --git a/vp9/encoder/vp9_dct.c b/vp9/encoder/vp9_dct.c index 8d0d10f81..f94540baa 100644 --- a/vp9/encoder/vp9_dct.c +++ b/vp9/encoder/vp9_dct.c @@ -17,7 +17,6 @@ #include "vp9/common/vp9_blockd.h" #include "vp9/common/vp9_idct.h" -#include "vp9/common/vp9_systemdependent.h" #include "vpx_dsp/fwd_txfm.h" #include "vpx_ports/mem.h" diff --git a/vp9/encoder/vp9_denoiser.c b/vp9/encoder/vp9_denoiser.c index f1d73790a..1f9a28c1e 100644 --- a/vp9/encoder/vp9_denoiser.c +++ b/vp9/encoder/vp9_denoiser.c @@ -63,10 +63,6 @@ static int sse_diff_thresh(BLOCK_SIZE bs, int increase_denoising, } } -int total_adj_strong_thresh(BLOCK_SIZE bs, int increase_denoising) { - return (1 << num_pels_log2_lookup[bs]) * (increase_denoising ? 3 : 2); -} - static int total_adj_weak_thresh(BLOCK_SIZE bs, int increase_denoising) { return (1 << num_pels_log2_lookup[bs]) * (increase_denoising ? 3 : 2); } diff --git a/vp9/encoder/vp9_denoiser.h b/vp9/encoder/vp9_denoiser.h index 8eb5da1b8..b2af792bb 100644 --- a/vp9/encoder/vp9_denoiser.h +++ b/vp9/encoder/vp9_denoiser.h @@ -57,7 +57,12 @@ int vp9_denoiser_alloc(VP9_DENOISER *denoiser, int width, int height, int border); #if CONFIG_VP9_TEMPORAL_DENOISING -int total_adj_strong_thresh(BLOCK_SIZE bs, int increase_denoising); +// This function is used by both c and sse2 denoiser implementations. +// Define it as a static function within the scope where vp9_denoiser.h +// is referenced. +static int total_adj_strong_thresh(BLOCK_SIZE bs, int increase_denoising) { + return (1 << num_pels_log2_lookup[bs]) * (increase_denoising ? 3 : 2); +} #endif void vp9_denoiser_free(VP9_DENOISER *denoiser); diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c index e7f14eb7e..63ab15773 100644 --- a/vp9/encoder/vp9_encodeframe.c +++ b/vp9/encoder/vp9_encodeframe.c @@ -18,6 +18,7 @@ #include "vpx_ports/mem.h" #include "vpx_ports/vpx_timer.h" +#include "vpx_ports/system_state.h" #include "vp9/common/vp9_common.h" #include "vp9/common/vp9_entropy.h" @@ -29,7 +30,6 @@ #include "vp9/common/vp9_reconintra.h" #include "vp9/common/vp9_reconinter.h" #include "vp9/common/vp9_seg_common.h" -#include "vp9/common/vp9_systemdependent.h" #include "vp9/common/vp9_tile_common.h" #include "vp9/encoder/vp9_aq_complexity.h" @@ -1152,7 +1152,7 @@ static int set_segment_rdmult(VP9_COMP *const cpi, int segment_qindex; VP9_COMMON *const cm = &cpi->common; vp9_init_plane_quantizers(cpi, x); - vp9_clear_system_state(); + vpx_clear_system_state(); segment_qindex = vp9_get_qindex(&cm->seg, segment_id, cm->base_qindex); return vp9_compute_rd_mult(cpi, segment_qindex + cm->y_dc_delta_q); @@ -1173,7 +1173,7 @@ static void rd_pick_sb_modes(VP9_COMP *cpi, const AQ_MODE aq_mode = cpi->oxcf.aq_mode; int i, orig_rdmult; - vp9_clear_system_state(); + vpx_clear_system_state(); // Use the lower precision, but faster, 32x32 fdct for mode selection. x->use_lp32x32fdct = 1; @@ -3925,7 +3925,9 @@ static void encode_frame_internal(VP9_COMP *cpi) { } vp9_zero(x->zcoeff_blk); - if (cm->frame_type != KEY_FRAME && cpi->rc.frames_since_golden == 0) + if (cm->frame_type != KEY_FRAME && + cpi->rc.frames_since_golden == 0 && + !cpi->use_svc) cpi->ref_frame_flags &= (~VP9_GOLD_FLAG); if (sf->partition_search_type == SOURCE_VAR_BASED_PARTITION) diff --git a/vp9/encoder/vp9_encodemb.c b/vp9/encoder/vp9_encodemb.c index 903839743..00e4c610b 100644 --- a/vp9/encoder/vp9_encodemb.c +++ b/vp9/encoder/vp9_encodemb.c @@ -21,7 +21,6 @@ #include "vp9/common/vp9_reconinter.h" #include "vp9/common/vp9_reconintra.h" #include "vp9/common/vp9_scan.h" -#include "vp9/common/vp9_systemdependent.h" #include "vp9/encoder/vp9_encodemb.h" #include "vp9/encoder/vp9_rd.h" @@ -58,7 +57,7 @@ typedef struct vp9_token_state { int error; int next; int16_t token; - short qc; + int16_t qc; } vp9_token_state; // TODO(jimbankoski): experiment to find optimal RD numbers. @@ -434,25 +433,25 @@ void vp9_xform_quant_dc(MACROBLOCK *x, int plane, int block, switch (tx_size) { case TX_32X32: vpx_highbd_fdct32x32_1(src_diff, coeff, diff_stride); - vp9_highbd_quantize_dc_32x32(coeff, x->skip_block, p->round, + vpx_highbd_quantize_dc_32x32(coeff, x->skip_block, p->round, p->quant_fp[0], qcoeff, dqcoeff, pd->dequant[0], eob); break; case TX_16X16: vpx_highbd_fdct16x16_1(src_diff, coeff, diff_stride); - vp9_highbd_quantize_dc(coeff, 256, x->skip_block, p->round, + vpx_highbd_quantize_dc(coeff, 256, x->skip_block, p->round, p->quant_fp[0], qcoeff, dqcoeff, pd->dequant[0], eob); break; case TX_8X8: vpx_highbd_fdct8x8_1(src_diff, coeff, diff_stride); - vp9_highbd_quantize_dc(coeff, 64, x->skip_block, p->round, + vpx_highbd_quantize_dc(coeff, 64, x->skip_block, p->round, p->quant_fp[0], qcoeff, dqcoeff, pd->dequant[0], eob); break; case TX_4X4: x->fwd_txm4x4(src_diff, coeff, diff_stride); - vp9_highbd_quantize_dc(coeff, 16, x->skip_block, p->round, + vpx_highbd_quantize_dc(coeff, 16, x->skip_block, p->round, p->quant_fp[0], qcoeff, dqcoeff, pd->dequant[0], eob); break; @@ -466,25 +465,25 @@ void vp9_xform_quant_dc(MACROBLOCK *x, int plane, int block, switch (tx_size) { case TX_32X32: vpx_fdct32x32_1(src_diff, coeff, diff_stride); - vp9_quantize_dc_32x32(coeff, x->skip_block, p->round, + vpx_quantize_dc_32x32(coeff, x->skip_block, p->round, p->quant_fp[0], qcoeff, dqcoeff, pd->dequant[0], eob); break; case TX_16X16: vpx_fdct16x16_1(src_diff, coeff, diff_stride); - vp9_quantize_dc(coeff, 256, x->skip_block, p->round, + vpx_quantize_dc(coeff, 256, x->skip_block, p->round, p->quant_fp[0], qcoeff, dqcoeff, pd->dequant[0], eob); break; case TX_8X8: vpx_fdct8x8_1(src_diff, coeff, diff_stride); - vp9_quantize_dc(coeff, 64, x->skip_block, p->round, + vpx_quantize_dc(coeff, 64, x->skip_block, p->round, p->quant_fp[0], qcoeff, dqcoeff, pd->dequant[0], eob); break; case TX_4X4: x->fwd_txm4x4(src_diff, coeff, diff_stride); - vp9_quantize_dc(coeff, 16, x->skip_block, p->round, + vpx_quantize_dc(coeff, 16, x->skip_block, p->round, p->quant_fp[0], qcoeff, dqcoeff, pd->dequant[0], eob); break; @@ -515,28 +514,28 @@ void vp9_xform_quant(MACROBLOCK *x, int plane, int block, switch (tx_size) { case TX_32X32: highbd_fdct32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride); - vp9_highbd_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin, + vpx_highbd_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); break; case TX_16X16: vpx_highbd_fdct16x16(src_diff, coeff, diff_stride); - vp9_highbd_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round, + vpx_highbd_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); break; case TX_8X8: vpx_highbd_fdct8x8(src_diff, coeff, diff_stride); - vp9_highbd_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round, + vpx_highbd_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); break; case TX_4X4: x->fwd_txm4x4(src_diff, coeff, diff_stride); - vp9_highbd_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round, + vpx_highbd_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); @@ -551,28 +550,28 @@ void vp9_xform_quant(MACROBLOCK *x, int plane, int block, switch (tx_size) { case TX_32X32: fdct32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride); - vp9_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin, p->round, + vpx_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); break; case TX_16X16: vpx_fdct16x16(src_diff, coeff, diff_stride); - vp9_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round, + vpx_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); break; case TX_8X8: vpx_fdct8x8(src_diff, coeff, diff_stride); - vp9_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round, + vpx_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); break; case TX_4X4: x->fwd_txm4x4(src_diff, coeff, diff_stride); - vp9_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round, + vpx_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); @@ -815,7 +814,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize, vpx_highbd_subtract_block(32, 32, src_diff, diff_stride, src, src_stride, dst, dst_stride, xd->bd); highbd_fdct32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride); - vp9_highbd_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin, + vpx_highbd_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); @@ -832,7 +831,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize, vpx_highbd_fdct16x16(src_diff, coeff, diff_stride); else vp9_highbd_fht16x16(src_diff, coeff, diff_stride, tx_type); - vp9_highbd_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round, + vpx_highbd_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); @@ -850,7 +849,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize, vpx_highbd_fdct8x8(src_diff, coeff, diff_stride); else vp9_highbd_fht8x8(src_diff, coeff, diff_stride, tx_type); - vp9_highbd_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round, + vpx_highbd_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); @@ -868,7 +867,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize, vp9_highbd_fht4x4(src_diff, coeff, diff_stride, tx_type); else x->fwd_txm4x4(src_diff, coeff, diff_stride); - vp9_highbd_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round, + vpx_highbd_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); @@ -901,7 +900,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize, vpx_subtract_block(32, 32, src_diff, diff_stride, src, src_stride, dst, dst_stride); fdct32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride); - vp9_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin, p->round, + vpx_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); @@ -914,7 +913,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize, vpx_subtract_block(16, 16, src_diff, diff_stride, src, src_stride, dst, dst_stride); vp9_fht16x16(src_diff, coeff, diff_stride, tx_type); - vp9_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round, + vpx_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); @@ -927,7 +926,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize, vpx_subtract_block(8, 8, src_diff, diff_stride, src, src_stride, dst, dst_stride); vp9_fht8x8(src_diff, coeff, diff_stride, tx_type); - vp9_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round, p->quant, + vpx_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); @@ -943,7 +942,7 @@ void vp9_encode_block_intra(int plane, int block, BLOCK_SIZE plane_bsize, vp9_fht4x4(src_diff, coeff, diff_stride, tx_type); else x->fwd_txm4x4(src_diff, coeff, diff_stride); - vp9_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round, p->quant, + vpx_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, qcoeff, dqcoeff, pd->dequant, eob, scan_order->scan, scan_order->iscan); diff --git a/vp9/encoder/vp9_encodemv.c b/vp9/encoder/vp9_encodemv.c index a1d77dba8..7848c93a8 100644 --- a/vp9/encoder/vp9_encodemv.c +++ b/vp9/encoder/vp9_encodemv.c @@ -12,7 +12,6 @@ #include "vp9/common/vp9_common.h" #include "vp9/common/vp9_entropymode.h" -#include "vp9/common/vp9_systemdependent.h" #include "vp9/encoder/vp9_cost.h" #include "vp9/encoder/vp9_encodemv.h" diff --git a/vp9/encoder/vp9_encoder.c b/vp9/encoder/vp9_encoder.c index 8aee22756..370a581fd 100644 --- a/vp9/encoder/vp9_encoder.c +++ b/vp9/encoder/vp9_encoder.c @@ -18,7 +18,11 @@ #include "./vpx_scale_rtcd.h" #include "vpx/internal/vpx_psnr.h" #include "vpx_dsp/vpx_filter.h" +#if CONFIG_INTERNAL_STATS +#include "vpx_dsp/ssim.h" +#endif #include "vpx_ports/mem.h" +#include "vpx_ports/system_state.h" #include "vpx_ports/vpx_timer.h" #include "vpx_scale/vpx_scale.h" @@ -30,7 +34,6 @@ #endif #include "vp9/common/vp9_reconinter.h" #include "vp9/common/vp9_reconintra.h" -#include "vp9/common/vp9_systemdependent.h" #include "vp9/common/vp9_tile_common.h" #include "vp9/encoder/vp9_aq_complexity.h" @@ -51,9 +54,6 @@ #include "vp9/encoder/vp9_segmentation.h" #include "vp9/encoder/vp9_skin_detection.h" #include "vp9/encoder/vp9_speed_features.h" -#if CONFIG_INTERNAL_STATS -#include "vp9/encoder/vp9_ssim.h" -#endif #include "vp9/encoder/vp9_svc_layercontext.h" #include "vp9/encoder/vp9_temporal_filter.h" @@ -1709,7 +1709,8 @@ VP9_COMP *vp9_create_compressor(VP9EncoderConfig *oxcf, } if (cpi->b_calculate_consistency) { - cpi->ssim_vars = vpx_malloc(sizeof(*cpi->ssim_vars)*720*480); + cpi->ssim_vars = vpx_malloc(sizeof(*cpi->ssim_vars) * + 4 * cpi->common.mi_rows * cpi->common.mi_cols); cpi->worst_consistency = 100.0; } @@ -1925,7 +1926,7 @@ void vp9_remove_compressor(VP9_COMP *cpi) { if (cpi && (cm->current_video_frame > 0)) { #if CONFIG_INTERNAL_STATS - vp9_clear_system_state(); + vpx_clear_system_state(); if (cpi->oxcf.pass != 1) { char headings[512] = {0}; @@ -2747,7 +2748,7 @@ static void loopfilter_frame(VP9_COMP *cpi, VP9_COMMON *cm) { } else { struct vpx_usec_timer timer; - vp9_clear_system_state(); + vpx_clear_system_state(); vpx_usec_timer_start(&timer); @@ -2931,7 +2932,7 @@ static void output_frame_level_debug_stats(VP9_COMP *cpi) { FILE *const f = fopen("tmp.stt", cm->current_video_frame ? "a" : "w"); int64_t recon_err; - vp9_clear_system_state(); + vpx_clear_system_state(); recon_err = vp9_get_y_sse(cpi->Source, get_frame_new_buffer(cm)); @@ -3183,7 +3184,7 @@ static void encode_without_recode_loop(VP9_COMP *cpi) { VP9_COMMON *const cm = &cpi->common; int q = 0, bottom_index = 0, top_index = 0; // Dummy variables. - vp9_clear_system_state(); + vpx_clear_system_state(); set_frame_size(cpi); @@ -3247,7 +3248,7 @@ static void encode_without_recode_loop(VP9_COMP *cpi) { // Update the skip mb flag probabilities based on the distribution // seen in the last encoder iteration. // update_base_skip_probs(cpi); - vp9_clear_system_state(); + vpx_clear_system_state(); } static void encode_with_recode_loop(VP9_COMP *cpi, @@ -3268,7 +3269,7 @@ static void encode_with_recode_loop(VP9_COMP *cpi, set_size_independent_vars(cpi); do { - vp9_clear_system_state(); + vpx_clear_system_state(); set_frame_size(cpi); @@ -3332,7 +3333,7 @@ static void encode_with_recode_loop(VP9_COMP *cpi, // seen in the last encoder iteration. // update_base_skip_probs(cpi); - vp9_clear_system_state(); + vpx_clear_system_state(); // Dummy pack of the bitstream using up to date stats to get an // accurate estimate of output frame size to determine if we need @@ -3636,7 +3637,7 @@ static void encode_frame_to_data_rate(VP9_COMP *cpi, TX_SIZE t; set_ext_overrides(cpi); - vp9_clear_system_state(); + vpx_clear_system_state(); // Set the arf sign bias for this frame. set_arf_sign_bias(cpi); @@ -3727,7 +3728,7 @@ static void encode_frame_to_data_rate(VP9_COMP *cpi, } } - vp9_clear_system_state(); + vpx_clear_system_state(); #if CONFIG_INTERNAL_STATS memset(cpi->mode_chosen_counts, 0, @@ -4247,7 +4248,7 @@ int vp9_get_compressed_data(VP9_COMP *cpi, unsigned int *frame_flags, } // Clear down mmx registers - vp9_clear_system_state(); + vpx_clear_system_state(); // adjust frame rates based on timestamps given if (cm->show_frame) { @@ -4399,7 +4400,7 @@ int vp9_get_compressed_data(VP9_COMP *cpi, unsigned int *frame_flags, vp9_deblock(cm->frame_to_show, &cm->post_proc_buffer, cm->lf.filter_level * 10 / 6); #endif - vp9_clear_system_state(); + vpx_clear_system_state(); #if CONFIG_VP9_HIGHBITDEPTH calc_highbd_psnr(orig, pp, &psnr2, cpi->td.mb.e_mbd.bd, @@ -4415,13 +4416,13 @@ int vp9_get_compressed_data(VP9_COMP *cpi, unsigned int *frame_flags, #if CONFIG_VP9_HIGHBITDEPTH if (cm->use_highbitdepth) { - frame_ssim2 = vp9_highbd_calc_ssim(orig, recon, &weight, + frame_ssim2 = vpx_highbd_calc_ssim(orig, recon, &weight, (int)cm->bit_depth); } else { - frame_ssim2 = vp9_calc_ssim(orig, recon, &weight); + frame_ssim2 = vpx_calc_ssim(orig, recon, &weight); } #else - frame_ssim2 = vp9_calc_ssim(orig, recon, &weight); + frame_ssim2 = vpx_calc_ssim(orig, recon, &weight); #endif // CONFIG_VP9_HIGHBITDEPTH cpi->worst_ssim= MIN(cpi->worst_ssim, frame_ssim2); @@ -4430,13 +4431,13 @@ int vp9_get_compressed_data(VP9_COMP *cpi, unsigned int *frame_flags, #if CONFIG_VP9_HIGHBITDEPTH if (cm->use_highbitdepth) { - frame_ssim2 = vp9_highbd_calc_ssim( + frame_ssim2 = vpx_highbd_calc_ssim( orig, &cm->post_proc_buffer, &weight, (int)cm->bit_depth); } else { - frame_ssim2 = vp9_calc_ssim(orig, &cm->post_proc_buffer, &weight); + frame_ssim2 = vpx_calc_ssim(orig, &cm->post_proc_buffer, &weight); } #else - frame_ssim2 = vp9_calc_ssim(orig, &cm->post_proc_buffer, &weight); + frame_ssim2 = vpx_calc_ssim(orig, &cm->post_proc_buffer, &weight); #endif // CONFIG_VP9_HIGHBITDEPTH cpi->summedp_quality += frame_ssim2 * weight; @@ -4471,7 +4472,7 @@ int vp9_get_compressed_data(VP9_COMP *cpi, unsigned int *frame_flags, if (!cm->use_highbitdepth) #endif { - double this_inconsistency = vp9_get_ssim_metrics( + double this_inconsistency = vpx_get_ssim_metrics( cpi->Source->y_buffer, cpi->Source->y_stride, cm->frame_to_show->y_buffer, cm->frame_to_show->y_stride, cpi->Source->y_width, cpi->Source->y_height, cpi->ssim_vars, @@ -4491,14 +4492,14 @@ int vp9_get_compressed_data(VP9_COMP *cpi, unsigned int *frame_flags, double y, u, v, frame_all; #if CONFIG_VP9_HIGHBITDEPTH if (cm->use_highbitdepth) { - frame_all = vp9_highbd_calc_ssimg(cpi->Source, cm->frame_to_show, &y, + frame_all = vpx_highbd_calc_ssimg(cpi->Source, cm->frame_to_show, &y, &u, &v, (int)cm->bit_depth); } else { - frame_all = vp9_calc_ssimg(cpi->Source, cm->frame_to_show, &y, &u, + frame_all = vpx_calc_ssimg(cpi->Source, cm->frame_to_show, &y, &u, &v); } #else - frame_all = vp9_calc_ssimg(cpi->Source, cm->frame_to_show, &y, &u, &v); + frame_all = vpx_calc_ssimg(cpi->Source, cm->frame_to_show, &y, &u, &v); #endif // CONFIG_VP9_HIGHBITDEPTH adjust_image_stat(y, u, v, frame_all, &cpi->ssimg); } @@ -4507,7 +4508,7 @@ int vp9_get_compressed_data(VP9_COMP *cpi, unsigned int *frame_flags, #endif { double y, u, v, frame_all; - frame_all = vp9_calc_fastssim(cpi->Source, cm->frame_to_show, &y, &u, + frame_all = vpx_calc_fastssim(cpi->Source, cm->frame_to_show, &y, &u, &v); adjust_image_stat(y, u, v, frame_all, &cpi->fastssim); /* TODO(JBB): add 10/12 bit support */ @@ -4517,7 +4518,7 @@ int vp9_get_compressed_data(VP9_COMP *cpi, unsigned int *frame_flags, #endif { double y, u, v, frame_all; - frame_all = vp9_psnrhvs(cpi->Source, cm->frame_to_show, &y, &u, &v); + frame_all = vpx_psnrhvs(cpi->Source, cm->frame_to_show, &y, &u, &v); adjust_image_stat(y, u, v, frame_all, &cpi->psnrhvs); } } @@ -4574,7 +4575,7 @@ int vp9_get_preview_raw_frame(VP9_COMP *cpi, YV12_BUFFER_CONFIG *dest, ret = -1; } #endif // !CONFIG_VP9_POSTPROC - vp9_clear_system_state(); + vpx_clear_system_state(); return ret; } } diff --git a/vp9/encoder/vp9_encoder.h b/vp9/encoder/vp9_encoder.h index 4d4da9283..78d55e1ea 100644 --- a/vp9/encoder/vp9_encoder.h +++ b/vp9/encoder/vp9_encoder.h @@ -16,6 +16,10 @@ #include "./vpx_config.h" #include "vpx/internal/vpx_codec_internal.h" #include "vpx/vp8cx.h" +#if CONFIG_INTERNAL_STATS +#include "vpx_dsp/ssim.h" +#endif +#include "vpx_dsp/variance.h" #include "vpx_util/vpx_thread.h" #include "vp9/common/vp9_alloccommon.h" @@ -34,13 +38,9 @@ #include "vp9/encoder/vp9_quantize.h" #include "vp9/encoder/vp9_ratectrl.h" #include "vp9/encoder/vp9_rd.h" -#if CONFIG_INTERNAL_STATS -#include "vp9/encoder/vp9_ssim.h" -#endif #include "vp9/encoder/vp9_speed_features.h" #include "vp9/encoder/vp9_svc_layercontext.h" #include "vp9/encoder/vp9_tokenize.h" -#include "vpx_dsp/variance.h" #if CONFIG_VP9_TEMPORAL_DENOISING #include "vp9/encoder/vp9_denoiser.h" diff --git a/vp9/encoder/vp9_fastssim.c b/vp9/encoder/vp9_fastssim.c deleted file mode 100644 index f1d408cbe..000000000 --- a/vp9/encoder/vp9_fastssim.c +++ /dev/null @@ -1,465 +0,0 @@ -/* - * Copyright (c) 2010 The WebM project authors. All Rights Reserved. - * - * Use of this source code is governed by a BSD-style license - * that can be found in the LICENSE file in the root of the source - * tree. An additional intellectual property rights grant can be found - * in the file PATENTS. All contributing project authors may - * be found in the AUTHORS file in the root of the source tree. - * - * This code was originally written by: Nathan E. Egge, at the Daala - * project. - */ -#include <math.h> -#include <string.h> -#include "./vpx_config.h" -#include "./vp9_rtcd.h" -#include "vp9/encoder/vp9_ssim.h" -/* TODO(jbb): High bit depth version of this code needed */ -typedef struct fs_level fs_level; -typedef struct fs_ctx fs_ctx; - -#define SSIM_C1 (255 * 255 * 0.01 * 0.01) -#define SSIM_C2 (255 * 255 * 0.03 * 0.03) - -#define FS_MINI(_a, _b) ((_a) < (_b) ? (_a) : (_b)) -#define FS_MAXI(_a, _b) ((_a) > (_b) ? (_a) : (_b)) - -struct fs_level { - uint16_t *im1; - uint16_t *im2; - double *ssim; - int w; - int h; -}; - -struct fs_ctx { - fs_level *level; - int nlevels; - unsigned *col_buf; -}; - -static void fs_ctx_init(fs_ctx *_ctx, int _w, int _h, int _nlevels) { - unsigned char *data; - size_t data_size; - int lw; - int lh; - int l; - lw = (_w + 1) >> 1; - lh = (_h + 1) >> 1; - data_size = _nlevels * sizeof(fs_level) - + 2 * (lw + 8) * 8 * sizeof(*_ctx->col_buf); - for (l = 0; l < _nlevels; l++) { - size_t im_size; - size_t level_size; - im_size = lw * (size_t) lh; - level_size = 2 * im_size * sizeof(*_ctx->level[l].im1); - level_size += sizeof(*_ctx->level[l].ssim) - 1; - level_size /= sizeof(*_ctx->level[l].ssim); - level_size += im_size; - level_size *= sizeof(*_ctx->level[l].ssim); - data_size += level_size; - lw = (lw + 1) >> 1; - lh = (lh + 1) >> 1; - } - data = (unsigned char *) malloc(data_size); - _ctx->level = (fs_level *) data; - _ctx->nlevels = _nlevels; - data += _nlevels * sizeof(*_ctx->level); - lw = (_w + 1) >> 1; - lh = (_h + 1) >> 1; - for (l = 0; l < _nlevels; l++) { - size_t im_size; - size_t level_size; - _ctx->level[l].w = lw; - _ctx->level[l].h = lh; - im_size = lw * (size_t) lh; - level_size = 2 * im_size * sizeof(*_ctx->level[l].im1); - level_size += sizeof(*_ctx->level[l].ssim) - 1; - level_size /= sizeof(*_ctx->level[l].ssim); - level_size *= sizeof(*_ctx->level[l].ssim); - _ctx->level[l].im1 = (uint16_t *) data; - _ctx->level[l].im2 = _ctx->level[l].im1 + im_size; - data += level_size; - _ctx->level[l].ssim = (double *) data; - data += im_size * sizeof(*_ctx->level[l].ssim); - lw = (lw + 1) >> 1; - lh = (lh + 1) >> 1; - } - _ctx->col_buf = (unsigned *) data; -} - -static void fs_ctx_clear(fs_ctx *_ctx) { - free(_ctx->level); -} - -static void fs_downsample_level(fs_ctx *_ctx, int _l) { - const uint16_t *src1; - const uint16_t *src2; - uint16_t *dst1; - uint16_t *dst2; - int w2; - int h2; - int w; - int h; - int i; - int j; - w = _ctx->level[_l].w; - h = _ctx->level[_l].h; - dst1 = _ctx->level[_l].im1; - dst2 = _ctx->level[_l].im2; - w2 = _ctx->level[_l - 1].w; - h2 = _ctx->level[_l - 1].h; - src1 = _ctx->level[_l - 1].im1; - src2 = _ctx->level[_l - 1].im2; - for (j = 0; j < h; j++) { - int j0offs; - int j1offs; - j0offs = 2 * j * w2; - j1offs = FS_MINI(2 * j + 1, h2) * w2; - for (i = 0; i < w; i++) { - int i0; - int i1; - i0 = 2 * i; - i1 = FS_MINI(i0 + 1, w2); - dst1[j * w + i] = src1[j0offs + i0] + src1[j0offs + i1] - + src1[j1offs + i0] + src1[j1offs + i1]; - dst2[j * w + i] = src2[j0offs + i0] + src2[j0offs + i1] - + src2[j1offs + i0] + src2[j1offs + i1]; - } - } -} - -static void fs_downsample_level0(fs_ctx *_ctx, const unsigned char *_src1, - int _s1ystride, const unsigned char *_src2, - int _s2ystride, int _w, int _h) { - uint16_t *dst1; - uint16_t *dst2; - int w; - int h; - int i; - int j; - w = _ctx->level[0].w; - h = _ctx->level[0].h; - dst1 = _ctx->level[0].im1; - dst2 = _ctx->level[0].im2; - for (j = 0; j < h; j++) { - int j0; - int j1; - j0 = 2 * j; - j1 = FS_MINI(j0 + 1, _h); - for (i = 0; i < w; i++) { - int i0; - int i1; - i0 = 2 * i; - i1 = FS_MINI(i0 + 1, _w); - dst1[j * w + i] = _src1[j0 * _s1ystride + i0] - + _src1[j0 * _s1ystride + i1] + _src1[j1 * _s1ystride + i0] - + _src1[j1 * _s1ystride + i1]; - dst2[j * w + i] = _src2[j0 * _s2ystride + i0] - + _src2[j0 * _s2ystride + i1] + _src2[j1 * _s2ystride + i0] - + _src2[j1 * _s2ystride + i1]; - } - } -} - -static void fs_apply_luminance(fs_ctx *_ctx, int _l) { - unsigned *col_sums_x; - unsigned *col_sums_y; - uint16_t *im1; - uint16_t *im2; - double *ssim; - double c1; - int w; - int h; - int j0offs; - int j1offs; - int i; - int j; - w = _ctx->level[_l].w; - h = _ctx->level[_l].h; - col_sums_x = _ctx->col_buf; - col_sums_y = col_sums_x + w; - im1 = _ctx->level[_l].im1; - im2 = _ctx->level[_l].im2; - for (i = 0; i < w; i++) - col_sums_x[i] = 5 * im1[i]; - for (i = 0; i < w; i++) - col_sums_y[i] = 5 * im2[i]; - for (j = 1; j < 4; j++) { - j1offs = FS_MINI(j, h - 1) * w; - for (i = 0; i < w; i++) - col_sums_x[i] += im1[j1offs + i]; - for (i = 0; i < w; i++) - col_sums_y[i] += im2[j1offs + i]; - } - ssim = _ctx->level[_l].ssim; - c1 = (double) (SSIM_C1 * 4096 * (1 << 4 * _l)); - for (j = 0; j < h; j++) { - unsigned mux; - unsigned muy; - int i0; - int i1; - mux = 5 * col_sums_x[0]; - muy = 5 * col_sums_y[0]; - for (i = 1; i < 4; i++) { - i1 = FS_MINI(i, w - 1); - mux += col_sums_x[i1]; - muy += col_sums_y[i1]; - } - for (i = 0; i < w; i++) { - ssim[j * w + i] *= (2 * mux * (double) muy + c1) - / (mux * (double) mux + muy * (double) muy + c1); - if (i + 1 < w) { - i0 = FS_MAXI(0, i - 4); - i1 = FS_MINI(i + 4, w - 1); - mux += col_sums_x[i1] - col_sums_x[i0]; - muy += col_sums_x[i1] - col_sums_x[i0]; - } - } - if (j + 1 < h) { - j0offs = FS_MAXI(0, j - 4) * w; - for (i = 0; i < w; i++) - col_sums_x[i] -= im1[j0offs + i]; - for (i = 0; i < w; i++) - col_sums_y[i] -= im2[j0offs + i]; - j1offs = FS_MINI(j + 4, h - 1) * w; - for (i = 0; i < w; i++) - col_sums_x[i] += im1[j1offs + i]; - for (i = 0; i < w; i++) - col_sums_y[i] += im2[j1offs + i]; - } - } -} - -#define FS_COL_SET(_col, _joffs, _ioffs) \ - do { \ - unsigned gx; \ - unsigned gy; \ - gx = gx_buf[((j + (_joffs)) & 7) * stride + i + (_ioffs)]; \ - gy = gy_buf[((j + (_joffs)) & 7) * stride + i + (_ioffs)]; \ - col_sums_gx2[(_col)] = gx * (double)gx; \ - col_sums_gy2[(_col)] = gy * (double)gy; \ - col_sums_gxgy[(_col)] = gx * (double)gy; \ - } \ - while (0) - -#define FS_COL_ADD(_col, _joffs, _ioffs) \ - do { \ - unsigned gx; \ - unsigned gy; \ - gx = gx_buf[((j + (_joffs)) & 7) * stride + i + (_ioffs)]; \ - gy = gy_buf[((j + (_joffs)) & 7) * stride + i + (_ioffs)]; \ - col_sums_gx2[(_col)] += gx * (double)gx; \ - col_sums_gy2[(_col)] += gy * (double)gy; \ - col_sums_gxgy[(_col)] += gx * (double)gy; \ - } \ - while (0) - -#define FS_COL_SUB(_col, _joffs, _ioffs) \ - do { \ - unsigned gx; \ - unsigned gy; \ - gx = gx_buf[((j + (_joffs)) & 7) * stride + i + (_ioffs)]; \ - gy = gy_buf[((j + (_joffs)) & 7) * stride + i + (_ioffs)]; \ - col_sums_gx2[(_col)] -= gx * (double)gx; \ - col_sums_gy2[(_col)] -= gy * (double)gy; \ - col_sums_gxgy[(_col)] -= gx * (double)gy; \ - } \ - while (0) - -#define FS_COL_COPY(_col1, _col2) \ - do { \ - col_sums_gx2[(_col1)] = col_sums_gx2[(_col2)]; \ - col_sums_gy2[(_col1)] = col_sums_gy2[(_col2)]; \ - col_sums_gxgy[(_col1)] = col_sums_gxgy[(_col2)]; \ - } \ - while (0) - -#define FS_COL_HALVE(_col1, _col2) \ - do { \ - col_sums_gx2[(_col1)] = col_sums_gx2[(_col2)] * 0.5; \ - col_sums_gy2[(_col1)] = col_sums_gy2[(_col2)] * 0.5; \ - col_sums_gxgy[(_col1)] = col_sums_gxgy[(_col2)] * 0.5; \ - } \ - while (0) - -#define FS_COL_DOUBLE(_col1, _col2) \ - do { \ - col_sums_gx2[(_col1)] = col_sums_gx2[(_col2)] * 2; \ - col_sums_gy2[(_col1)] = col_sums_gy2[(_col2)] * 2; \ - col_sums_gxgy[(_col1)] = col_sums_gxgy[(_col2)] * 2; \ - } \ - while (0) - -static void fs_calc_structure(fs_ctx *_ctx, int _l) { - uint16_t *im1; - uint16_t *im2; - unsigned *gx_buf; - unsigned *gy_buf; - double *ssim; - double col_sums_gx2[8]; - double col_sums_gy2[8]; - double col_sums_gxgy[8]; - double c2; - int stride; - int w; - int h; - int i; - int j; - w = _ctx->level[_l].w; - h = _ctx->level[_l].h; - im1 = _ctx->level[_l].im1; - im2 = _ctx->level[_l].im2; - ssim = _ctx->level[_l].ssim; - gx_buf = _ctx->col_buf; - stride = w + 8; - gy_buf = gx_buf + 8 * stride; - memset(gx_buf, 0, 2 * 8 * stride * sizeof(*gx_buf)); - c2 = SSIM_C2 * (1 << 4 * _l) * 16 * 104; - for (j = 0; j < h + 4; j++) { - if (j < h - 1) { - for (i = 0; i < w - 1; i++) { - unsigned g1; - unsigned g2; - unsigned gx; - unsigned gy; - g1 = abs(im1[(j + 1) * w + i + 1] - im1[j * w + i]); - g2 = abs(im1[(j + 1) * w + i] - im1[j * w + i + 1]); - gx = 4 * FS_MAXI(g1, g2) + FS_MINI(g1, g2); - g1 = abs(im2[(j + 1) * w + i + 1] - im2[j * w + i]); - g2 = abs(im2[(j + 1) * w + i] - im2[j * w + i + 1]); - gy = 4 * FS_MAXI(g1, g2) + FS_MINI(g1, g2); - gx_buf[(j & 7) * stride + i + 4] = gx; - gy_buf[(j & 7) * stride + i + 4] = gy; - } - } else { - memset(gx_buf + (j & 7) * stride, 0, stride * sizeof(*gx_buf)); - memset(gy_buf + (j & 7) * stride, 0, stride * sizeof(*gy_buf)); - } - if (j >= 4) { - int k; - col_sums_gx2[3] = col_sums_gx2[2] = col_sums_gx2[1] = col_sums_gx2[0] = 0; - col_sums_gy2[3] = col_sums_gy2[2] = col_sums_gy2[1] = col_sums_gy2[0] = 0; - col_sums_gxgy[3] = col_sums_gxgy[2] = col_sums_gxgy[1] = - col_sums_gxgy[0] = 0; - for (i = 4; i < 8; i++) { - FS_COL_SET(i, -1, 0); - FS_COL_ADD(i, 0, 0); - for (k = 1; k < 8 - i; k++) { - FS_COL_DOUBLE(i, i); - FS_COL_ADD(i, -k - 1, 0); - FS_COL_ADD(i, k, 0); - } - } - for (i = 0; i < w; i++) { - double mugx2; - double mugy2; - double mugxgy; - mugx2 = col_sums_gx2[0]; - for (k = 1; k < 8; k++) - mugx2 += col_sums_gx2[k]; - mugy2 = col_sums_gy2[0]; - for (k = 1; k < 8; k++) - mugy2 += col_sums_gy2[k]; - mugxgy = col_sums_gxgy[0]; - for (k = 1; k < 8; k++) - mugxgy += col_sums_gxgy[k]; - ssim[(j - 4) * w + i] = (2 * mugxgy + c2) / (mugx2 + mugy2 + c2); - if (i + 1 < w) { - FS_COL_SET(0, -1, 1); - FS_COL_ADD(0, 0, 1); - FS_COL_SUB(2, -3, 2); - FS_COL_SUB(2, 2, 2); - FS_COL_HALVE(1, 2); - FS_COL_SUB(3, -4, 3); - FS_COL_SUB(3, 3, 3); - FS_COL_HALVE(2, 3); - FS_COL_COPY(3, 4); - FS_COL_DOUBLE(4, 5); - FS_COL_ADD(4, -4, 5); - FS_COL_ADD(4, 3, 5); - FS_COL_DOUBLE(5, 6); - FS_COL_ADD(5, -3, 6); - FS_COL_ADD(5, 2, 6); - FS_COL_DOUBLE(6, 7); - FS_COL_ADD(6, -2, 7); - FS_COL_ADD(6, 1, 7); - FS_COL_SET(7, -1, 8); - FS_COL_ADD(7, 0, 8); - } - } - } - } -} - -#define FS_NLEVELS (4) - -/*These weights were derived from the default weights found in Wang's original - Matlab implementation: {0.0448, 0.2856, 0.2363, 0.1333}. - We drop the finest scale and renormalize the rest to sum to 1.*/ - -static const double FS_WEIGHTS[FS_NLEVELS] = {0.2989654541015625, - 0.3141326904296875, 0.2473602294921875, 0.1395416259765625}; - -static double fs_average(fs_ctx *_ctx, int _l) { - double *ssim; - double ret; - int w; - int h; - int i; - int j; - w = _ctx->level[_l].w; - h = _ctx->level[_l].h; - ssim = _ctx->level[_l].ssim; - ret = 0; - for (j = 0; j < h; j++) - for (i = 0; i < w; i++) - ret += ssim[j * w + i]; - return pow(ret / (w * h), FS_WEIGHTS[_l]); -} - -static double calc_ssim(const unsigned char *_src, int _systride, - const unsigned char *_dst, int _dystride, int _w, int _h) { - fs_ctx ctx; - double ret; - int l; - ret = 1; - fs_ctx_init(&ctx, _w, _h, FS_NLEVELS); - fs_downsample_level0(&ctx, _src, _systride, _dst, _dystride, _w, _h); - for (l = 0; l < FS_NLEVELS - 1; l++) { - fs_calc_structure(&ctx, l); - ret *= fs_average(&ctx, l); - fs_downsample_level(&ctx, l + 1); - } - fs_calc_structure(&ctx, l); - fs_apply_luminance(&ctx, l); - ret *= fs_average(&ctx, l); - fs_ctx_clear(&ctx); - return ret; -} - -static double convert_ssim_db(double _ssim, double _weight) { - return 10 * (log10(_weight) - log10(_weight - _ssim)); -} - -double vp9_calc_fastssim(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest, - double *ssim_y, double *ssim_u, double *ssim_v) { - double ssimv; - vp9_clear_system_state(); - - *ssim_y = calc_ssim(source->y_buffer, source->y_stride, dest->y_buffer, - dest->y_stride, source->y_crop_width, - source->y_crop_height); - - *ssim_u = calc_ssim(source->u_buffer, source->uv_stride, dest->u_buffer, - dest->uv_stride, source->uv_crop_width, - source->uv_crop_height); - - *ssim_v = calc_ssim(source->v_buffer, source->uv_stride, dest->v_buffer, - dest->uv_stride, source->uv_crop_width, - source->uv_crop_height); - ssimv = (*ssim_y) * .8 + .1 * ((*ssim_u) + (*ssim_v)); - - return convert_ssim_db(ssimv, 1.0); -} diff --git a/vp9/encoder/vp9_firstpass.c b/vp9/encoder/vp9_firstpass.c index 5caf2cbd8..3c2545452 100644 --- a/vp9/encoder/vp9_firstpass.c +++ b/vp9/encoder/vp9_firstpass.c @@ -17,13 +17,13 @@ #include "vpx_mem/vpx_mem.h" #include "vpx_ports/mem.h" +#include "vpx_ports/system_state.h" #include "vpx_scale/vpx_scale.h" #include "vpx_scale/yv12config.h" #include "vp9/common/vp9_entropymv.h" #include "vp9/common/vp9_quant_common.h" #include "vp9/common/vp9_reconinter.h" // vp9_setup_dst_planes() -#include "vp9/common/vp9_systemdependent.h" #include "vp9/encoder/vp9_aq_variance.h" #include "vp9/encoder/vp9_block.h" #include "vp9/encoder/vp9_encodeframe.h" @@ -237,8 +237,7 @@ static void subtract_stats(FIRSTPASS_STATS *section, #define MIN_ACTIVE_AREA 0.5 #define MAX_ACTIVE_AREA 1.0 static double calculate_active_area(const VP9_COMP *cpi, - const FIRSTPASS_STATS *this_frame) -{ + const FIRSTPASS_STATS *this_frame) { double active_pct; active_pct = 1.0 - @@ -546,7 +545,7 @@ void vp9_first_pass(VP9_COMP *cpi, const struct lookahead_entry *source) { } #endif - vp9_clear_system_state(); + vpx_clear_system_state(); intra_factor = 0.0; brightness_factor = 0.0; @@ -657,7 +656,7 @@ void vp9_first_pass(VP9_COMP *cpi, const struct lookahead_entry *source) { const int mb_index = mb_row * cm->mb_cols + mb_col; #endif - vp9_clear_system_state(); + vpx_clear_system_state(); xd->plane[0].dst.buf = new_yv12->y_buffer + recon_yoffset; xd->plane[1].dst.buf = new_yv12->u_buffer + recon_uvoffset; @@ -708,7 +707,7 @@ void vp9_first_pass(VP9_COMP *cpi, const struct lookahead_entry *source) { } #endif // CONFIG_VP9_HIGHBITDEPTH - vp9_clear_system_state(); + vpx_clear_system_state(); log_intra = log(this_error + 1.0); if (log_intra < 10.0) intra_factor += 1.0 + ((10.0 - log_intra) * 0.05); @@ -878,7 +877,7 @@ void vp9_first_pass(VP9_COMP *cpi, const struct lookahead_entry *source) { #endif if (motion_error <= this_error) { - vp9_clear_system_state(); + vpx_clear_system_state(); // Keep a count of cases where the inter and intra were very close // and very low. This helps with scene cut detection for example in @@ -1013,7 +1012,7 @@ void vp9_first_pass(VP9_COMP *cpi, const struct lookahead_entry *source) { x->plane[2].src.buf += uv_mb_height * x->plane[1].src.stride - uv_mb_height * cm->mb_cols; - vp9_clear_system_state(); + vpx_clear_system_state(); } // Clamp the image start to rows/2. This number of rows is discarded top @@ -1891,7 +1890,7 @@ static void define_gf_group(VP9_COMP *cpi, FIRSTPASS_STATS *this_frame) { vp9_zero(twopass->gf_group); } - vp9_clear_system_state(); + vpx_clear_system_state(); vp9_zero(next_frame); // Load stats for the current frame. @@ -2632,7 +2631,7 @@ void vp9_rc_get_second_pass_params(VP9_COMP *cpi) { return; } - vp9_clear_system_state(); + vpx_clear_system_state(); if (cpi->oxcf.rc_mode == VPX_Q) { twopass->active_worst_quality = cpi->oxcf.cq_level; diff --git a/vp9/encoder/vp9_mbgraph.c b/vp9/encoder/vp9_mbgraph.c index acbd7dd43..d59f3157a 100644 --- a/vp9/encoder/vp9_mbgraph.c +++ b/vp9/encoder/vp9_mbgraph.c @@ -14,12 +14,12 @@ #include "./vpx_dsp_rtcd.h" #include "vpx_mem/vpx_mem.h" +#include "vpx_ports/system_state.h" #include "vp9/encoder/vp9_segmentation.h" #include "vp9/encoder/vp9_mcomp.h" #include "vp9/common/vp9_blockd.h" #include "vp9/common/vp9_reconinter.h" #include "vp9/common/vp9_reconintra.h" -#include "vp9/common/vp9_systemdependent.h" static unsigned int do_16x16_motion_iteration(VP9_COMP *cpi, @@ -410,7 +410,7 @@ void vp9_update_mbgraph_stats(VP9_COMP *cpi) { golden_ref, cpi->Source); } - vp9_clear_system_state(); + vpx_clear_system_state(); separate_arf_mbs(cpi); } diff --git a/vp9/encoder/vp9_mcomp.c b/vp9/encoder/vp9_mcomp.c index 081b99f9f..aa3e51ceb 100644 --- a/vp9/encoder/vp9_mcomp.c +++ b/vp9/encoder/vp9_mcomp.c @@ -256,6 +256,27 @@ static INLINE const uint8_t *pre(const uint8_t *buf, int stride, int r, int c) { } \ } +// TODO(yunqingwang): SECOND_LEVEL_CHECKS_BEST was a rewrote of +// SECOND_LEVEL_CHECKS, and SECOND_LEVEL_CHECKS should be rewritten +// later in the same way. +#define SECOND_LEVEL_CHECKS_BEST \ + { \ + unsigned int second; \ + int br0 = br; \ + int bc0 = bc; \ + assert(tr == br || tc == bc); \ + if (tr == br && tc != bc) { \ + kc = bc - tc; \ + } else if (tr != br && tc == bc) { \ + kr = br - tr; \ + } \ + CHECK_BETTER(second, br0 + kr, bc0); \ + CHECK_BETTER(second, br0, bc0 + kc); \ + if (br0 != br || bc0 != bc) { \ + CHECK_BETTER(second, br0 + kr, bc0 + kc); \ + } \ + } + #define SETUP_SUBPEL_SEARCH \ const uint8_t *const z = x->plane[0].src.buf; \ const int src_stride = x->plane[0].src.stride; \ @@ -636,7 +657,6 @@ int vp9_find_best_sub_pixel_tree(const MACROBLOCK *x, const MACROBLOCKD *xd = &x->e_mbd; unsigned int besterr = INT_MAX; unsigned int sse; - unsigned int whichdir = 0; int thismse; const int y_stride = xd->plane[0].pre[0].stride; const int offset = bestmv->row * y_stride + bestmv->col; @@ -657,6 +677,7 @@ int vp9_find_best_sub_pixel_tree(const MACROBLOCK *x, const MV *search_step = search_step_table; int idx, best_idx = -1; unsigned int cost_array[5]; + int kr, kc; if (!(allow_hp && vp9_use_mv_hp(ref_mv))) if (round == 3) @@ -703,8 +724,11 @@ int vp9_find_best_sub_pixel_tree(const MACROBLOCK *x, } // Check diagonal sub-pixel position - tc = bc + (cost_array[0] < cost_array[1] ? -hstep : hstep); - tr = br + (cost_array[2] < cost_array[3] ? -hstep : hstep); + kc = (cost_array[0] <= cost_array[1] ? -hstep : hstep); + kr = (cost_array[2] <= cost_array[3] ? -hstep : hstep); + + tc = bc + kc; + tr = br + kr; if (tc >= minc && tc <= maxc && tr >= minr && tr <= maxr) { const uint8_t *const pre_address = y + (tr >> 3) * y_stride + (tc >> 3); MV this_mv = {tr, tc}; @@ -735,8 +759,8 @@ int vp9_find_best_sub_pixel_tree(const MACROBLOCK *x, bc = tc; } - if (iters_per_step > 1) - SECOND_LEVEL_CHECKS; + if (iters_per_step > 1 && best_idx != -1) + SECOND_LEVEL_CHECKS_BEST; tr = br; tc = bc; @@ -1704,7 +1728,7 @@ int vp9_diamond_search_sad_c(const MACROBLOCK *x, } } break; - }; + } #endif } else if (best_address == in_what) { (*num00)++; diff --git a/vp9/encoder/vp9_pickmode.c b/vp9/encoder/vp9_pickmode.c index 6831be8d5..cc018fcbe 100644 --- a/vp9/encoder/vp9_pickmode.c +++ b/vp9/encoder/vp9_pickmode.c @@ -1050,6 +1050,16 @@ static const REF_MODE ref_mode_set[RT_INTER_MODES] = { {GOLDEN_FRAME, NEARMV}, {GOLDEN_FRAME, NEWMV} }; +static const REF_MODE ref_mode_set_svc[RT_INTER_MODES] = { + {LAST_FRAME, ZEROMV}, + {GOLDEN_FRAME, ZEROMV}, + {LAST_FRAME, NEARESTMV}, + {LAST_FRAME, NEARMV}, + {GOLDEN_FRAME, NEARESTMV}, + {GOLDEN_FRAME, NEARMV}, + {LAST_FRAME, NEWMV}, + {GOLDEN_FRAME, NEWMV} +}; // TODO(jingning) placeholder for inter-frame non-RD mode decision. // this needs various further optimizations. to be continued.. @@ -1157,12 +1167,11 @@ void vp9_pick_inter_mode(VP9_COMP *cpi, MACROBLOCK *x, vp9_denoiser_reset_frame_stats(ctx); #endif - if (cpi->rc.frames_since_golden == 0) { + if (cpi->rc.frames_since_golden == 0 && !cpi->use_svc) { usable_ref_frame = LAST_FRAME; } else { usable_ref_frame = GOLDEN_FRAME; } - for (ref_frame = LAST_FRAME; ref_frame <= usable_ref_frame; ++ref_frame) { const YV12_BUFFER_CONFIG *yv12 = get_ref_frame_buffer(cpi, ref_frame); @@ -1204,15 +1213,19 @@ void vp9_pick_inter_mode(VP9_COMP *cpi, MACROBLOCK *x, int mode_rd_thresh; int mode_index; int i; - PREDICTION_MODE this_mode = ref_mode_set[idx].pred_mode; int64_t this_sse; int is_skippable; int this_early_term = 0; + PREDICTION_MODE this_mode = ref_mode_set[idx].pred_mode; + if (cpi->use_svc) + this_mode = ref_mode_set_svc[idx].pred_mode; if (!(cpi->sf.inter_mode_mask[bsize] & (1 << this_mode))) continue; ref_frame = ref_mode_set[idx].ref_frame; + if (cpi->use_svc) + ref_frame = ref_mode_set_svc[idx].ref_frame; if (!(cpi->ref_frame_flags & flag_list[ref_frame])) continue; if (const_motion[ref_frame] && this_mode == NEARMV) @@ -1240,7 +1253,7 @@ void vp9_pick_inter_mode(VP9_COMP *cpi, MACROBLOCK *x, continue; if (this_mode == NEWMV) { - if (ref_frame > LAST_FRAME) { + if (ref_frame > LAST_FRAME && !cpi->use_svc) { int tmp_sad; int dis, cost_list[5]; @@ -1289,6 +1302,21 @@ void vp9_pick_inter_mode(VP9_COMP *cpi, MACROBLOCK *x, x->pred_mv_sad[LAST_FRAME] = best_pred_sad; } + if (cpi->use_svc) { + if (this_mode == NEWMV && ref_frame == GOLDEN_FRAME && + frame_mv[NEWMV][GOLDEN_FRAME].as_int != INVALID_MV) { + const int pre_stride = xd->plane[0].pre[0].stride; + const uint8_t * const pre_buf = xd->plane[0].pre[0].buf + + (frame_mv[NEWMV][GOLDEN_FRAME].as_mv.row >> 3) * pre_stride + + (frame_mv[NEWMV][GOLDEN_FRAME].as_mv.col >> 3); + best_pred_sad = cpi->fn_ptr[bsize].sdf(x->plane[0].src.buf, + x->plane[0].src.stride, + pre_buf, pre_stride); + x->pred_mv_sad[GOLDEN_FRAME] = best_pred_sad; + } + } + + if (this_mode != NEARESTMV && frame_mv[this_mode][ref_frame].as_int == frame_mv[NEARESTMV][ref_frame].as_int) @@ -1311,7 +1339,8 @@ void vp9_pick_inter_mode(VP9_COMP *cpi, MACROBLOCK *x, } if ((this_mode == NEWMV || filter_ref == SWITCHABLE) && pred_filter_search - && (ref_frame == LAST_FRAME) + && (ref_frame == LAST_FRAME || + (ref_frame == GOLDEN_FRAME && cpi->use_svc)) && (((mbmi->mv[0].as_mv.row | mbmi->mv[0].as_mv.col) & 0x07) != 0)) { int pf_rate[3]; int64_t pf_dist[3]; diff --git a/vp9/encoder/vp9_psnrhvs.c b/vp9/encoder/vp9_psnrhvs.c deleted file mode 100644 index 5104b9af6..000000000 --- a/vp9/encoder/vp9_psnrhvs.c +++ /dev/null @@ -1,224 +0,0 @@ -/* - * Copyright (c) 2010 The WebM project authors. All Rights Reserved. - * - * Use of this source code is governed by a BSD-style license - * that can be found in the LICENSE file in the root of the source - * tree. An additional intellectual property rights grant can be found - * in the file PATENTS. All contributing project authors may - * be found in the AUTHORS file in the root of the source tree. - * - * This code was originally written by: Gregory Maxwell, at the Daala - * project. - */ -#include <stdio.h> -#include <stdlib.h> -#include <math.h> - -#include "./vpx_config.h" -#include "./vp9_rtcd.h" -#include "./vpx_dsp_rtcd.h" -#include "vp9/encoder/vp9_ssim.h" - -#if !defined(M_PI) -# define M_PI (3.141592653589793238462643) -#endif -#include <string.h> - -void od_bin_fdct8x8(tran_low_t *y, int ystride, const int16_t *x, int xstride) { - (void) xstride; - vpx_fdct8x8(x, y, ystride); -} - -/* Normalized inverse quantization matrix for 8x8 DCT at the point of - * transparency. This is not the JPEG based matrix from the paper, - this one gives a slightly higher MOS agreement.*/ -float csf_y[8][8] = {{1.6193873005, 2.2901594831, 2.08509755623, 1.48366094411, - 1.00227514334, 0.678296995242, 0.466224900598, 0.3265091542}, {2.2901594831, - 1.94321815382, 2.04793073064, 1.68731108984, 1.2305666963, 0.868920337363, - 0.61280991668, 0.436405793551}, {2.08509755623, 2.04793073064, - 1.34329019223, 1.09205635862, 0.875748795257, 0.670882927016, - 0.501731932449, 0.372504254596}, {1.48366094411, 1.68731108984, - 1.09205635862, 0.772819797575, 0.605636379554, 0.48309405692, - 0.380429446972, 0.295774038565}, {1.00227514334, 1.2305666963, - 0.875748795257, 0.605636379554, 0.448996256676, 0.352889268808, - 0.283006984131, 0.226951348204}, {0.678296995242, 0.868920337363, - 0.670882927016, 0.48309405692, 0.352889268808, 0.27032073436, - 0.215017739696, 0.17408067321}, {0.466224900598, 0.61280991668, - 0.501731932449, 0.380429446972, 0.283006984131, 0.215017739696, - 0.168869545842, 0.136153931001}, {0.3265091542, 0.436405793551, - 0.372504254596, 0.295774038565, 0.226951348204, 0.17408067321, - 0.136153931001, 0.109083846276}}; -float csf_cb420[8][8] = { - {1.91113096927, 2.46074210438, 1.18284184739, 1.14982565193, 1.05017074788, - 0.898018824055, 0.74725392039, 0.615105596242}, {2.46074210438, - 1.58529308355, 1.21363250036, 1.38190029285, 1.33100189972, - 1.17428548929, 0.996404342439, 0.830890433625}, {1.18284184739, - 1.21363250036, 0.978712413627, 1.02624506078, 1.03145147362, - 0.960060382087, 0.849823426169, 0.731221236837}, {1.14982565193, - 1.38190029285, 1.02624506078, 0.861317501629, 0.801821139099, - 0.751437590932, 0.685398513368, 0.608694761374}, {1.05017074788, - 1.33100189972, 1.03145147362, 0.801821139099, 0.676555426187, - 0.605503172737, 0.55002013668, 0.495804539034}, {0.898018824055, - 1.17428548929, 0.960060382087, 0.751437590932, 0.605503172737, - 0.514674450957, 0.454353482512, 0.407050308965}, {0.74725392039, - 0.996404342439, 0.849823426169, 0.685398513368, 0.55002013668, - 0.454353482512, 0.389234902883, 0.342353999733}, {0.615105596242, - 0.830890433625, 0.731221236837, 0.608694761374, 0.495804539034, - 0.407050308965, 0.342353999733, 0.295530605237}}; -float csf_cr420[8][8] = { - {2.03871978502, 2.62502345193, 1.26180942886, 1.11019789803, 1.01397751469, - 0.867069376285, 0.721500455585, 0.593906509971}, {2.62502345193, - 1.69112867013, 1.17180569821, 1.3342742857, 1.28513006198, - 1.13381474809, 0.962064122248, 0.802254508198}, {1.26180942886, - 1.17180569821, 0.944981930573, 0.990876405848, 0.995903384143, - 0.926972725286, 0.820534991409, 0.706020324706}, {1.11019789803, - 1.3342742857, 0.990876405848, 0.831632933426, 0.77418706195, - 0.725539939514, 0.661776842059, 0.587716619023}, {1.01397751469, - 1.28513006198, 0.995903384143, 0.77418706195, 0.653238524286, - 0.584635025748, 0.531064164893, 0.478717061273}, {0.867069376285, - 1.13381474809, 0.926972725286, 0.725539939514, 0.584635025748, - 0.496936637883, 0.438694579826, 0.393021669543}, {0.721500455585, - 0.962064122248, 0.820534991409, 0.661776842059, 0.531064164893, - 0.438694579826, 0.375820256136, 0.330555063063}, {0.593906509971, - 0.802254508198, 0.706020324706, 0.587716619023, 0.478717061273, - 0.393021669543, 0.330555063063, 0.285345396658}}; - -static double convert_score_db(double _score, double _weight) { - return 10 * (log10(255 * 255) - log10(_weight * _score)); -} - -static double calc_psnrhvs(const unsigned char *_src, int _systride, - const unsigned char *_dst, int _dystride, - double _par, int _w, int _h, int _step, - float _csf[8][8]) { - float ret; - int16_t dct_s[8 * 8], dct_d[8 * 8]; - tran_low_t dct_s_coef[8 * 8], dct_d_coef[8 * 8]; - float mask[8][8]; - int pixels; - int x; - int y; - (void) _par; - ret = pixels = 0; - /*In the PSNR-HVS-M paper[1] the authors describe the construction of - their masking table as "we have used the quantization table for the - color component Y of JPEG [6] that has been also obtained on the - basis of CSF. Note that the values in quantization table JPEG have - been normalized and then squared." Their CSF matrix (from PSNR-HVS) - was also constructed from the JPEG matrices. I can not find any obvious - scheme of normalizing to produce their table, but if I multiply their - CSF by 0.38857 and square the result I get their masking table. - I have no idea where this constant comes from, but deviating from it - too greatly hurts MOS agreement. - - [1] Nikolay Ponomarenko, Flavia Silvestri, Karen Egiazarian, Marco Carli, - Jaakko Astola, Vladimir Lukin, "On between-coefficient contrast masking - of DCT basis functions", CD-ROM Proceedings of the Third - International Workshop on Video Processing and Quality Metrics for Consumer - Electronics VPQM-07, Scottsdale, Arizona, USA, 25-26 January, 2007, 4 p.*/ - for (x = 0; x < 8; x++) - for (y = 0; y < 8; y++) - mask[x][y] = (_csf[x][y] * 0.3885746225901003) - * (_csf[x][y] * 0.3885746225901003); - for (y = 0; y < _h - 7; y += _step) { - for (x = 0; x < _w - 7; x += _step) { - int i; - int j; - float s_means[4]; - float d_means[4]; - float s_vars[4]; - float d_vars[4]; - float s_gmean = 0; - float d_gmean = 0; - float s_gvar = 0; - float d_gvar = 0; - float s_mask = 0; - float d_mask = 0; - for (i = 0; i < 4; i++) - s_means[i] = d_means[i] = s_vars[i] = d_vars[i] = 0; - for (i = 0; i < 8; i++) { - for (j = 0; j < 8; j++) { - int sub = ((i & 12) >> 2) + ((j & 12) >> 1); - dct_s[i * 8 + j] = _src[(y + i) * _systride + (j + x)]; - dct_d[i * 8 + j] = _dst[(y + i) * _dystride + (j + x)]; - s_gmean += dct_s[i * 8 + j]; - d_gmean += dct_d[i * 8 + j]; - s_means[sub] += dct_s[i * 8 + j]; - d_means[sub] += dct_d[i * 8 + j]; - } - } - s_gmean /= 64.f; - d_gmean /= 64.f; - for (i = 0; i < 4; i++) - s_means[i] /= 16.f; - for (i = 0; i < 4; i++) - d_means[i] /= 16.f; - for (i = 0; i < 8; i++) { - for (j = 0; j < 8; j++) { - int sub = ((i & 12) >> 2) + ((j & 12) >> 1); - s_gvar += (dct_s[i * 8 + j] - s_gmean) * (dct_s[i * 8 + j] - s_gmean); - d_gvar += (dct_d[i * 8 + j] - d_gmean) * (dct_d[i * 8 + j] - d_gmean); - s_vars[sub] += (dct_s[i * 8 + j] - s_means[sub]) - * (dct_s[i * 8 + j] - s_means[sub]); - d_vars[sub] += (dct_d[i * 8 + j] - d_means[sub]) - * (dct_d[i * 8 + j] - d_means[sub]); - } - } - s_gvar *= 1 / 63.f * 64; - d_gvar *= 1 / 63.f * 64; - for (i = 0; i < 4; i++) - s_vars[i] *= 1 / 15.f * 16; - for (i = 0; i < 4; i++) - d_vars[i] *= 1 / 15.f * 16; - if (s_gvar > 0) - s_gvar = (s_vars[0] + s_vars[1] + s_vars[2] + s_vars[3]) / s_gvar; - if (d_gvar > 0) - d_gvar = (d_vars[0] + d_vars[1] + d_vars[2] + d_vars[3]) / d_gvar; - od_bin_fdct8x8(dct_s_coef, 8, dct_s, 8); - od_bin_fdct8x8(dct_d_coef, 8, dct_d, 8); - for (i = 0; i < 8; i++) - for (j = (i == 0); j < 8; j++) - s_mask += dct_s_coef[i * 8 + j] * dct_s_coef[i * 8 + j] * mask[i][j]; - for (i = 0; i < 8; i++) - for (j = (i == 0); j < 8; j++) - d_mask += dct_d_coef[i * 8 + j] * dct_d_coef[i * 8 + j] * mask[i][j]; - s_mask = sqrt(s_mask * s_gvar) / 32.f; - d_mask = sqrt(d_mask * d_gvar) / 32.f; - if (d_mask > s_mask) - s_mask = d_mask; - for (i = 0; i < 8; i++) { - for (j = 0; j < 8; j++) { - float err; - err = fabs(dct_s_coef[i * 8 + j] - dct_d_coef[i * 8 + j]); - if (i != 0 || j != 0) - err = err < s_mask / mask[i][j] ? 0 : err - s_mask / mask[i][j]; - ret += (err * _csf[i][j]) * (err * _csf[i][j]); - pixels++; - } - } - } - } - ret /= pixels; - return ret; -} -double vp9_psnrhvs(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest, - double *y_psnrhvs, double *u_psnrhvs, double *v_psnrhvs) { - double psnrhvs; - double par = 1.0; - int step = 7; - vp9_clear_system_state(); - *y_psnrhvs = calc_psnrhvs(source->y_buffer, source->y_stride, dest->y_buffer, - dest->y_stride, par, source->y_crop_width, - source->y_crop_height, step, csf_y); - - *u_psnrhvs = calc_psnrhvs(source->u_buffer, source->uv_stride, dest->u_buffer, - dest->uv_stride, par, source->uv_crop_width, - source->uv_crop_height, step, csf_cb420); - - *v_psnrhvs = calc_psnrhvs(source->v_buffer, source->uv_stride, dest->v_buffer, - dest->uv_stride, par, source->uv_crop_width, - source->uv_crop_height, step, csf_cr420); - psnrhvs = (*y_psnrhvs) * .8 + .1 * ((*u_psnrhvs) + (*v_psnrhvs)); - - return convert_score_db(psnrhvs, 1.0); -} diff --git a/vp9/encoder/vp9_quantize.c b/vp9/encoder/vp9_quantize.c index d53d95d29..cb3e21a56 100644 --- a/vp9/encoder/vp9_quantize.c +++ b/vp9/encoder/vp9_quantize.c @@ -199,7 +199,7 @@ void vp9_regular_quantize_b_4x4(MACROBLOCK *x, int plane, int block, #if CONFIG_VP9_HIGHBITDEPTH if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) { - vp9_highbd_quantize_b(BLOCK_OFFSET(p->coeff, block), + vpx_highbd_quantize_b(BLOCK_OFFSET(p->coeff, block), 16, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, BLOCK_OFFSET(p->qcoeff, block), @@ -209,7 +209,7 @@ void vp9_regular_quantize_b_4x4(MACROBLOCK *x, int plane, int block, return; } #endif - vp9_quantize_b(BLOCK_OFFSET(p->coeff, block), + vpx_quantize_b(BLOCK_OFFSET(p->coeff, block), 16, x->skip_block, p->zbin, p->round, p->quant, p->quant_shift, BLOCK_OFFSET(p->qcoeff, block), diff --git a/vp9/encoder/vp9_ratectrl.c b/vp9/encoder/vp9_ratectrl.c index 7427ccfb9..a0b3f8a75 100644 --- a/vp9/encoder/vp9_ratectrl.c +++ b/vp9/encoder/vp9_ratectrl.c @@ -17,6 +17,7 @@ #include "vpx_mem/vpx_mem.h" #include "vpx_ports/mem.h" +#include "vpx_ports/system_state.h" #include "vp9/common/vp9_alloccommon.h" #include "vp9/encoder/vp9_aq_cyclicrefresh.h" @@ -24,7 +25,6 @@ #include "vp9/common/vp9_entropymode.h" #include "vp9/common/vp9_quant_common.h" #include "vp9/common/vp9_seg_common.h" -#include "vp9/common/vp9_systemdependent.h" #include "vp9/encoder/vp9_encodemv.h" #include "vp9/encoder/vp9_ratectrl.h" @@ -455,7 +455,7 @@ void vp9_rc_update_rate_correction_factors(VP9_COMP *cpi) { return; // Clear down mmx registers to allow floating point in what follows - vp9_clear_system_state(); + vpx_clear_system_state(); // Work out how big we would have expected the frame to be at this Q given // the current correction factor. @@ -635,6 +635,7 @@ static int calc_active_worst_quality_one_pass_cbr(const VP9_COMP *cpi) { int adjustment = 0; int active_worst_quality; int ambient_qp; + unsigned int num_frames_weight_key = 5 * cpi->svc.number_temporal_layers; if (cm->frame_type == KEY_FRAME) return rc->worst_quality; // For ambient_qp we use minimum of avg_frame_qindex[KEY_FRAME/INTER_FRAME] @@ -642,7 +643,7 @@ static int calc_active_worst_quality_one_pass_cbr(const VP9_COMP *cpi) { // to worst_quality and updated with (3/4, 1/4) average in postencode_update. // So for first few frames following key, the qp of that key frame is weighted // into the active_worst_quality setting. - ambient_qp = (cm->current_video_frame < 5) ? + ambient_qp = (cm->current_video_frame < num_frames_weight_key) ? MIN(rc->avg_frame_qindex[INTER_FRAME], rc->avg_frame_qindex[KEY_FRAME]) : rc->avg_frame_qindex[INTER_FRAME]; active_worst_quality = MIN(rc->worst_quality, @@ -764,7 +765,7 @@ static int rc_pick_q_and_bounds_one_pass_cbr(const VP9_COMP *cpi, !rc->this_key_frame_forced && !(cm->current_video_frame == 0)) { int qdelta = 0; - vp9_clear_system_state(); + vpx_clear_system_state(); qdelta = vp9_compute_qdelta_by_rate(&cpi->rc, cm->frame_type, active_worst_quality, 2.0, cm->bit_depth); @@ -823,7 +824,6 @@ static int rc_pick_q_and_bounds_one_pass_vbr(const VP9_COMP *cpi, ASSIGN_MINQ_TABLE(cm->bit_depth, inter_minq); if (frame_is_intra_only(cm)) { - // Handle the special case for key frames forced when we have reached // the maximum key frame interval. Here force the Q to a range // based on the ambient Q to reduce the risk of popping. @@ -915,7 +915,7 @@ static int rc_pick_q_and_bounds_one_pass_vbr(const VP9_COMP *cpi, #if LIMIT_QRANGE_FOR_ALTREF_AND_KEY { int qdelta = 0; - vp9_clear_system_state(); + vpx_clear_system_state(); // Limit Q range for the adaptive loop. if (cm->frame_type == KEY_FRAME && @@ -1109,7 +1109,7 @@ static int rc_pick_q_and_bounds_two_pass(const VP9_COMP *cpi, } #if LIMIT_QRANGE_FOR_ALTREF_AND_KEY - vp9_clear_system_state(); + vpx_clear_system_state(); // Static forced key frames Q restrictions dealt with elsewhere. if (!((frame_is_intra_only(cm) || vp9_is_upper_layer_key_frame(cpi))) || !rc->this_key_frame_forced || @@ -1286,6 +1286,18 @@ void vp9_rc_postencode_update(VP9_COMP *cpi, uint64_t bytes_used) { rc->last_q[KEY_FRAME] = qindex; rc->avg_frame_qindex[KEY_FRAME] = ROUND_POWER_OF_TWO(3 * rc->avg_frame_qindex[KEY_FRAME] + qindex, 2); + if (cpi->use_svc) { + int i = 0; + SVC *svc = &cpi->svc; + for (i = 0; i < svc->number_temporal_layers; ++i) { + const int layer = LAYER_IDS_TO_IDX(svc->spatial_layer_id, i, + svc->number_temporal_layers); + LAYER_CONTEXT *lc = &svc->layer_context[layer]; + RATE_CONTROL *lrc = &lc->rc; + lrc->last_q[KEY_FRAME] = rc->last_q[KEY_FRAME]; + lrc->avg_frame_qindex[KEY_FRAME] = rc->avg_frame_qindex[KEY_FRAME]; + } + } } else { if (rc->is_src_frame_alt_ref || !(cpi->refresh_golden_frame || cpi->refresh_alt_ref_frame) || @@ -1339,13 +1351,15 @@ void vp9_rc_postencode_update(VP9_COMP *cpi, uint64_t bytes_used) { rc->total_target_vs_actual = rc->total_actual_bits - rc->total_target_bits; - if (is_altref_enabled(cpi) && cpi->refresh_alt_ref_frame && - (cm->frame_type != KEY_FRAME)) - // Update the alternate reference frame stats as appropriate. - update_alt_ref_frame_stats(cpi); - else - // Update the Golden frame stats as appropriate. - update_golden_frame_stats(cpi); + if (!cpi->use_svc) { + if (is_altref_enabled(cpi) && cpi->refresh_alt_ref_frame && + (cm->frame_type != KEY_FRAME)) + // Update the alternate reference frame stats as appropriate. + update_alt_ref_frame_stats(cpi); + else + // Update the Golden frame stats as appropriate. + update_golden_frame_stats(cpi); + } if (cm->frame_type == KEY_FRAME) rc->frames_since_key = 0; diff --git a/vp9/encoder/vp9_rd.c b/vp9/encoder/vp9_rd.c index 3f6de42e3..3ee6fbec3 100644 --- a/vp9/encoder/vp9_rd.c +++ b/vp9/encoder/vp9_rd.c @@ -15,7 +15,9 @@ #include "./vp9_rtcd.h" #include "vpx_mem/vpx_mem.h" +#include "vpx_ports/bitops.h" #include "vpx_ports/mem.h" +#include "vpx_ports/system_state.h" #include "vp9/common/vp9_common.h" #include "vp9/common/vp9_entropy.h" @@ -26,7 +28,6 @@ #include "vp9/common/vp9_reconinter.h" #include "vp9/common/vp9_reconintra.h" #include "vp9/common/vp9_seg_common.h" -#include "vp9/common/vp9_systemdependent.h" #include "vp9/encoder/vp9_cost.h" #include "vp9/encoder/vp9_encodemb.h" @@ -268,7 +269,7 @@ void vp9_initialize_rd_consts(VP9_COMP *cpi) { RD_OPT *const rd = &cpi->rd; int i; - vp9_clear_system_state(); + vpx_clear_system_state(); rd->RDDIV = RDDIV_BITS; // In bits (to multiply D by 128). rd->RDMULT = vp9_compute_rd_mult(cpi, cm->base_qindex + cm->y_dc_delta_q); diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c index b6cb8b57f..9f3873600 100644 --- a/vp9/encoder/vp9_rdopt.c +++ b/vp9/encoder/vp9_rdopt.c @@ -16,6 +16,7 @@ #include "vpx_mem/vpx_mem.h" #include "vpx_ports/mem.h" +#include "vpx_ports/system_state.h" #include "vp9/common/vp9_common.h" #include "vp9/common/vp9_entropy.h" @@ -28,7 +29,6 @@ #include "vp9/common/vp9_reconintra.h" #include "vp9/common/vp9_scan.h" #include "vp9/common/vp9_seg_common.h" -#include "vp9/common/vp9_systemdependent.h" #include "vp9/encoder/vp9_cost.h" #include "vp9/encoder/vp9_encodemb.h" @@ -1045,7 +1045,6 @@ static int64_t rd_pick_intra_sby_mode(VP9_COMP *cpi, MACROBLOCK *x, memset(x->skip_txfm, SKIP_TXFM_NONE, sizeof(x->skip_txfm)); /* Y Search for intra prediction mode */ for (mode = DC_PRED; mode <= TM_PRED; mode++) { - if (cpi->sf.use_nonrd_pick_mode) { // These speed features are turned on in hybrid non-RD and RD mode // for key frame coding in the context of real-time setting. @@ -2489,9 +2488,8 @@ static int64_t handle_inter_mode(VP9_COMP *cpi, MACROBLOCK *x, } // We don't include the cost of the second reference here, because there - // are only three options: Last/Golden, ARF/Last or Golden/ARF, or in other - // words if you present them in that order, the second one is always known - // if the first is known. + // are only two options: Last/ARF or Golden/ARF; The second one is always + // known, which is ARF. // // Under some circumstances we discount the cost of new mv mode to encourage // initiation of a motion field. diff --git a/vp9/encoder/vp9_ssim.c b/vp9/encoder/vp9_ssim.c deleted file mode 100644 index 172de5d1d..000000000 --- a/vp9/encoder/vp9_ssim.c +++ /dev/null @@ -1,500 +0,0 @@ -/* - * Copyright (c) 2010 The WebM project authors. All Rights Reserved. - * - * Use of this source code is governed by a BSD-style license - * that can be found in the LICENSE file in the root of the source - * tree. An additional intellectual property rights grant can be found - * in the file PATENTS. All contributing project authors may - * be found in the AUTHORS file in the root of the source tree. - */ - -#include <math.h> -#include "./vp9_rtcd.h" -#include "vpx_ports/mem.h" -#include "vp9/encoder/vp9_ssim.h" - -void vp9_ssim_parms_16x16_c(uint8_t *s, int sp, uint8_t *r, - int rp, unsigned long *sum_s, unsigned long *sum_r, - unsigned long *sum_sq_s, unsigned long *sum_sq_r, - unsigned long *sum_sxr) { - int i, j; - for (i = 0; i < 16; i++, s += sp, r += rp) { - for (j = 0; j < 16; j++) { - *sum_s += s[j]; - *sum_r += r[j]; - *sum_sq_s += s[j] * s[j]; - *sum_sq_r += r[j] * r[j]; - *sum_sxr += s[j] * r[j]; - } - } -} -void vp9_ssim_parms_8x8_c(uint8_t *s, int sp, uint8_t *r, int rp, - unsigned long *sum_s, unsigned long *sum_r, - unsigned long *sum_sq_s, unsigned long *sum_sq_r, - unsigned long *sum_sxr) { - int i, j; - for (i = 0; i < 8; i++, s += sp, r += rp) { - for (j = 0; j < 8; j++) { - *sum_s += s[j]; - *sum_r += r[j]; - *sum_sq_s += s[j] * s[j]; - *sum_sq_r += r[j] * r[j]; - *sum_sxr += s[j] * r[j]; - } - } -} - -#if CONFIG_VP9_HIGHBITDEPTH -void vp9_highbd_ssim_parms_8x8_c(uint16_t *s, int sp, uint16_t *r, int rp, - uint32_t *sum_s, uint32_t *sum_r, - uint32_t *sum_sq_s, uint32_t *sum_sq_r, - uint32_t *sum_sxr) { - int i, j; - for (i = 0; i < 8; i++, s += sp, r += rp) { - for (j = 0; j < 8; j++) { - *sum_s += s[j]; - *sum_r += r[j]; - *sum_sq_s += s[j] * s[j]; - *sum_sq_r += r[j] * r[j]; - *sum_sxr += s[j] * r[j]; - } - } -} -#endif // CONFIG_VP9_HIGHBITDEPTH - -static const int64_t cc1 = 26634; // (64^2*(.01*255)^2 -static const int64_t cc2 = 239708; // (64^2*(.03*255)^2 - -static double similarity(unsigned long sum_s, unsigned long sum_r, - unsigned long sum_sq_s, unsigned long sum_sq_r, - unsigned long sum_sxr, int count) { - int64_t ssim_n, ssim_d; - int64_t c1, c2; - - // scale the constants by number of pixels - c1 = (cc1 * count * count) >> 12; - c2 = (cc2 * count * count) >> 12; - - ssim_n = (2 * sum_s * sum_r + c1) * ((int64_t) 2 * count * sum_sxr - - (int64_t) 2 * sum_s * sum_r + c2); - - ssim_d = (sum_s * sum_s + sum_r * sum_r + c1) * - ((int64_t)count * sum_sq_s - (int64_t)sum_s * sum_s + - (int64_t)count * sum_sq_r - (int64_t) sum_r * sum_r + c2); - - return ssim_n * 1.0 / ssim_d; -} - -static double ssim_8x8(uint8_t *s, int sp, uint8_t *r, int rp) { - unsigned long sum_s = 0, sum_r = 0, sum_sq_s = 0, sum_sq_r = 0, sum_sxr = 0; - vp9_ssim_parms_8x8(s, sp, r, rp, &sum_s, &sum_r, &sum_sq_s, &sum_sq_r, - &sum_sxr); - return similarity(sum_s, sum_r, sum_sq_s, sum_sq_r, sum_sxr, 64); -} - -#if CONFIG_VP9_HIGHBITDEPTH -static double highbd_ssim_8x8(uint16_t *s, int sp, uint16_t *r, int rp, - unsigned int bd) { - uint32_t sum_s = 0, sum_r = 0, sum_sq_s = 0, sum_sq_r = 0, sum_sxr = 0; - const int oshift = bd - 8; - vp9_highbd_ssim_parms_8x8(s, sp, r, rp, &sum_s, &sum_r, &sum_sq_s, &sum_sq_r, - &sum_sxr); - return similarity(sum_s >> oshift, - sum_r >> oshift, - sum_sq_s >> (2 * oshift), - sum_sq_r >> (2 * oshift), - sum_sxr >> (2 * oshift), - 64); -} -#endif // CONFIG_VP9_HIGHBITDEPTH - -// We are using a 8x8 moving window with starting location of each 8x8 window -// on the 4x4 pixel grid. Such arrangement allows the windows to overlap -// block boundaries to penalize blocking artifacts. -double vp9_ssim2(uint8_t *img1, uint8_t *img2, int stride_img1, - int stride_img2, int width, int height) { - int i, j; - int samples = 0; - double ssim_total = 0; - - // sample point start with each 4x4 location - for (i = 0; i <= height - 8; - i += 4, img1 += stride_img1 * 4, img2 += stride_img2 * 4) { - for (j = 0; j <= width - 8; j += 4) { - double v = ssim_8x8(img1 + j, stride_img1, img2 + j, stride_img2); - ssim_total += v; - samples++; - } - } - ssim_total /= samples; - return ssim_total; -} - -#if CONFIG_VP9_HIGHBITDEPTH -double vp9_highbd_ssim2(uint8_t *img1, uint8_t *img2, int stride_img1, - int stride_img2, int width, int height, - unsigned int bd) { - int i, j; - int samples = 0; - double ssim_total = 0; - - // sample point start with each 4x4 location - for (i = 0; i <= height - 8; - i += 4, img1 += stride_img1 * 4, img2 += stride_img2 * 4) { - for (j = 0; j <= width - 8; j += 4) { - double v = highbd_ssim_8x8(CONVERT_TO_SHORTPTR(img1 + j), stride_img1, - CONVERT_TO_SHORTPTR(img2 + j), stride_img2, - bd); - ssim_total += v; - samples++; - } - } - ssim_total /= samples; - return ssim_total; -} -#endif // CONFIG_VP9_HIGHBITDEPTH - -double vp9_calc_ssim(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest, - double *weight) { - double a, b, c; - double ssimv; - - a = vp9_ssim2(source->y_buffer, dest->y_buffer, - source->y_stride, dest->y_stride, - source->y_crop_width, source->y_crop_height); - - b = vp9_ssim2(source->u_buffer, dest->u_buffer, - source->uv_stride, dest->uv_stride, - source->uv_crop_width, source->uv_crop_height); - - c = vp9_ssim2(source->v_buffer, dest->v_buffer, - source->uv_stride, dest->uv_stride, - source->uv_crop_width, source->uv_crop_height); - - ssimv = a * .8 + .1 * (b + c); - - *weight = 1; - - return ssimv; -} - -double vp9_calc_ssimg(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest, - double *ssim_y, double *ssim_u, double *ssim_v) { - double ssim_all = 0; - double a, b, c; - - a = vp9_ssim2(source->y_buffer, dest->y_buffer, - source->y_stride, dest->y_stride, - source->y_crop_width, source->y_crop_height); - - b = vp9_ssim2(source->u_buffer, dest->u_buffer, - source->uv_stride, dest->uv_stride, - source->uv_crop_width, source->uv_crop_height); - - c = vp9_ssim2(source->v_buffer, dest->v_buffer, - source->uv_stride, dest->uv_stride, - source->uv_crop_width, source->uv_crop_height); - *ssim_y = a; - *ssim_u = b; - *ssim_v = c; - ssim_all = (a * 4 + b + c) / 6; - - return ssim_all; -} - -// traditional ssim as per: http://en.wikipedia.org/wiki/Structural_similarity -// -// Re working out the math -> -// -// ssim(x,y) = (2*mean(x)*mean(y) + c1)*(2*cov(x,y)+c2) / -// ((mean(x)^2+mean(y)^2+c1)*(var(x)+var(y)+c2)) -// -// mean(x) = sum(x) / n -// -// cov(x,y) = (n*sum(xi*yi)-sum(x)*sum(y))/(n*n) -// -// var(x) = (n*sum(xi*xi)-sum(xi)*sum(xi))/(n*n) -// -// ssim(x,y) = -// (2*sum(x)*sum(y)/(n*n) + c1)*(2*(n*sum(xi*yi)-sum(x)*sum(y))/(n*n)+c2) / -// (((sum(x)*sum(x)+sum(y)*sum(y))/(n*n) +c1) * -// ((n*sum(xi*xi) - sum(xi)*sum(xi))/(n*n)+ -// (n*sum(yi*yi) - sum(yi)*sum(yi))/(n*n)+c2))) -// -// factoring out n*n -// -// ssim(x,y) = -// (2*sum(x)*sum(y) + n*n*c1)*(2*(n*sum(xi*yi)-sum(x)*sum(y))+n*n*c2) / -// (((sum(x)*sum(x)+sum(y)*sum(y)) + n*n*c1) * -// (n*sum(xi*xi)-sum(xi)*sum(xi)+n*sum(yi*yi)-sum(yi)*sum(yi)+n*n*c2)) -// -// Replace c1 with n*n * c1 for the final step that leads to this code: -// The final step scales by 12 bits so we don't lose precision in the constants. - -double ssimv_similarity(Ssimv *sv, int64_t n) { - // Scale the constants by number of pixels. - const int64_t c1 = (cc1 * n * n) >> 12; - const int64_t c2 = (cc2 * n * n) >> 12; - - const double l = 1.0 * (2 * sv->sum_s * sv->sum_r + c1) / - (sv->sum_s * sv->sum_s + sv->sum_r * sv->sum_r + c1); - - // Since these variables are unsigned sums, convert to double so - // math is done in double arithmetic. - const double v = (2.0 * n * sv->sum_sxr - 2 * sv->sum_s * sv->sum_r + c2) - / (n * sv->sum_sq_s - sv->sum_s * sv->sum_s + n * sv->sum_sq_r - - sv->sum_r * sv->sum_r + c2); - - return l * v; -} - -// The first term of the ssim metric is a luminance factor. -// -// (2*mean(x)*mean(y) + c1)/ (mean(x)^2+mean(y)^2+c1) -// -// This luminance factor is super sensitive to the dark side of luminance -// values and completely insensitive on the white side. check out 2 sets -// (1,3) and (250,252) the term gives ( 2*1*3/(1+9) = .60 -// 2*250*252/ (250^2+252^2) => .99999997 -// -// As a result in this tweaked version of the calculation in which the -// luminance is taken as percentage off from peak possible. -// -// 255 * 255 - (sum_s - sum_r) / count * (sum_s - sum_r) / count -// -double ssimv_similarity2(Ssimv *sv, int64_t n) { - // Scale the constants by number of pixels. - const int64_t c1 = (cc1 * n * n) >> 12; - const int64_t c2 = (cc2 * n * n) >> 12; - - const double mean_diff = (1.0 * sv->sum_s - sv->sum_r) / n; - const double l = (255 * 255 - mean_diff * mean_diff + c1) / (255 * 255 + c1); - - // Since these variables are unsigned, sums convert to double so - // math is done in double arithmetic. - const double v = (2.0 * n * sv->sum_sxr - 2 * sv->sum_s * sv->sum_r + c2) - / (n * sv->sum_sq_s - sv->sum_s * sv->sum_s + - n * sv->sum_sq_r - sv->sum_r * sv->sum_r + c2); - - return l * v; -} -void ssimv_parms(uint8_t *img1, int img1_pitch, uint8_t *img2, int img2_pitch, - Ssimv *sv) { - vp9_ssim_parms_8x8(img1, img1_pitch, img2, img2_pitch, - &sv->sum_s, &sv->sum_r, &sv->sum_sq_s, &sv->sum_sq_r, - &sv->sum_sxr); -} - -double vp9_get_ssim_metrics(uint8_t *img1, int img1_pitch, - uint8_t *img2, int img2_pitch, - int width, int height, - Ssimv *sv2, Metrics *m, - int do_inconsistency) { - double dssim_total = 0; - double ssim_total = 0; - double ssim2_total = 0; - double inconsistency_total = 0; - int i, j; - int c = 0; - double norm; - double old_ssim_total = 0; - vp9_clear_system_state(); - // We can sample points as frequently as we like start with 1 per 4x4. - for (i = 0; i < height; i += 4, - img1 += img1_pitch * 4, img2 += img2_pitch * 4) { - for (j = 0; j < width; j += 4, ++c) { - Ssimv sv = {0}; - double ssim; - double ssim2; - double dssim; - uint32_t var_new; - uint32_t var_old; - uint32_t mean_new; - uint32_t mean_old; - double ssim_new; - double ssim_old; - - // Not sure there's a great way to handle the edge pixels - // in ssim when using a window. Seems biased against edge pixels - // however you handle this. This uses only samples that are - // fully in the frame. - if (j + 8 <= width && i + 8 <= height) { - ssimv_parms(img1 + j, img1_pitch, img2 + j, img2_pitch, &sv); - } - - ssim = ssimv_similarity(&sv, 64); - ssim2 = ssimv_similarity2(&sv, 64); - - sv.ssim = ssim2; - - // dssim is calculated to use as an actual error metric and - // is scaled up to the same range as sum square error. - // Since we are subsampling every 16th point maybe this should be - // *16 ? - dssim = 255 * 255 * (1 - ssim2) / 2; - - // Here I introduce a new error metric: consistency-weighted - // SSIM-inconsistency. This metric isolates frames where the - // SSIM 'suddenly' changes, e.g. if one frame in every 8 is much - // sharper or blurrier than the others. Higher values indicate a - // temporally inconsistent SSIM. There are two ideas at work: - // - // 1) 'SSIM-inconsistency': the total inconsistency value - // reflects how much SSIM values are changing between this - // source / reference frame pair and the previous pair. - // - // 2) 'consistency-weighted': weights de-emphasize areas in the - // frame where the scene content has changed. Changes in scene - // content are detected via changes in local variance and local - // mean. - // - // Thus the overall measure reflects how inconsistent the SSIM - // values are, over consistent regions of the frame. - // - // The metric has three terms: - // - // term 1 -> uses change in scene Variance to weight error score - // 2 * var(Fi)*var(Fi-1) / (var(Fi)^2+var(Fi-1)^2) - // larger changes from one frame to the next mean we care - // less about consistency. - // - // term 2 -> uses change in local scene luminance to weight error - // 2 * avg(Fi)*avg(Fi-1) / (avg(Fi)^2+avg(Fi-1)^2) - // larger changes from one frame to the next mean we care - // less about consistency. - // - // term3 -> measures inconsistency in ssim scores between frames - // 1 - ( 2 * ssim(Fi)*ssim(Fi-1)/(ssim(Fi)^2+sssim(Fi-1)^2). - // - // This term compares the ssim score for the same location in 2 - // subsequent frames. - var_new = sv.sum_sq_s - sv.sum_s * sv.sum_s / 64; - var_old = sv2[c].sum_sq_s - sv2[c].sum_s * sv2[c].sum_s / 64; - mean_new = sv.sum_s; - mean_old = sv2[c].sum_s; - ssim_new = sv.ssim; - ssim_old = sv2[c].ssim; - - if (do_inconsistency) { - // We do the metric once for every 4x4 block in the image. Since - // we are scaling the error to SSE for use in a psnr calculation - // 1.0 = 4x4x255x255 the worst error we can possibly have. - static const double kScaling = 4. * 4 * 255 * 255; - - // The constants have to be non 0 to avoid potential divide by 0 - // issues other than that they affect kind of a weighting between - // the terms. No testing of what the right terms should be has been - // done. - static const double c1 = 1, c2 = 1, c3 = 1; - - // This measures how much consistent variance is in two consecutive - // source frames. 1.0 means they have exactly the same variance. - const double variance_term = (2.0 * var_old * var_new + c1) / - (1.0 * var_old * var_old + 1.0 * var_new * var_new + c1); - - // This measures how consistent the local mean are between two - // consecutive frames. 1.0 means they have exactly the same mean. - const double mean_term = (2.0 * mean_old * mean_new + c2) / - (1.0 * mean_old * mean_old + 1.0 * mean_new * mean_new + c2); - - // This measures how consistent the ssims of two - // consecutive frames is. 1.0 means they are exactly the same. - double ssim_term = pow((2.0 * ssim_old * ssim_new + c3) / - (ssim_old * ssim_old + ssim_new * ssim_new + c3), - 5); - - double this_inconsistency; - - // Floating point math sometimes makes this > 1 by a tiny bit. - // We want the metric to scale between 0 and 1.0 so we can convert - // it to an snr scaled value. - if (ssim_term > 1) - ssim_term = 1; - - // This converts the consistency metric to an inconsistency metric - // ( so we can scale it like psnr to something like sum square error. - // The reason for the variance and mean terms is the assumption that - // if there are big changes in the source we shouldn't penalize - // inconsistency in ssim scores a bit less as it will be less visible - // to the user. - this_inconsistency = (1 - ssim_term) * variance_term * mean_term; - - this_inconsistency *= kScaling; - inconsistency_total += this_inconsistency; - } - sv2[c] = sv; - ssim_total += ssim; - ssim2_total += ssim2; - dssim_total += dssim; - - old_ssim_total += ssim_old; - } - old_ssim_total += 0; - } - - norm = 1. / (width / 4) / (height / 4); - ssim_total *= norm; - ssim2_total *= norm; - m->ssim2 = ssim2_total; - m->ssim = ssim_total; - if (old_ssim_total == 0) - inconsistency_total = 0; - - m->ssimc = inconsistency_total; - - m->dssim = dssim_total; - return inconsistency_total; -} - - -#if CONFIG_VP9_HIGHBITDEPTH -double vp9_highbd_calc_ssim(YV12_BUFFER_CONFIG *source, - YV12_BUFFER_CONFIG *dest, - double *weight, unsigned int bd) { - double a, b, c; - double ssimv; - - a = vp9_highbd_ssim2(source->y_buffer, dest->y_buffer, - source->y_stride, dest->y_stride, - source->y_crop_width, source->y_crop_height, bd); - - b = vp9_highbd_ssim2(source->u_buffer, dest->u_buffer, - source->uv_stride, dest->uv_stride, - source->uv_crop_width, source->uv_crop_height, bd); - - c = vp9_highbd_ssim2(source->v_buffer, dest->v_buffer, - source->uv_stride, dest->uv_stride, - source->uv_crop_width, source->uv_crop_height, bd); - - ssimv = a * .8 + .1 * (b + c); - - *weight = 1; - - return ssimv; -} - -double vp9_highbd_calc_ssimg(YV12_BUFFER_CONFIG *source, - YV12_BUFFER_CONFIG *dest, double *ssim_y, - double *ssim_u, double *ssim_v, unsigned int bd) { - double ssim_all = 0; - double a, b, c; - - a = vp9_highbd_ssim2(source->y_buffer, dest->y_buffer, - source->y_stride, dest->y_stride, - source->y_crop_width, source->y_crop_height, bd); - - b = vp9_highbd_ssim2(source->u_buffer, dest->u_buffer, - source->uv_stride, dest->uv_stride, - source->uv_crop_width, source->uv_crop_height, bd); - - c = vp9_highbd_ssim2(source->v_buffer, dest->v_buffer, - source->uv_stride, dest->uv_stride, - source->uv_crop_width, source->uv_crop_height, bd); - *ssim_y = a; - *ssim_u = b; - *ssim_v = c; - ssim_all = (a * 4 + b + c) / 6; - - return ssim_all; -} -#endif // CONFIG_VP9_HIGHBITDEPTH diff --git a/vp9/encoder/vp9_ssim.h b/vp9/encoder/vp9_ssim.h deleted file mode 100644 index 10f14c4d2..000000000 --- a/vp9/encoder/vp9_ssim.h +++ /dev/null @@ -1,96 +0,0 @@ -/* - * Copyright (c) 2014 The WebM project authors. All Rights Reserved. - * - * Use of this source code is governed by a BSD-style license - * that can be found in the LICENSE file in the root of the source - * tree. An additional intellectual property rights grant can be found - * in the file PATENTS. All contributing project authors may - * be found in the AUTHORS file in the root of the source tree. - */ - -#ifndef VP9_ENCODER_VP9_SSIM_H_ -#define VP9_ENCODER_VP9_SSIM_H_ - -#ifdef __cplusplus -extern "C" { -#endif - -#include "vpx_scale/yv12config.h" - -// metrics used for calculating ssim, ssim2, dssim, and ssimc -typedef struct { - // source sum ( over 8x8 region ) - uint64_t sum_s; - - // reference sum (over 8x8 region ) - uint64_t sum_r; - - // source sum squared ( over 8x8 region ) - uint64_t sum_sq_s; - - // reference sum squared (over 8x8 region ) - uint64_t sum_sq_r; - - // sum of source times reference (over 8x8 region) - uint64_t sum_sxr; - - // calculated ssim score between source and reference - double ssim; -} Ssimv; - -// metrics collected on a frame basis -typedef struct { - // ssim consistency error metric ( see code for explanation ) - double ssimc; - - // standard ssim - double ssim; - - // revised ssim ( see code for explanation) - double ssim2; - - // ssim restated as an error metric like sse - double dssim; - - // dssim converted to decibels - double dssimd; - - // ssimc converted to decibels - double ssimcd; -} Metrics; - -double vp9_get_ssim_metrics(uint8_t *img1, int img1_pitch, uint8_t *img2, - int img2_pitch, int width, int height, Ssimv *sv2, - Metrics *m, int do_inconsistency); - -double vp9_calc_ssim(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest, - double *weight); - -double vp9_calc_ssimg(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest, - double *ssim_y, double *ssim_u, double *ssim_v); - -double vp9_calc_fastssim(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest, - double *ssim_y, double *ssim_u, double *ssim_v); - -double vp9_psnrhvs(YV12_BUFFER_CONFIG *source, YV12_BUFFER_CONFIG *dest, - double *ssim_y, double *ssim_u, double *ssim_v); - -#if CONFIG_VP9_HIGHBITDEPTH -double vp9_highbd_calc_ssim(YV12_BUFFER_CONFIG *source, - YV12_BUFFER_CONFIG *dest, - double *weight, - unsigned int bd); - -double vp9_highbd_calc_ssimg(YV12_BUFFER_CONFIG *source, - YV12_BUFFER_CONFIG *dest, - double *ssim_y, - double *ssim_u, - double *ssim_v, - unsigned int bd); -#endif // CONFIG_VP9_HIGHBITDEPTH - -#ifdef __cplusplus -} // extern "C" -#endif - -#endif // VP9_ENCODER_VP9_SSIM_H_ diff --git a/vp9/encoder/vp9_temporal_filter.c b/vp9/encoder/vp9_temporal_filter.c index 3475d58da..c4d66e021 100644 --- a/vp9/encoder/vp9_temporal_filter.c +++ b/vp9/encoder/vp9_temporal_filter.c @@ -15,7 +15,6 @@ #include "vp9/common/vp9_onyxc_int.h" #include "vp9/common/vp9_quant_common.h" #include "vp9/common/vp9_reconinter.h" -#include "vp9/common/vp9_systemdependent.h" #include "vp9/encoder/vp9_extend.h" #include "vp9/encoder/vp9_firstpass.h" #include "vp9/encoder/vp9_mcomp.h" diff --git a/vp9/encoder/x86/vp9_dct_mmx.asm b/vp9/encoder/x86/vp9_dct_mmx.asm index b41fbc8b3..7a7a6b655 100644 --- a/vp9/encoder/x86/vp9_dct_mmx.asm +++ b/vp9/encoder/x86/vp9_dct_mmx.asm @@ -7,6 +7,9 @@ ; in the file PATENTS. All contributing project authors may ; be found in the AUTHORS file in the root of the source tree. ; + +%define private_prefix vp9 + %include "third_party/x86inc/x86inc.asm" SECTION .text diff --git a/vp9/encoder/x86/vp9_dct_ssse3_x86_64.asm b/vp9/encoder/x86/vp9_dct_ssse3_x86_64.asm index 29074e713..74c52df19 100644 --- a/vp9/encoder/x86/vp9_dct_ssse3_x86_64.asm +++ b/vp9/encoder/x86/vp9_dct_ssse3_x86_64.asm @@ -7,6 +7,9 @@ ; in the file PATENTS. All contributing project authors may ; be found in the AUTHORS file in the root of the source tree. ; + +%define private_prefix vp9 + %include "third_party/x86inc/x86inc.asm" ; This file provides SSSE3 version of the forward transformation. Part diff --git a/vp9/encoder/x86/vp9_error_sse2.asm b/vp9/encoder/x86/vp9_error_sse2.asm index 56373e897..5b0238272 100644 --- a/vp9/encoder/x86/vp9_error_sse2.asm +++ b/vp9/encoder/x86/vp9_error_sse2.asm @@ -8,6 +8,8 @@ ; be found in the AUTHORS file in the root of the source tree. ; +%define private_prefix vp9 + %include "third_party/x86inc/x86inc.asm" SECTION .text diff --git a/vp9/encoder/x86/vp9_quantize_ssse3_x86_64.asm b/vp9/encoder/x86/vp9_quantize_ssse3_x86_64.asm index ec2e87cb1..ec61c0c3a 100644 --- a/vp9/encoder/x86/vp9_quantize_ssse3_x86_64.asm +++ b/vp9/encoder/x86/vp9_quantize_ssse3_x86_64.asm @@ -8,6 +8,8 @@ ; be found in the AUTHORS file in the root of the source tree. ; +%define private_prefix vp9 + %include "third_party/x86inc/x86inc.asm" SECTION_RODATA diff --git a/vp9/encoder/x86/vp9_ssim_opt_x86_64.asm b/vp9/encoder/x86/vp9_ssim_opt_x86_64.asm deleted file mode 100644 index 455d10d2c..000000000 --- a/vp9/encoder/x86/vp9_ssim_opt_x86_64.asm +++ /dev/null @@ -1,216 +0,0 @@ -; -; Copyright (c) 2010 The WebM project authors. All Rights Reserved. -; -; Use of this source code is governed by a BSD-style license -; that can be found in the LICENSE file in the root of the source -; tree. An additional intellectual property rights grant can be found -; in the file PATENTS. All contributing project authors may -; be found in the AUTHORS file in the root of the source tree. -; - -%include "vpx_ports/x86_abi_support.asm" - -; tabulate_ssim - sums sum_s,sum_r,sum_sq_s,sum_sq_r, sum_sxr -%macro TABULATE_SSIM 0 - paddusw xmm15, xmm3 ; sum_s - paddusw xmm14, xmm4 ; sum_r - movdqa xmm1, xmm3 - pmaddwd xmm1, xmm1 - paddd xmm13, xmm1 ; sum_sq_s - movdqa xmm2, xmm4 - pmaddwd xmm2, xmm2 - paddd xmm12, xmm2 ; sum_sq_r - pmaddwd xmm3, xmm4 - paddd xmm11, xmm3 ; sum_sxr -%endmacro - -; Sum across the register %1 starting with q words -%macro SUM_ACROSS_Q 1 - movdqa xmm2,%1 - punpckldq %1,xmm0 - punpckhdq xmm2,xmm0 - paddq %1,xmm2 - movdqa xmm2,%1 - punpcklqdq %1,xmm0 - punpckhqdq xmm2,xmm0 - paddq %1,xmm2 -%endmacro - -; Sum across the register %1 starting with q words -%macro SUM_ACROSS_W 1 - movdqa xmm1, %1 - punpcklwd %1,xmm0 - punpckhwd xmm1,xmm0 - paddd %1, xmm1 - SUM_ACROSS_Q %1 -%endmacro -;void ssim_parms_sse2( -; unsigned char *s, -; int sp, -; unsigned char *r, -; int rp -; unsigned long *sum_s, -; unsigned long *sum_r, -; unsigned long *sum_sq_s, -; unsigned long *sum_sq_r, -; unsigned long *sum_sxr); -; -; TODO: Use parm passing through structure, probably don't need the pxors -; ( calling app will initialize to 0 ) could easily fit everything in sse2 -; without too much hastle, and can probably do better estimates with psadw -; or pavgb At this point this is just meant to be first pass for calculating -; all the parms needed for 16x16 ssim so we can play with dssim as distortion -; in mode selection code. -global sym(vp9_ssim_parms_16x16_sse2) PRIVATE -sym(vp9_ssim_parms_16x16_sse2): - push rbp - mov rbp, rsp - SHADOW_ARGS_TO_STACK 9 - SAVE_XMM 15 - push rsi - push rdi - ; end prolog - - mov rsi, arg(0) ;s - mov rcx, arg(1) ;sp - mov rdi, arg(2) ;r - mov rax, arg(3) ;rp - - pxor xmm0, xmm0 - pxor xmm15,xmm15 ;sum_s - pxor xmm14,xmm14 ;sum_r - pxor xmm13,xmm13 ;sum_sq_s - pxor xmm12,xmm12 ;sum_sq_r - pxor xmm11,xmm11 ;sum_sxr - - mov rdx, 16 ;row counter -.NextRow: - - ;grab source and reference pixels - movdqu xmm5, [rsi] - movdqu xmm6, [rdi] - movdqa xmm3, xmm5 - movdqa xmm4, xmm6 - punpckhbw xmm3, xmm0 ; high_s - punpckhbw xmm4, xmm0 ; high_r - - TABULATE_SSIM - - movdqa xmm3, xmm5 - movdqa xmm4, xmm6 - punpcklbw xmm3, xmm0 ; low_s - punpcklbw xmm4, xmm0 ; low_r - - TABULATE_SSIM - - add rsi, rcx ; next s row - add rdi, rax ; next r row - - dec rdx ; counter - jnz .NextRow - - SUM_ACROSS_W xmm15 - SUM_ACROSS_W xmm14 - SUM_ACROSS_Q xmm13 - SUM_ACROSS_Q xmm12 - SUM_ACROSS_Q xmm11 - - mov rdi,arg(4) - movd [rdi], xmm15; - mov rdi,arg(5) - movd [rdi], xmm14; - mov rdi,arg(6) - movd [rdi], xmm13; - mov rdi,arg(7) - movd [rdi], xmm12; - mov rdi,arg(8) - movd [rdi], xmm11; - - ; begin epilog - pop rdi - pop rsi - RESTORE_XMM - UNSHADOW_ARGS - pop rbp - ret - -;void ssim_parms_sse2( -; unsigned char *s, -; int sp, -; unsigned char *r, -; int rp -; unsigned long *sum_s, -; unsigned long *sum_r, -; unsigned long *sum_sq_s, -; unsigned long *sum_sq_r, -; unsigned long *sum_sxr); -; -; TODO: Use parm passing through structure, probably don't need the pxors -; ( calling app will initialize to 0 ) could easily fit everything in sse2 -; without too much hastle, and can probably do better estimates with psadw -; or pavgb At this point this is just meant to be first pass for calculating -; all the parms needed for 16x16 ssim so we can play with dssim as distortion -; in mode selection code. -global sym(vp9_ssim_parms_8x8_sse2) PRIVATE -sym(vp9_ssim_parms_8x8_sse2): - push rbp - mov rbp, rsp - SHADOW_ARGS_TO_STACK 9 - SAVE_XMM 15 - push rsi - push rdi - ; end prolog - - mov rsi, arg(0) ;s - mov rcx, arg(1) ;sp - mov rdi, arg(2) ;r - mov rax, arg(3) ;rp - - pxor xmm0, xmm0 - pxor xmm15,xmm15 ;sum_s - pxor xmm14,xmm14 ;sum_r - pxor xmm13,xmm13 ;sum_sq_s - pxor xmm12,xmm12 ;sum_sq_r - pxor xmm11,xmm11 ;sum_sxr - - mov rdx, 8 ;row counter -.NextRow: - - ;grab source and reference pixels - movq xmm3, [rsi] - movq xmm4, [rdi] - punpcklbw xmm3, xmm0 ; low_s - punpcklbw xmm4, xmm0 ; low_r - - TABULATE_SSIM - - add rsi, rcx ; next s row - add rdi, rax ; next r row - - dec rdx ; counter - jnz .NextRow - - SUM_ACROSS_W xmm15 - SUM_ACROSS_W xmm14 - SUM_ACROSS_Q xmm13 - SUM_ACROSS_Q xmm12 - SUM_ACROSS_Q xmm11 - - mov rdi,arg(4) - movd [rdi], xmm15; - mov rdi,arg(5) - movd [rdi], xmm14; - mov rdi,arg(6) - movd [rdi], xmm13; - mov rdi,arg(7) - movd [rdi], xmm12; - mov rdi,arg(8) - movd [rdi], xmm11; - - ; begin epilog - pop rdi - pop rsi - RESTORE_XMM - UNSHADOW_ARGS - pop rbp - ret diff --git a/vp9/vp9_common.mk b/vp9/vp9_common.mk index e95da4404..d0135c6f8 100644 --- a/vp9/vp9_common.mk +++ b/vp9/vp9_common.mk @@ -45,7 +45,6 @@ VP9_COMMON_SRCS-yes += common/vp9_scale.h VP9_COMMON_SRCS-yes += common/vp9_scale.c VP9_COMMON_SRCS-yes += common/vp9_seg_common.h VP9_COMMON_SRCS-yes += common/vp9_seg_common.c -VP9_COMMON_SRCS-yes += common/vp9_systemdependent.h VP9_COMMON_SRCS-yes += common/vp9_textblit.h VP9_COMMON_SRCS-yes += common/vp9_tile_common.h VP9_COMMON_SRCS-yes += common/vp9_tile_common.c @@ -71,16 +70,10 @@ VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_mfqe_sse2.asm VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_postproc_sse2.asm endif -# common (c) -VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_common_dspr2.h -VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_common_dspr2.c - ifneq ($(CONFIG_VP9_HIGHBITDEPTH),yes) VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_itrans4_dspr2.c VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_itrans8_dspr2.c VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_itrans16_dspr2.c -VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_itrans32_cols_dspr2.c -VP9_COMMON_SRCS-$(HAVE_DSPR2) += common/mips/dspr2/vp9_itrans32_dspr2.c endif # common (msa) @@ -94,8 +87,6 @@ endif VP9_COMMON_SRCS-$(HAVE_SSE2) += common/x86/vp9_idct_intrin_sse2.c -VP9_COMMON_SRCS-$(HAVE_NEON_ASM) += common/arm/neon/vp9_save_reg_neon$(ASM) - ifneq ($(CONFIG_VP9_HIGHBITDEPTH),yes) VP9_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/vp9_iht4x4_add_neon.c VP9_COMMON_SRCS-$(HAVE_NEON) += common/arm/neon/vp9_iht8x8_add_neon.c diff --git a/vp9/vp9cx.mk b/vp9/vp9cx.mk index 186ce11f0..84b12d78e 100644 --- a/vp9/vp9cx.mk +++ b/vp9/vp9cx.mk @@ -33,7 +33,6 @@ VP9_CX_SRCS-yes += encoder/vp9_encodemv.c VP9_CX_SRCS-yes += encoder/vp9_ethread.h VP9_CX_SRCS-yes += encoder/vp9_ethread.c VP9_CX_SRCS-yes += encoder/vp9_extend.c -VP9_CX_SRCS-$(CONFIG_INTERNAL_STATS) += encoder/vp9_fastssim.c VP9_CX_SRCS-yes += encoder/vp9_firstpass.c VP9_CX_SRCS-yes += encoder/vp9_block.h VP9_CX_SRCS-yes += encoder/vp9_bitstream.h @@ -57,7 +56,6 @@ VP9_CX_SRCS-yes += encoder/vp9_mcomp.c VP9_CX_SRCS-yes += encoder/vp9_encoder.c VP9_CX_SRCS-yes += encoder/vp9_picklpf.c VP9_CX_SRCS-yes += encoder/vp9_picklpf.h -VP9_CX_SRCS-$(CONFIG_INTERNAL_STATS) += encoder/vp9_psnrhvs.c VP9_CX_SRCS-yes += encoder/vp9_quantize.c VP9_CX_SRCS-yes += encoder/vp9_ratectrl.c VP9_CX_SRCS-yes += encoder/vp9_rd.c @@ -72,8 +70,6 @@ VP9_CX_SRCS-yes += encoder/vp9_subexp.h VP9_CX_SRCS-yes += encoder/vp9_svc_layercontext.c VP9_CX_SRCS-yes += encoder/vp9_resize.c VP9_CX_SRCS-yes += encoder/vp9_resize.h -VP9_CX_SRCS-$(CONFIG_INTERNAL_STATS) += encoder/vp9_ssim.c -VP9_CX_SRCS-$(CONFIG_INTERNAL_STATS) += encoder/vp9_ssim.h VP9_CX_SRCS-$(CONFIG_INTERNAL_STATS) += encoder/vp9_blockiness.c VP9_CX_SRCS-yes += encoder/vp9_tokenize.c @@ -113,7 +109,6 @@ VP9_CX_SRCS-$(HAVE_SSSE3) += encoder/x86/vp9_quantize_ssse3_x86_64.asm VP9_CX_SRCS-$(HAVE_SSSE3) += encoder/x86/vp9_dct_ssse3_x86_64.asm endif endif -VP9_CX_SRCS-$(ARCH_X86_64) += encoder/x86/vp9_ssim_opt_x86_64.asm VP9_CX_SRCS-$(HAVE_SSE2) += encoder/x86/vp9_dct_sse2.c VP9_CX_SRCS-$(HAVE_SSSE3) += encoder/x86/vp9_dct_ssse3.c |