diff options
author | Angie Chiang <angiebird@google.com> | 2015-09-04 14:51:54 -0700 |
---|---|---|
committer | Angie Chiang <angiebird@google.com> | 2015-09-10 15:08:37 -0700 |
commit | 87175ed592a5eea774842615b914bf1e0798e57e (patch) | |
tree | f54c732172f662efe5bad6dcc26dacf4a86ebca7 /vp10/common/vp10_inv_txfm.h | |
parent | 9511b948ef5ee7f318f15f3744497640bdbcc69e (diff) | |
download | libvpx-87175ed592a5eea774842615b914bf1e0798e57e.tar libvpx-87175ed592a5eea774842615b914bf1e0798e57e.tar.gz libvpx-87175ed592a5eea774842615b914bf1e0798e57e.tar.bz2 libvpx-87175ed592a5eea774842615b914bf1e0798e57e.zip |
Isolate vp10's inv_txfm from vp9
1) copy following files from vpx_dsp/ to vp10/common/
vp10_inv_txfm.c
vp10_inv_txfm.h
vp10_inv_txfm_sse2.c
vp10_inv_txfm_sse2.h
2) change the function prefix "vpx_" to "vp10_" in above files
3) add unit test at vp10_inv_txfm_test.cc
Change-Id: I206f10f60c8b27d872c84b7482c3bb1d1cb4b913
Diffstat (limited to 'vp10/common/vp10_inv_txfm.h')
-rw-r--r-- | vp10/common/vp10_inv_txfm.h | 122 |
1 files changed, 122 insertions, 0 deletions
diff --git a/vp10/common/vp10_inv_txfm.h b/vp10/common/vp10_inv_txfm.h new file mode 100644 index 000000000..52611acbd --- /dev/null +++ b/vp10/common/vp10_inv_txfm.h @@ -0,0 +1,122 @@ +/* + * 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 VPX_DSP_INV_TXFM_H_ +#define VPX_DSP_INV_TXFM_H_ + +#include <assert.h> + +#include "./vpx_config.h" +#include "vpx_dsp/txfm_common.h" +#include "vpx_ports/mem.h" + +#ifdef __cplusplus +extern "C" { +#endif + +static INLINE tran_low_t check_range(tran_high_t input) { +#if CONFIG_COEFFICIENT_RANGE_CHECKING + // For valid VP9 input streams, intermediate stage coefficients should always + // stay within the range of a signed 16 bit integer. Coefficients can go out + // of this range for invalid/corrupt VP9 streams. However, strictly checking + // this range for every intermediate coefficient can burdensome for a decoder, + // therefore the following assertion is only enabled when configured with + // --enable-coefficient-range-checking. + assert(INT16_MIN <= input); + assert(input <= INT16_MAX); +#endif // CONFIG_COEFFICIENT_RANGE_CHECKING + return (tran_low_t)input; +} + +static INLINE tran_low_t dct_const_round_shift(tran_high_t input) { + tran_high_t rv = ROUND_POWER_OF_TWO(input, DCT_CONST_BITS); + return check_range(rv); +} + +#if CONFIG_VP9_HIGHBITDEPTH +static INLINE tran_low_t highbd_check_range(tran_high_t input, + int bd) { +#if CONFIG_COEFFICIENT_RANGE_CHECKING + // For valid highbitdepth VP9 streams, intermediate stage coefficients will + // stay within the ranges: + // - 8 bit: signed 16 bit integer + // - 10 bit: signed 18 bit integer + // - 12 bit: signed 20 bit integer + const int32_t int_max = (1 << (7 + bd)) - 1; + const int32_t int_min = -int_max - 1; + assert(int_min <= input); + assert(input <= int_max); + (void) int_min; +#endif // CONFIG_COEFFICIENT_RANGE_CHECKING + (void) bd; + return (tran_low_t)input; +} + +static INLINE tran_low_t highbd_dct_const_round_shift(tran_high_t input, + int bd) { + tran_high_t rv = ROUND_POWER_OF_TWO(input, DCT_CONST_BITS); + return highbd_check_range(rv, bd); +} +#endif // CONFIG_VP9_HIGHBITDEPTH + +#if CONFIG_EMULATE_HARDWARE +// When CONFIG_EMULATE_HARDWARE is 1 the transform performs a +// non-normative method to handle overflows. A stream that causes +// overflows in the inverse transform is considered invalid in VP9, +// and a hardware implementer is free to choose any reasonable +// method to handle overflows. However to aid in hardware +// verification they can use a specific implementation of the +// WRAPLOW() macro below that is identical to their intended +// hardware implementation (and also use configure options to trigger +// the C-implementation of the transform). +// +// The particular WRAPLOW implementation below performs strict +// overflow wrapping to match common hardware implementations. +// bd of 8 uses trans_low with 16bits, need to remove 16bits +// bd of 10 uses trans_low with 18bits, need to remove 14bits +// bd of 12 uses trans_low with 20bits, need to remove 12bits +// bd of x uses trans_low with 8+x bits, need to remove 24-x bits +#define WRAPLOW(x, bd) ((((int32_t)(x)) << (24 - bd)) >> (24 - bd)) +#else +#define WRAPLOW(x, bd) ((int32_t)(x)) +#endif // CONFIG_EMULATE_HARDWARE + +void vp10_idct4_c(const tran_low_t *input, tran_low_t *output); +void vp10_idct8_c(const tran_low_t *input, tran_low_t *output); +void vp10_idct16_c(const tran_low_t *input, tran_low_t *output); +void vp10_idct32_c(const tran_low_t *input, tran_low_t *output); +void vp10_iadst4_c(const tran_low_t *input, tran_low_t *output); +void vp10_iadst8_c(const tran_low_t *input, tran_low_t *output); +void vp10_iadst16_c(const tran_low_t *input, tran_low_t *output); + +#if CONFIG_VP9_HIGHBITDEPTH +void vp10_highbd_idct4_c(const tran_low_t *input, tran_low_t *output, int bd); +void vp10_highbd_idct8_c(const tran_low_t *input, tran_low_t *output, int bd); +void vp10_highbd_idct16_c(const tran_low_t *input, tran_low_t *output, int bd); + +void vp10_highbd_iadst4_c(const tran_low_t *input, tran_low_t *output, int bd); +void vp10_highbd_iadst8_c(const tran_low_t *input, tran_low_t *output, int bd); +void vp10_highbd_iadst16_c(const tran_low_t *input, tran_low_t *output, int bd); + +static INLINE uint16_t highbd_clip_pixel_add(uint16_t dest, tran_high_t trans, + int bd) { + trans = WRAPLOW(trans, bd); + return clip_pixel_highbd(WRAPLOW(dest + trans, bd), bd); +} +#endif + +static INLINE uint8_t clip_pixel_add(uint8_t dest, tran_high_t trans) { + trans = WRAPLOW(trans, 8); + return clip_pixel(WRAPLOW(dest + trans, 8)); +} +#ifdef __cplusplus +} // extern "C" +#endif +#endif // VPX_DSP_INV_TXFM_H_ |