summaryrefslogtreecommitdiff
path: root/vp8
diff options
context:
space:
mode:
Diffstat (limited to 'vp8')
-rw-r--r--vp8/common/arm/arm_systemdependent.c17
-rw-r--r--vp8/common/generic/systemdependent.c6
-rw-r--r--vp8/common/x86/idct_x86.h12
-rw-r--r--vp8/common/x86/idctllm_mmx.asm368
-rw-r--r--vp8/common/x86/x86_systemdependent.c14
-rw-r--r--vp8/decoder/arm/arm_dsystemdependent.c11
-rw-r--r--vp8/decoder/generic/dsystemdependent.c6
-rw-r--r--vp8/decoder/x86/dequantize_mmx.asm598
-rw-r--r--vp8/decoder/x86/dequantize_x86.h6
-rw-r--r--vp8/decoder/x86/idct_blk_mmx.c48
-rw-r--r--vp8/decoder/x86/x86_dsystemdependent.c14
11 files changed, 480 insertions, 620 deletions
diff --git a/vp8/common/arm/arm_systemdependent.c b/vp8/common/arm/arm_systemdependent.c
index 1b6f27970..72f044ab6 100644
--- a/vp8/common/arm/arm_systemdependent.c
+++ b/vp8/common/arm/arm_systemdependent.c
@@ -33,6 +33,7 @@ void vp8_arch_arm_common_init(VP8_COMMON *ctx)
}
#endif
+// The commented functions need to be re-written for vpx.
#if HAVE_ARMV6
if (flags & HAS_MEDIA)
{
@@ -46,10 +47,10 @@ void vp8_arch_arm_common_init(VP8_COMMON *ctx)
rtcd->subpix.bilinear8x4 = vp8_bilinear_predict8x4_armv6;
rtcd->subpix.bilinear4x4 = vp8_bilinear_predict4x4_armv6;
- rtcd->idct.idct1 = vp8_short_idct4x4llm_1_v6;
- rtcd->idct.idct16 = vp8_short_idct4x4llm_v6_dual;
- rtcd->idct.iwalsh1 = vp8_short_inv_walsh4x4_1_v6;
- rtcd->idct.iwalsh16 = vp8_short_inv_walsh4x4_v6;
+ //rtcd->idct.idct1 = vp8_short_idct4x4llm_1_v6;
+ //rtcd->idct.idct16 = vp8_short_idct4x4llm_v6_dual;
+ //rtcd->idct.iwalsh1 = vp8_short_inv_walsh4x4_1_v6;
+ //rtcd->idct.iwalsh16 = vp8_short_inv_walsh4x4_v6;
rtcd->loopfilter.normal_mb_v = vp8_loop_filter_mbv_armv6;
rtcd->loopfilter.normal_b_v = vp8_loop_filter_bv_armv6;
@@ -84,10 +85,10 @@ void vp8_arch_arm_common_init(VP8_COMMON *ctx)
rtcd->subpix.bilinear8x4 = vp8_bilinear_predict8x4_neon;
rtcd->subpix.bilinear4x4 = vp8_bilinear_predict4x4_neon;
- rtcd->idct.idct1 = vp8_short_idct4x4llm_1_neon;
- rtcd->idct.idct16 = vp8_short_idct4x4llm_neon;
- rtcd->idct.iwalsh1 = vp8_short_inv_walsh4x4_1_neon;
- rtcd->idct.iwalsh16 = vp8_short_inv_walsh4x4_neon;
+ //rtcd->idct.idct1 = vp8_short_idct4x4llm_1_neon;
+ //rtcd->idct.idct16 = vp8_short_idct4x4llm_neon;
+ //rtcd->idct.iwalsh1 = vp8_short_inv_walsh4x4_1_neon;
+ //rtcd->idct.iwalsh16 = vp8_short_inv_walsh4x4_neon;
rtcd->loopfilter.normal_mb_v = vp8_loop_filter_mbv_neon;
rtcd->loopfilter.normal_b_v = vp8_loop_filter_bv_neon;
diff --git a/vp8/common/generic/systemdependent.c b/vp8/common/generic/systemdependent.c
index 79622487e..e6ae6b5ed 100644
--- a/vp8/common/generic/systemdependent.c
+++ b/vp8/common/generic/systemdependent.c
@@ -139,10 +139,4 @@ void vp8_machine_specific_config(VP8_COMMON *ctx)
#if ARCH_ARM
vp8_arch_arm_common_init(ctx);
#endif
- rtcd->idct.idct1 = vp8_short_idct4x4llm_1_c;
- rtcd->idct.idct16 = vp8_short_idct4x4llm_c;
- rtcd->idct.idct1_scalar_add = vp8_dc_only_idct_add_c;
- rtcd->idct.iwalsh1 = vp8_short_inv_walsh4x4_1_c;
- rtcd->idct.iwalsh16 = vp8_short_inv_walsh4x4_c;
-
}
diff --git a/vp8/common/x86/idct_x86.h b/vp8/common/x86/idct_x86.h
index f6e568cdc..062149368 100644
--- a/vp8/common/x86/idct_x86.h
+++ b/vp8/common/x86/idct_x86.h
@@ -20,22 +20,22 @@
*/
#if HAVE_MMX
-extern prototype_idct(vp8_short_idct4x4llm_1_mmx);
-extern prototype_idct(vp8_short_idct4x4llm_mmx);
-extern prototype_idct_scalar_add(vp8_dc_only_idct_add_mmx);
+extern prototype_idct(vpx_short_idct4x4llm_1_mmx);
+extern prototype_idct(vpx_short_idct4x4llm_mmx);
+extern prototype_idct_scalar_add(vpx_dc_only_idct_add_mmx);
extern prototype_second_order(vp8_short_inv_walsh4x4_mmx);
extern prototype_second_order(vp8_short_inv_walsh4x4_1_mmx);
#if !CONFIG_RUNTIME_CPU_DETECT
#undef vp8_idct_idct1
-#define vp8_idct_idct1 vp8_short_idct4x4llm_1_mmx
+#define vp8_idct_idct1 vpx_short_idct4x4llm_1_mmx
#undef vp8_idct_idct16
-#define vp8_idct_idct16 vp8_short_idct4x4llm_mmx
+#define vp8_idct_idct16 vpx_short_idct4x4llm_mmx
#undef vp8_idct_idct1_scalar_add
-#define vp8_idct_idct1_scalar_add vp8_dc_only_idct_add_mmx
+#define vp8_idct_idct1_scalar_add vpx_dc_only_idct_add_mmx
#undef vp8_idct_iwalsh16
#define vp8_idct_iwalsh16 vp8_short_inv_walsh4x4_mmx
diff --git a/vp8/common/x86/idctllm_mmx.asm b/vp8/common/x86/idctllm_mmx.asm
index 465626b8f..ab9eedbef 100644
--- a/vp8/common/x86/idctllm_mmx.asm
+++ b/vp8/common/x86/idctllm_mmx.asm
@@ -1,5 +1,5 @@
;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+; Copyright (c) 2012 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
@@ -8,8 +8,18 @@
; be found in the AUTHORS file in the root of the source tree.
;
+%include "vpx_ports/x86inc.asm"
+
+SECTION_RODATA
+align 16
+x_s1sqr2: times 4 dw 0x8A8C
+align 16
+x_c1sqr2less1: times 4 dw 0x4E7B
+align 16
+pw_16: times 4 dw 16
+
+SECTION .text
-%include "vpx_ports/x86_abi_support.asm"
; /****************************************************************************
; * Notes:
@@ -31,262 +41,196 @@
; *
; **************************************************************************/
+INIT_MMX
;void short_idct4x4llm_mmx(short *input, short *output, int pitch)
-global sym(vp8_short_idct4x4llm_mmx)
-sym(vp8_short_idct4x4llm_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 3
- GET_GOT rbx
- ; end prolog
-
- mov rax, arg(0) ;input
- mov rdx, arg(1) ;output
-
- movq mm0, [rax ]
- movq mm1, [rax+ 8]
+cglobal short_idct4x4llm_mmx, 3,3,0, inp, out, pit
+ mova m0, [inpq +0]
+ mova m1, [inpq +8]
- movq mm2, [rax+16]
- movq mm3, [rax+24]
+ mova m2, [inpq+16]
+ mova m3, [inpq+24]
- movsxd rax, dword ptr arg(2) ;pitch
+ psubw m0, m2 ; b1= 0-2
+ paddw m2, m2 ;
- psubw mm0, mm2 ; b1= 0-2
- paddw mm2, mm2 ;
+ mova m5, m1
+ paddw m2, m0 ; a1 =0+2
- movq mm5, mm1
- paddw mm2, mm0 ; a1 =0+2
+ pmulhw m5, [x_s1sqr2] ;
+ paddw m5, m1 ; ip1 * sin(pi/8) * sqrt(2)
- pmulhw mm5, [GLOBAL(x_s1sqr2)] ;
- paddw mm5, mm1 ; ip1 * sin(pi/8) * sqrt(2)
+ mova m7, m3 ;
+ pmulhw m7, [x_c1sqr2less1] ;
- movq mm7, mm3 ;
- pmulhw mm7, [GLOBAL(x_c1sqr2less1)] ;
+ paddw m7, m3 ; ip3 * cos(pi/8) * sqrt(2)
+ psubw m7, m5 ; c1
- paddw mm7, mm3 ; ip3 * cos(pi/8) * sqrt(2)
- psubw mm7, mm5 ; c1
+ mova m5, m1
+ mova m4, m3
- movq mm5, mm1
- movq mm4, mm3
+ pmulhw m5, [x_c1sqr2less1]
+ paddw m5, m1
- pmulhw mm5, [GLOBAL(x_c1sqr2less1)]
- paddw mm5, mm1
+ pmulhw m3, [x_s1sqr2]
+ paddw m3, m4
- pmulhw mm3, [GLOBAL(x_s1sqr2)]
- paddw mm3, mm4
+ paddw m3, m5 ; d1
+ mova m6, m2 ; a1
- paddw mm3, mm5 ; d1
- movq mm6, mm2 ; a1
+ mova m4, m0 ; b1
+ paddw m2, m3 ;0
- movq mm4, mm0 ; b1
- paddw mm2, mm3 ;0
+ paddw m4, m7 ;1
+ psubw m0, m7 ;2
- paddw mm4, mm7 ;1
- psubw mm0, mm7 ;2
+ psubw m6, m3 ;3
- psubw mm6, mm3 ;3
+ mova m1, m2 ; 03 02 01 00
+ mova m3, m4 ; 23 22 21 20
- movq mm1, mm2 ; 03 02 01 00
- movq mm3, mm4 ; 23 22 21 20
+ punpcklwd m1, m0 ; 11 01 10 00
+ punpckhwd m2, m0 ; 13 03 12 02
- punpcklwd mm1, mm0 ; 11 01 10 00
- punpckhwd mm2, mm0 ; 13 03 12 02
+ punpcklwd m3, m6 ; 31 21 30 20
+ punpckhwd m4, m6 ; 33 23 32 22
- punpcklwd mm3, mm6 ; 31 21 30 20
- punpckhwd mm4, mm6 ; 33 23 32 22
+ mova m0, m1 ; 11 01 10 00
+ mova m5, m2 ; 13 03 12 02
- movq mm0, mm1 ; 11 01 10 00
- movq mm5, mm2 ; 13 03 12 02
+ punpckldq m0, m3 ; 30 20 10 00
+ punpckhdq m1, m3 ; 31 21 11 01
- punpckldq mm0, mm3 ; 30 20 10 00
- punpckhdq mm1, mm3 ; 31 21 11 01
+ punpckldq m2, m4 ; 32 22 12 02
+ punpckhdq m5, m4 ; 33 23 13 03
- punpckldq mm2, mm4 ; 32 22 12 02
- punpckhdq mm5, mm4 ; 33 23 13 03
+ mova m3, m5 ; 33 23 13 03
- movq mm3, mm5 ; 33 23 13 03
+ psubw m0, m2 ; b1= 0-2
+ paddw m2, m2 ;
- psubw mm0, mm2 ; b1= 0-2
- paddw mm2, mm2 ;
+ mova m5, m1
+ paddw m2, m0 ; a1 =0+2
- movq mm5, mm1
- paddw mm2, mm0 ; a1 =0+2
+ pmulhw m5, [x_s1sqr2] ;
+ paddw m5, m1 ; ip1 * sin(pi/8) * sqrt(2)
- pmulhw mm5, [GLOBAL(x_s1sqr2)] ;
- paddw mm5, mm1 ; ip1 * sin(pi/8) * sqrt(2)
+ mova m7, m3 ;
+ pmulhw m7, [x_c1sqr2less1] ;
- movq mm7, mm3 ;
- pmulhw mm7, [GLOBAL(x_c1sqr2less1)] ;
+ paddw m7, m3 ; ip3 * cos(pi/8) * sqrt(2)
+ psubw m7, m5 ; c1
- paddw mm7, mm3 ; ip3 * cos(pi/8) * sqrt(2)
- psubw mm7, mm5 ; c1
+ mova m5, m1
+ mova m4, m3
- movq mm5, mm1
- movq mm4, mm3
+ pmulhw m5, [x_c1sqr2less1]
+ paddw m5, m1
- pmulhw mm5, [GLOBAL(x_c1sqr2less1)]
- paddw mm5, mm1
+ pmulhw m3, [x_s1sqr2]
+ paddw m3, m4
- pmulhw mm3, [GLOBAL(x_s1sqr2)]
- paddw mm3, mm4
+ paddw m3, m5 ; d1
+ paddw m0, [pw_16]
- paddw mm3, mm5 ; d1
- paddw mm0, [GLOBAL(fours)]
+ paddw m2, [pw_16]
+ mova m6, m2 ; a1
- paddw mm2, [GLOBAL(fours)]
- movq mm6, mm2 ; a1
+ mova m4, m0 ; b1
+ paddw m2, m3 ;0
- movq mm4, mm0 ; b1
- paddw mm2, mm3 ;0
+ paddw m4, m7 ;1
+ psubw m0, m7 ;2
- paddw mm4, mm7 ;1
- psubw mm0, mm7 ;2
+ psubw m6, m3 ;3
+ psraw m2, 5
- psubw mm6, mm3 ;3
- psraw mm2, 3
+ psraw m0, 5
+ psraw m4, 5
- psraw mm0, 3
- psraw mm4, 3
+ psraw m6, 5
- psraw mm6, 3
+ mova m1, m2 ; 03 02 01 00
+ mova m3, m4 ; 23 22 21 20
- movq mm1, mm2 ; 03 02 01 00
- movq mm3, mm4 ; 23 22 21 20
+ punpcklwd m1, m0 ; 11 01 10 00
+ punpckhwd m2, m0 ; 13 03 12 02
- punpcklwd mm1, mm0 ; 11 01 10 00
- punpckhwd mm2, mm0 ; 13 03 12 02
+ punpcklwd m3, m6 ; 31 21 30 20
+ punpckhwd m4, m6 ; 33 23 32 22
- punpcklwd mm3, mm6 ; 31 21 30 20
- punpckhwd mm4, mm6 ; 33 23 32 22
+ mova m0, m1 ; 11 01 10 00
+ mova m5, m2 ; 13 03 12 02
- movq mm0, mm1 ; 11 01 10 00
- movq mm5, mm2 ; 13 03 12 02
+ punpckldq m0, m3 ; 30 20 10 00
+ punpckhdq m1, m3 ; 31 21 11 01
- punpckldq mm0, mm3 ; 30 20 10 00
- punpckhdq mm1, mm3 ; 31 21 11 01
+ punpckldq m2, m4 ; 32 22 12 02
+ punpckhdq m5, m4 ; 33 23 13 03
- punpckldq mm2, mm4 ; 32 22 12 02
- punpckhdq mm5, mm4 ; 33 23 13 03
+ mova [outq], m0
- movq [rdx], mm0
-
- movq [rdx+rax], mm1
- movq [rdx+rax*2], mm2
-
- add rdx, rax
- movq [rdx+rax*2], mm5
-
- ; begin epilog
- RESTORE_GOT
- UNSHADOW_ARGS
- pop rbp
- ret
+ mova [outq+r2], m1
+ mova [outq+pitq*2], m2
+ add outq, pitq
+ mova [outq+pitq*2], m5
+ RET
;void short_idct4x4llm_1_mmx(short *input, short *output, int pitch)
-global sym(vp8_short_idct4x4llm_1_mmx)
-sym(vp8_short_idct4x4llm_1_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 3
- GET_GOT rbx
- ; end prolog
-
- mov rax, arg(0) ;input
- movd mm0, [rax]
-
- paddw mm0, [GLOBAL(fours)]
- mov rdx, arg(1) ;output
-
- psraw mm0, 3
- movsxd rax, dword ptr arg(2) ;pitch
-
- punpcklwd mm0, mm0
- punpckldq mm0, mm0
-
- movq [rdx], mm0
- movq [rdx+rax], mm0
-
- movq [rdx+rax*2], mm0
- add rdx, rax
-
- movq [rdx+rax*2], mm0
-
-
- ; begin epilog
- RESTORE_GOT
- UNSHADOW_ARGS
- pop rbp
- ret
-
-;void vp8_dc_only_idct_add_mmx(short input_dc, unsigned char *pred_ptr, unsigned char *dst_ptr, int pitch, int stride)
-global sym(vp8_dc_only_idct_add_mmx)
-sym(vp8_dc_only_idct_add_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 5
- GET_GOT rbx
- push rsi
- push rdi
- ; end prolog
-
- mov rsi, arg(1) ;s -- prediction
- mov rdi, arg(2) ;d -- destination
- movsxd rax, dword ptr arg(4) ;stride
- movsxd rdx, dword ptr arg(3) ;pitch
- pxor mm0, mm0
-
- movd mm5, arg(0) ;input_dc
-
- paddw mm5, [GLOBAL(fours)]
-
- psraw mm5, 3
-
- punpcklwd mm5, mm5
- punpckldq mm5, mm5
-
- movd mm1, [rsi]
- punpcklbw mm1, mm0
- paddsw mm1, mm5
- packuswb mm1, mm0 ; pack and unpack to saturate
- movd [rdi], mm1
-
- movd mm2, [rsi+rdx]
- punpcklbw mm2, mm0
- paddsw mm2, mm5
- packuswb mm2, mm0 ; pack and unpack to saturate
- movd [rdi+rax], mm2
-
- movd mm3, [rsi+2*rdx]
- punpcklbw mm3, mm0
- paddsw mm3, mm5
- packuswb mm3, mm0 ; pack and unpack to saturate
- movd [rdi+2*rax], mm3
-
- add rdi, rax
- add rsi, rdx
- movd mm4, [rsi+2*rdx]
- punpcklbw mm4, mm0
- paddsw mm4, mm5
- packuswb mm4, mm0 ; pack and unpack to saturate
- movd [rdi+2*rax], mm4
-
- ; begin epilog
- pop rdi
- pop rsi
- RESTORE_GOT
- UNSHADOW_ARGS
- pop rbp
- ret
+cglobal short_idct4x4llm_1_mmx,3,3,0,inp,out,pit
+ movh m0, [inpq]
+ paddw m0, [pw_16]
+ psraw m0, 5
+ punpcklwd m0, m0
+ punpckldq m0, m0
+
+ mova [outq], m0
+ mova [outq+pitq], m0
+
+ mova [outq+pitq*2], m0
+ add r1, r2
+
+ mova [outq+pitq*2], m0
+ RET
+
+
+;void dc_only_idct_add_mmx(short input_dc, unsigned char *pred_ptr, unsigned char *dst_ptr, int pitch, int stride)
+cglobal dc_only_idct_add_mmx, 5,5,0,in_dc,pred,dst,pit,stride
+ pxor m0, m0
+
+ movh m5, in_dcq ; dc
+ paddw m5, [pw_16]
+
+ psraw m5, 5
+
+ punpcklwd m5, m5
+ punpckldq m5, m5
+
+ movh m1, [predq]
+ punpcklbw m1, m0
+ paddsw m1, m5
+ packuswb m1, m0 ; pack and unpack to saturate
+ movh [dstq], m1
+
+ movh m2, [predq+pitq]
+ punpcklbw m2, m0
+ paddsw m2, m5
+ packuswb m2, m0 ; pack and unpack to saturate
+ movh [dstq+strideq], m2
+
+ movh m3, [predq+2*pitq]
+ punpcklbw m3, m0
+ paddsw m3, m5
+ packuswb m3, m0 ; pack and unpack to saturate
+ movh [dstq+2*strideq], m3
+
+ add dstq, strideq
+ add predq, pitq
+ movh m4, [predq+2*pitq]
+ punpcklbw m4, m0
+ paddsw m4, m5
+ packuswb m4, m0 ; pack and unpack to saturate
+ movh [dstq+2*strideq], m4
+ RET
-SECTION_RODATA
-align 16
-x_s1sqr2:
- times 4 dw 0x8A8C
-align 16
-x_c1sqr2less1:
- times 4 dw 0x4E7B
-align 16
-fours:
- times 4 dw 0x0004
diff --git a/vp8/common/x86/x86_systemdependent.c b/vp8/common/x86/x86_systemdependent.c
index b88eb3dc8..fae8632a8 100644
--- a/vp8/common/x86/x86_systemdependent.c
+++ b/vp8/common/x86/x86_systemdependent.c
@@ -34,14 +34,14 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
/* Override default functions with fastest ones for this CPU. */
#if HAVE_MMX
-
+// The commented functions need to be re-written for vpx.
if (flags & HAS_MMX)
{
- rtcd->idct.idct1 = vp8_short_idct4x4llm_1_mmx;
- rtcd->idct.idct16 = vp8_short_idct4x4llm_mmx;
- rtcd->idct.idct1_scalar_add = vp8_dc_only_idct_add_mmx;
- rtcd->idct.iwalsh16 = vp8_short_inv_walsh4x4_mmx;
- rtcd->idct.iwalsh1 = vp8_short_inv_walsh4x4_1_mmx;
+ rtcd->idct.idct1 = vpx_short_idct4x4llm_1_mmx;
+ rtcd->idct.idct16 = vpx_short_idct4x4llm_mmx;
+ rtcd->idct.idct1_scalar_add = vpx_dc_only_idct_add_mmx;
+ //rtcd->idct.iwalsh16 = vp8_short_inv_walsh4x4_mmx;
+ //rtcd->idct.iwalsh1 = vp8_short_inv_walsh4x4_1_mmx;
rtcd->recon.recon = vp8_recon_b_mmx;
rtcd->recon.copy8x8 = vp8_copy_mem8x8_mmx;
@@ -91,7 +91,7 @@ void vp8_arch_x86_common_init(VP8_COMMON *ctx)
vp8_build_intra_predictors_mbuv_s_sse2;
#endif
- rtcd->idct.iwalsh16 = vp8_short_inv_walsh4x4_sse2;
+ //rtcd->idct.iwalsh16 = vp8_short_inv_walsh4x4_sse2;
#if CONFIG_ENHANCED_INTERP == 0 && CONFIG_HIGH_PRECISION_MV == 0 && CONFIG_SIXTEENTH_SUBPEL_UV == 0
rtcd->subpix.sixtap16x16 = vp8_sixtap_predict16x16_sse2;
diff --git a/vp8/decoder/arm/arm_dsystemdependent.c b/vp8/decoder/arm/arm_dsystemdependent.c
index 6ce471217..a6e9d2447 100644
--- a/vp8/decoder/arm/arm_dsystemdependent.c
+++ b/vp8/decoder/arm/arm_dsystemdependent.c
@@ -27,15 +27,16 @@ void vp8_arch_arm_decode_init(VP8D_COMP *pbi)
}
#endif
+//The commented functions need to be re-written for vpx.
#if HAVE_ARMV6
if (flags & HAS_MEDIA)
{
pbi->dequant.block = vp8_dequantize_b_v6;
- pbi->dequant.idct_add = vp8_dequant_idct_add_v6;
+ /*pbi->dequant.idct_add = vp8_dequant_idct_add_v6;
pbi->dequant.dc_idct_add = vp8_dequant_dc_idct_add_v6;
pbi->dequant.dc_idct_add_y_block = vp8_dequant_dc_idct_add_y_block_v6;
pbi->dequant.idct_add_y_block = vp8_dequant_idct_add_y_block_v6;
- pbi->dequant.idct_add_uv_block = vp8_dequant_idct_add_uv_block_v6;
+ pbi->dequant.idct_add_uv_block = vp8_dequant_idct_add_uv_block_v6;*/
}
#endif
@@ -43,12 +44,12 @@ void vp8_arch_arm_decode_init(VP8D_COMP *pbi)
if (flags & HAS_NEON)
{
pbi->dequant.block = vp8_dequantize_b_neon;
- pbi->dequant.idct_add = vp8_dequant_idct_add_neon;
+ //pbi->dequant.idct_add = vp8_dequant_idct_add_neon;
/*This is not used: NEON always dequants two blocks at once.
pbi->dequant.dc_idct_add = vp8_dequant_dc_idct_add_neon;*/
- pbi->dequant.dc_idct_add_y_block = vp8_dequant_dc_idct_add_y_block_neon;
+ /*pbi->dequant.dc_idct_add_y_block = vp8_dequant_dc_idct_add_y_block_neon;
pbi->dequant.idct_add_y_block = vp8_dequant_idct_add_y_block_neon;
- pbi->dequant.idct_add_uv_block = vp8_dequant_idct_add_uv_block_neon;
+ pbi->dequant.idct_add_uv_block = vp8_dequant_idct_add_uv_block_neon;*/
}
#endif
#endif
diff --git a/vp8/decoder/generic/dsystemdependent.c b/vp8/decoder/generic/dsystemdependent.c
index fbba1b50a..4b4a2d105 100644
--- a/vp8/decoder/generic/dsystemdependent.c
+++ b/vp8/decoder/generic/dsystemdependent.c
@@ -43,10 +43,4 @@ void vp8_dmachine_specific_config(VP8D_COMP *pbi)
#if ARCH_ARM
vp8_arch_arm_decode_init(pbi);
#endif
-
- pbi->dequant.idct_add = vp8_dequant_idct_add_c;
- pbi->dequant.dc_idct_add = vp8_dequant_dc_idct_add_c;
- pbi->dequant.dc_idct_add_y_block = vp8_dequant_dc_idct_add_y_block_c;
- pbi->dequant.idct_add_y_block = vp8_dequant_idct_add_y_block_c;
- pbi->dequant.idct_add_uv_block = vp8_dequant_idct_add_uv_block_c;
}
diff --git a/vp8/decoder/x86/dequantize_mmx.asm b/vp8/decoder/x86/dequantize_mmx.asm
index 0d6133a46..9b53e6d92 100644
--- a/vp8/decoder/x86/dequantize_mmx.asm
+++ b/vp8/decoder/x86/dequantize_mmx.asm
@@ -1,5 +1,5 @@
;
-; Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+; Copyright (c) 2012 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
@@ -8,454 +8,380 @@
; be found in the AUTHORS file in the root of the source tree.
;
+%include "vpx_ports/x86inc.asm"
-%include "vpx_ports/x86_abi_support.asm"
-
+SECTION_RODATA
+align 16
+x_s1sqr2: times 4 dw 0x8A8C
+align 16
+x_c1sqr2less1: times 4 dw 0x4E7B
+align 16
+pw_16: times 4 dw 16
-;void vp8_dequantize_b_impl_mmx(short *sq, short *dq, short *q)
-global sym(vp8_dequantize_b_impl_mmx)
-sym(vp8_dequantize_b_impl_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 3
- push rsi
- push rdi
- ; end prolog
+SECTION .text
- mov rsi, arg(0) ;sq
- mov rdi, arg(1) ;dq
- mov rax, arg(2) ;q
+INIT_MMX
- movq mm1, [rsi]
- pmullw mm1, [rax+0] ; mm4 *= kernel 0 modifiers.
- movq [rdi], mm1
- movq mm1, [rsi+8]
- pmullw mm1, [rax+8] ; mm4 *= kernel 0 modifiers.
- movq [rdi+8], mm1
+;void dequantize_b_impl_mmx(short *sq, short *dq, short *q)
+cglobal dequantize_b_impl_mmx, 3,3,0,sq,dq,arg3
+ mova m1, [sqq]
+ pmullw m1, [arg3q+0] ; mm4 *= kernel 0 modifiers.
+ mova [dqq+ 0], m1
- movq mm1, [rsi+16]
- pmullw mm1, [rax+16] ; mm4 *= kernel 0 modifiers.
- movq [rdi+16], mm1
+ mova m1, [sqq+8]
+ pmullw m1, [arg3q+8] ; mm4 *= kernel 0 modifiers.
+ mova [dqq+ 8], m1
- movq mm1, [rsi+24]
- pmullw mm1, [rax+24] ; mm4 *= kernel 0 modifiers.
- movq [rdi+24], mm1
+ mova m1, [sqq+16]
+ pmullw m1, [arg3q+16] ; mm4 *= kernel 0 modifiers.
+ mova [dqq+16], m1
- ; begin epilog
- pop rdi
- pop rsi
- UNSHADOW_ARGS
- pop rbp
- ret
+ mova m1, [sqq+24]
+ pmullw m1, [arg3q+24] ; mm4 *= kernel 0 modifiers.
+ mova [dqq+24], m1
+ RET
;void dequant_idct_add_mmx(short *input, short *dq, unsigned char *pred, unsigned char *dest, int pitch, int stride)
-global sym(vp8_dequant_idct_add_mmx)
-sym(vp8_dequant_idct_add_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 6
- GET_GOT rbx
- push rsi
- push rdi
- ; end prolog
-
- mov rax, arg(0) ;input
- mov rdx, arg(1) ;dq
-
-
- movq mm0, [rax ]
- pmullw mm0, [rdx]
-
- movq mm1, [rax +8]
- pmullw mm1, [rdx +8]
-
- movq mm2, [rax+16]
- pmullw mm2, [rdx+16]
-
- movq mm3, [rax+24]
- pmullw mm3, [rdx+24]
-
- mov rdx, arg(3) ;dest
- mov rsi, arg(2) ;pred
- pxor mm7, mm7
-
+cglobal dequant_idct_add_mmx, 6,6,0,inp,dq,pred,dest,pit,stride
+ mova m0, [inpq+ 0]
+ pmullw m0, [dqq]
- movq [rax], mm7
- movq [rax+8], mm7
+ mova m1, [inpq+ 8]
+ pmullw m1, [dqq+ 8]
- movq [rax+16],mm7
- movq [rax+24],mm7
+ mova m2, [inpq+16]
+ pmullw m2, [dqq+16]
+ mova m3, [inpq+24]
+ pmullw m3, [dqq+24]
- movsxd rax, dword ptr arg(4) ;pitch
- movsxd rdi, dword ptr arg(5) ;stride
+ pxor m7, m7
+ mova [inpq], m7
+ mova [inpq+8], m7
+ mova [inpq+16], m7
+ mova [inpq+24], m7
- psubw mm0, mm2 ; b1= 0-2
- paddw mm2, mm2 ;
- movq mm5, mm1
- paddw mm2, mm0 ; a1 =0+2
+ psubw m0, m2 ; b1= 0-2
+ paddw m2, m2 ;
- pmulhw mm5, [GLOBAL(x_s1sqr2)];
- paddw mm5, mm1 ; ip1 * sin(pi/8) * sqrt(2)
+ mova m5, m1
+ paddw m2, m0 ; a1 =0+2
- movq mm7, mm3 ;
- pmulhw mm7, [GLOBAL(x_c1sqr2less1)];
+ pmulhw m5, [x_s1sqr2];
+ paddw m5, m1 ; ip1 * sin(pi/8) * sqrt(2)
- paddw mm7, mm3 ; ip3 * cos(pi/8) * sqrt(2)
- psubw mm7, mm5 ; c1
+ mova m7, m3 ;
+ pmulhw m7, [x_c1sqr2less1];
- movq mm5, mm1
- movq mm4, mm3
+ paddw m7, m3 ; ip3 * cos(pi/8) * sqrt(2)
+ psubw m7, m5 ; c1
- pmulhw mm5, [GLOBAL(x_c1sqr2less1)]
- paddw mm5, mm1
+ mova m5, m1
+ mova m4, m3
- pmulhw mm3, [GLOBAL(x_s1sqr2)]
- paddw mm3, mm4
+ pmulhw m5, [x_c1sqr2less1]
+ paddw m5, m1
- paddw mm3, mm5 ; d1
- movq mm6, mm2 ; a1
+ pmulhw m3, [x_s1sqr2]
+ paddw m3, m4
- movq mm4, mm0 ; b1
- paddw mm2, mm3 ;0
+ paddw m3, m5 ; d1
+ mova m6, m2 ; a1
- paddw mm4, mm7 ;1
- psubw mm0, mm7 ;2
+ mova m4, m0 ; b1
+ paddw m2, m3 ;0
- psubw mm6, mm3 ;3
+ paddw m4, m7 ;1
+ psubw m0, m7 ;2
- movq mm1, mm2 ; 03 02 01 00
- movq mm3, mm4 ; 23 22 21 20
+ psubw m6, m3 ;3
- punpcklwd mm1, mm0 ; 11 01 10 00
- punpckhwd mm2, mm0 ; 13 03 12 02
+ mova m1, m2 ; 03 02 01 00
+ mova m3, m4 ; 23 22 21 20
- punpcklwd mm3, mm6 ; 31 21 30 20
- punpckhwd mm4, mm6 ; 33 23 32 22
+ punpcklwd m1, m0 ; 11 01 10 00
+ punpckhwd m2, m0 ; 13 03 12 02
- movq mm0, mm1 ; 11 01 10 00
- movq mm5, mm2 ; 13 03 12 02
+ punpcklwd m3, m6 ; 31 21 30 20
+ punpckhwd m4, m6 ; 33 23 32 22
- punpckldq mm0, mm3 ; 30 20 10 00
- punpckhdq mm1, mm3 ; 31 21 11 01
+ mova m0, m1 ; 11 01 10 00
+ mova m5, m2 ; 13 03 12 02
- punpckldq mm2, mm4 ; 32 22 12 02
- punpckhdq mm5, mm4 ; 33 23 13 03
+ punpckldq m0, m3 ; 30 20 10 00
+ punpckhdq m1, m3 ; 31 21 11 01
- movq mm3, mm5 ; 33 23 13 03
+ punpckldq m2, m4 ; 32 22 12 02
+ punpckhdq m5, m4 ; 33 23 13 03
- psubw mm0, mm2 ; b1= 0-2
- paddw mm2, mm2 ;
+ mova m3, m5 ; 33 23 13 03
- movq mm5, mm1
- paddw mm2, mm0 ; a1 =0+2
+ psubw m0, m2 ; b1= 0-2
+ paddw m2, m2 ;
- pmulhw mm5, [GLOBAL(x_s1sqr2)];
- paddw mm5, mm1 ; ip1 * sin(pi/8) * sqrt(2)
+ mova m5, m1
+ paddw m2, m0 ; a1 =0+2
- movq mm7, mm3 ;
- pmulhw mm7, [GLOBAL(x_c1sqr2less1)];
+ pmulhw m5, [x_s1sqr2];
+ paddw m5, m1 ; ip1 * sin(pi/8) * sqrt(2)
- paddw mm7, mm3 ; ip3 * cos(pi/8) * sqrt(2)
- psubw mm7, mm5 ; c1
+ mova m7, m3 ;
+ pmulhw m7, [x_c1sqr2less1];
- movq mm5, mm1
- movq mm4, mm3
+ paddw m7, m3 ; ip3 * cos(pi/8) * sqrt(2)
+ psubw m7, m5 ; c1
- pmulhw mm5, [GLOBAL(x_c1sqr2less1)]
- paddw mm5, mm1
+ mova m5, m1
+ mova m4, m3
- pmulhw mm3, [GLOBAL(x_s1sqr2)]
- paddw mm3, mm4
+ pmulhw m5, [x_c1sqr2less1]
+ paddw m5, m1
- paddw mm3, mm5 ; d1
- paddw mm0, [GLOBAL(fours)]
+ pmulhw m3, [x_s1sqr2]
+ paddw m3, m4
- paddw mm2, [GLOBAL(fours)]
- movq mm6, mm2 ; a1
+ paddw m3, m5 ; d1
+ paddw m0, [pw_16]
- movq mm4, mm0 ; b1
- paddw mm2, mm3 ;0
+ paddw m2, [pw_16]
+ mova m6, m2 ; a1
- paddw mm4, mm7 ;1
- psubw mm0, mm7 ;2
+ mova m4, m0 ; b1
+ paddw m2, m3 ;0
- psubw mm6, mm3 ;3
- psraw mm2, 3
+ paddw m4, m7 ;1
+ psubw m0, m7 ;2
- psraw mm0, 3
- psraw mm4, 3
+ psubw m6, m3 ;3
+ psraw m2, 5
- psraw mm6, 3
+ psraw m0, 5
+ psraw m4, 5
- movq mm1, mm2 ; 03 02 01 00
- movq mm3, mm4 ; 23 22 21 20
+ psraw m6, 5
- punpcklwd mm1, mm0 ; 11 01 10 00
- punpckhwd mm2, mm0 ; 13 03 12 02
+ mova m1, m2 ; 03 02 01 00
+ mova m3, m4 ; 23 22 21 20
- punpcklwd mm3, mm6 ; 31 21 30 20
- punpckhwd mm4, mm6 ; 33 23 32 22
+ punpcklwd m1, m0 ; 11 01 10 00
+ punpckhwd m2, m0 ; 13 03 12 02
- movq mm0, mm1 ; 11 01 10 00
- movq mm5, mm2 ; 13 03 12 02
+ punpcklwd m3, m6 ; 31 21 30 20
+ punpckhwd m4, m6 ; 33 23 32 22
- punpckldq mm0, mm3 ; 30 20 10 00
- punpckhdq mm1, mm3 ; 31 21 11 01
+ mova m0, m1 ; 11 01 10 00
+ mova m5, m2 ; 13 03 12 02
- punpckldq mm2, mm4 ; 32 22 12 02
- punpckhdq mm5, mm4 ; 33 23 13 03
+ punpckldq m0, m3 ; 30 20 10 00
+ punpckhdq m1, m3 ; 31 21 11 01
- pxor mm7, mm7
+ punpckldq m2, m4 ; 32 22 12 02
+ punpckhdq m5, m4 ; 33 23 13 03
- movd mm4, [rsi]
- punpcklbw mm4, mm7
- paddsw mm0, mm4
- packuswb mm0, mm7
- movd [rdx], mm0
+ pxor m7, m7
- movd mm4, [rsi+rax]
- punpcklbw mm4, mm7
- paddsw mm1, mm4
- packuswb mm1, mm7
- movd [rdx+rdi], mm1
+ movh m4, [predq]
+ punpcklbw m4, m7
+ paddsw m0, m4
+ packuswb m0, m7
+ movh [destq], m0
- movd mm4, [rsi+2*rax]
- punpcklbw mm4, mm7
- paddsw mm2, mm4
- packuswb mm2, mm7
- movd [rdx+rdi*2], mm2
+ movh m4, [predq+pitq]
+ punpcklbw m4, m7
+ paddsw m1, m4
+ packuswb m1, m7
+ movh [destq+strideq], m1
- add rdx, rdi
- add rsi, rax
+ movh m4, [predq+2*pitq]
+ punpcklbw m4, m7
+ paddsw m2, m4
+ packuswb m2, m7
+ movh [destq+strideq*2], m2
- movd mm4, [rsi+2*rax]
- punpcklbw mm4, mm7
- paddsw mm5, mm4
- packuswb mm5, mm7
- movd [rdx+rdi*2], mm5
+ add destq, strideq
+ add predq, pitq
- ; begin epilog
- pop rdi
- pop rsi
- RESTORE_GOT
- UNSHADOW_ARGS
- pop rbp
- ret
+ movh m4, [predq+2*pitq]
+ punpcklbw m4, m7
+ paddsw m5, m4
+ packuswb m5, m7
+ movh [destq+strideq*2], m5
+ RET
;void dequant_dc_idct_add_mmx(short *input, short *dq, unsigned char *pred, unsigned char *dest, int pitch, int stride, int Dc)
-global sym(vp8_dequant_dc_idct_add_mmx)
-sym(vp8_dequant_dc_idct_add_mmx):
- push rbp
- mov rbp, rsp
- SHADOW_ARGS_TO_STACK 7
- GET_GOT rbx
- push rsi
- push rdi
- ; end prolog
+cglobal dequant_dc_idct_add_mmx, 7,7,0,inp,dq,pred,dest,pit,stride,Dc
+ mova m0, [inpq+ 0]
+ pmullw m0, [dqq+ 0]
- mov rax, arg(0) ;input
- mov rdx, arg(1) ;dq
+ mova m1, [inpq+ 8]
+ pmullw m1, [dqq+ 8]
- movq mm0, [rax ]
- pmullw mm0, [rdx]
+ mova m2, [inpq+16]
+ pmullw m2, [dqq+16]
- movq mm1, [rax +8]
- pmullw mm1, [rdx +8]
+ mova m3, [inpq+24]
+ pmullw m3, [dqq+24]
- movq mm2, [rax+16]
- pmullw mm2, [rdx+16]
+ pxor m7, m7
+ mova [inpq+ 0], m7
+ mova [inpq+ 8], m7
+ mova [inpq+16], m7
+ mova [inpq+24], m7
- movq mm3, [rax+24]
- pmullw mm3, [rdx+24]
+ ; move lower word of Dc to lower word of m0
+ psrlq m0, 16
+ psllq m0, 16
+ and Dcq, 0xFFFF ; If Dc < 0, we don't want the full dword precision.
+ movh m7, Dcq
+ por m0, m7
+ psubw m0, m2 ; b1= 0-2
+ paddw m2, m2 ;
- mov rdx, arg(3) ;dest
- mov rsi, arg(2) ;pred
- pxor mm7, mm7
+ mova m5, m1
+ paddw m2, m0 ; a1 =0+2
+ pmulhw m5, [x_s1sqr2];
+ paddw m5, m1 ; ip1 * sin(pi/8) * sqrt(2)
- movq [rax], mm7
- movq [rax+8], mm7
+ mova m7, m3 ;
+ pmulhw m7, [x_c1sqr2less1];
- movq [rax+16],mm7
- movq [rax+24],mm7
+ paddw m7, m3 ; ip3 * cos(pi/8) * sqrt(2)
+ psubw m7, m5 ; c1
- ; move lower word of Dc to lower word of mm0
- psrlq mm0, 16
- movzx rcx, word ptr arg(6) ;Dc
- psllq mm0, 16
- movq mm7, rcx
- por mm0, mm7
+ mova m5, m1
+ mova m4, m3
- movsxd rax, dword ptr arg(4) ;pitch
- movsxd rdi, dword ptr arg(5) ;stride
+ pmulhw m5, [x_c1sqr2less1]
+ paddw m5, m1
- psubw mm0, mm2 ; b1= 0-2
- paddw mm2, mm2 ;
+ pmulhw m3, [x_s1sqr2]
+ paddw m3, m4
- movq mm5, mm1
- paddw mm2, mm0 ; a1 =0+2
+ paddw m3, m5 ; d1
+ mova m6, m2 ; a1
- pmulhw mm5, [GLOBAL(x_s1sqr2)];
- paddw mm5, mm1 ; ip1 * sin(pi/8) * sqrt(2)
+ mova m4, m0 ; b1
+ paddw m2, m3 ;0
- movq mm7, mm3 ;
- pmulhw mm7, [GLOBAL(x_c1sqr2less1)];
+ paddw m4, m7 ;1
+ psubw m0, m7 ;2
- paddw mm7, mm3 ; ip3 * cos(pi/8) * sqrt(2)
- psubw mm7, mm5 ; c1
+ psubw m6, m3 ;3
- movq mm5, mm1
- movq mm4, mm3
+ mova m1, m2 ; 03 02 01 00
+ mova m3, m4 ; 23 22 21 20
- pmulhw mm5, [GLOBAL(x_c1sqr2less1)]
- paddw mm5, mm1
+ punpcklwd m1, m0 ; 11 01 10 00
+ punpckhwd m2, m0 ; 13 03 12 02
- pmulhw mm3, [GLOBAL(x_s1sqr2)]
- paddw mm3, mm4
+ punpcklwd m3, m6 ; 31 21 30 20
+ punpckhwd m4, m6 ; 33 23 32 22
- paddw mm3, mm5 ; d1
- movq mm6, mm2 ; a1
+ mova m0, m1 ; 11 01 10 00
+ mova m5, m2 ; 13 03 12 02
- movq mm4, mm0 ; b1
- paddw mm2, mm3 ;0
+ punpckldq m0, m3 ; 30 20 10 00
+ punpckhdq m1, m3 ; 31 21 11 01
- paddw mm4, mm7 ;1
- psubw mm0, mm7 ;2
+ punpckldq m2, m4 ; 32 22 12 02
+ punpckhdq m5, m4 ; 33 23 13 03
- psubw mm6, mm3 ;3
+ mova m3, m5 ; 33 23 13 03
- movq mm1, mm2 ; 03 02 01 00
- movq mm3, mm4 ; 23 22 21 20
+ psubw m0, m2 ; b1= 0-2
+ paddw m2, m2 ;
- punpcklwd mm1, mm0 ; 11 01 10 00
- punpckhwd mm2, mm0 ; 13 03 12 02
+ mova m5, m1
+ paddw m2, m0 ; a1 =0+2
- punpcklwd mm3, mm6 ; 31 21 30 20
- punpckhwd mm4, mm6 ; 33 23 32 22
+ pmulhw m5, [x_s1sqr2];
+ paddw m5, m1 ; ip1 * sin(pi/8) * sqrt(2)
- movq mm0, mm1 ; 11 01 10 00
- movq mm5, mm2 ; 13 03 12 02
+ mova m7, m3 ;
+ pmulhw m7, [x_c1sqr2less1];
- punpckldq mm0, mm3 ; 30 20 10 00
- punpckhdq mm1, mm3 ; 31 21 11 01
+ paddw m7, m3 ; ip3 * cos(pi/8) * sqrt(2)
+ psubw m7, m5 ; c1
- punpckldq mm2, mm4 ; 32 22 12 02
- punpckhdq mm5, mm4 ; 33 23 13 03
+ mova m5, m1
+ mova m4, m3
- movq mm3, mm5 ; 33 23 13 03
+ pmulhw m5, [x_c1sqr2less1]
+ paddw m5, m1
- psubw mm0, mm2 ; b1= 0-2
- paddw mm2, mm2 ;
+ pmulhw m3, [x_s1sqr2]
+ paddw m3, m4
- movq mm5, mm1
- paddw mm2, mm0 ; a1 =0+2
+ paddw m3, m5 ; d1
+ paddw m0, [pw_16]
- pmulhw mm5, [GLOBAL(x_s1sqr2)];
- paddw mm5, mm1 ; ip1 * sin(pi/8) * sqrt(2)
+ paddw m2, [pw_16]
+ mova m6, m2 ; a1
- movq mm7, mm3 ;
- pmulhw mm7, [GLOBAL(x_c1sqr2less1)];
+ mova m4, m0 ; b1
+ paddw m2, m3 ;0
- paddw mm7, mm3 ; ip3 * cos(pi/8) * sqrt(2)
- psubw mm7, mm5 ; c1
+ paddw m4, m7 ;1
+ psubw m0, m7 ;2
- movq mm5, mm1
- movq mm4, mm3
+ psubw m6, m3 ;3
+ psraw m2, 5
- pmulhw mm5, [GLOBAL(x_c1sqr2less1)]
- paddw mm5, mm1
+ psraw m0, 5
+ psraw m4, 5
- pmulhw mm3, [GLOBAL(x_s1sqr2)]
- paddw mm3, mm4
+ psraw m6, 5
- paddw mm3, mm5 ; d1
- paddw mm0, [GLOBAL(fours)]
+ mova m1, m2 ; 03 02 01 00
+ mova m3, m4 ; 23 22 21 20
- paddw mm2, [GLOBAL(fours)]
- movq mm6, mm2 ; a1
+ punpcklwd m1, m0 ; 11 01 10 00
+ punpckhwd m2, m0 ; 13 03 12 02
- movq mm4, mm0 ; b1
- paddw mm2, mm3 ;0
+ punpcklwd m3, m6 ; 31 21 30 20
+ punpckhwd m4, m6 ; 33 23 32 22
- paddw mm4, mm7 ;1
- psubw mm0, mm7 ;2
+ mova m0, m1 ; 11 01 10 00
+ mova m5, m2 ; 13 03 12 02
- psubw mm6, mm3 ;3
- psraw mm2, 3
+ punpckldq m0, m3 ; 30 20 10 00
+ punpckhdq m1, m3 ; 31 21 11 01
- psraw mm0, 3
- psraw mm4, 3
+ punpckldq m2, m4 ; 32 22 12 02
+ punpckhdq m5, m4 ; 33 23 13 03
- psraw mm6, 3
+ pxor m7, m7
- movq mm1, mm2 ; 03 02 01 00
- movq mm3, mm4 ; 23 22 21 20
+ movh m4, [predq]
+ punpcklbw m4, m7
+ paddsw m0, m4
+ packuswb m0, m7
+ movh [destq], m0
- punpcklwd mm1, mm0 ; 11 01 10 00
- punpckhwd mm2, mm0 ; 13 03 12 02
+ movh m4, [predq+pitq]
+ punpcklbw m4, m7
+ paddsw m1, m4
+ packuswb m1, m7
+ movh [destq+strideq], m1
- punpcklwd mm3, mm6 ; 31 21 30 20
- punpckhwd mm4, mm6 ; 33 23 32 22
+ movh m4, [predq+2*pitq]
+ punpcklbw m4, m7
+ paddsw m2, m4
+ packuswb m2, m7
+ movh [destq+strideq*2], m2
- movq mm0, mm1 ; 11 01 10 00
- movq mm5, mm2 ; 13 03 12 02
+ add destq, strideq
+ add predq, pitq
- punpckldq mm0, mm3 ; 30 20 10 00
- punpckhdq mm1, mm3 ; 31 21 11 01
+ movh m4, [predq+2*pitq]
+ punpcklbw m4, m7
+ paddsw m5, m4
+ packuswb m5, m7
+ movh [destq+strideq*2], m5
+ RET
- punpckldq mm2, mm4 ; 32 22 12 02
- punpckhdq mm5, mm4 ; 33 23 13 03
-
- pxor mm7, mm7
-
- movd mm4, [rsi]
- punpcklbw mm4, mm7
- paddsw mm0, mm4
- packuswb mm0, mm7
- movd [rdx], mm0
-
- movd mm4, [rsi+rax]
- punpcklbw mm4, mm7
- paddsw mm1, mm4
- packuswb mm1, mm7
- movd [rdx+rdi], mm1
-
- movd mm4, [rsi+2*rax]
- punpcklbw mm4, mm7
- paddsw mm2, mm4
- packuswb mm2, mm7
- movd [rdx+rdi*2], mm2
-
- add rdx, rdi
- add rsi, rax
-
- movd mm4, [rsi+2*rax]
- punpcklbw mm4, mm7
- paddsw mm5, mm4
- packuswb mm5, mm7
- movd [rdx+rdi*2], mm5
-
- ; begin epilog
- pop rdi
- pop rsi
- RESTORE_GOT
- UNSHADOW_ARGS
- pop rbp
- ret
-
-
-SECTION_RODATA
-align 16
-x_s1sqr2:
- times 4 dw 0x8A8C
-align 16
-x_c1sqr2less1:
- times 4 dw 0x4E7B
-align 16
-fours:
- times 4 dw 0x0004
diff --git a/vp8/decoder/x86/dequantize_x86.h b/vp8/decoder/x86/dequantize_x86.h
index dc68daab3..32f25cced 100644
--- a/vp8/decoder/x86/dequantize_x86.h
+++ b/vp8/decoder/x86/dequantize_x86.h
@@ -21,8 +21,8 @@
*/
#if HAVE_MMX
extern prototype_dequant_block(vp8_dequantize_b_mmx);
-extern prototype_dequant_idct_add(vp8_dequant_idct_add_mmx);
-extern prototype_dequant_dc_idct_add(vp8_dequant_dc_idct_add_mmx);
+extern prototype_dequant_idct_add(vpx_dequant_idct_add_mmx);
+extern prototype_dequant_dc_idct_add(vpx_dequant_dc_idct_add_mmx);
extern prototype_dequant_dc_idct_add_y_block(vp8_dequant_dc_idct_add_y_block_mmx);
extern prototype_dequant_idct_add_y_block(vp8_dequant_idct_add_y_block_mmx);
extern prototype_dequant_idct_add_uv_block(vp8_dequant_idct_add_uv_block_mmx);
@@ -32,7 +32,7 @@ extern prototype_dequant_idct_add_uv_block(vp8_dequant_idct_add_uv_block_mmx);
#define vp8_dequant_block vp8_dequantize_b_mmx
#undef vp8_dequant_idct_add
-#define vp8_dequant_idct_add vp8_dequant_idct_add_mmx
+#define vp8_dequant_idct_add vpx_dequant_idct_add_mmx
#undef vp8_dequant_dc_idct_add
#define vp8_dequant_dc_idct_add vp8_dequant_dc_idct_add_mmx
diff --git a/vp8/decoder/x86/idct_blk_mmx.c b/vp8/decoder/x86/idct_blk_mmx.c
index 8f1a363cd..b9c11bc65 100644
--- a/vp8/decoder/x86/idct_blk_mmx.c
+++ b/vp8/decoder/x86/idct_blk_mmx.c
@@ -21,24 +21,24 @@ void vp8_dequant_dc_idct_add_y_block_mmx
for (i = 0; i < 4; i++)
{
if (eobs[0] > 1)
- vp8_dequant_dc_idct_add_mmx (q, dq, pre, dst, 16, stride, dc[0]);
+ vpx_dequant_dc_idct_add_mmx (q, dq, pre, dst, 16, stride, dc[0]);
else
- vp8_dc_only_idct_add_mmx (dc[0], pre, dst, 16, stride);
+ vpx_dc_only_idct_add_mmx (dc[0], pre, dst, 16, stride);
if (eobs[1] > 1)
- vp8_dequant_dc_idct_add_mmx (q+16, dq, pre+4, dst+4, 16, stride, dc[1]);
+ vpx_dequant_dc_idct_add_mmx (q+16, dq, pre+4, dst+4, 16, stride, dc[1]);
else
- vp8_dc_only_idct_add_mmx (dc[1], pre+4, dst+4, 16, stride);
+ vpx_dc_only_idct_add_mmx (dc[1], pre+4, dst+4, 16, stride);
if (eobs[2] > 1)
- vp8_dequant_dc_idct_add_mmx (q+32, dq, pre+8, dst+8, 16, stride, dc[2]);
+ vpx_dequant_dc_idct_add_mmx (q+32, dq, pre+8, dst+8, 16, stride, dc[2]);
else
- vp8_dc_only_idct_add_mmx (dc[2], pre+8, dst+8, 16, stride);
+ vpx_dc_only_idct_add_mmx (dc[2], pre+8, dst+8, 16, stride);
if (eobs[3] > 1)
- vp8_dequant_dc_idct_add_mmx (q+48, dq, pre+12, dst+12, 16, stride, dc[3]);
+ vpx_dequant_dc_idct_add_mmx (q+48, dq, pre+12, dst+12, 16, stride, dc[3]);
else
- vp8_dc_only_idct_add_mmx (dc[3], pre+12, dst+12, 16, stride);
+ vpx_dc_only_idct_add_mmx (dc[3], pre+12, dst+12, 16, stride);
q += 64;
dc += 4;
@@ -57,34 +57,34 @@ void vp8_dequant_idct_add_y_block_mmx
for (i = 0; i < 4; i++)
{
if (eobs[0] > 1)
- vp8_dequant_idct_add_mmx (q, dq, pre, dst, 16, stride);
+ vpx_dequant_idct_add_mmx (q, dq, pre, dst, 16, stride);
else
{
- vp8_dc_only_idct_add_mmx (q[0]*dq[0], pre, dst, 16, stride);
+ vpx_dc_only_idct_add_mmx (q[0]*dq[0], pre, dst, 16, stride);
((int *)q)[0] = 0;
}
if (eobs[1] > 1)
- vp8_dequant_idct_add_mmx (q+16, dq, pre+4, dst+4, 16, stride);
+ vpx_dequant_idct_add_mmx (q+16, dq, pre+4, dst+4, 16, stride);
else
{
- vp8_dc_only_idct_add_mmx (q[16]*dq[0], pre+4, dst+4, 16, stride);
+ vpx_dc_only_idct_add_mmx (q[16]*dq[0], pre+4, dst+4, 16, stride);
((int *)(q+16))[0] = 0;
}
if (eobs[2] > 1)
- vp8_dequant_idct_add_mmx (q+32, dq, pre+8, dst+8, 16, stride);
+ vpx_dequant_idct_add_mmx (q+32, dq, pre+8, dst+8, 16, stride);
else
{
- vp8_dc_only_idct_add_mmx (q[32]*dq[0], pre+8, dst+8, 16, stride);
+ vpx_dc_only_idct_add_mmx (q[32]*dq[0], pre+8, dst+8, 16, stride);
((int *)(q+32))[0] = 0;
}
if (eobs[3] > 1)
- vp8_dequant_idct_add_mmx (q+48, dq, pre+12, dst+12, 16, stride);
+ vpx_dequant_idct_add_mmx (q+48, dq, pre+12, dst+12, 16, stride);
else
{
- vp8_dc_only_idct_add_mmx (q[48]*dq[0], pre+12, dst+12, 16, stride);
+ vpx_dc_only_idct_add_mmx (q[48]*dq[0], pre+12, dst+12, 16, stride);
((int *)(q+48))[0] = 0;
}
@@ -104,18 +104,18 @@ void vp8_dequant_idct_add_uv_block_mmx
for (i = 0; i < 2; i++)
{
if (eobs[0] > 1)
- vp8_dequant_idct_add_mmx (q, dq, pre, dstu, 8, stride);
+ vpx_dequant_idct_add_mmx (q, dq, pre, dstu, 8, stride);
else
{
- vp8_dc_only_idct_add_mmx (q[0]*dq[0], pre, dstu, 8, stride);
+ vpx_dc_only_idct_add_mmx (q[0]*dq[0], pre, dstu, 8, stride);
((int *)q)[0] = 0;
}
if (eobs[1] > 1)
- vp8_dequant_idct_add_mmx (q+16, dq, pre+4, dstu+4, 8, stride);
+ vpx_dequant_idct_add_mmx (q+16, dq, pre+4, dstu+4, 8, stride);
else
{
- vp8_dc_only_idct_add_mmx (q[16]*dq[0], pre+4, dstu+4, 8, stride);
+ vpx_dc_only_idct_add_mmx (q[16]*dq[0], pre+4, dstu+4, 8, stride);
((int *)(q+16))[0] = 0;
}
@@ -128,18 +128,18 @@ void vp8_dequant_idct_add_uv_block_mmx
for (i = 0; i < 2; i++)
{
if (eobs[0] > 1)
- vp8_dequant_idct_add_mmx (q, dq, pre, dstv, 8, stride);
+ vpx_dequant_idct_add_mmx (q, dq, pre, dstv, 8, stride);
else
{
- vp8_dc_only_idct_add_mmx (q[0]*dq[0], pre, dstv, 8, stride);
+ vpx_dc_only_idct_add_mmx (q[0]*dq[0], pre, dstv, 8, stride);
((int *)q)[0] = 0;
}
if (eobs[1] > 1)
- vp8_dequant_idct_add_mmx (q+16, dq, pre+4, dstv+4, 8, stride);
+ vpx_dequant_idct_add_mmx (q+16, dq, pre+4, dstv+4, 8, stride);
else
{
- vp8_dc_only_idct_add_mmx (q[16]*dq[0], pre+4, dstv+4, 8, stride);
+ vpx_dc_only_idct_add_mmx (q[16]*dq[0], pre+4, dstv+4, 8, stride);
((int *)(q+16))[0] = 0;
}
diff --git a/vp8/decoder/x86/x86_dsystemdependent.c b/vp8/decoder/x86/x86_dsystemdependent.c
index a244a3a98..c46cfaec7 100644
--- a/vp8/decoder/x86/x86_dsystemdependent.c
+++ b/vp8/decoder/x86/x86_dsystemdependent.c
@@ -15,14 +15,14 @@
#if HAVE_MMX
-void vp8_dequantize_b_impl_mmx(short *sq, short *dq, short *q);
+void vpx_dequantize_b_impl_mmx(short *sq, short *dq, short *q);
void vp8_dequantize_b_mmx(BLOCKD *d)
{
short *sq = (short *) d->qcoeff;
short *dq = (short *) d->dqcoeff;
short *q = (short *) d->dequant;
- vp8_dequantize_b_impl_mmx(sq, dq, q);
+ vpx_dequantize_b_impl_mmx(sq, dq, q);
}
#endif
@@ -42,8 +42,8 @@ void vp8_arch_x86_decode_init(VP8D_COMP *pbi)
if (flags & HAS_MMX)
{
pbi->dequant.block = vp8_dequantize_b_mmx;
- pbi->dequant.idct_add = vp8_dequant_idct_add_mmx;
- pbi->dequant.dc_idct_add = vp8_dequant_dc_idct_add_mmx;
+ pbi->dequant.idct_add = vpx_dequant_idct_add_mmx;
+ pbi->dequant.dc_idct_add = vpx_dequant_dc_idct_add_mmx;
pbi->dequant.dc_idct_add_y_block = vp8_dequant_dc_idct_add_y_block_mmx;
pbi->dequant.idct_add_y_block = vp8_dequant_idct_add_y_block_mmx;
pbi->dequant.idct_add_uv_block = vp8_dequant_idct_add_uv_block_mmx;
@@ -52,9 +52,9 @@ void vp8_arch_x86_decode_init(VP8D_COMP *pbi)
#if HAVE_SSE2
if (flags & HAS_SSE2)
{
- pbi->dequant.dc_idct_add_y_block = vp8_dequant_dc_idct_add_y_block_sse2;
- pbi->dequant.idct_add_y_block = vp8_dequant_idct_add_y_block_sse2;
- pbi->dequant.idct_add_uv_block = vp8_dequant_idct_add_uv_block_sse2;
+ //pbi->dequant.dc_idct_add_y_block = vp8_dequant_dc_idct_add_y_block_sse2;
+ //pbi->dequant.idct_add_y_block = vp8_dequant_idct_add_y_block_sse2;
+ //pbi->dequant.idct_add_uv_block = vp8_dequant_idct_add_uv_block_sse2;
}
#endif
generated by cgit v1.2.3 (git 2.25.1) at 2024-10-11 01:23:44 +0000