Imported Debian version 2.5.0~trusty1.1

[deb_ffmpeg.git] / ffmpeg / libavcodec / x86 / idctdsp.asm

diff --git a/ffmpeg/libavcodec/x86/idctdsp.asm b/ffmpeg/libavcodec/x86/idctdsp.asm
index 44a1a6ef856d0ae56013937fb99b156354dd0a81..0aa73459e23a2c6f9e0e88718d2551b8b1619e08 100644 (file)
--- a/ffmpeg/libavcodec/x86/idctdsp.asm
+++ b/ffmpeg/libavcodec/x86/idctdsp.asm
@@ -31,7 +31,7 @@ SECTION_TEXT
 
 ;--------------------------------------------------------------------------
 ;void ff_put_signed_pixels_clamped(const int16_t *block, uint8_t *pixels,
-;                                  int line_size)
+;                                  ptrdiff_t line_size)
 ;--------------------------------------------------------------------------
 
 %macro PUT_SIGNED_PIXELS_CLAMPED_HALF 1
@@ -78,3 +78,106 @@ INIT_MMX mmx
 PUT_SIGNED_PIXELS_CLAMPED 0
 INIT_XMM sse2
 PUT_SIGNED_PIXELS_CLAMPED 3
+
+;--------------------------------------------------------------------------
+; void ff_put_pixels_clamped(const int16_t *block, uint8_t *pixels,
+;                            ptrdiff_t line_size);
+;--------------------------------------------------------------------------
+; %1 = block offset
+%macro PUT_PIXELS_CLAMPED_HALF 1
+    mova     m0, [blockq+mmsize*0+%1]
+    mova     m1, [blockq+mmsize*2+%1]
+%if mmsize == 8
+    mova     m2, [blockq+mmsize*4+%1]
+    mova     m3, [blockq+mmsize*6+%1]
+%endif
+    packuswb m0, [blockq+mmsize*1+%1]
+    packuswb m1, [blockq+mmsize*3+%1]
+%if mmsize == 8
+    packuswb m2, [blockq+mmsize*5+%1]
+    packuswb m3, [blockq+mmsize*7+%1]
+    movq           [pixelsq], m0
+    movq    [lsizeq+pixelsq], m1
+    movq  [2*lsizeq+pixelsq], m2
+    movq   [lsize3q+pixelsq], m3
+%else
+    movq           [pixelsq], m0
+    movhps  [lsizeq+pixelsq], m0
+    movq  [2*lsizeq+pixelsq], m1
+    movhps [lsize3q+pixelsq], m1
+%endif
+%endmacro
+
+%macro PUT_PIXELS_CLAMPED 0
+cglobal put_pixels_clamped, 3, 4, 2, block, pixels, lsize, lsize3
+    lea lsize3q, [lsizeq*3]
+    PUT_PIXELS_CLAMPED_HALF 0
+    lea pixelsq, [pixelsq+lsizeq*4]
+    PUT_PIXELS_CLAMPED_HALF 64
+    RET
+%endmacro
+
+INIT_MMX mmx
+PUT_PIXELS_CLAMPED
+INIT_XMM sse2
+PUT_PIXELS_CLAMPED
+
+;--------------------------------------------------------------------------
+; void ff_add_pixels_clamped(const int16_t *block, uint8_t *pixels,
+;                            ptrdiff_t line_size);
+;--------------------------------------------------------------------------
+; %1 = block offset
+%macro ADD_PIXELS_CLAMPED 1
+    mova       m0, [blockq+mmsize*0+%1]
+    mova       m1, [blockq+mmsize*1+%1]
+%if mmsize == 8
+    mova       m5, [blockq+mmsize*2+%1]
+    mova       m6, [blockq+mmsize*3+%1]
+%endif
+    movq       m2, [pixelsq]
+    movq       m3, [pixelsq+lsizeq]
+%if mmsize == 8
+    mova       m7, m2
+    punpcklbw  m2, m4
+    punpckhbw  m7, m4
+    paddsw     m0, m2
+    paddsw     m1, m7
+    mova       m7, m3
+    punpcklbw  m3, m4
+    punpckhbw  m7, m4
+    paddsw     m5, m3
+    paddsw     m6, m7
+%else
+    punpcklbw  m2, m4
+    punpcklbw  m3, m4
+    paddsw     m0, m2
+    paddsw     m1, m3
+%endif
+    packuswb   m0, m1
+%if mmsize == 8
+    packuswb   m5, m6
+    movq       [pixelsq], m0
+    movq       [pixelsq+lsizeq], m5
+%else
+    movq       [pixelsq], m0
+    movhps     [pixelsq+lsizeq], m0
+%endif
+%endmacro
+
+%macro ADD_PIXELS_CLAMPED 0
+cglobal add_pixels_clamped, 3, 3, 5, block, pixels, lsize
+    pxor       m4, m4
+    ADD_PIXELS_CLAMPED 0
+    lea        pixelsq, [pixelsq+lsizeq*2]
+    ADD_PIXELS_CLAMPED 32
+    lea        pixelsq, [pixelsq+lsizeq*2]
+    ADD_PIXELS_CLAMPED 64
+    lea        pixelsq, [pixelsq+lsizeq*2]
+    ADD_PIXELS_CLAMPED 96
+    RET
+%endmacro
+
+INIT_MMX mmx
+ADD_PIXELS_CLAMPED
+INIT_XMM sse2
+ADD_PIXELS_CLAMPED