Merge remote-tracking branch 'qatar/master'

[ffmpeg] / libavcodec / x86 / dsputil_yasm.asm

diff --git a/libavcodec/x86/dsputil_yasm.asm b/libavcodec/x86/dsputil_yasm.asm
index 4607ff15d77d6c9de7323127500aed099d72932f..7ddc770a2eedc884172a4ac782609ae4c1e8e9aa 100644 (file)
--- a/libavcodec/x86/dsputil_yasm.asm
+++ b/libavcodec/x86/dsputil_yasm.asm
@@ -2,24 +2,24 @@
 ;* MMX optimized DSP utils
 ;* Copyright (c) 2008 Loren Merritt
 ;*
-;* This file is part of Libav.
+;* This file is part of FFmpeg.
 ;*
-;* Libav is free software; you can redistribute it and/or
+;* FFmpeg is free software; you can redistribute it and/or
 ;* modify it under the terms of the GNU Lesser General Public
 ;* License as published by the Free Software Foundation; either
 ;* version 2.1 of the License, or (at your option) any later version.
 ;*
-;* Libav is distributed in the hope that it will be useful,
+;* FFmpeg is distributed in the hope that it will be useful,
 ;* but WITHOUT ANY WARRANTY; without even the implied warranty of
 ;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 ;* Lesser General Public License for more details.
 ;*
 ;* You should have received a copy of the GNU Lesser General Public
-;* License along with Libav; if not, write to the Free Software
+;* License along with FFmpeg; if not, write to the Free Software
 ;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 ;******************************************************************************
 
-%include "x86inc.asm"
+%include "libavutil/x86/x86inc.asm"
 %include "x86util.asm"
 
 SECTION_RODATA
@@ -30,17 +30,17 @@ pb_zzzz3333zzzzbbbb: db -1,-1,-1,-1,3,3,3,3,-1,-1,-1,-1,11,11,11,11
 pb_zz11zz55zz99zzdd: db -1,-1,1,1,-1,-1,5,5,-1,-1,9,9,-1,-1,13,13
 pb_revwords: db 14, 15, 12, 13, 10, 11, 8, 9, 6, 7, 4, 5, 2, 3, 0, 1
 pd_16384: times 4 dd 16384
+pb_bswap32: db 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12
 
 SECTION_TEXT
 
 %macro SCALARPRODUCT 1
-; int scalarproduct_int16(int16_t *v1, int16_t *v2, int order, int shift)
-cglobal scalarproduct_int16_%1, 3,3,4, v1, v2, order, shift
+; int scalarproduct_int16(int16_t *v1, int16_t *v2, int order)
+cglobal scalarproduct_int16_%1, 3,3,3, v1, v2, order
     shl orderq, 1
     add v1q, orderq
     add v2q, orderq
     neg orderq
-    movd    m3, shiftm
     pxor    m2, m2
 .loop:
     movu    m0, [v1q + orderq]
@@ -54,10 +54,8 @@ cglobal scalarproduct_int16_%1, 3,3,4, v1, v2, order, shift
 %if mmsize == 16
     movhlps m0, m2
     paddd   m2, m0
-    psrad   m2, m3
     pshuflw m0, m2, 0x4e
 %else
-    psrad   m2, m3
     pshufw  m0, m2, 0x4e
 %endif
     paddd   m2, m0
@@ -1062,7 +1060,7 @@ emu_edge mmx
 ; %4 = CLIPD function takes min/max as float instead of int (CLIPD_SSE2)
 ; %5 = suffix
 %macro VECTOR_CLIP_INT32 4-5
-cglobal vector_clip_int32%5, 5,5,%2, dst, src, min, max, len
+cglobal vector_clip_int32%5, 5,5,%1, dst, src, min, max, len
 %if %4
     cvtsi2ss  m4, minm
     cvtsi2ss  m5, maxm
@@ -1176,5 +1174,130 @@ cglobal butterflies_float_interleave, 4,4,3, dst, src0, src1, len
 
 INIT_XMM sse
 BUTTERFLIES_FLOAT_INTERLEAVE
+%if HAVE_AVX
 INIT_YMM avx
 BUTTERFLIES_FLOAT_INTERLEAVE
+%endif
+
+INIT_XMM sse2
+; %1 = aligned/unaligned
+%macro BSWAP_LOOPS_SSE2  1
+    mov      r3, r2
+    sar      r2, 3
+    jz       .left4_%1
+.loop8_%1:
+    mov%1    m0, [r1 +  0]
+    mov%1    m1, [r1 + 16]
+    pshuflw  m0, m0, 10110001b
+    pshuflw  m1, m1, 10110001b
+    pshufhw  m0, m0, 10110001b
+    pshufhw  m1, m1, 10110001b
+    mova     m2, m0
+    mova     m3, m1
+    psllw    m0, 8
+    psllw    m1, 8
+    psrlw    m2, 8
+    psrlw    m3, 8
+    por      m2, m0
+    por      m3, m1
+    mova     [r0 +  0], m2
+    mova     [r0 + 16], m3
+    add      r1, 32
+    add      r0, 32
+    dec      r2
+    jnz      .loop8_%1
+.left4_%1:
+    mov      r2, r3
+    and      r3, 4
+    jz       .left
+    mov%1    m0, [r1]
+    pshuflw  m0, m0, 10110001b
+    pshufhw  m0, m0, 10110001b
+    mova     m2, m0
+    psllw    m0, 8
+    psrlw    m2, 8
+    por      m2, m0
+    mova     [r0], m2
+    add      r1, 16
+    add      r0, 16
+%endmacro
+
+; void bswap_buf(uint32_t *dst, const uint32_t *src, int w);
+cglobal bswap32_buf, 3,4,5
+    mov      r3, r1
+    and      r3, 15
+    jz       .start_align
+    BSWAP_LOOPS_SSE2  u
+    jmp      .left
+.start_align:
+    BSWAP_LOOPS_SSE2  a
+.left:
+    and      r2, 3
+    jz       .end
+.loop2:
+    mov      r3d, [r1]
+    bswap    r3d
+    mov      [r0], r3d
+    add      r1, 4
+    add      r0, 4
+    dec      r2
+    jnz      .loop2
+.end
+    RET
+
+; %1 = aligned/unaligned
+%macro BSWAP_LOOPS_SSSE3  1
+    mov      r3, r2
+    sar      r2, 3
+    jz       .left4_%1
+.loop8_%1:
+    mov%1    m0, [r1 +  0]
+    mov%1    m1, [r1 + 16]
+    pshufb   m0, m2
+    pshufb   m1, m2
+    mova     [r0 +  0], m0
+    mova     [r0 + 16], m1
+    add      r0, 32
+    add      r1, 32
+    dec      r2
+    jnz      .loop8_%1
+.left4_%1:
+    mov      r2, r3
+    and      r3, 4
+    jz       .left2
+    mov%1    m0, [r1]
+    pshufb   m0, m2
+    mova     [r0], m0
+    add      r1, 16
+    add      r0, 16
+%endmacro
+
+INIT_XMM ssse3
+; void bswap_buf(uint32_t *dst, const uint32_t *src, int w);
+cglobal bswap32_buf, 3,4,3
+    mov      r3, r1
+    mova     m2, [pb_bswap32]
+    and      r3, 15
+    jz       .start_align
+    BSWAP_LOOPS_SSSE3  u
+    jmp      .left2
+.start_align:
+    BSWAP_LOOPS_SSSE3  a
+.left2:
+    mov      r3, r2
+    and      r2, 2
+    jz       .left1
+    movq     m0, [r1]
+    pshufb   m0, m2
+    movq     [r0], m0
+    add      r1, 8
+    add      r0, 8
+.left1:
+    and      r3, 1
+    jz       .end
+    mov      r2d, [r1]
+    bswap    r2d
+    mov      [r0], r2d
+.end:
+    RET
+