[FFmpeg-cvslog] png: add SSE2 version for add_bytes_l2.

[Ronald S. Bultje]

ffmpeg | branch: master | Ronald S. Bultje <rsbultje at gmail.com> | Fri Jan 27 23:23:05 2012 +0800| [f91c4b7824a558b472bb0acd610c772cd9904003] | committer: Ronald S. Bultje

png: add SSE2 version for add_bytes_l2.

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=f91c4b7824a558b472bb0acd610c772cd9904003
---

 libavcodec/x86/pngdsp-init.c |    6 ++++++
 libavcodec/x86/pngdsp.asm    |   20 ++++++++++++++++++++
 2 files changed, 26 insertions(+), 0 deletions(-)

diff --git a/libavcodec/x86/pngdsp-init.c b/libavcodec/x86/pngdsp-init.c
index 9c7d696..136e92e 100644
--- a/libavcodec/x86/pngdsp-init.c
+++ b/libavcodec/x86/pngdsp-init.c
@@ -29,16 +29,22 @@ void ff_add_png_paeth_prediction_ssse3(uint8_t *dst, uint8_t *src,
                                        uint8_t *top, int w, int bpp);
 void ff_add_bytes_l2_mmx (uint8_t *dst, uint8_t *src1,
                           uint8_t *src2, int w);
+void ff_add_bytes_l2_sse2(uint8_t *dst, uint8_t *src1,
+                          uint8_t *src2, int w);
 
 void ff_pngdsp_init_x86(PNGDSPContext *dsp)
 {
 #if HAVE_YASM
     int flags = av_get_cpu_flags();
 
+#if ARCH_X86_32
     if (flags & AV_CPU_FLAG_MMX)
         dsp->add_bytes_l2         = ff_add_bytes_l2_mmx;
+#endif
     if (flags & AV_CPU_FLAG_MMX2)
         dsp->add_paeth_prediction = ff_add_png_paeth_prediction_mmx2;
+    if (flags & AV_CPU_FLAG_SSE2)
+        dsp->add_bytes_l2         = ff_add_bytes_l2_sse2;
     if (flags & AV_CPU_FLAG_SSSE3)
         dsp->add_paeth_prediction = ff_add_png_paeth_prediction_ssse3;
 #endif
diff --git a/libavcodec/x86/pngdsp.asm b/libavcodec/x86/pngdsp.asm
index 92425ad..bff76e0 100644
--- a/libavcodec/x86/pngdsp.asm
+++ b/libavcodec/x86/pngdsp.asm
@@ -53,6 +53,21 @@ cglobal add_bytes_l2, 4, 6, %1, dst, src1, src2, wa, w, i
     cmp                 iq, waq
     jl .loop_v
 
+%if mmsize == 16
+    ; vector loop
+    mov                 wq, waq
+    and                waq, ~7
+    jmp .end_l
+.loop_l:
+    movq               mm0, [src1q+iq]
+    paddb              mm0, [src2q+iq]
+    movq  [dstq+iq       ], mm0
+    add                 iq, 8
+.end_l:
+    cmp                 iq, waq
+    jl .loop_l
+%endif
+
     ; scalar loop for leftover
     jmp .end_s
 .loop_s:
@@ -66,8 +81,13 @@ cglobal add_bytes_l2, 4, 6, %1, dst, src1, src2, wa, w, i
     REP_RET
 %endmacro
 
+%if ARCH_X86_32
 INIT_MMX mmx
 ADD_BYTES_FN 0
+%endif
+
+INIT_XMM sse2
+ADD_BYTES_FN 2
 
 %macro ADD_PAETH_PRED_FN 1
 cglobal add_png_paeth_prediction, 5, 7, %1, dst, src, top, w, bpp, end, cntr