avcodec/x86/h264_qpel: Add ff_{avg,put}_pixels16_l2_shift5_sse2

Up until now this function was emulated via two calls
to ff_{avg,pull}_pixels8_l2_shift5_mmxext(). Adding a dedicated
function proved beneficial both size wise and performance wise:
The new functions take 192B, yet the simplified calls save
256B with GCC and 320B with Clang here.

This change will also allow further optimizations.

Old benchmarks:
avg_h264_qpel_16_mc12_8_c:                            1735.8 ( 1.00x)
avg_h264_qpel_16_mc12_8_sse2:                          300.8 ( 5.77x)
avg_h264_qpel_16_mc12_8_ssse3:                         233.3 ( 7.44x)
avg_h264_qpel_16_mc32_8_c:                            1777.9 ( 1.00x)
avg_h264_qpel_16_mc32_8_sse2:                          275.6 ( 6.45x)
avg_h264_qpel_16_mc32_8_ssse3:                         235.7 ( 7.54x)
put_h264_qpel_16_mc12_8_c:                            1808.2 ( 1.00x)
put_h264_qpel_16_mc12_8_sse2:                          267.2 ( 6.77x)
put_h264_qpel_16_mc12_8_ssse3:                         231.9 ( 7.80x)
put_h264_qpel_16_mc32_8_c:                            1766.9 ( 1.00x)
put_h264_qpel_16_mc32_8_sse2:                          272.9 ( 6.47x)
put_h264_qpel_16_mc32_8_ssse3:                         229.5 ( 7.70x)

New benchmarks:
avg_h264_qpel_16_mc12_8_c:                            1742.3 ( 1.00x)
avg_h264_qpel_16_mc12_8_sse2:                          240.3 ( 7.25x)
avg_h264_qpel_16_mc12_8_ssse3:                         214.8 ( 8.11x)
avg_h264_qpel_16_mc32_8_c:                            1748.0 ( 1.00x)
avg_h264_qpel_16_mc32_8_sse2:                          238.0 ( 7.35x)
avg_h264_qpel_16_mc32_8_ssse3:                         209.2 ( 8.35x)
put_h264_qpel_16_mc12_8_c:                            2014.4 ( 1.00x)
put_h264_qpel_16_mc12_8_sse2:                          243.7 ( 8.27x)
put_h264_qpel_16_mc12_8_ssse3:                         211.5 ( 9.52x)
put_h264_qpel_16_mc32_8_c:                            1800.0 ( 1.00x)
put_h264_qpel_16_mc32_8_sse2:                          238.8 ( 7.54x)
put_h264_qpel_16_mc32_8_ssse3:                         206.7 ( 8.71x)

Reviewed-by: James Almer <[email protected]>
Signed-off-by: Andreas Rheinhardt <[email protected]>

2 files changed, 22 insertions, 17 deletions

diff --git a/libavcodec/x86/h264_qpel.c b/libavcodec/x86/h264_qpel.c
index 649bfabda8..b782d32bea 100644
--- a/libavcodec/x86/h264_qpel.c
+++ b/libavcodec/x86/h264_qpel.c
@@ -68,7 +68,8 @@ void ff_ ## OPNAME ## _h264_qpel8or16_hv1_lowpass_op_sse2(const uint8_t *src, in
 void ff_ ## OPNAME ## _h264_qpel8or16_hv2_lowpass_op_mmxext(uint8_t *dst, int16_t *tmp, int dstStride, int unused, int h);\
 void ff_ ## OPNAME ## _h264_qpel8or16_hv2_lowpass_ssse3(uint8_t *dst, int16_t *tmp, int dstStride, int tmpStride, int size);\
 void ff_ ## OPNAME ## _pixels4_l2_shift5_mmxext(uint8_t *dst, const int16_t *src16, const uint8_t *src8, int dstStride, int src8Stride, int h);\
-void ff_ ## OPNAME ## _pixels8_l2_shift5_mmxext(uint8_t *dst, const int16_t *src16, const uint8_t *src8, int dstStride, int src8Stride, int h);
+void ff_ ## OPNAME ## _pixels8_l2_shift5_mmxext(uint8_t *dst, const int16_t *src16, const uint8_t *src8, int dstStride, int src8Stride, int h);\
+void ff_ ## OPNAME ## _pixels16_l2_shift5_sse2(uint8_t *dst, const int16_t *src16, const uint8_t *src8, int dstStride, int src8Stride, int h);\
 
 DEF_QPEL(avg)
 DEF_QPEL(put)
@@ -104,12 +105,6 @@ static av_always_inline void ff_ ## OPNAME ## h264_qpel16_h_lowpass_l2_ ## MMX(u
     ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst  , src  , src2  , dstStride, src2Stride);\
     ff_ ## OPNAME ## h264_qpel8_h_lowpass_l2_ ## MMX(dst+8, src+8, src2+8, dstStride, src2Stride);\
 }\
-\
-static av_always_inline void ff_ ## OPNAME ## pixels16_l2_shift5_ ## MMX(uint8_t *dst, const int16_t *src16, const uint8_t *src8, int dstStride, int src8Stride, int h)\
-{\
-    ff_ ## OPNAME ## pixels8_l2_shift5_ ## MMX(dst  , src16  , src8  , dstStride, src8Stride, h);\
-    ff_ ## OPNAME ## pixels8_l2_shift5_ ## MMX(dst+8, src16+8, src8+8, dstStride, src8Stride, h);\
-}\
 
 
 #if ARCH_X86_64
@@ -191,6 +186,9 @@ static av_always_inline void ff_ ## OPNAME ## h264_qpel16_hv_lowpass_ ## MMX(uin
 #define ff_put_h264_qpel8or16_hv2_lowpass_sse2 ff_put_h264_qpel8or16_hv2_lowpass_mmxext
 #define ff_avg_h264_qpel8or16_hv2_lowpass_sse2 ff_avg_h264_qpel8or16_hv2_lowpass_mmxext
 
+#define ff_put_pixels16_l2_shift5_mmxext ff_put_pixels16_l2_shift5_sse2
+#define ff_avg_pixels16_l2_shift5_mmxext ff_avg_pixels16_l2_shift5_sse2
+
 #define H264_MC_V_H_HV(OPNAME, SIZE, MMX, ALIGN) \
 H264_MC_V(OPNAME, SIZE, MMX, ALIGN)\
 H264_MC_H(OPNAME, SIZE, MMX, ALIGN)\
diff --git a/libavcodec/x86/h264_qpel_8bit.asm b/libavcodec/x86/h264_qpel_8bit.asm
index 07056f1215..fefa3aff01 100644
--- a/libavcodec/x86/h264_qpel_8bit.asm
+++ b/libavcodec/x86/h264_qpel_8bit.asm
@@ -734,15 +734,19 @@ PIXELS4_L2_SHIFT5 put
 PIXELS4_L2_SHIFT5 avg
 
 
-%macro PIXELS8_L2_SHIFT5 1
-cglobal %1_pixels8_l2_shift5, 6, 6 ; dst, src16, src8, dstStride, src8Stride, h
+%macro PIXELS_L2_SHIFT5 2
+%if cpuflag(sse2)
+cglobal %1_pixels%2_l2_shift5, 6, 6, 4 ; dst, src16, src8, dstStride, src8Stride, h
+%else
+cglobal %1_pixels%2_l2_shift5, 6, 6 ; dst, src16, src8, dstStride, src8Stride, h
+%endif
     movsxdifnidn  r3, r3d
     movsxdifnidn  r4, r4d
 .loop:
-    mova          m0, [r1]
-    mova          m1, [r1+8]
-    mova          m2, [r1+48]
-    mova          m3, [r1+48+8]
+    movu          m0, [r1]
+    movu          m1, [r1+%2]
+    movu          m2, [r1+48]
+    movu          m3, [r1+48+%2]
     psraw         m0, 5
     psraw         m1, 5
     psraw         m2, 5
@@ -751,8 +755,8 @@ cglobal %1_pixels8_l2_shift5, 6, 6 ; dst, src16, src8, dstStride, src8Stride, h
     packuswb      m2, m3
     pavgb         m0, [r2]
     pavgb         m2, [r2+r4]
-    op_%1         m0, [r0], m4
-    op_%1         m2, [r0+r3], m5
+    op_%1         m0, [r0], m1
+    op_%1         m2, [r0+r3], m1
     lea           r2, [r2+2*r4]
     add           r1, 48*2
     lea           r0, [r0+2*r3]
@@ -762,9 +766,12 @@ cglobal %1_pixels8_l2_shift5, 6, 6 ; dst, src16, src8, dstStride, src8Stride, h
 %endmacro
 
 INIT_MMX mmxext
-PIXELS8_L2_SHIFT5 put
-PIXELS8_L2_SHIFT5 avg
+PIXELS_L2_SHIFT5 put, 8
+PIXELS_L2_SHIFT5 avg, 8
 
+INIT_XMM sse2
+PIXELS_L2_SHIFT5 put, 16
+PIXELS_L2_SHIFT5 avg, 16
 
 %if ARCH_X86_64
 %macro QPEL16_H_LOWPASS_L2_OP 1