[x265] [PATCH Review Only] chroma_hpp[32x32] for colorspace i420 in avx2: improve 6189c->4856c

[aasaipriya at multicorewareinc.com]

# HG changeset patch
# User Aasaipriya Chandran <aasaipriya at multicorewareinc.com>
# Date 1417685186 -19800
#      Thu Dec 04 14:56:26 2014 +0530
# Node ID 0813637ebb45b461dc882c5d578540273a0924af
# Parent  ada9f34f4a7cc59e101d8af0e6dcac016048308c
chroma_hpp[32x32] for colorspace i420 in avx2: improve 6189c->4856c

diff -r ada9f34f4a7c -r 0813637ebb45 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Thu Dec 04 14:48:06 2014 +0530
+++ b/source/common/x86/asm-primitives.cpp	Thu Dec 04 14:56:26 2014 +0530
@@ -1824,6 +1824,8 @@
 
         p.chroma[X265_CSP_I420].filter_hpp[CHROMA_16x16] = x265_interp_4tap_horiz_pp_16x16_avx2;
 
+        p.chroma[X265_CSP_I420].filter_hpp[CHROMA_32x32] = x265_interp_4tap_horiz_pp_32x32_avx2;
+
         p.luma_vpp[LUMA_4x4] = x265_interp_8tap_vert_pp_4x4_avx2;
     }
 #endif // if HIGH_BIT_DEPTH
diff -r ada9f34f4a7c -r 0813637ebb45 source/common/x86/ipfilter8.asm
--- a/source/common/x86/ipfilter8.asm	Thu Dec 04 14:48:06 2014 +0530
+++ b/source/common/x86/ipfilter8.asm	Thu Dec 04 14:56:26 2014 +0530
@@ -1290,6 +1290,67 @@
     jnz .loop
     RET
 
+INIT_YMM avx2
+cglobal interp_4tap_horiz_pp_32x32, 4,6,6
+    mov               r4d,          r4m
+
+%ifdef PIC
+    lea               r5,           [tab_ChromaCoeff]
+    vpbroadcastd      m0,           [r5 + r4 * 4]
+%else
+    vpbroadcastd      m0,           [tab_ChromaCoeff + r4 * 4]
+%endif
+
+    movu              m1,           [tab_Tm]
+    vpbroadcastd      m2,           [pw_1]
+
+    ; register map
+    ; m0 - interpolate coeff
+    ; m1 - shuffle order table
+    ; m2 - constant word 1
+
+    mov               r4d,          32
+    sub               r0,           1
+    
+    .loop
+    ; Row 0
+    vbroadcasti128    m3,           [r0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 8]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    pmulhrsw          m3,           [pw_512]
+    vextracti128      xm4,          m3,     1
+    packuswb          xm3,          xm4
+    pshufd            xm3,          xm3,     11011000b
+    movu              [r2],         xm3
+
+    vbroadcasti128    m3,           [r0 + 16]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m3,           m1
+    pmaddubsw         m3,           m0
+    pmaddwd           m3,           m2
+    vbroadcasti128    m4,           [r0 + 24]                        ; [x x x x x A 9 8 7 6 5 4 3 2 1 0]
+    pshufb            m4,           m1
+    pmaddubsw         m4,           m0
+    pmaddwd           m4,           m2
+
+    packssdw          m3,           m4
+    pmulhrsw          m3,           [pw_512]
+    vextracti128      xm4,          m3,     1
+    packuswb          xm3,          xm4
+    pshufd            xm3,          xm3,    11011000b
+    movu              [r2 + 16],    xm3
+
+    lea               r0,           [r0 + r1]
+    lea               r2,           [r2 + r3]
+    dec r4d
+    jnz .loop
+    RET
 
 ;--------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_horiz_pp_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)