[x265] [PATCH] asm: luma_vpp[8x16, 8x32] in avx2: improve 1139c->774c, 1968c->1452c

[Divya Manivannan]

# HG changeset patch
# User Divya Manivannan <divya at multicorewareinc.com>
# Date 1416467833 -19800
#      Thu Nov 20 12:47:13 2014 +0530
# Node ID 02bc16b116ebfdb61c91a516291f1b19b259bcbf
# Parent  a6e1b125424acc727f9ba464ccc530550203d407
asm: luma_vpp[8x16, 8x32] in avx2: improve 1139c->774c, 1968c->1452c

diff -r a6e1b125424a -r 02bc16b116eb source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Thu Nov 20 12:23:05 2014 +0530
+++ b/source/common/x86/asm-primitives.cpp	Thu Nov 20 12:47:13 2014 +0530
@@ -1802,6 +1802,8 @@
         p.luma_hpp[LUMA_4x4] = x265_interp_8tap_horiz_pp_4x4_avx2;
         p.luma_vpp[LUMA_4x4] = x265_interp_8tap_vert_pp_4x4_avx2;
         p.luma_vpp[LUMA_8x8] = x265_interp_8tap_vert_pp_8x8_avx2;
+        p.luma_vpp[LUMA_8x16] = x265_interp_8tap_vert_pp_8x16_avx2;
+        p.luma_vpp[LUMA_8x32] = x265_interp_8tap_vert_pp_8x32_avx2;
     }
 #endif // if HIGH_BIT_DEPTH
 }
diff -r a6e1b125424a -r 02bc16b116eb source/common/x86/ipfilter8.asm
--- a/source/common/x86/ipfilter8.asm	Thu Nov 20 12:23:05 2014 +0530
+++ b/source/common/x86/ipfilter8.asm	Thu Nov 20 12:47:13 2014 +0530
@@ -3729,6 +3729,53 @@
     movhps          [r2 + r4], xm4
     RET
 
+%macro FILTER_VER_LUMA_AVX2_8xN 2
+INIT_YMM avx2
+cglobal interp_8tap_vert_pp_%1x%2, 4, 7, 8, 0-gprsize
+    mov             r4d, r4m
+    shl             r4d, 7
+
+%ifdef PIC
+    lea             r5, [tab_LumaCoeffVer_32]
+    add             r5, r4
+%else
+    lea             r5, [tab_LumaCoeffVer_32 + r4]
+%endif
+
+    lea             r4, [r1 * 3]
+    sub             r0, r4
+    lea             r6, [r1 * 4]
+    mov             word [rsp], %2 / 8
+    mova            m7, [pw_512]
+
+.loop:
+    PROCESS_LUMA_AVX2_W8_8R
+    pmulhrsw        m5, m7                          ; m5 = word: row 0, row 1
+    pmulhrsw        m2, m7                          ; m2 = word: row 2, row 3
+    pmulhrsw        m1, m7                          ; m1 = word: row 4, row 5
+    pmulhrsw        m4, m7                          ; m4 = word: row 6, row 7
+    packuswb        m5, m2
+    packuswb        m1, m4
+    vextracti128    xm2, m5, 1
+    vextracti128    xm4, m1, 1
+    movq            [r2], xm5
+    movq            [r2 + r3], xm2
+    lea             r2, [r2 + r3 * 2]
+    movhps          [r2], xm5
+    movhps          [r2 + r3], xm2
+    lea             r2, [r2 + r3 * 2]
+    movq            [r2], xm1
+    movq            [r2 + r3], xm4
+    lea             r2, [r2 + r3 * 2]
+    movhps          [r2], xm1
+    movhps          [r2 + r3], xm4
+    lea             r2, [r2 + r3 * 2]
+    sub             r0, r6
+    dec             word [rsp]
+    jnz             .loop
+    RET
+%endmacro
+
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_pp_8x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
@@ -3743,11 +3790,13 @@
 ; void interp_8tap_vert_pp_8x16(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
 FILTER_VER_LUMA_8xN 8, 16, pp
+FILTER_VER_LUMA_AVX2_8xN 8, 16
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_pp_8x32(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)
 ;-------------------------------------------------------------------------------------------------------------
 FILTER_VER_LUMA_8xN 8, 32, pp
+FILTER_VER_LUMA_AVX2_8xN 8, 32
 
 ;-------------------------------------------------------------------------------------------------------------
 ; void interp_8tap_vert_ps_8x4(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx)