[x265] [PATCH] asm: assembly code for pixel_sa8d_16x64

[yuvaraj at multicorewareinc.com]

# HG changeset patch
# User Yuvaraj Venkatesh <yuvaraj at multicorewareinc.com>
# Date 1385022121 -19800
#      Thu Nov 21 13:52:01 2013 +0530
# Node ID 184896d5cd2086e0f6a3017b1a88d5605255405c
# Parent  5a49867b288cc485dcc450344eb2c9f2cb6c7b92
asm: assembly code for pixel_sa8d_16x64

diff -r 5a49867b288c -r 184896d5cd20 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Thu Nov 21 13:32:11 2013 +0530
+++ b/source/common/x86/asm-primitives.cpp	Thu Nov 21 13:52:01 2013 +0530
@@ -107,7 +107,7 @@
     p.sa8d_inter[LUMA_64x48] = x265_pixel_sa8d_64x48_ ## cpu; \
     p.sa8d_inter[LUMA_48x64] = x265_pixel_sa8d_48x64_ ## cpu; \
     p.sa8d_inter[LUMA_64x16] = x265_pixel_sa8d_64x16_ ## cpu; \
-    p.sa8d_inter[LUMA_16x64] = cmp<16, 64, 16, 16, x265_pixel_sa8d_16x16_ ## cpu>
+    p.sa8d_inter[LUMA_16x64] = x265_pixel_sa8d_16x64_ ## cpu;
 
 #define PIXEL_AVG(cpu) \
     p.pixelavg_pp[LUMA_64x64] = x265_pixel_avg_64x64_ ## cpu; \
diff -r 5a49867b288c -r 184896d5cd20 source/common/x86/pixel-a.asm
--- a/source/common/x86/pixel-a.asm	Thu Nov 21 13:32:11 2013 +0530
+++ b/source/common/x86/pixel-a.asm	Thu Nov 21 13:52:01 2013 +0530
@@ -3598,6 +3598,27 @@
     movd eax, m12
     RET
 
+cglobal pixel_sa8d_16x64, 4,8,12
+    FIX_STRIDES r1, r3
+    lea  r4, [3*r1]
+    lea  r5, [3*r3]
+    pxor m12, m12
+%if vertical == 0
+    mova m7, [hmul_8p]
+%endif
+    SA8D_16x16
+    lea r0, [r0+8*r1]
+    lea r2, [r2+8*r3]
+    SA8D_16x16
+    lea r0, [r0+8*r1]
+    lea r2, [r2+8*r3]
+    SA8D_16x16
+    lea r0, [r0+8*r1]
+    lea r2, [r2+8*r3]
+    SA8D_16x16
+    movd eax, m12
+    RET
+
 cglobal pixel_sa8d_32x16, 4,8,12
     FIX_STRIDES r1, r3
     lea  r4, [3*r1]
@@ -4349,6 +4370,121 @@
     mov esp, r6
     RET
 
+cglobal pixel_sa8d_16x64, 4,7,8
+    FIX_STRIDES r1, r3
+    mov  r6, esp
+    and  esp, ~15
+    sub  esp, 64
+
+    lea  r4, [r1 + 2*r1]
+    lea  r5, [r3 + 2*r3]
+    call pixel_sa8d_8x8_internal2
+    mova [rsp+48], m0
+    call pixel_sa8d_8x8_internal2
+    SA8D_INTER
+    mova [esp+48], m0
+
+    mov  r0, [r6+20]
+    mov  r2, [r6+28]
+    add  r0, 8*SIZEOF_PIXEL
+    add  r2, 8*SIZEOF_PIXEL
+    call pixel_sa8d_8x8_internal2
+    SA8D_INTER
+    mova [esp+48], m0
+    call pixel_sa8d_8x8_internal2
+    paddusw m0, [esp+48]
+    HADDUW m0, m1
+    movd r4d, m0
+    add  r4d, 1
+    shr  r4d, 1
+    mov dword [esp+36], r4d
+
+    mov  r0, [r6+20]
+    mov  r2, [r6+28]
+    lea  r0, [r0 + r1*8]
+    lea  r2, [r2 + r3*8]
+    lea  r0, [r0 + r1*8]
+    lea  r2, [r2 + r3*8]
+    mov  [r6+20], r0
+    mov  [r6+28], r2
+
+    lea  r4, [r1 + 2*r1]
+    call pixel_sa8d_8x8_internal2
+    mova [esp+48], m0
+    call pixel_sa8d_8x8_internal2
+    SA8D_INTER
+    mova [esp+48], m0
+
+    mov  r0, [r6+20]
+    mov  r2, [r6+28]
+    add  r0, 8*SIZEOF_PIXEL
+    add  r2, 8*SIZEOF_PIXEL
+    call pixel_sa8d_8x8_internal2
+    SA8D_INTER
+    mova [esp+64-mmsize], m0
+    call pixel_sa8d_8x8_internal2
+    AVG_16x16
+
+    mov  r0, [r6+20]
+    mov  r2, [r6+28]
+    lea  r0, [r0 + r1*8]
+    lea  r2, [r2 + r3*8]
+    lea  r0, [r0 + r1*8]
+    lea  r2, [r2 + r3*8]
+    mov  [r6+20], r0
+    mov  [r6+28], r2
+
+    lea  r4, [r1 + 2*r1]
+    call pixel_sa8d_8x8_internal2
+    mova [esp+48], m0
+    call pixel_sa8d_8x8_internal2
+    SA8D_INTER
+    mova [esp+48], m0
+
+    mov  r0, [r6+20]
+    mov  r2, [r6+28]
+    add  r0, 8*SIZEOF_PIXEL
+    add  r2, 8*SIZEOF_PIXEL
+    call pixel_sa8d_8x8_internal2
+    SA8D_INTER
+    mova [esp+64-mmsize], m0
+    call pixel_sa8d_8x8_internal2
+    AVG_16x16
+
+    mov  r0, [r6+20]
+    mov  r2, [r6+28]
+    lea  r0, [r0 + r1*8]
+    lea  r2, [r2 + r3*8]
+    lea  r0, [r0 + r1*8]
+    lea  r2, [r2 + r3*8]
+    mov  [r6+20], r0
+    mov  [r6+28], r2
+
+    lea  r4, [r1 + 2*r1]
+    call pixel_sa8d_8x8_internal2
+    mova [esp+48], m0
+    call pixel_sa8d_8x8_internal2
+    SA8D_INTER
+    mova [esp+48], m0
+
+    mov  r0, [r6+20]
+    mov  r2, [r6+28]
+    add  r0, 8*SIZEOF_PIXEL
+    add  r2, 8*SIZEOF_PIXEL
+    call pixel_sa8d_8x8_internal2
+    SA8D_INTER
+    mova [esp+64-mmsize], m0
+    call pixel_sa8d_8x8_internal2
+    paddusw m0, [esp+48]
+    HADDUW m0, m1
+    movd r4d, m0
+    add  r4d, 1
+    shr  r4d, 1
+    add r4d, dword [esp+36]
+    mov eax, r4d
+    mov esp, r6
+    RET
+
 cglobal pixel_sa8d_32x16, 4,7,8
     FIX_STRIDES r1, r3
     mov  r6, esp