[x265] [PATCH] asm: 10bpp code for pixel_sub_24x32

[murugan at multicorewareinc.com]

# HG changeset patch
# User Murugan Vairavel <murugan at multicorewareinc.com>
# Date 1386314852 -19800
#      Fri Dec 06 12:57:32 2013 +0530
# Node ID 5feafa073b2990c5e8a2945dd923bbfa058cad6f
# Parent  e638bc57a626d23a4276dba81023469f35d396c5
asm: 10bpp code for pixel_sub_24x32

diff -r e638bc57a626 -r 5feafa073b29 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Fri Dec 06 11:52:41 2013 +0530
+++ b/source/common/x86/asm-primitives.cpp	Fri Dec 06 12:57:32 2013 +0530
@@ -651,6 +651,7 @@
         p.chroma[X265_CSP_I420].sub_ps[LUMA_32x24] = x265_pixel_sub_ps_16x12_sse2;
         p.chroma[X265_CSP_I420].sub_ps[LUMA_32x32] = x265_pixel_sub_ps_16x16_sse2;
         p.chroma[X265_CSP_I420].sub_ps[LUMA_32x64] = x265_pixel_sub_ps_16x32_sse2;
+        p.chroma[X265_CSP_I420].sub_ps[LUMA_48x64] = x265_pixel_sub_ps_24x32_sse2;
         p.luma_sub_ps[LUMA_16x64] = x265_pixel_sub_ps_16x64_sse2;
     }
     if (cpuMask & X265_CPU_SSSE3)
diff -r e638bc57a626 -r 5feafa073b29 source/common/x86/pixel-util8.asm
--- a/source/common/x86/pixel-util8.asm	Fri Dec 06 11:52:41 2013 +0530
+++ b/source/common/x86/pixel-util8.asm	Fri Dec 06 12:57:32 2013 +0530
@@ -3067,15 +3067,46 @@
 ; void pixel_sub_ps_c_%1x%2(int16_t *dest, intptr_t destride, pixel *src0, pixel *src1, intptr_t srcstride0, intptr_t srcstride1);
 ;-----------------------------------------------------------------------------
 %macro PIXELSUB_PS_W24_H2 2
-INIT_XMM sse4
+%if HIGH_BIT_DEPTH
+cglobal pixel_sub_ps_%1x%2, 6, 7, 6, dest, deststride, src0, src1, srcstride0, srcstride1
+    add     r1,     r1
+    mov     r6d,    %2/2
+    add     r4,     r4
+    add     r5,     r5
+.loop
+    movu     m0,    [r2]
+    movu     m1,    [r3]
+    movu     m2,    [r2 + 16]
+    movu     m3,    [r3 + 16]
+    movu     m4,    [r2 + 32]
+    movu     m5,    [r3 + 32]
+    psubw    m0,    m1
+    psubw    m2,    m3
+    psubw    m4,    m5
+
+    movu    [r0],              m0
+    movu    [r0 + 16],         m2
+    movu    [r0 + 32],         m4
+
+    movu     m0,    [r2 + r4]
+    movu     m1,    [r3 + r5]
+    movu     m2,    [r2 + r4 + 16]
+    movu     m3,    [r3 + r5 + 16]
+    movu     m4,    [r2 + r4 + 32]
+    movu     m5,    [r3 + r5 + 32]
+    psubw    m0,    m1
+    psubw    m2,    m3
+    psubw    m4,    m5
+
+    movu    [r0 + r1],         m0
+    movu    [r0 + r1 + 16],    m2
+    movu    [r0 + r1 + 32],    m4
+%else
 cglobal pixel_sub_ps_%1x%2, 6, 7, 7, dest, deststride, src0, src1, srcstride0, srcstride1
-
-add    r1,     r1
-mov    r6d,    %2/2
-pxor   m6,     m6
-
+    add    r1,     r1
+    mov    r6d,    %2/2
+    pxor   m6,     m6
 .loop
-
     movu         m1,    [r2]
     pmovzxbw     m0,    m1
     punpckhbw    m1,    m6
@@ -3086,7 +3117,6 @@
     punpckhbw    m4,    m6
     movh         m5,    [r3 + 16]
     pmovzxbw     m5,    m5
-
     psubw        m0,    m3
     psubw        m1,    m4
     psubw        m2,    m5
@@ -3105,7 +3135,6 @@
     punpckhbw    m4,    m6
     movh         m5,    [r3 + r5 + 16]
     pmovzxbw     m5,    m5
-
     psubw        m0,    m3
     psubw        m1,    m4
     psubw        m2,    m5
@@ -3113,19 +3142,24 @@
     movu    [r0 + r1],         m0
     movu    [r0 + r1 + 16],    m1
     movu    [r0 + r1 + 32],    m2
-
+%endif
+
+    dec    r6d
     lea    r2,    [r2 + 2 * r4]
     lea    r3,    [r3 + 2 * r5]
     lea    r0,    [r0 + 2 * r1]
-
-    dec    r6d
-
-jnz    .loop
+    jnz    .loop
 
 RET
 %endmacro
 
+%if HIGH_BIT_DEPTH
+INIT_XMM sse2
 PIXELSUB_PS_W24_H2 24, 32
+%else
+INIT_XMM sse4
+PIXELSUB_PS_W24_H2 24, 32
+%endif
 
 ;-----------------------------------------------------------------------------
 ; void pixel_sub_ps_c_%1x%2(int16_t *dest, intptr_t destride, pixel *src0, pixel *src1, intptr_t srcstride0, intptr_t srcstride1);