[x265] [PATCH] asm: sse4 10bit code for sign primitive

[rajesh at multicorewareinc.com]

# HG changeset patch
# User Rajesh Paulraj<rajesh at multicorewareinc.com>
# Date 1435229751 -19800
#      Thu Jun 25 16:25:51 2015 +0530
# Node ID a0de1e88f3b1a10d6f8cf656a95e6ec37e1bc134
# Parent  b1af4c36f48a4500a4912373ebcda9a5540b5c15
asm: sse4 10bit code for sign primitive

     calSign  6.16x    356.91          2197.63

diff -r b1af4c36f48a -r a0de1e88f3b1 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Wed Jun 24 10:36:15 2015 -0500
+++ b/source/common/x86/asm-primitives.cpp	Thu Jun 25 16:25:51 2015 +0530
@@ -1097,6 +1097,7 @@
         p.saoCuOrgE3[0] = PFX(saoCuOrgE3_sse4);
         p.saoCuOrgE3[1] = PFX(saoCuOrgE3_sse4);
         p.saoCuOrgB0 = PFX(saoCuOrgB0_sse4);
+        p.sign = PFX(calSign_sse4);
 
         LUMA_ADDAVG(sse4);
         CHROMA_420_ADDAVG(sse4);
diff -r b1af4c36f48a -r a0de1e88f3b1 source/common/x86/loopfilter.asm
--- a/source/common/x86/loopfilter.asm	Wed Jun 24 10:36:15 2015 -0500
+++ b/source/common/x86/loopfilter.asm	Thu Jun 25 16:25:51 2015 +0530
@@ -40,6 +40,7 @@
 cextern pw_2
 cextern pw_1023
 cextern pb_movemask
+cextern pw_1
 
 
 ;============================================================================================================
@@ -1321,6 +1322,48 @@
 ; void calSign(int8_t *dst, const Pixel *src1, const Pixel *src2, const int width)
 ;============================================================================================================
 INIT_XMM sse4
+%if HIGH_BIT_DEPTH
+cglobal calSign, 4, 7, 5
+    mova            m0, [pw_1]
+    mov             r4d, r3d
+    shr             r3d, 4
+    add             r3d, 1
+    mov             r5, r0
+    movu            m4, [r0 + r4]
+.loop
+    movu            m1, [r1]        ; m2 = pRec[x]
+    movu            m2, [r2]        ; m3 = pTmpU[x]
+
+    pcmpgtw         m3, m1, m2
+    pcmpgtw         m2, m1
+    pand            m3, m0
+    por             m3, m2
+    packsswb        m3, m3
+    movh            [r0], xm3
+
+    movu            m1, [r1 + 16]   ; m2 = pRec[x]
+    movu            m2, [r2 + 16]   ; m3 = pTmpU[x]
+
+    pcmpgtw         m3, m1, m2
+    pcmpgtw         m2, m1
+    pand            m3, m0
+    por             m3, m2
+    packsswb        m3, m3
+    movh            [r0 + 8], xm3
+
+    add             r0, 16
+    add             r1, 32
+    add             r2, 32
+    dec             r3d
+    jnz             .loop
+
+    mov             r6, r0
+    sub             r6, r5
+    sub             r4, r6
+    movu            [r0 + r4], m4
+    RET
+%else ; HIGH_BIT_DEPTH
+
 cglobal calSign, 4,5,6
     mova        m0,     [pb_128]
     mova        m1,     [pb_1]
@@ -1369,6 +1412,7 @@
 
 .end:
     RET
+%endif
 
 INIT_YMM avx2
 cglobal calSign, 4, 5, 6