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

[rajesh at multicorewareinc.com]

# HG changeset patch
# User Rajesh Paulraj<rajesh at multicorewareinc.com>
# Date 1435219198 -19800
#      Thu Jun 25 13:29:58 2015 +0530
# Node ID a03487d6295cf89b065eff36e5c1ec4ee4253243
# Parent  b1af4c36f48a4500a4912373ebcda9a5540b5c15
asm: sse4 10bit code for sign primitive

     calSign  6.16x    356.91          2197.63

diff -r b1af4c36f48a -r a03487d6295c 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 13:29:58 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 = x265_calculateSign_sse4;
 
         LUMA_ADDAVG(sse4);
         CHROMA_420_ADDAVG(sse4);
diff -r b1af4c36f48a -r a03487d6295c 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 13:29:58 2015 +0530
@@ -40,6 +40,7 @@
 cextern pw_2
 cextern pw_1023
 cextern pb_movemask
+cextern pw_1
 
 
 ;============================================================================================================
@@ -1419,3 +1420,49 @@
 
 .end:
     RET
+
+;-----------------------------------------------------------------------------
+; void calSign(int8_t *dst, const pixel *src1, const pixel *src2, const int endX)
+;-----------------------------------------------------------------------------
+%if HIGH_BIT_DEPTH
+INIT_XMM sse4
+cglobal calculateSign, 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
+%endif
diff -r b1af4c36f48a -r a03487d6295c source/common/x86/loopfilter.h
--- a/source/common/x86/loopfilter.h	Wed Jun 24 10:36:15 2015 -0500
+++ b/source/common/x86/loopfilter.h	Thu Jun 25 13:29:58 2015 +0530
@@ -37,7 +37,8 @@
     void PFX(saoCuOrgB0_ ## cpu)(pixel* rec, const int8_t* offsetBo, int ctuWidth, int ctuHeight, intptr_t stride); \
     void PFX(saoCuStatsE2_ ## cpu)(const pixel *fenc, const pixel *rec, intptr_t stride, int8_t *upBuff1, int8_t *upBufft, int endX, int endY, int32_t *stats, int32_t *count); \
     void PFX(saoCuStatsE3_ ## cpu)(const pixel *fenc, const pixel *rec, intptr_t stride, int8_t *upBuff1, int endX, int endY, int32_t *stats, int32_t *count); \
-    void PFX(calSign_ ## cpu)(int8_t *dst, const pixel *src1, const pixel *src2, const int endX);
+    void PFX(calSign_ ## cpu)(int8_t *dst, const pixel *src1, const pixel *src2, const int endX); \
+    void PFX(calculateSign_ ## cpu)(int8_t *dst, const pixel *src1, const pixel *src2, const int endX);
 
 DECL_SAO(sse4);
 DECL_SAO(avx2);