[x265] [PATCH] asm: intrapred dc4 sse2

[dtyx265 at gmail.com]

# HG changeset patch
# User David T Yuen <dtyx265 at gmail.com>
# Date 1424810386 28800
# Node ID 8528fea56a4202c871544621d5806ee686e671fc
# Parent  7252c10278a1e1aeddbe489180aad35196262919
asm: intrapred dc4 sse2

This replaces c code for systems using ssse3 to sse2 processors
The code is backported from intrapred dc4 sse4

./test/TestBench --testbench intrapred
Testing only harnesses that match name <intrapred>
Using random seed 54ECE101 8bpp
Testing primitives: SSE2
Testing primitives: SSE3

Test performance improvement with full optimizations
== intrapred primitives ==
intra_dc_4x4[f=0]	1.26x 	 192.96   	 242.35
intra_dc_4x4[f=1]	1.25x 	 409.99   	 512.69

diff -r 7252c10278a1 -r 8528fea56a42 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp	Mon Feb 23 23:11:49 2015 -0600
+++ b/source/common/x86/asm-primitives.cpp	Tue Feb 24 12:39:46 2015 -0800
@@ -1071,6 +1071,8 @@
         ALL_LUMA_TU_S(cpy1Dto2D_shr, cpy1Dto2D_shr_, sse2);
         ALL_LUMA_TU_S(ssd_s, pixel_ssd_s_, sse2);
 
+        p.cu[BLOCK_4x4].intra_pred[DC_IDX] = x265_intra_pred_dc4_sse2;
+
         p.cu[BLOCK_4x4].calcresidual = x265_getResidual4_sse2;
         p.cu[BLOCK_8x8].calcresidual = x265_getResidual8_sse2;
 
diff -r 7252c10278a1 -r 8528fea56a42 source/common/x86/intrapred.h
--- a/source/common/x86/intrapred.h	Mon Feb 23 23:11:49 2015 -0600
+++ b/source/common/x86/intrapred.h	Tue Feb 24 12:39:46 2015 -0800
@@ -26,6 +26,7 @@
 #ifndef X265_INTRAPRED_H
 #define X265_INTRAPRED_H
 
+void x265_intra_pred_dc4_sse2 (pixel* dst, intptr_t dstStride, const pixel*srcPix, int, int filter);
 void x265_intra_pred_dc4_sse4 (pixel* dst, intptr_t dstStride, const pixel*srcPix, int, int filter);
 void x265_intra_pred_dc8_sse4(pixel* dst, intptr_t dstStride, const pixel* srcPix, int, int filter);
 void x265_intra_pred_dc16_sse4(pixel* dst, intptr_t dstStride, const pixel* srcPix, int, int filter);
diff -r 7252c10278a1 -r 8528fea56a42 source/common/x86/intrapred8.asm
--- a/source/common/x86/intrapred8.asm	Mon Feb 23 23:11:49 2015 -0600
+++ b/source/common/x86/intrapred8.asm	Tue Feb 24 12:39:46 2015 -0800
@@ -65,6 +65,8 @@
 pw_planar32_L:        dw 31, 30, 29, 28, 27, 26, 25, 24
 pw_planar32_H:        dw 23, 22, 21, 20, 19, 18, 17, 16
 
+pw_257:         times 8 dw 257
+
 const ang_table
 %assign x 0
 %rep 32
@@ -89,6 +91,71 @@
 ;---------------------------------------------------------------------------------------------
 ; void intra_pred_dc(pixel* dst, intptr_t dstStride, pixel *srcPix, int dirMode, int bFilter)
 ;---------------------------------------------------------------------------------------------
+INIT_XMM sse2
+cglobal intra_pred_dc4, 5,5,3
+    inc         r2
+    pxor        m0, m0
+    movu        m1, [r2]
+    pshufd      m1, m1, 0xF8
+    psadbw      m1, m0              ; m1 = sum
+
+    test        r4d, r4d
+
+    paddw       m1, [pw_4]
+    psraw       m1, 3
+    movd        r4d, m1             ; r4d = dc_val
+    pmullw      m1, [pw_257]
+    pshuflw     m1, m1, 0x00
+
+    ; store DC 4x4
+    lea         r3, [r1 * 3]
+    movd        [r0], m1
+    movd        [r0 + r1], m1
+    movd        [r0 + r1 * 2], m1
+    movd        [r0 + r3], m1
+
+    ; do DC filter
+    jz         .end
+    lea         r3d, [r4d * 2 + 2]  ; r3d = DC * 2 + 2
+    add         r4d, r3d            ; r4d = DC * 3 + 2
+    movd        m1, r4d
+    pshuflw     m1, m1, 0           ; m1 = pixDCx3
+
+    ; filter top
+    movd        m2, [r2]
+    punpcklbw   m2, m0
+    paddw       m2, m1
+    psraw       m2, 2
+    packuswb    m2, m2
+    movd        [r0], m2            ; overwrite top-left pixel, we will update it later
+
+    ; filter top-left
+    movzx       r4d, byte [r2 + 8]
+    add         r3d, r4d
+    movzx       r4d, byte [r2]
+    add         r3d, r4d
+    shr         r3d, 2
+    mov         [r0], r3b
+
+    ; filter left
+    add         r0, r1
+    movq        m2, [r2 + 9]
+    punpcklbw   m2, m0
+    paddw       m2, m1
+    psraw       m2, 2
+    packuswb    m2, m2
+    movq        r4, m2
+    mov         [r0], r4b
+    shr         r4, 8
+    mov         [r0 + r1], r4b
+    shr         r4, 8
+    mov         [r0 + r1 * 2], r4b
+.end:
+    RET
+
+;---------------------------------------------------------------------------------------------
+; void intra_pred_dc(pixel* dst, intptr_t dstStride, pixel *srcPix, int dirMode, int bFilter)
+;---------------------------------------------------------------------------------------------
 INIT_XMM sse4
 cglobal intra_pred_dc4, 5,5,3
     inc         r2