# HG changeset patch
# User David T Yuen <[email protected]>
# Date 1427503866 25200
# Node ID d7fee58a8466d6862e72cced5a4aa12e37294d4e
# Parent dc7a6c5fbda13173004e6193f185ea857d7a22c8
asm: intra_pred_ang4_10_sse2
This is backported from sse4 code and replaces c code.
64-bit
./test/TestBench --testbench intrapred | grep "intra_ang_4x4\[10\]"
intra_ang_4x4[10] 6.16x 215.12 1326.19
32-bit
./test/TestBench --testbench intrapred | grep "intra_ang_4x4\[10\]"
intra_ang_4x4[10] 6.91x 290.06 2004.46
diff -r dc7a6c5fbda1 -r d7fee58a8466 source/common/x86/asm-primitives.cpp
--- a/source/common/x86/asm-primitives.cpp Thu Mar 26 11:24:23 2015 -0500
+++ b/source/common/x86/asm-primitives.cpp Fri Mar 27 17:51:06 2015 -0700
@@ -1204,6 +1204,7 @@
p.cu[BLOCK_4x4].intra_pred[7] = x265_intra_pred_ang4_7_sse2;
p.cu[BLOCK_4x4].intra_pred[8] = x265_intra_pred_ang4_8_sse2;
p.cu[BLOCK_4x4].intra_pred[9] = x265_intra_pred_ang4_9_sse2;
+ p.cu[BLOCK_4x4].intra_pred[10] = x265_intra_pred_ang4_10_sse2;
p.cu[BLOCK_4x4].calcresidual = x265_getResidual4_sse2;
p.cu[BLOCK_8x8].calcresidual = x265_getResidual8_sse2;
diff -r dc7a6c5fbda1 -r d7fee58a8466 source/common/x86/intrapred.h
--- a/source/common/x86/intrapred.h Thu Mar 26 11:24:23 2015 -0500
+++ b/source/common/x86/intrapred.h Fri Mar 27 17:51:06 2015 -0700
@@ -55,6 +55,7 @@
DECL_ANG(4, 7, sse2);
DECL_ANG(4, 8, sse2);
DECL_ANG(4, 9, sse2);
+DECL_ANG(4, 10, sse2);
DECL_ANG(4, 2, ssse3);
DECL_ANG(4, 3, sse4);
diff -r dc7a6c5fbda1 -r d7fee58a8466 source/common/x86/intrapred8.asm
--- a/source/common/x86/intrapred8.asm Thu Mar 26 11:24:23 2015 -0500
+++ b/source/common/x86/intrapred8.asm Fri Mar 27 17:51:06 2015 -0700
@@ -1482,6 +1482,36 @@
mova m7, [r3 + 4 * 16] ; [ 8]
jmp mangle(private_prefix %+ _ %+ intra_pred_ang4_3 %+ SUFFIX %+
.do_filter4x4)
+cglobal intra_pred_ang4_10, 5,5,4
+ movd m0, [r2 + 9] ; [8 7 6 5 4 3 2 1]
+ punpcklbw m0, m0
+ punpcklwd m0, m0
+ pshufd m1, m0, 1
+ movhlps m2, m0
+ pshufd m3, m0, 3
+ movd [r0 + r1], m1
+ movd [r0 + r1 * 2], m2
+ lea r1, [r1 * 3]
+ movd [r0 + r1], m3
+ cmp r4m, byte 0
+ jz .quit
+
+ ; filter
+ pxor m3, m3
+ punpcklbw m0, m3
+ movh m1, [r2] ; [4 3 2 1 0]
+ punpcklbw m1, m3
+ pshuflw m2, m1, 0x00
+ psrldq m1, 2
+ psubw m1, m2
+ psraw m1, 1
+ paddw m0, m1
+ packuswb m0, m0
+
+.quit:
+ movd [r0], m0
+ RET
+
;---------------------------------------------------------------------------------------------
; void intra_pred_dc(pixel* dst, intptr_t dstStride, pixel *srcPix, int
dirMode, int bFilter)
;---------------------------------------------------------------------------------------------
_______________________________________________
x265-devel mailing list
[email protected]
https://mailman.videolan.org/listinfo/x265-devel
