[x265] [PATCH Review Only] Generic Luma filter macro to generate necessary code for any width and height
nabajit at multicorewareinc.com
nabajit at multicorewareinc.com
Fri Oct 11 17:43:08 CEST 2013
# HG changeset patch
# User Nabajit Deka
# Date 1381506174 -19800
# Fri Oct 11 21:12:54 2013 +0530
# Node ID edc41b0972547e4493898ddf021248d8c204e200
# Parent 6054812606c2d45306824a30298f87410f2994e3
Generic Luma filter macro to generate necessary code for any width and height.
diff -r 6054812606c2 -r edc41b097254 source/common/x86/ipfilter8.asm
--- a/source/common/x86/ipfilter8.asm Fri Oct 11 20:53:15 2013 +0530
+++ b/source/common/x86/ipfilter8.asm Fri Oct 11 21:12:54 2013 +0530
@@ -1406,3 +1406,108 @@
dec r4d
jnz .loop
RET
+
+SECTION_RODATA 32
+
+tab_Tm: db 0, 1, 2, 3, 4, 5, 6, 7, 1, 2, 3, 4, 5, 6, 7, 8
+ db 2, 3, 4, 5, 6, 7, 8, 9, 3, 4, 5, 6, 7, 8, 9, 10
+ db 4, 5, 6, 7, 8, 9, 10, 11, 5, 6, 7, 8, 9, 10, 11, 12
+ db 6, 7, 8, 9, 10, 11, 12, 13, 7, 8, 9, 10, 11, 12, 13, 14
+
+tab_c_512: times 8 dw 512
+
+SECTION .text
+
+%macro FILTER_H8_W8 3
+ movu %1, [r0 - 3 + r5]
+ pshufb %2, %1, [tab_Tm]
+ pmaddubsw %2, m3
+ pshufb m7, %1, [tab_Tm + 16]
+ pmaddubsw m7, m3
+ phaddw %2, m7
+ pshufb m7, %1, [tab_Tm + 32]
+ pmaddubsw m7, m3
+ pshufb %1, %1, [tab_Tm + 48]
+ pmaddubsw %1, m3
+ phaddw m7, %1
+ phaddw %2, m7
+ pmulhrsw %2, %3
+ packuswb %2, %2
+ movh [r2 + r5], %2
+%endmacro
+
+%macro FILTER_H8_W4 3
+ movu %1, [r0 - 3 + r5]
+ pshufb %2, %1, [tab_Tm]
+ pmaddubsw %2, m3
+ pshufb m7, %1, [tab_Tm + 16]
+ pmaddubsw m7, m3
+ phaddw %2, m7
+ phaddw %2, %2
+ pmulhrsw %2, %3
+ packuswb %2, %2
+ movd [r2 + r5], %2
+%endmacro
+
+%macro FILTER_H8_W1 3
+ movu %1, [r0 - 3 + r5]
+ pshufb %2, %1, [tab_Tm]
+ pmaddubsw %2, m3
+ phaddw %2, %2
+ phaddw %2, %2
+ pmulhrsw %2, %3
+ packuswb %2, %2
+ pextrb [r2 + r5], %2, 0
+%endmacro
+
+;-----------------------------------------------------------------------------
+; void filterHorizontal_p_p_luma_%1x%2(pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int width, int height, short const *coeff)
+;-----------------------------------------------------------------------------
+%macro IPFILTER_LUMA 2
+cglobal filterHorizontal_p_p_luma_%1x%2, 4, 6, 5
+ mov r4, r6m
+ movu m3, [r4]
+ packsswb m3, m3
+ mova m2, [tab_c_512]
+ mov r4, %2
+
+.loop
+ xor r5, r5
+%rep %1 / 8
+ FILTER_H8_W8 m0, m1, m2
+ add r5, 8
+%endrep
+
+%rep (%1 % 8) / 4
+ FILTER_H8_W4 m0, m1, m2
+ add r5, 8
+%endrep
+
+ %rep(%1 % 4)
+ FILTER_H8_W1 m0, m1, m2
+ %endrep
+
+ add r0, r1
+ add r2, r3
+
+ dec r4d
+ jnz .loop
+ RET
+%endmacro
+
+IPFILTER_LUMA 12, 16
+IPFILTER_LUMA 12, 17
+IPFILTER_LUMA 13, 16
+IPFILTER_LUMA 16, 12
+IPFILTER_LUMA 16, 13
+IPFILTER_LUMA 16, 16
+IPFILTER_LUMA 16, 17
+IPFILTER_LUMA 16, 32
+IPFILTER_LUMA 16, 33
+IPFILTER_LUMA 16, 4
+IPFILTER_LUMA 16, 5
+IPFILTER_LUMA 16, 64
+IPFILTER_LUMA 16, 8
+IPFILTER_LUMA 16, 9
+IPFILTER_LUMA 64, 64
+IPFILTER_LUMA 32, 32
More information about the x265-devel
mailing list