Commit dfd865e5 authored by Christophe Gisquet's avatar Christophe Gisquet Committed by Michael Niedermayer

x86/synth_filter: remove the main loop when it's not needed

Signed-off-by: 's avatarMichael Niedermayer <michaelni@gmx.at>
parent 3aac5fcf
...@@ -311,9 +311,14 @@ cglobal synth_filter_inner, 0, 6 + 4 * ARCH_X86_64, 7 + 6 * ARCH_X86_64, \ ...@@ -311,9 +311,14 @@ cglobal synth_filter_inner, 0, 6 + 4 * ARCH_X86_64, 7 + 6 * ARCH_X86_64, \
sub r5q, offmp sub r5q, offmp
and r5q, -64 and r5q, -64
shl r5q, 2 shl r5q, 2
%if ARCH_X86_32 || mmsize < 32
mov OFFQ, r5q mov OFFQ, r5q
%define i r5q %define i r5q
mov i, 16 * 4 - (ARCH_X86_64 + 1) * mmsize ; main loop counter mov i, 16 * 4 - (ARCH_X86_64 + 1) * mmsize ; main loop counter
%else
%define i 0
%define OFFQ r5q
%endif
%define buf2 synth_buf2q %define buf2 synth_buf2q
%if ARCH_X86_32 %if ARCH_X86_32
...@@ -332,8 +337,10 @@ cglobal synth_filter_inner, 0, 6 + 4 * ARCH_X86_64, 7 + 6 * ARCH_X86_64, \ ...@@ -332,8 +337,10 @@ cglobal synth_filter_inner, 0, 6 + 4 * ARCH_X86_64, 7 + 6 * ARCH_X86_64, \
%define j r3q %define j r3q
mov win, windowm mov win, windowm
mov ptr1, synth_bufm mov ptr1, synth_bufm
%if ARCH_X86_32 || mmsize < 32
add win, i add win, i
add ptr1, i add ptr1, i
%endif
%else ; ARCH_X86_64 %else ; ARCH_X86_64
%define ptr1 r6q %define ptr1 r6q
%define ptr2 r7q ; must be loaded %define ptr2 r7q ; must be loaded
...@@ -349,7 +356,9 @@ cglobal synth_filter_inner, 0, 6 + 4 * ARCH_X86_64, 7 + 6 * ARCH_X86_64, \ ...@@ -349,7 +356,9 @@ cglobal synth_filter_inner, 0, 6 + 4 * ARCH_X86_64, 7 + 6 * ARCH_X86_64, \
mov ptr2, synth_bufmp mov ptr2, synth_bufmp
; prepare the inner loop counter ; prepare the inner loop counter
mov j, OFFQ mov j, OFFQ
%if ARCH_X86_32 || mmsize < 32
sub ptr2, i sub ptr2, i
%endif
.loop1: .loop1:
INNER_LOOP 0 INNER_LOOP 0
jge .loop1 jge .loop1
...@@ -394,8 +403,10 @@ cglobal synth_filter_inner, 0, 6 + 4 * ARCH_X86_64, 7 + 6 * ARCH_X86_64, \ ...@@ -394,8 +403,10 @@ cglobal synth_filter_inner, 0, 6 + 4 * ARCH_X86_64, 7 + 6 * ARCH_X86_64, \
mova [outq + i + 0 * 4 + mmsize], m7 mova [outq + i + 0 * 4 + mmsize], m7
mova [outq + i + 16 * 4 + mmsize], m8 mova [outq + i + 16 * 4 + mmsize], m8
%endif %endif
%if ARCH_X86_32 || mmsize < 32
sub i, (ARCH_X86_64 + 1) * mmsize sub i, (ARCH_X86_64 + 1) * mmsize
jge .mainloop jge .mainloop
%endif
RET RET
%endmacro %endmacro
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment