1
0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-12-23 12:43:46 +02:00

Merge remote-tracking branch 'qatar/master'

* qatar/master:
  x86: lavr: use the x86inc.asm automatic stack alignment in mixing functions

Merged-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
Michael Niedermayer 2013-01-06 13:20:14 +01:00
commit 7456164b50

View File

@ -266,29 +266,20 @@ MIX_1_TO_2_S16P_FLT
%else
%assign matrix_elements_stack 0
%endif
%assign matrix_stack_size matrix_elements_stack * mmsize
cglobal mix_%1_to_%2_%3_flt, 3,in_channels+2,needed_mmregs+matrix_elements_mm, src0, src1, len, src2, src3, src4, src5, src6, src7
%assign needed_stack_size -1 * matrix_stack_size
%if ARCH_X86_32 && in_channels >= 7
%assign needed_stack_size needed_stack_size - 16
%endif
; get aligned stack space if needed
%if matrix_elements_stack > 0
%if mmsize == 32
%assign bkpreg %1 + 1
%define bkpq r %+ bkpreg %+ q
mov bkpq, rsp
and rsp, ~(mmsize-1)
sub rsp, matrix_elements_stack * mmsize
%else
%assign matrix_stack_size matrix_elements_stack * mmsize
%assign pad matrix_stack_size + (mmsize - gprsize) - (stack_offset & (mmsize - gprsize))
; on x86-32 for 7 and 8 channels we need more stack space for src pointers
%if ARCH_X86_32 && in_channels >= 7
%assign pad pad + 0x10
cglobal mix_%1_to_%2_%3_flt, 3,in_channels+2,needed_mmregs+matrix_elements_mm, needed_stack_size, src0, src1, len, src2, src3, src4, src5, src6, src7
; define src pointers on stack if needed
%if matrix_elements_stack > 0 && ARCH_X86_32 && in_channels >= 7
%define src5m [rsp+matrix_stack_size+0]
%define src6m [rsp+matrix_stack_size+4]
%define src7m [rsp+matrix_stack_size+8]
%endif
SUB rsp, pad
%endif
%endif
; load matrix pointers
@ -469,14 +460,6 @@ cglobal mix_%1_to_%2_%3_flt, 3,in_channels+2,needed_mmregs+matrix_elements_mm, s
add lenq, mmsize
jl .loop
; restore stack pointer
%if matrix_elements_stack > 0
%if mmsize == 32
mov rsp, bkpq
%else
ADD rsp, pad
%endif
%endif
; zero ymm high halves
%if mmsize == 32
vzeroupper