x86inc improvements for 64-bit

Add support for all x86-64 registers
Prefer caller-saved register over callee-saved on WIN64
Support up to 15 function arguments

Also (by Ronald S. Bultje)
Fix up our asm to work with new x86inc.asm.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
Signed-off-by: Justin Ruggles <justin.ruggles@gmail.com>
This commit is contained in:
Henrik Gramner
2012-04-11 15:47:00 -04:00
committed by Justin Ruggles
parent e1ce756844
commit 729f90e268
13 changed files with 316 additions and 339 deletions
+2 -2
View File
@@ -62,11 +62,11 @@ SECTION .text
%define cntr_reg fltsizeq
%define movsx mov
%else
%define cntr_reg r11
%define cntr_reg r7
%define movsx movsxd
%endif
cglobal yuv2planeX_%1, %3, 7, %2, filter, fltsize, src, dst, w, dither, offset
cglobal yuv2planeX_%1, %3, 8, %2, filter, fltsize, src, dst, w, dither, offset
%if %1 == 8 || %1 == 9 || %1 == 10
pxor m6, m6
%endif ; %1 == 8/9/10
+4 -14
View File
@@ -53,7 +53,7 @@ SECTION .text
%ifnidn %3, X
cglobal hscale%1to%2_%4, %5, 7, %6, pos0, dst, w, src, filter, fltpos, pos1
%else
cglobal hscale%1to%2_%4, %5, 7, %6, pos0, dst, w, srcmem, filter, fltpos, fltsize
cglobal hscale%1to%2_%4, %5, 10, %6, pos0, dst, w, srcmem, filter, fltpos, fltsize
%endif
%if ARCH_X86_64
movsxd wq, wd
@@ -245,10 +245,9 @@ cglobal hscale%1to%2_%4, %5, 7, %6, pos0, dst, w, srcmem, filter, fltpos, fltsiz
%define dlt 0
%endif ; %4 ==/!= X4
%if ARCH_X86_64
push r12
%define srcq r11
%define pos1q r10
%define srcendq r12
%define srcq r8
%define pos1q r7
%define srcendq r9
movsxd fltsizeq, fltsized ; filterSize
lea srcendq, [srcmemq+(fltsizeq-dlt)*srcmul] ; &src[filterSize&~4]
%else ; x86-32
@@ -388,16 +387,7 @@ cglobal hscale%1to%2_%4, %5, 7, %6, pos0, dst, w, srcmem, filter, fltpos, fltsiz
add wq, 2
%endif ; %3 ==/!= X
jl .loop
%ifnidn %3, X
REP_RET
%else ; %3 == X
%if ARCH_X86_64
pop r12
RET
%else ; x86-32
REP_RET
%endif ; x86-32/64
%endif ; %3 ==/!= X
%endmacro
; SCALE_FUNCS source_width, intermediate_nbits, n_xmm