Lay the groundwork for 32-bit AVX2 SIMD support
This commit is contained in:
@@ -2,6 +2,7 @@
|
||||
; jfdctfst.asm - fast integer FDCT (SSE2)
|
||||
;
|
||||
; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
|
||||
; Copyright (C) 2016, D. R. Commander.
|
||||
;
|
||||
; Based on the x86 SIMD extension for IJG JPEG library
|
||||
; Copyright (C) 1999-2006, MIYASAKA Masaru.
|
||||
@@ -50,7 +51,7 @@ F_1_306 equ DESCALE(1402911301, 30-CONST_BITS) ; FIX(1.306562965)
|
||||
%define PRE_MULTIPLY_SCALE_BITS 2
|
||||
%define CONST_SHIFT (16 - PRE_MULTIPLY_SCALE_BITS - CONST_BITS)
|
||||
|
||||
alignz 16
|
||||
alignz 32
|
||||
global EXTN(jconst_fdct_ifast_sse2)
|
||||
|
||||
EXTN(jconst_fdct_ifast_sse2):
|
||||
@@ -60,7 +61,7 @@ PW_F0382 times 8 dw F_0_382 << CONST_SHIFT
|
||||
PW_F0541 times 8 dw F_0_541 << CONST_SHIFT
|
||||
PW_F1306 times 8 dw F_1_306 << CONST_SHIFT
|
||||
|
||||
alignz 16
|
||||
alignz 32
|
||||
|
||||
; --------------------------------------------------------------------------
|
||||
SECTION SEG_TEXT
|
||||
@@ -78,7 +79,7 @@ PW_F1306 times 8 dw F_1_306 << CONST_SHIFT
|
||||
%define wk(i) ebp-(WK_NUM-(i))*SIZEOF_XMMWORD ; xmmword wk[WK_NUM]
|
||||
%define WK_NUM 2
|
||||
|
||||
align 16
|
||||
align 32
|
||||
global EXTN(jsimd_fdct_ifast_sse2)
|
||||
|
||||
EXTN(jsimd_fdct_ifast_sse2):
|
||||
@@ -400,4 +401,4 @@ EXTN(jsimd_fdct_ifast_sse2):
|
||||
|
||||
; For some reason, the OS X linker does not honor the request to align the
|
||||
; segment unless we do this.
|
||||
align 16
|
||||
align 32
|
||||
|
||||
Reference in New Issue
Block a user