ryan/mozjpeg
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
031e16ecffd21c6d648bbd94e222146cb1a24f1a
mozjpeg/simd/i386/jquant-3dn.asm
DRC 19c791cdac Improve code formatting consistency 
With rare exceptions ...
- Always separate line continuation characters by one space from
  preceding code.
- Always use two-space indentation.  Never use tabs.
- Always use K&R-style conditional blocks.
- Always surround operators with spaces, except in raw assembly code.
- Always put a space after, but not before, a comma.
- Never put a space between type casts and variables/function calls.
- Never put a space between the function name and the argument list in
  function declarations and prototypes.
- Always surround braces ('{' and '}') with spaces.
- Always surround statements (if, for, else, catch, while, do, switch)
  with spaces.
- Always attach pointer symbols ('*' and '**') to the variable or
  function name.
- Always precede pointer symbols ('*' and '**') by a space in type
  casts.
- Use the MIN() macro from jpegint.h within the libjpeg and TurboJPEG
  API libraries (using min() from tjutil.h is still necessary for
  TJBench.)
- Where it makes sense (particularly in the TurboJPEG code), put a blank
  line after variable declaration blocks.
- Always separate statements in one-liners by two spaces.

The purpose of this was to ease maintenance on my part and also to make
it easier for contributors to figure out how to format patch
submissions.  This was admittedly confusing (even to me sometimes) when
we had 3 or 4 different style conventions in the same source tree.  The
new convention is more consistent with the formatting of other OSS code
bases.

This commit corrects deviations from the chosen formatting style in the
libjpeg API code and reformats the TurboJPEG API code such that it
conforms to the same standard.

NOTES:
- Although it is no longer necessary for the function name in function
  declarations to begin in Column 1 (this was historically necessary
  because of the ansi2knr utility, which allowed libjpeg to be built
  with non-ANSI compilers), we retain that formatting for the libjpeg
  code because it improves readability when using libjpeg's function
  attribute macros (GLOBAL(), etc.)
- This reformatting project was accomplished with the help of AStyle and
  Uncrustify, although neither was completely up to the task, and thus
  a great deal of manual tweaking was required.  Note to developers of
  code formatting utilities:  the libjpeg-turbo code base is an
  excellent test bed, because AFAICT, it breaks every single one of the
  utilities that are currently available.
- The legacy (MMX, SSE, 3DNow!) assembly code for i386 has been
  formatted to match the SSE2 code (refer to
  ff5685d5344273df321eb63a005eaae19d2496e3.)  I hadn't intended to
  bother with this, but the Loongson MMI implementation demonstrated
  that there is still academic value to the MMX implementation, as an
  algorithmic model for other 64-bit vector implementations.  Thus, it
  is desirable to improve its readability in the same manner as that of
  the SSE2 implementation.
2018-03-16 02:14:34 -05:00

233 lines
8.8 KiB
NASM
Raw Blame History

;
; jquant.asm - sample data conversion and quantization (3DNow! & MMX)
;
; 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.
; For conditions of distribution and use, see copyright notice in jsimdext.inc
;
; This file should be assembled with NASM (Netwide Assembler),
; can *not* be assembled with Microsoft's MASM or any compatible
; assembler (including Borland's Turbo Assembler).
; NASM is available from http://nasm.sourceforge.net/ or
; http://sourceforge.net/project/showfiles.php?group_id=6208
;
; [TAB8]
%include "jsimdext.inc"
%include "jdct.inc"
; --------------------------------------------------------------------------
SECTION SEG_TEXT
BITS 32
;
; Load data into workspace, applying unsigned->signed conversion
;
; GLOBAL(void)
; jsimd_convsamp_float_3dnow(JSAMPARRAY sample_data, JDIMENSION start_col,
; FAST_FLOAT *workspace);
;
%define sample_data ebp + 8 ; JSAMPARRAY sample_data
%define start_col ebp + 12 ; JDIMENSION start_col
%define workspace ebp + 16 ; FAST_FLOAT *workspace
align 32
GLOBAL_FUNCTION(jsimd_convsamp_float_3dnow)
EXTN(jsimd_convsamp_float_3dnow):
push ebp
mov ebp, esp
push ebx
; push ecx ; need not be preserved
; push edx ; need not be preserved
push esi
push edi
pcmpeqw mm7, mm7
psllw mm7, 7
packsswb mm7, mm7 ; mm7 = PB_CENTERJSAMPLE (0x808080..)
mov esi, JSAMPARRAY [sample_data] ; (JSAMPROW *)
mov eax, JDIMENSION [start_col]
mov edi, POINTER [workspace] ; (DCTELEM *)
mov ecx, DCTSIZE/2
alignx 16, 7
.convloop:
mov ebx, JSAMPROW [esi+0*SIZEOF_JSAMPROW] ; (JSAMPLE *)
mov edx, JSAMPROW [esi+1*SIZEOF_JSAMPROW] ; (JSAMPLE *)
movq mm0, MMWORD [ebx+eax*SIZEOF_JSAMPLE]
movq mm1, MMWORD [edx+eax*SIZEOF_JSAMPLE]
psubb mm0, mm7 ; mm0=(01234567)
psubb mm1, mm7 ; mm1=(89ABCDEF)
punpcklbw mm2, mm0 ; mm2=(*0*1*2*3)
punpckhbw mm0, mm0 ; mm0=(*4*5*6*7)
punpcklbw mm3, mm1 ; mm3=(*8*9*A*B)
punpckhbw mm1, mm1 ; mm1=(*C*D*E*F)
punpcklwd mm4, mm2 ; mm4=(***0***1)
punpckhwd mm2, mm2 ; mm2=(***2***3)
punpcklwd mm5, mm0 ; mm5=(***4***5)
punpckhwd mm0, mm0 ; mm0=(***6***7)
psrad mm4, (DWORD_BIT-BYTE_BIT) ; mm4=(01)
psrad mm2, (DWORD_BIT-BYTE_BIT) ; mm2=(23)
pi2fd mm4, mm4
pi2fd mm2, mm2
psrad mm5, (DWORD_BIT-BYTE_BIT) ; mm5=(45)
psrad mm0, (DWORD_BIT-BYTE_BIT) ; mm0=(67)
pi2fd mm5, mm5
pi2fd mm0, mm0
movq MMWORD [MMBLOCK(0,0,edi,SIZEOF_FAST_FLOAT)], mm4
movq MMWORD [MMBLOCK(0,1,edi,SIZEOF_FAST_FLOAT)], mm2
movq MMWORD [MMBLOCK(0,2,edi,SIZEOF_FAST_FLOAT)], mm5
movq MMWORD [MMBLOCK(0,3,edi,SIZEOF_FAST_FLOAT)], mm0
punpcklwd mm6, mm3 ; mm6=(***8***9)
punpckhwd mm3, mm3 ; mm3=(***A***B)
punpcklwd mm4, mm1 ; mm4=(***C***D)
punpckhwd mm1, mm1 ; mm1=(***E***F)
psrad mm6, (DWORD_BIT-BYTE_BIT) ; mm6=(89)
psrad mm3, (DWORD_BIT-BYTE_BIT) ; mm3=(AB)
pi2fd mm6, mm6
pi2fd mm3, mm3
psrad mm4, (DWORD_BIT-BYTE_BIT) ; mm4=(CD)
psrad mm1, (DWORD_BIT-BYTE_BIT) ; mm1=(EF)
pi2fd mm4, mm4
pi2fd mm1, mm1
movq MMWORD [MMBLOCK(1,0,edi,SIZEOF_FAST_FLOAT)], mm6
movq MMWORD [MMBLOCK(1,1,edi,SIZEOF_FAST_FLOAT)], mm3
movq MMWORD [MMBLOCK(1,2,edi,SIZEOF_FAST_FLOAT)], mm4
movq MMWORD [MMBLOCK(1,3,edi,SIZEOF_FAST_FLOAT)], mm1
add esi, byte 2*SIZEOF_JSAMPROW
add edi, byte 2*DCTSIZE*SIZEOF_FAST_FLOAT
dec ecx
jnz near .convloop
femms ; empty MMX/3DNow! state
pop edi
pop esi
; pop edx ; need not be preserved
; pop ecx ; need not be preserved
pop ebx
pop ebp
ret
; --------------------------------------------------------------------------
;
; Quantize/descale the coefficients, and store into coef_block
;
; GLOBAL(void)
; jsimd_quantize_float_3dnow(JCOEFPTR coef_block, FAST_FLOAT *divisors,
; FAST_FLOAT *workspace);
;
%define coef_block ebp + 8 ; JCOEFPTR coef_block
%define divisors ebp + 12 ; FAST_FLOAT *divisors
%define workspace ebp + 16 ; FAST_FLOAT *workspace
align 32
GLOBAL_FUNCTION(jsimd_quantize_float_3dnow)
EXTN(jsimd_quantize_float_3dnow):
push ebp
mov ebp, esp
; push ebx ; unused
; push ecx ; unused
; push edx ; need not be preserved
push esi
push edi
mov eax, 0x4B400000 ; (float)0x00C00000 (rndint_magic)
movd mm7, eax
punpckldq mm7, mm7 ; mm7={12582912.0F 12582912.0F}
mov esi, POINTER [workspace]
mov edx, POINTER [divisors]
mov edi, JCOEFPTR [coef_block]
mov eax, DCTSIZE2/16
alignx 16, 7
.quantloop:
movq mm0, MMWORD [MMBLOCK(0,0,esi,SIZEOF_FAST_FLOAT)]
movq mm1, MMWORD [MMBLOCK(0,1,esi,SIZEOF_FAST_FLOAT)]
pfmul mm0, MMWORD [MMBLOCK(0,0,edx,SIZEOF_FAST_FLOAT)]
pfmul mm1, MMWORD [MMBLOCK(0,1,edx,SIZEOF_FAST_FLOAT)]
movq mm2, MMWORD [MMBLOCK(0,2,esi,SIZEOF_FAST_FLOAT)]
movq mm3, MMWORD [MMBLOCK(0,3,esi,SIZEOF_FAST_FLOAT)]
pfmul mm2, MMWORD [MMBLOCK(0,2,edx,SIZEOF_FAST_FLOAT)]
pfmul mm3, MMWORD [MMBLOCK(0,3,edx,SIZEOF_FAST_FLOAT)]
pfadd mm0, mm7 ; mm0=(00 ** 01 **)
pfadd mm1, mm7 ; mm1=(02 ** 03 **)
pfadd mm2, mm7 ; mm0=(04 ** 05 **)
pfadd mm3, mm7 ; mm1=(06 ** 07 **)
movq mm4, mm0
punpcklwd mm0, mm1 ; mm0=(00 02 ** **)
punpckhwd mm4, mm1 ; mm4=(01 03 ** **)
movq mm5, mm2
punpcklwd mm2, mm3 ; mm2=(04 06 ** **)
punpckhwd mm5, mm3 ; mm5=(05 07 ** **)
punpcklwd mm0, mm4 ; mm0=(00 01 02 03)
punpcklwd mm2, mm5 ; mm2=(04 05 06 07)
movq mm6, MMWORD [MMBLOCK(1,0,esi,SIZEOF_FAST_FLOAT)]
movq mm1, MMWORD [MMBLOCK(1,1,esi,SIZEOF_FAST_FLOAT)]
pfmul mm6, MMWORD [MMBLOCK(1,0,edx,SIZEOF_FAST_FLOAT)]
pfmul mm1, MMWORD [MMBLOCK(1,1,edx,SIZEOF_FAST_FLOAT)]
movq mm3, MMWORD [MMBLOCK(1,2,esi,SIZEOF_FAST_FLOAT)]
movq mm4, MMWORD [MMBLOCK(1,3,esi,SIZEOF_FAST_FLOAT)]
pfmul mm3, MMWORD [MMBLOCK(1,2,edx,SIZEOF_FAST_FLOAT)]
pfmul mm4, MMWORD [MMBLOCK(1,3,edx,SIZEOF_FAST_FLOAT)]
pfadd mm6, mm7 ; mm0=(10 ** 11 **)
pfadd mm1, mm7 ; mm4=(12 ** 13 **)
pfadd mm3, mm7 ; mm0=(14 ** 15 **)
pfadd mm4, mm7 ; mm4=(16 ** 17 **)
movq mm5, mm6
punpcklwd mm6, mm1 ; mm6=(10 12 ** **)
punpckhwd mm5, mm1 ; mm5=(11 13 ** **)
movq mm1, mm3
punpcklwd mm3, mm4 ; mm3=(14 16 ** **)
punpckhwd mm1, mm4 ; mm1=(15 17 ** **)
punpcklwd mm6, mm5 ; mm6=(10 11 12 13)
punpcklwd mm3, mm1 ; mm3=(14 15 16 17)
movq MMWORD [MMBLOCK(0,0,edi,SIZEOF_JCOEF)], mm0
movq MMWORD [MMBLOCK(0,1,edi,SIZEOF_JCOEF)], mm2
movq MMWORD [MMBLOCK(1,0,edi,SIZEOF_JCOEF)], mm6
movq MMWORD [MMBLOCK(1,1,edi,SIZEOF_JCOEF)], mm3
add esi, byte 16*SIZEOF_FAST_FLOAT
add edx, byte 16*SIZEOF_FAST_FLOAT
add edi, byte 16*SIZEOF_JCOEF
dec eax
jnz near .quantloop
femms ; empty MMX/3DNow! state
pop edi
pop esi
; pop edx ; need not be preserved
; pop ecx ; unused
; pop ebx ; unused
pop ebp
ret
; For some reason, the OS X linker does not honor the request to align the
; segment unless we do this.
align 32
Reference in New Issue View Git Blame Copy Permalink