jquanti-avx2.asm

Show log
Commit

Hash : 19c791cd
Author :
Date : 2018-03-08T10:55:20

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.

Download

simd/x86_64/jquanti-avx2.asm

;
; jquanti.asm - sample data conversion and quantization (64-bit AVX2)
;
; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
; Copyright (C) 2009, 2016, 2018, D. R. Commander.
; Copyright (C) 2016, Matthieu Darbois.
;
; 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        64
;
; Load data into workspace, applying unsigned->signed conversion
;
; GLOBAL(void)
; jsimd_convsamp_avx2(JSAMPARRAY sample_data, JDIMENSION start_col,
;                     DCTELEM *workspace);
;

; r10 = JSAMPARRAY sample_data
; r11d = JDIMENSION start_col
; r12 = DCTELEM *workspace

    align       32
    GLOBAL_FUNCTION(jsimd_convsamp_avx2)

EXTN(jsimd_convsamp_avx2):
    push        rbp
    mov         rax, rsp
    mov         rbp, rsp
    collect_args 3

    mov         eax, r11d

    mov         rsi, JSAMPROW [r10+0*SIZEOF_JSAMPROW]  ; (JSAMPLE *)
    mov         rdi, JSAMPROW [r10+1*SIZEOF_JSAMPROW]  ; (JSAMPLE *)
    movq        xmm0, XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE]
    pinsrq      xmm0, XMM_MMWORD [rdi+rax*SIZEOF_JSAMPLE], 1

    mov         rsi, JSAMPROW [r10+2*SIZEOF_JSAMPROW]  ; (JSAMPLE *)
    mov         rdi, JSAMPROW [r10+3*SIZEOF_JSAMPROW]  ; (JSAMPLE *)
    movq        xmm1, XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE]
    pinsrq      xmm1, XMM_MMWORD [rdi+rax*SIZEOF_JSAMPLE], 1

    mov         rsi, JSAMPROW [r10+4*SIZEOF_JSAMPROW]  ; (JSAMPLE *)
    mov         rdi, JSAMPROW [r10+5*SIZEOF_JSAMPROW]  ; (JSAMPLE *)
    movq        xmm2, XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE]
    pinsrq      xmm2, XMM_MMWORD [rdi+rax*SIZEOF_JSAMPLE], 1

    mov         rsi, JSAMPROW [r10+6*SIZEOF_JSAMPROW]  ; (JSAMPLE *)
    mov         rdi, JSAMPROW [r10+7*SIZEOF_JSAMPROW]  ; (JSAMPLE *)
    movq        xmm3, XMM_MMWORD [rsi+rax*SIZEOF_JSAMPLE]
    pinsrq      xmm3, XMM_MMWORD [rdi+rax*SIZEOF_JSAMPLE], 1

    vpmovzxbw   ymm0, xmm0              ; ymm0=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
    vpmovzxbw   ymm1, xmm1              ; ymm1=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
    vpmovzxbw   ymm2, xmm2              ; ymm2=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57)
    vpmovzxbw   ymm3, xmm3              ; ymm3=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77)

    vpcmpeqw    ymm7, ymm7, ymm7
    vpsllw      ymm7, ymm7, 7           ; ymm7={0xFF80 0xFF80 0xFF80 0xFF80 ..}

    vpaddw      ymm0, ymm0, ymm7
    vpaddw      ymm1, ymm1, ymm7
    vpaddw      ymm2, ymm2, ymm7
    vpaddw      ymm3, ymm3, ymm7

    vmovdqu     YMMWORD [YMMBLOCK(0,0,r12,SIZEOF_DCTELEM)], ymm0
    vmovdqu     YMMWORD [YMMBLOCK(2,0,r12,SIZEOF_DCTELEM)], ymm1
    vmovdqu     YMMWORD [YMMBLOCK(4,0,r12,SIZEOF_DCTELEM)], ymm2
    vmovdqu     YMMWORD [YMMBLOCK(6,0,r12,SIZEOF_DCTELEM)], ymm3

    vzeroupper
    uncollect_args 3
    pop         rbp
    ret

; --------------------------------------------------------------------------
;
; Quantize/descale the coefficients, and store into coef_block
;
; This implementation is based on an algorithm described in
;   "How to optimize for the Pentium family of microprocessors"
;   (http://www.agner.org/assem/).
;
; GLOBAL(void)
; jsimd_quantize_avx2(JCOEFPTR coef_block, DCTELEM *divisors,
;                     DCTELEM *workspace);
;

%define RECIPROCAL(m, n, b) \
  YMMBLOCK(DCTSIZE * 0 + (m), (n), (b), SIZEOF_DCTELEM)
%define CORRECTION(m, n, b) \
  YMMBLOCK(DCTSIZE * 1 + (m), (n), (b), SIZEOF_DCTELEM)
%define SCALE(m, n, b) \
  YMMBLOCK(DCTSIZE * 2 + (m), (n), (b), SIZEOF_DCTELEM)

; r10 = JCOEFPTR coef_block
; r11 = DCTELEM *divisors
; r12 = DCTELEM *workspace

    align       32
    GLOBAL_FUNCTION(jsimd_quantize_avx2)

EXTN(jsimd_quantize_avx2):
    push        rbp
    mov         rax, rsp
    mov         rbp, rsp
    collect_args 3

    vmovdqu     ymm4, [YMMBLOCK(0,0,r12,SIZEOF_DCTELEM)]
    vmovdqu     ymm5, [YMMBLOCK(2,0,r12,SIZEOF_DCTELEM)]
    vmovdqu     ymm6, [YMMBLOCK(4,0,r12,SIZEOF_DCTELEM)]
    vmovdqu     ymm7, [YMMBLOCK(6,0,r12,SIZEOF_DCTELEM)]
    vpabsw      ymm0, ymm4
    vpabsw      ymm1, ymm5
    vpabsw      ymm2, ymm6
    vpabsw      ymm3, ymm7

    vpaddw      ymm0, YMMWORD [CORRECTION(0,0,r11)]  ; correction + roundfactor
    vpaddw      ymm1, YMMWORD [CORRECTION(2,0,r11)]
    vpaddw      ymm2, YMMWORD [CORRECTION(4,0,r11)]
    vpaddw      ymm3, YMMWORD [CORRECTION(6,0,r11)]
    vpmulhuw    ymm0, YMMWORD [RECIPROCAL(0,0,r11)]  ; reciprocal
    vpmulhuw    ymm1, YMMWORD [RECIPROCAL(2,0,r11)]
    vpmulhuw    ymm2, YMMWORD [RECIPROCAL(4,0,r11)]
    vpmulhuw    ymm3, YMMWORD [RECIPROCAL(6,0,r11)]
    vpmulhuw    ymm0, YMMWORD [SCALE(0,0,r11)]       ; scale
    vpmulhuw    ymm1, YMMWORD [SCALE(2,0,r11)]
    vpmulhuw    ymm2, YMMWORD [SCALE(4,0,r11)]
    vpmulhuw    ymm3, YMMWORD [SCALE(6,0,r11)]

    vpsignw     ymm0, ymm0, ymm4
    vpsignw     ymm1, ymm1, ymm5
    vpsignw     ymm2, ymm2, ymm6
    vpsignw     ymm3, ymm3, ymm7

    vmovdqu     [YMMBLOCK(0,0,r10,SIZEOF_DCTELEM)], ymm0
    vmovdqu     [YMMBLOCK(2,0,r10,SIZEOF_DCTELEM)], ymm1
    vmovdqu     [YMMBLOCK(4,0,r10,SIZEOF_DCTELEM)], ymm2
    vmovdqu     [YMMBLOCK(6,0,r10,SIZEOF_DCTELEM)], ymm3

    vzeroupper
    uncollect_args 3
    pop         rbp
    ret

; For some reason, the OS X linker does not honor the request to align the
; segment unless we do this.
    align       32

Download

kc3-lang/libjpeg-turbo/simd/x86_64/jquanti-avx2.asm

Commit

simd/x86_64/jquanti-avx2.asm