ryan/mozjpeg
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
4f51f36eb347e27b51da84b511f25f8b9c8672fe
mozjpeg/simd/i386/jfdctint-avx2.asm
DRC 6e632af9f6 Demote "fast" [I]DCT algorithms to legacy status 
- Refer to the "slow" [I]DCT algorithms as "accurate" instead, since
  they are not slow under libjpeg-turbo.
- Adjust documentation claims to reflect the fact that the "slow" and
  "fast" algorithms produce about the same performance on AVX2-equipped
  CPUs (because of the dual-lane nature of AVX2, it was not possible to
  accelerate the "fast" algorithm beyond what was achievable with SSE2.)
  Also adjust the claims to reflect the fact that the "fast" algorithm
  tends to be ~5-15% faster than the "slow" algorithm on
  non-AVX2-equipped CPUs, regardless of the use of the libjpeg-turbo
  SIMD extensions.
- Indicate the legacy status of the "fast" and float algorithms in the
  documentation and cjpeg/djpeg usage info.
- Remove obsolete paragraph in the djpeg man page that suggested that
  the float algorithm could be faster than the "fast" algorithm on some
  CPUs.
2020-11-05 15:59:31 -06:00

332 lines
13 KiB
NASM
Raw Blame History

;
; jfdctint.asm - accurate integer FDCT (AVX2)
;
; Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
; Copyright (C) 2009, 2016, 2018, 2020, 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
;
; This file contains a slower but more accurate integer implementation of the
; forward DCT (Discrete Cosine Transform). The following code is based
; directly on the IJG's original jfdctint.c; see the jfdctint.c for
; more details.
%include "jsimdext.inc"
%include "jdct.inc"
; --------------------------------------------------------------------------
%define CONST_BITS 13
%define PASS1_BITS 2
%define DESCALE_P1 (CONST_BITS - PASS1_BITS)
%define DESCALE_P2 (CONST_BITS + PASS1_BITS)
%if CONST_BITS == 13
F_0_298 equ 2446 ; FIX(0.298631336)
F_0_390 equ 3196 ; FIX(0.390180644)
F_0_541 equ 4433 ; FIX(0.541196100)
F_0_765 equ 6270 ; FIX(0.765366865)
F_0_899 equ 7373 ; FIX(0.899976223)
F_1_175 equ 9633 ; FIX(1.175875602)
F_1_501 equ 12299 ; FIX(1.501321110)
F_1_847 equ 15137 ; FIX(1.847759065)
F_1_961 equ 16069 ; FIX(1.961570560)
F_2_053 equ 16819 ; FIX(2.053119869)
F_2_562 equ 20995 ; FIX(2.562915447)
F_3_072 equ 25172 ; FIX(3.072711026)
%else
; NASM cannot do compile-time arithmetic on floating-point constants.
%define DESCALE(x, n) (((x) + (1 << ((n) - 1))) >> (n))
F_0_298 equ DESCALE( 320652955, 30 - CONST_BITS) ; FIX(0.298631336)
F_0_390 equ DESCALE( 418953276, 30 - CONST_BITS) ; FIX(0.390180644)
F_0_541 equ DESCALE( 581104887, 30 - CONST_BITS) ; FIX(0.541196100)
F_0_765 equ DESCALE( 821806413, 30 - CONST_BITS) ; FIX(0.765366865)
F_0_899 equ DESCALE( 966342111, 30 - CONST_BITS) ; FIX(0.899976223)
F_1_175 equ DESCALE(1262586813, 30 - CONST_BITS) ; FIX(1.175875602)
F_1_501 equ DESCALE(1612031267, 30 - CONST_BITS) ; FIX(1.501321110)
F_1_847 equ DESCALE(1984016188, 30 - CONST_BITS) ; FIX(1.847759065)
F_1_961 equ DESCALE(2106220350, 30 - CONST_BITS) ; FIX(1.961570560)
F_2_053 equ DESCALE(2204520673, 30 - CONST_BITS) ; FIX(2.053119869)
F_2_562 equ DESCALE(2751909506, 30 - CONST_BITS) ; FIX(2.562915447)
F_3_072 equ DESCALE(3299298341, 30 - CONST_BITS) ; FIX(3.072711026)
%endif
; --------------------------------------------------------------------------
; In-place 8x8x16-bit matrix transpose using AVX2 instructions
; %1-%4: Input/output registers
; %5-%8: Temp registers
%macro dotranspose 8
; %1=(00 01 02 03 04 05 06 07 40 41 42 43 44 45 46 47)
; %2=(10 11 12 13 14 15 16 17 50 51 52 53 54 55 56 57)
; %3=(20 21 22 23 24 25 26 27 60 61 62 63 64 65 66 67)
; %4=(30 31 32 33 34 35 36 37 70 71 72 73 74 75 76 77)
vpunpcklwd %5, %1, %2
vpunpckhwd %6, %1, %2
vpunpcklwd %7, %3, %4
vpunpckhwd %8, %3, %4
; transpose coefficients(phase 1)
; %5=(00 10 01 11 02 12 03 13 40 50 41 51 42 52 43 53)
; %6=(04 14 05 15 06 16 07 17 44 54 45 55 46 56 47 57)
; %7=(20 30 21 31 22 32 23 33 60 70 61 71 62 72 63 73)
; %8=(24 34 25 35 26 36 27 37 64 74 65 75 66 76 67 77)
vpunpckldq %1, %5, %7
vpunpckhdq %2, %5, %7
vpunpckldq %3, %6, %8
vpunpckhdq %4, %6, %8
; transpose coefficients(phase 2)
; %1=(00 10 20 30 01 11 21 31 40 50 60 70 41 51 61 71)
; %2=(02 12 22 32 03 13 23 33 42 52 62 72 43 53 63 73)
; %3=(04 14 24 34 05 15 25 35 44 54 64 74 45 55 65 75)
; %4=(06 16 26 36 07 17 27 37 46 56 66 76 47 57 67 77)
vpermq %1, %1, 0x8D
vpermq %2, %2, 0x8D
vpermq %3, %3, 0xD8
vpermq %4, %4, 0xD8
; transpose coefficients(phase 3)
; %1=(01 11 21 31 41 51 61 71 00 10 20 30 40 50 60 70)
; %2=(03 13 23 33 43 53 63 73 02 12 22 32 42 52 62 72)
; %3=(04 14 24 34 44 54 64 74 05 15 25 35 45 55 65 75)
; %4=(06 16 26 36 46 56 66 76 07 17 27 37 47 57 67 77)
%endmacro
; --------------------------------------------------------------------------
; In-place 8x8x16-bit accurate integer forward DCT using AVX2 instructions
; %1-%4: Input/output registers
; %5-%8: Temp registers
; %9: Pass (1 or 2)
%macro dodct 9
vpsubw %5, %1, %4 ; %5=data1_0-data6_7=tmp6_7
vpaddw %6, %1, %4 ; %6=data1_0+data6_7=tmp1_0
vpaddw %7, %2, %3 ; %7=data3_2+data4_5=tmp3_2
vpsubw %8, %2, %3 ; %8=data3_2-data4_5=tmp4_5
; -- Even part
vperm2i128 %6, %6, %6, 0x01 ; %6=tmp0_1
vpaddw %1, %6, %7 ; %1=tmp0_1+tmp3_2=tmp10_11
vpsubw %6, %6, %7 ; %6=tmp0_1-tmp3_2=tmp13_12
vperm2i128 %7, %1, %1, 0x01 ; %7=tmp11_10
vpsignw %1, %1, [GOTOFF(ebx, PW_1_NEG1)] ; %1=tmp10_neg11
vpaddw %7, %7, %1 ; %7=(tmp10+tmp11)_(tmp10-tmp11)
%if %9 == 1
vpsllw %1, %7, PASS1_BITS ; %1=data0_4
%else
vpaddw %7, %7, [GOTOFF(ebx, PW_DESCALE_P2X)]
vpsraw %1, %7, PASS1_BITS ; %1=data0_4
%endif
; (Original)
; z1 = (tmp12 + tmp13) * 0.541196100;
; data2 = z1 + tmp13 * 0.765366865;
; data6 = z1 + tmp12 * -1.847759065;
;
; (This implementation)
; data2 = tmp13 * (0.541196100 + 0.765366865) + tmp12 * 0.541196100;
; data6 = tmp13 * 0.541196100 + tmp12 * (0.541196100 - 1.847759065);
vperm2i128 %7, %6, %6, 0x01 ; %7=tmp12_13
vpunpcklwd %2, %6, %7
vpunpckhwd %6, %6, %7
vpmaddwd %2, %2, [GOTOFF(ebx, PW_F130_F054_MF130_F054)] ; %2=data2_6L
vpmaddwd %6, %6, [GOTOFF(ebx, PW_F130_F054_MF130_F054)] ; %6=data2_6H
vpaddd %2, %2, [GOTOFF(ebx, PD_DESCALE_P %+ %9)]
vpaddd %6, %6, [GOTOFF(ebx, PD_DESCALE_P %+ %9)]
vpsrad %2, %2, DESCALE_P %+ %9
vpsrad %6, %6, DESCALE_P %+ %9
vpackssdw %3, %2, %6 ; %6=data2_6
; -- Odd part
vpaddw %7, %8, %5 ; %7=tmp4_5+tmp6_7=z3_4
; (Original)
; z5 = (z3 + z4) * 1.175875602;
; z3 = z3 * -1.961570560; z4 = z4 * -0.390180644;
; z3 += z5; z4 += z5;
;
; (This implementation)
; z3 = z3 * (1.175875602 - 1.961570560) + z4 * 1.175875602;
; z4 = z3 * 1.175875602 + z4 * (1.175875602 - 0.390180644);
vperm2i128 %2, %7, %7, 0x01 ; %2=z4_3
vpunpcklwd %6, %7, %2
vpunpckhwd %7, %7, %2
vpmaddwd %6, %6, [GOTOFF(ebx, PW_MF078_F117_F078_F117)] ; %6=z3_4L
vpmaddwd %7, %7, [GOTOFF(ebx, PW_MF078_F117_F078_F117)] ; %7=z3_4H
; (Original)
; z1 = tmp4 + tmp7; z2 = tmp5 + tmp6;
; tmp4 = tmp4 * 0.298631336; tmp5 = tmp5 * 2.053119869;
; tmp6 = tmp6 * 3.072711026; tmp7 = tmp7 * 1.501321110;
; z1 = z1 * -0.899976223; z2 = z2 * -2.562915447;
; data7 = tmp4 + z1 + z3; data5 = tmp5 + z2 + z4;
; data3 = tmp6 + z2 + z3; data1 = tmp7 + z1 + z4;
;
; (This implementation)
; tmp4 = tmp4 * (0.298631336 - 0.899976223) + tmp7 * -0.899976223;
; tmp5 = tmp5 * (2.053119869 - 2.562915447) + tmp6 * -2.562915447;
; tmp6 = tmp5 * -2.562915447 + tmp6 * (3.072711026 - 2.562915447);
; tmp7 = tmp4 * -0.899976223 + tmp7 * (1.501321110 - 0.899976223);
; data7 = tmp4 + z3; data5 = tmp5 + z4;
; data3 = tmp6 + z3; data1 = tmp7 + z4;
vperm2i128 %4, %5, %5, 0x01 ; %4=tmp7_6
vpunpcklwd %2, %8, %4
vpunpckhwd %4, %8, %4
vpmaddwd %2, %2, [GOTOFF(ebx, PW_MF060_MF089_MF050_MF256)] ; %2=tmp4_5L
vpmaddwd %4, %4, [GOTOFF(ebx, PW_MF060_MF089_MF050_MF256)] ; %4=tmp4_5H
vpaddd %2, %2, %6 ; %2=data7_5L
vpaddd %4, %4, %7 ; %4=data7_5H
vpaddd %2, %2, [GOTOFF(ebx, PD_DESCALE_P %+ %9)]
vpaddd %4, %4, [GOTOFF(ebx, PD_DESCALE_P %+ %9)]
vpsrad %2, %2, DESCALE_P %+ %9
vpsrad %4, %4, DESCALE_P %+ %9
vpackssdw %4, %2, %4 ; %4=data7_5
vperm2i128 %2, %8, %8, 0x01 ; %2=tmp5_4
vpunpcklwd %8, %5, %2
vpunpckhwd %5, %5, %2
vpmaddwd %8, %8, [GOTOFF(ebx, PW_F050_MF256_F060_MF089)] ; %8=tmp6_7L
vpmaddwd %5, %5, [GOTOFF(ebx, PW_F050_MF256_F060_MF089)] ; %5=tmp6_7H
vpaddd %8, %8, %6 ; %8=data3_1L
vpaddd %5, %5, %7 ; %5=data3_1H
vpaddd %8, %8, [GOTOFF(ebx, PD_DESCALE_P %+ %9)]
vpaddd %5, %5, [GOTOFF(ebx, PD_DESCALE_P %+ %9)]
vpsrad %8, %8, DESCALE_P %+ %9
vpsrad %5, %5, DESCALE_P %+ %9
vpackssdw %2, %8, %5 ; %2=data3_1
%endmacro
; --------------------------------------------------------------------------
SECTION SEG_CONST
alignz 32
GLOBAL_DATA(jconst_fdct_islow_avx2)
EXTN(jconst_fdct_islow_avx2):
PW_F130_F054_MF130_F054 times 4 dw (F_0_541 + F_0_765), F_0_541
times 4 dw (F_0_541 - F_1_847), F_0_541
PW_MF078_F117_F078_F117 times 4 dw (F_1_175 - F_1_961), F_1_175
times 4 dw (F_1_175 - F_0_390), F_1_175
PW_MF060_MF089_MF050_MF256 times 4 dw (F_0_298 - F_0_899), -F_0_899
times 4 dw (F_2_053 - F_2_562), -F_2_562
PW_F050_MF256_F060_MF089 times 4 dw (F_3_072 - F_2_562), -F_2_562
times 4 dw (F_1_501 - F_0_899), -F_0_899
PD_DESCALE_P1 times 8 dd 1 << (DESCALE_P1 - 1)
PD_DESCALE_P2 times 8 dd 1 << (DESCALE_P2 - 1)
PW_DESCALE_P2X times 16 dw 1 << (PASS1_BITS - 1)
PW_1_NEG1 times 8 dw 1
times 8 dw -1
alignz 32
; --------------------------------------------------------------------------
SECTION SEG_TEXT
BITS 32
;
; Perform the forward DCT on one block of samples.
;
; GLOBAL(void)
; jsimd_fdct_islow_avx2(DCTELEM *data)
;
%define data(b) (b) + 8 ; DCTELEM *data
align 32
GLOBAL_FUNCTION(jsimd_fdct_islow_avx2)
EXTN(jsimd_fdct_islow_avx2):
push ebp
mov ebp, esp
pushpic ebx
; push ecx ; unused
; push edx ; need not be preserved
; push esi ; unused
; push edi ; unused
get_GOT ebx ; get GOT address
; ---- Pass 1: process rows.
mov edx, POINTER [data(ebp)] ; (DCTELEM *)
vmovdqu ymm4, YMMWORD [YMMBLOCK(0,0,edx,SIZEOF_DCTELEM)]
vmovdqu ymm5, YMMWORD [YMMBLOCK(2,0,edx,SIZEOF_DCTELEM)]
vmovdqu ymm6, YMMWORD [YMMBLOCK(4,0,edx,SIZEOF_DCTELEM)]
vmovdqu ymm7, YMMWORD [YMMBLOCK(6,0,edx,SIZEOF_DCTELEM)]
; ymm4=(00 01 02 03 04 05 06 07 10 11 12 13 14 15 16 17)
; ymm5=(20 21 22 23 24 25 26 27 30 31 32 33 34 35 36 37)
; ymm6=(40 41 42 43 44 45 46 47 50 51 52 53 54 55 56 57)
; ymm7=(60 61 62 63 64 65 66 67 70 71 72 73 74 75 76 77)
vperm2i128 ymm0, ymm4, ymm6, 0x20
vperm2i128 ymm1, ymm4, ymm6, 0x31
vperm2i128 ymm2, ymm5, ymm7, 0x20
vperm2i128 ymm3, ymm5, ymm7, 0x31
; ymm0=(00 01 02 03 04 05 06 07 40 41 42 43 44 45 46 47)
; ymm1=(10 11 12 13 14 15 16 17 50 51 52 53 54 55 56 57)
; ymm2=(20 21 22 23 24 25 26 27 60 61 62 63 64 65 66 67)
; ymm3=(30 31 32 33 34 35 36 37 70 71 72 73 74 75 76 77)
dotranspose ymm0, ymm1, ymm2, ymm3, ymm4, ymm5, ymm6, ymm7
dodct ymm0, ymm1, ymm2, ymm3, ymm4, ymm5, ymm6, ymm7, 1
; ymm0=data0_4, ymm1=data3_1, ymm2=data2_6, ymm3=data7_5
; ---- Pass 2: process columns.
vperm2i128 ymm4, ymm1, ymm3, 0x20 ; ymm4=data3_7
vperm2i128 ymm1, ymm1, ymm3, 0x31 ; ymm1=data1_5
dotranspose ymm0, ymm1, ymm2, ymm4, ymm3, ymm5, ymm6, ymm7
dodct ymm0, ymm1, ymm2, ymm4, ymm3, ymm5, ymm6, ymm7, 2
; ymm0=data0_4, ymm1=data3_1, ymm2=data2_6, ymm4=data7_5
vperm2i128 ymm3, ymm0, ymm1, 0x30 ; ymm3=data0_1
vperm2i128 ymm5, ymm2, ymm1, 0x20 ; ymm5=data2_3
vperm2i128 ymm6, ymm0, ymm4, 0x31 ; ymm6=data4_5
vperm2i128 ymm7, ymm2, ymm4, 0x21 ; ymm7=data6_7
vmovdqu YMMWORD [YMMBLOCK(0,0,edx,SIZEOF_DCTELEM)], ymm3
vmovdqu YMMWORD [YMMBLOCK(2,0,edx,SIZEOF_DCTELEM)], ymm5
vmovdqu YMMWORD [YMMBLOCK(4,0,edx,SIZEOF_DCTELEM)], ymm6
vmovdqu YMMWORD [YMMBLOCK(6,0,edx,SIZEOF_DCTELEM)], ymm7
vzeroupper
; pop edi ; unused
; pop esi ; unused
; pop edx ; need not be preserved
; pop ecx ; unused
poppic ebx
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