Chiebot-Mirror
/
FFmpeg
mirror of https://github.com/FFmpeg/FFmpeg.git
8
0
Fork 0
Code Issues Projects Releases Wiki Activity

x86/tx_float: use all arguments on vperm2f and vpermilps and reindent comments

Browse Source 
Apparently even old nasm isn't required to accept incomplete instructions.
pull/362/head
Lynne 4 years ago
parent fddddc7ec2
commit e2cf0a1f68
No known key found for this signature in database
GPG Key ID: A2FEA5F03F034464
1 changed files with 78 additions and 78 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 46
      libavutil/x86/tx_float.asm

46
libavutil/x86/tx_float.asm
Unescape View File

@ -214,20 +214,20 @@ SECTION .text
addsubps %3, %3, %2 ; z1234, z5678
addsubps %1, %1, %4 ; s3142, s7586
mulps %3, %3, [s8_mult_odd] ; z * s8_mult_odd
vpermilps %1, [s8_perm_even] ; s1234, s5687 !
mulps %3, [s8_mult_odd] ; z * s8_mult_odd
vpermilps %1, %1, [s8_perm_even] ; s1234, s5687 !
shufps %2, %3, %3, q2332 ; junk, z7887
xorps %4, %1, [mask_mmmmpppm] ; e1234, e5687 !
vpermilps %3, %3, [s8_perm_odd2] ; z2314, z6556
vperm2f128 %1, %4, 0x03 ; e5687, s1234
vperm2f128 %1, %1, %4, 0x03 ; e5687, s1234
addsubps %2, %2, %3 ; junk, t5678
subps %1, %1, %4 ; w1234, w5678 even
vperm2f128 %2, %2, 0x11 ; t5678, t5678
vperm2f128 %3, %3, 0x00 ; z2314, z2314
vperm2f128 %2, %2, %2, 0x11 ; t5678, t5678
vperm2f128 %3, %3, %3, 0x00 ; z2314, z2314
xorps %2, %2, [mask_ppmpmmpm] ; t * ppmpmmpm
addps %2, %3, %2 ; u1234, u5678 odd
@ -279,14 +279,14 @@ SECTION .text
xorps %4, %6, mask ; s[8..15]*mpmppmpm
xorps %3, %5, mask ; s[0...7]*mpmppmpm
vperm2f128 %4, %4, 0x01 ; s[12..15, 8..11]
vperm2f128 %3, %3, 0x01 ; s[4..7, 0..3]
vperm2f128 %4, %4, %4, 0x01 ; s[12..15, 8..11]
vperm2f128 %3, %3, %3, 0x01 ; s[4..7, 0..3]
addps %6, %6, %4 ; y56, u56, y34, u34
addps %5, %5, %3 ; w56, x56, w34, x34
vpermilps %6, perm ; y56, u56, y43, u43
vpermilps %5, perm ; w56, x56, w43, x43
vpermilps %6, %6, perm ; y56, u56, y43, u43
vpermilps %5, %5, perm ; w56, x56, w43, x43
subps %4, %2, %6 ; odd part 2
addps %3, %2, %6 ; odd part 1
@ -453,7 +453,7 @@ SECTION .text
fmsubaddps %5, %5, %8, %9 ; j[0..8] even
%else
mulps %5, %5, %8 ; m2,3[23]reim * cos0246
xorps %9, %9, [mask_pmpmpmpm]; +-m2,3[23]imre * wim7531
xorps %9, %9, [mask_pmpmpmpm] ; +-m2,3[23]imre * wim7531
addps %5, %5, %9 ; j[0..8]
%endif
@ -486,7 +486,7 @@ SECTION .text
movaps [outq + 14*mmsize], tx2_o0
movaps tw_e, [ff_cos_64_float + mmsize]
vperm2f128 tw_o, [ff_cos_64_float + 64 - 4*7 - mmsize], 0x23
vperm2f128 tw_o, tw_o, [ff_cos_64_float + 64 - 4*7 - mmsize], 0x23
movaps m0, [outq + 1*mmsize]
movaps m1, [outq + 3*mmsize]
@ -513,7 +513,7 @@ SECTION .text
; %1 must contain len*2, %2 must contain len*4, %3 must contain len*6
%macro SPLIT_RADIX_LOAD_COMBINE_4 8
movaps m8, [rtabq + (%5)*mmsize + %7]
vperm2f128 m9, [itabq - (%5)*mmsize + %8], 0x23
vperm2f128 m9, m9, [itabq - (%5)*mmsize + %8], 0x23
movaps m0, [outq + (0 + %4)*mmsize + %6]
movaps m2, [outq + (2 + %4)*mmsize + %6]
@ -570,7 +570,7 @@ SECTION .text
; %3 must contain len*2, %4 must contain len*4, %5 must contain len*6
%macro SPLIT_RADIX_COMBINE_DEINTERLEAVE_2 6
movaps m8, [rtabq + (0 + %2)*mmsize]
vperm2f128 m9, [itabq - (0 + %2)*mmsize], 0x23
vperm2f128 m9, m9, [itabq - (0 + %2)*mmsize], 0x23
movaps m0, [outq + (0 + 0 + %1)*mmsize + %6]
movaps m2, [outq + (2 + 0 + %1)*mmsize + %6]
@ -612,7 +612,7 @@ SECTION .text
vperm2f128 m13, m13, m5, 0x13
movaps m8, [rtabq + (1 + %2)*mmsize]
vperm2f128 m9, [itabq - (1 + %2)*mmsize], 0x23
vperm2f128 m9, m9, [itabq - (1 + %2)*mmsize], 0x23
movaps m0, [outq + (0 + 1 + %1)*mmsize + %6]
movaps m2, [outq + (2 + 1 + %1)*mmsize + %6]
@ -801,7 +801,7 @@ cglobal fft32_float, 4, 4, 16, ctx, out, in, tmp
LOAD64_LUT m3, inq, ctxq, (mmsize/2)*3, tmpq, m14, m15
movaps m8, [ff_cos_32_float]
vperm2f128 m9, [ff_cos_32_float + 4*8 - 4*7], 0x23
vperm2f128 m9, m9, [ff_cos_32_float + 4*8 - 4*7], 0x23
FFT16 m0, m1, m2, m3, m10, m11, m12, m13
@ -859,7 +859,7 @@ ALIGN 16
sub outq, (%1*4) + (%1*2) + (%1/2)
lea rtabq, [ff_cos_ %+ %1 %+ _float]
lea itabq, [rtabq + %1 - 4*7]
lea itabq, [ff_cos_ %+ %1 %+ _float + %1 - 4*7]
%if %0 > 1
cmp tgtq, %1
@ -904,7 +904,7 @@ ALIGN 16
LOAD64_LUT m3, inq, lutq, (mmsize/2)*3, tmpq, m14, m15
movaps m8, [ff_cos_32_float]
vperm2f128 m9, [ff_cos_32_float + 32 - 4*7], 0x23
vperm2f128 m9, m9, [ff_cos_32_float + 32 - 4*7], 0x23
FFT16 m0, m1, m2, m3, m10, m11, m12, m13
@ -962,7 +962,7 @@ ALIGN 16
FFT16 tx2_e0, tx2_e1, tx2_o0, tx2_o1, tmp1, tmp2, tw_e, tw_o
movaps tw_e, [ff_cos_64_float]
vperm2f128 tw_o, [ff_cos_64_float + 64 - 4*7], 0x23
vperm2f128 tw_o, tw_o, [ff_cos_64_float + 64 - 4*7], 0x23
add lutq, (mmsize/2)*8
cmp tgtq, 64
@ -990,7 +990,7 @@ ALIGN 16
sub outq, 24*mmsize
lea rtabq, [ff_cos_128_float]
lea itabq, [rtabq + 128 - 4*7]
lea itabq, [ff_cos_128_float + 128 - 4*7]
cmp tgtq, 128
je .deinterleave
@ -1017,7 +1017,7 @@ ALIGN 16
sub outq, 48*mmsize
lea rtabq, [ff_cos_256_float]
lea itabq, [rtabq + 256 - 4*7]
lea itabq, [ff_cos_256_float + 256 - 4*7]
cmp tgtq, 256
je .deinterleave
@ -1045,7 +1045,7 @@ ALIGN 16
sub outq, 96*mmsize
lea rtabq, [ff_cos_512_float]
lea itabq, [rtabq + 512 - 4*7]
lea itabq, [ff_cos_512_float + 512 - 4*7]
cmp tgtq, 512
je .deinterleave
@ -1080,7 +1080,7 @@ ALIGN 16
sub outq, 192*mmsize
lea rtabq, [ff_cos_1024_float]
lea itabq, [rtabq + 1024 - 4*7]
lea itabq, [ff_cos_1024_float + 1024 - 4*7]
cmp tgtq, 1024
je .deinterleave
@ -1161,7 +1161,7 @@ FFT_SPLIT_RADIX_DEF 131072
vextractf128 [outq + 13*mmsize + 16], tx2_e0, 1
movaps tw_e, [ff_cos_64_float + mmsize]
vperm2f128 tw_o, [ff_cos_64_float + 64 - 4*7 - mmsize], 0x23
vperm2f128 tw_o, tw_o, [ff_cos_64_float + 64 - 4*7 - mmsize], 0x23
movaps m0, [outq + 1*mmsize]
movaps m1, [outq + 3*mmsize]

Write Preview
Loading…
Cancel
Save