virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-11-21 10:55:51 +02:00
Code Issues Releases Activity
f3980b75f8
FFmpeg/libavcodec/x86/vp3dsp.asm

710 lines
23 KiB
NASM
Raw Normal View History

Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
;******************************************************************************
;* MMX/SSE2-optimized functions for the VP3 decoder
;* Copyright (c) 2007 Aurelien Jacobs <aurel@gnuage.org>
;*
;* This file is part of FFmpeg.
;*
;* FFmpeg is free software; you can redistribute it and/or
;* modify it under the terms of the GNU Lesser General Public
;* License as published by the Free Software Foundation; either
;* version 2.1 of the License, or (at your option) any later version.
;*
;* FFmpeg is distributed in the hope that it will be useful,
;* but WITHOUT ANY WARRANTY; without even the implied warranty of
;* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
;* Lesser General Public License for more details.
;*
;* You should have received a copy of the GNU Lesser General Public
;* License along with FFmpeg; if not, write to the Free Software
Fix FSF address copy paste error in some license headers.
2011-05-14 22:32:31 +03:00
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
;******************************************************************************
Move x264asm to libavutil. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2011-06-05 17:19:16 +03:00
%include "libavutil/x86/x86util.asm"
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
; MMX-optimized functions cribbed from the original VP3 source code.
SECTION_RODATA
vp3_idct_data: times 8 dw 64277
times 8 dw 60547
times 8 dw 54491
times 8 dw 46341
times 8 dw 36410
times 8 dw 25080
times 8 dw 12785
vp3: use hpeldsp instead of dsputil for half-pel functions. This makes vp3 independent of dsputil.
2013-03-10 08:26:09 +03:00
pb_7: times 8 db 7
pb_1F: times 8 db 0x1f
pb_81: times 8 db 0x81
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
cextern pb_1
cextern pb_3
vp3: don't use calls to inline asm in yasm code. Mixing yasm and inline asm is a bad idea, since if either yasm or inline asm is not supported by your toolchain, all of the asm stops working. Thus, better to use either one or the other alone. Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
2012-07-22 23:38:56 +03:00
cextern pb_80
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
cextern pw_8
SECTION .text
; this is off by one or two for some cases when filter_limit is greater than 63
; in: p0 in mm6, p1 in mm4, p2 in mm2, p3 in mm1
; out: p1 in mm4, p2 in mm3
%macro VP3_LOOP_FILTER 0
movq m7, m6
pand m6, [pb_7] ; p0&7
psrlw m7, 3
pand m7, [pb_1F] ; p0>>3
movq m3, m2 ; p2
pxor m2, m4
pand m2, [pb_1] ; (p2^p1)&1
movq m5, m2
paddb m2, m2
paddb m2, m5 ; 3*(p2^p1)&1
paddb m2, m6 ; extra bits lost in shifts
pcmpeqb m0, m0
pxor m1, m0 ; 255 - p3
pavgb m1, m2 ; (256 - p3 + extrabits) >> 1
pxor m0, m4 ; 255 - p1
pavgb m0, m3 ; (256 + p2-p1) >> 1
paddb m1, [pb_3]
pavgb m1, m0 ; 128+2+( p2-p1 - p3) >> 2
pavgb m1, m0 ; 128+1+(3*(p2-p1) - p3) >> 3
paddusb m7, m1 ; d+128+1
movq m6, [pb_81]
psubusb m6, m7
psubusb m7, [pb_81]
movq m5, [r2+516] ; flim
pminub m6, m5
pminub m7, m5
movq m0, m6
movq m1, m7
paddb m6, m6
paddb m7, m7
pminub m6, m5
pminub m7, m5
psubb m6, m0
psubb m7, m1
paddusb m4, m7
psubusb m4, m6
psubusb m3, m7
paddusb m3, m6
%endmacro
%macro STORE_4_WORDS 1
Use "d" suffix for general-purpose registers used with movd. This increases compatibilty with nasm and is also more consistent, e.g. with h264_intrapred.asm and h264_chromamc.asm that already do it that way. Originally committed as revision 25042 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-09-05 13:10:16 +03:00
movd r2d, %1
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
mov [r0 -1], r2w
psrlq %1, 32
shr r2, 16
mov [r0+r1 -1], r2w
Use "d" suffix for general-purpose registers used with movd. This increases compatibilty with nasm and is also more consistent, e.g. with h264_intrapred.asm and h264_chromamc.asm that already do it that way. Originally committed as revision 25042 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-09-05 13:10:16 +03:00
movd r2d, %1
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
mov [r0+r1*2-1], r2w
shr r2, 16
mov [r0+r3 -1], r2w
%endmacro
x86: mmx2 ---> mmxext in asm constructs
2012-07-10 01:04:18 +03:00
INIT_MMX mmxext
vp3: port x86 SIMD to cpuflags.
2012-07-28 01:17:27 +03:00
cglobal vp3_v_loop_filter, 3, 4
config.asm: change %ifdef directives to %if directives. This allows combining multiple conditionals in a single statement.
2012-01-23 12:45:58 +03:00
%if ARCH_X86_64
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
movsxd r1, r1d
%endif
mov r3, r1
neg r1
movq m6, [r0+r1*2]
movq m4, [r0+r1 ]
movq m2, [r0 ]
movq m1, [r0+r3 ]
VP3_LOOP_FILTER
movq [r0+r1], m4
movq [r0 ], m3
RET
vp3: port x86 SIMD to cpuflags.
2012-07-28 01:17:27 +03:00
cglobal vp3_h_loop_filter, 3, 4
config.asm: change %ifdef directives to %if directives. This allows combining multiple conditionals in a single statement.
2012-01-23 12:45:58 +03:00
%if ARCH_X86_64
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
movsxd r1, r1d
%endif
lea r3, [r1*3]
movd m6, [r0 -2]
movd m4, [r0+r1 -2]
movd m2, [r0+r1*2-2]
movd m1, [r0+r3 -2]
lea r0, [r0+r1*4 ]
punpcklbw m6, [r0 -2]
punpcklbw m4, [r0+r1 -2]
punpcklbw m2, [r0+r1*2-2]
punpcklbw m1, [r0+r3 -2]
sub r0, r3
sub r0, r1
TRANSPOSE4x4B 6, 4, 2, 1, 0
VP3_LOOP_FILTER
SBUTTERFLY bw, 4, 3, 5
STORE_4_WORDS m4
lea r0, [r0+r1*4 ]
STORE_4_WORDS m3
RET
; from original comments: The Macro does IDct on 4 1-D Dcts
%macro BeginIDCT 0
movq m2, I(3)
movq m6, C(3)
movq m4, m2
movq m7, J(5)
pmulhw m4, m6 ; r4 = c3*i3 - i3
movq m1, C(5)
pmulhw m6, m7 ; r6 = c3*i5 - i5
movq m5, m1
pmulhw m1, m2 ; r1 = c5*i3 - i3
movq m3, I(1)
pmulhw m5, m7 ; r5 = c5*i5 - i5
movq m0, C(1)
paddw m4, m2 ; r4 = c3*i3
paddw m6, m7 ; r6 = c3*i5
paddw m2, m1 ; r2 = c5*i3
movq m1, J(7)
paddw m7, m5 ; r7 = c5*i5
movq m5, m0 ; r5 = c1
pmulhw m0, m3 ; r0 = c1*i1 - i1
paddsw m4, m7 ; r4 = C = c3*i3 + c5*i5
pmulhw m5, m1 ; r5 = c1*i7 - i7
movq m7, C(7)
psubsw m6, m2 ; r6 = D = c3*i5 - c5*i3
paddw m0, m3 ; r0 = c1*i1
pmulhw m3, m7 ; r3 = c7*i1
movq m2, I(2)
pmulhw m7, m1 ; r7 = c7*i7
paddw m5, m1 ; r5 = c1*i7
movq m1, m2 ; r1 = i2
pmulhw m2, C(2) ; r2 = c2*i2 - i2
psubsw m3, m5 ; r3 = B = c7*i1 - c1*i7
movq m5, J(6)
paddsw m0, m7 ; r0 = A = c1*i1 + c7*i7
movq m7, m5 ; r7 = i6
psubsw m0, m4 ; r0 = A - C
pmulhw m5, C(2) ; r5 = c2*i6 - i6
paddw m2, m1 ; r2 = c2*i2
pmulhw m1, C(6) ; r1 = c6*i2
paddsw m4, m4 ; r4 = C + C
paddsw m4, m0 ; r4 = C. = A + C
psubsw m3, m6 ; r3 = B - D
paddw m5, m7 ; r5 = c2*i6
paddsw m6, m6 ; r6 = D + D
pmulhw m7, C(6) ; r7 = c6*i6
paddsw m6, m3 ; r6 = D. = B + D
movq I(1), m4 ; save C. at I(1)
psubsw m1, m5 ; r1 = H = c6*i2 - c2*i6
movq m4, C(4)
movq m5, m3 ; r5 = B - D
pmulhw m3, m4 ; r3 = (c4 - 1) * (B - D)
paddsw m7, m2 ; r3 = (c4 - 1) * (B - D)
movq I(2), m6 ; save D. at I(2)
movq m2, m0 ; r2 = A - C
movq m6, I(0)
pmulhw m0, m4 ; r0 = (c4 - 1) * (A - C)
paddw m5, m3 ; r5 = B. = c4 * (B - D)
movq m3, J(4)
psubsw m5, m1 ; r5 = B.. = B. - H
paddw m2, m0 ; r0 = A. = c4 * (A - C)
psubsw m6, m3 ; r6 = i0 - i4
movq m0, m6
pmulhw m6, m4 ; r6 = (c4 - 1) * (i0 - i4)
paddsw m3, m3 ; r3 = i4 + i4
paddsw m1, m1 ; r1 = H + H
paddsw m3, m0 ; r3 = i0 + i4
paddsw m1, m5 ; r1 = H. = B + H
pmulhw m4, m3 ; r4 = (c4 - 1) * (i0 + i4)
paddsw m6, m0 ; r6 = F = c4 * (i0 - i4)
psubsw m6, m2 ; r6 = F. = F - A.
paddsw m2, m2 ; r2 = A. + A.
movq m0, I(1) ; r0 = C.
paddsw m2, m6 ; r2 = A.. = F + A.
paddw m4, m3 ; r4 = E = c4 * (i0 + i4)
psubsw m2, m1 ; r2 = R2 = A.. - H.
%endmacro
; RowIDCT gets ready to transpose
%macro RowIDCT 0
BeginIDCT
movq m3, I(2) ; r3 = D.
psubsw m4, m7 ; r4 = E. = E - G
paddsw m1, m1 ; r1 = H. + H.
paddsw m7, m7 ; r7 = G + G
paddsw m1, m2 ; r1 = R1 = A.. + H.
paddsw m7, m4 ; r1 = R1 = A.. + H.
psubsw m4, m3 ; r4 = R4 = E. - D.
paddsw m3, m3
psubsw m6, m5 ; r6 = R6 = F. - B..
paddsw m5, m5
paddsw m3, m4 ; r3 = R3 = E. + D.
paddsw m5, m6 ; r5 = R5 = F. + B..
psubsw m7, m0 ; r7 = R7 = G. - C.
paddsw m0, m0
movq I(1), m1 ; save R1
paddsw m0, m7 ; r0 = R0 = G. + C.
%endmacro
; Column IDCT normalizes and stores final results
%macro ColumnIDCT 0
BeginIDCT
paddsw m2, OC_8 ; adjust R2 (and R1) for shift
paddsw m1, m1 ; r1 = H. + H.
paddsw m1, m2 ; r1 = R1 = A.. + H.
psraw m2, 4 ; r2 = NR2
psubsw m4, m7 ; r4 = E. = E - G
psraw m1, 4 ; r1 = NR2
movq m3, I(2) ; r3 = D.
paddsw m7, m7 ; r7 = G + G
movq I(2), m2 ; store NR2 at I2
paddsw m7, m4 ; r7 = G. = E + G
movq I(1), m1 ; store NR1 at I1
psubsw m4, m3 ; r4 = R4 = E. - D.
paddsw m4, OC_8 ; adjust R4 (and R3) for shift
paddsw m3, m3 ; r3 = D. + D.
paddsw m3, m4 ; r3 = R3 = E. + D.
psraw m4, 4 ; r4 = NR4
psubsw m6, m5 ; r6 = R6 = F. - B..
psraw m3, 4 ; r3 = NR3
paddsw m6, OC_8 ; adjust R6 (and R5) for shift
paddsw m5, m5 ; r5 = B.. + B..
paddsw m5, m6 ; r5 = R5 = F. + B..
psraw m6, 4 ; r6 = NR6
movq J(4), m4 ; store NR4 at J4
psraw m5, 4 ; r5 = NR5
movq I(3), m3 ; store NR3 at I3
psubsw m7, m0 ; r7 = R7 = G. - C.
paddsw m7, OC_8 ; adjust R7 (and R0) for shift
paddsw m0, m0 ; r0 = C. + C.
paddsw m0, m7 ; r0 = R0 = G. + C.
psraw m7, 4 ; r7 = NR7
movq J(6), m6 ; store NR6 at J6
psraw m0, 4 ; r0 = NR0
movq J(5), m5 ; store NR5 at J5
movq J(7), m7 ; store NR7 at J7
movq I(0), m0 ; store NR0 at I0
%endmacro
; Following macro does two 4x4 transposes in place.
;
; At entry (we assume):
;
; r0 = a3 a2 a1 a0
; I(1) = b3 b2 b1 b0
; r2 = c3 c2 c1 c0
; r3 = d3 d2 d1 d0
;
; r4 = e3 e2 e1 e0
; r5 = f3 f2 f1 f0
; r6 = g3 g2 g1 g0
; r7 = h3 h2 h1 h0
;
; At exit, we have:
;
; I(0) = d0 c0 b0 a0
; I(1) = d1 c1 b1 a1
; I(2) = d2 c2 b2 a2
; I(3) = d3 c3 b3 a3
;
; J(4) = h0 g0 f0 e0
; J(5) = h1 g1 f1 e1
; J(6) = h2 g2 f2 e2
; J(7) = h3 g3 f3 e3
;
; I(0) I(1) I(2) I(3) is the transpose of r0 I(1) r2 r3.
; J(4) J(5) J(6) J(7) is the transpose of r4 r5 r6 r7.
;
; Since r1 is free at entry, we calculate the Js first.
%macro Transpose 0
movq m1, m4 ; r1 = e3 e2 e1 e0
punpcklwd m4, m5 ; r4 = f1 e1 f0 e0
movq I(0), m0 ; save a3 a2 a1 a0
punpckhwd m1, m5 ; r1 = f3 e3 f2 e2
movq m0, m6 ; r0 = g3 g2 g1 g0
punpcklwd m6, m7 ; r6 = h1 g1 h0 g0
movq m5, m4 ; r5 = f1 e1 f0 e0
punpckldq m4, m6 ; r4 = h0 g0 f0 e0 = R4
punpckhdq m5, m6 ; r5 = h1 g1 f1 e1 = R5
movq m6, m1 ; r6 = f3 e3 f2 e2
movq J(4), m4
punpckhwd m0, m7 ; r0 = h3 g3 h2 g2
movq J(5), m5
punpckhdq m6, m0 ; r6 = h3 g3 f3 e3 = R7
movq m4, I(0) ; r4 = a3 a2 a1 a0
punpckldq m1, m0 ; r1 = h2 g2 f2 e2 = R6
movq m5, I(1) ; r5 = b3 b2 b1 b0
movq m0, m4 ; r0 = a3 a2 a1 a0
movq J(7), m6
punpcklwd m0, m5 ; r0 = b1 a1 b0 a0
movq J(6), m1
punpckhwd m4, m5 ; r4 = b3 a3 b2 a2
movq m5, m2 ; r5 = c3 c2 c1 c0
punpcklwd m2, m3 ; r2 = d1 c1 d0 c0
movq m1, m0 ; r1 = b1 a1 b0 a0
punpckldq m0, m2 ; r0 = d0 c0 b0 a0 = R0
punpckhdq m1, m2 ; r1 = d1 c1 b1 a1 = R1
movq m2, m4 ; r2 = b3 a3 b2 a2
movq I(0), m0
punpckhwd m5, m3 ; r5 = d3 c3 d2 c2
movq I(1), m1
punpckhdq m4, m5 ; r4 = d3 c3 b3 a3 = R3
punpckldq m2, m5 ; r2 = d2 c2 b2 a2 = R2
movq I(3), m4
movq I(2), m2
%endmacro
%macro VP3_1D_IDCT_SSE2 0
movdqa m2, I(3) ; xmm2 = i3
movdqa m6, C(3) ; xmm6 = c3
movdqa m4, m2 ; xmm4 = i3
movdqa m7, I(5) ; xmm7 = i5
pmulhw m4, m6 ; xmm4 = c3 * i3 - i3
movdqa m1, C(5) ; xmm1 = c5
pmulhw m6, m7 ; xmm6 = c3 * i5 - i5
movdqa m5, m1 ; xmm5 = c5
pmulhw m1, m2 ; xmm1 = c5 * i3 - i3
movdqa m3, I(1) ; xmm3 = i1
pmulhw m5, m7 ; xmm5 = c5 * i5 - i5
movdqa m0, C(1) ; xmm0 = c1
paddw m4, m2 ; xmm4 = c3 * i3
paddw m6, m7 ; xmm6 = c3 * i5
paddw m2, m1 ; xmm2 = c5 * i3
movdqa m1, I(7) ; xmm1 = i7
paddw m7, m5 ; xmm7 = c5 * i5
movdqa m5, m0 ; xmm5 = c1
pmulhw m0, m3 ; xmm0 = c1 * i1 - i1
paddsw m4, m7 ; xmm4 = c3 * i3 + c5 * i5 = C
pmulhw m5, m1 ; xmm5 = c1 * i7 - i7
movdqa m7, C(7) ; xmm7 = c7
psubsw m6, m2 ; xmm6 = c3 * i5 - c5 * i3 = D
paddw m0, m3 ; xmm0 = c1 * i1
pmulhw m3, m7 ; xmm3 = c7 * i1
movdqa m2, I(2) ; xmm2 = i2
pmulhw m7, m1 ; xmm7 = c7 * i7
paddw m5, m1 ; xmm5 = c1 * i7
movdqa m1, m2 ; xmm1 = i2
pmulhw m2, C(2) ; xmm2 = i2 * c2 -i2
psubsw m3, m5 ; xmm3 = c7 * i1 - c1 * i7 = B
movdqa m5, I(6) ; xmm5 = i6
paddsw m0, m7 ; xmm0 = c1 * i1 + c7 * i7 = A
movdqa m7, m5 ; xmm7 = i6
psubsw m0, m4 ; xmm0 = A - C
pmulhw m5, C(2) ; xmm5 = c2 * i6 - i6
paddw m2, m1 ; xmm2 = i2 * c2
pmulhw m1, C(6) ; xmm1 = c6 * i2
paddsw m4, m4 ; xmm4 = C + C
paddsw m4, m0 ; xmm4 = A + C = C.
psubsw m3, m6 ; xmm3 = B - D
paddw m5, m7 ; xmm5 = c2 * i6
paddsw m6, m6 ; xmm6 = D + D
pmulhw m7, C(6) ; xmm7 = c6 * i6
paddsw m6, m3 ; xmm6 = B + D = D.
movdqa I(1), m4 ; Save C. at I(1)
psubsw m1, m5 ; xmm1 = c6 * i2 - c2 * i6 = H
movdqa m4, C(4) ; xmm4 = C4
movdqa m5, m3 ; xmm5 = B - D
pmulhw m3, m4 ; xmm3 = ( c4 -1 ) * ( B - D )
paddsw m7, m2 ; xmm7 = c2 * i2 + c6 * i6 = G
movdqa I(2), m6 ; save D. at I(2)
movdqa m2, m0 ; xmm2 = A - C
movdqa m6, I(0) ; xmm6 = i0
pmulhw m0, m4 ; xmm0 = ( c4 - 1 ) * ( A - C ) = A.
paddw m5, m3 ; xmm5 = c4 * ( B - D ) = B.
movdqa m3, I(4) ; xmm3 = i4
psubsw m5, m1 ; xmm5 = B. - H = B..
paddw m2, m0 ; xmm2 = c4 * ( A - C) = A.
psubsw m6, m3 ; xmm6 = i0 - i4
movdqa m0, m6 ; xmm0 = i0 - i4
pmulhw m6, m4 ; xmm6 = (c4 - 1) * (i0 - i4) = F
paddsw m3, m3 ; xmm3 = i4 + i4
paddsw m1, m1 ; xmm1 = H + H
paddsw m3, m0 ; xmm3 = i0 + i4
paddsw m1, m5 ; xmm1 = B. + H = H.
pmulhw m4, m3 ; xmm4 = ( c4 - 1 ) * ( i0 + i4 )
paddw m6, m0 ; xmm6 = c4 * ( i0 - i4 )
psubsw m6, m2 ; xmm6 = F - A. = F.
paddsw m2, m2 ; xmm2 = A. + A.
movdqa m0, I(1) ; Load C. from I(1)
paddsw m2, m6 ; xmm2 = F + A. = A..
paddw m4, m3 ; xmm4 = c4 * ( i0 + i4 ) = 3
psubsw m2, m1 ; xmm2 = A.. - H. = R2
ADD(m2) ; Adjust R2 and R1 before shifting
paddsw m1, m1 ; xmm1 = H. + H.
paddsw m1, m2 ; xmm1 = A.. + H. = R1
SHIFT(m2) ; xmm2 = op2
psubsw m4, m7 ; xmm4 = E - G = E.
SHIFT(m1) ; xmm1 = op1
movdqa m3, I(2) ; Load D. from I(2)
paddsw m7, m7 ; xmm7 = G + G
paddsw m7, m4 ; xmm7 = E + G = G.
psubsw m4, m3 ; xmm4 = E. - D. = R4
ADD(m4) ; Adjust R4 and R3 before shifting
paddsw m3, m3 ; xmm3 = D. + D.
paddsw m3, m4 ; xmm3 = E. + D. = R3
SHIFT(m4) ; xmm4 = op4
psubsw m6, m5 ; xmm6 = F. - B..= R6
SHIFT(m3) ; xmm3 = op3
ADD(m6) ; Adjust R6 and R5 before shifting
paddsw m5, m5 ; xmm5 = B.. + B..
paddsw m5, m6 ; xmm5 = F. + B.. = R5
SHIFT(m6) ; xmm6 = op6
SHIFT(m5) ; xmm5 = op5
psubsw m7, m0 ; xmm7 = G. - C. = R7
ADD(m7) ; Adjust R7 and R0 before shifting
paddsw m0, m0 ; xmm0 = C. + C.
paddsw m0, m7 ; xmm0 = G. + C.
SHIFT(m7) ; xmm7 = op7
SHIFT(m0) ; xmm0 = op0
%endmacro
%macro PUT_BLOCK 8
movdqa O(0), m%1
movdqa O(1), m%2
movdqa O(2), m%3
movdqa O(3), m%4
movdqa O(4), m%5
movdqa O(5), m%6
movdqa O(6), m%7
movdqa O(7), m%8
%endmacro
vp3: port x86 SIMD to cpuflags.
2012-07-28 01:17:27 +03:00
%macro VP3_IDCT 1
%if mmsize == 16
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
%define I(x) [%1+16*x]
%define O(x) [%1+16*x]
%define C(x) [vp3_idct_data+16*(x-1)]
%define SHIFT(x)
%define ADD(x)
VP3_1D_IDCT_SSE2
config.asm: change %ifdef directives to %if directives. This allows combining multiple conditionals in a single statement.
2012-01-23 12:45:58 +03:00
%if ARCH_X86_64
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
TRANSPOSE8x8W 0, 1, 2, 3, 4, 5, 6, 7, 8
%else
TRANSPOSE8x8W 0, 1, 2, 3, 4, 5, 6, 7, [%1], [%1+16]
%endif
PUT_BLOCK 0, 1, 2, 3, 4, 5, 6, 7
%define SHIFT(x) psraw x, 4
%define ADD(x) paddsw x, [pw_8]
VP3_1D_IDCT_SSE2
PUT_BLOCK 0, 1, 2, 3, 4, 5, 6, 7
vp3: port x86 SIMD to cpuflags.
2012-07-28 01:17:27 +03:00
%else ; mmsize == 8
; eax = quantized input
; ebx = dequantizer matrix
; ecx = IDCT constants
; M(I) = ecx + MaskOffset(0) + I * 8
; C(I) = ecx + CosineOffset(32) + (I-1) * 8
; edx = output
; r0..r7 = mm0..mm7
%define OC_8 [pw_8]
%define C(x) [vp3_idct_data+16*(x-1)]
; at this point, function has completed dequantization + dezigzag +
; partial transposition; now do the idct itself
vp3/x86: use full transpose for all IDCTs. This way, the special IDCT permutations are no longer needed. Bfin code is disabled until someone updates it. This is similar to how H264 does it, and removes the dsputil dependency imposed by the scantable code. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2013-03-12 17:28:12 +03:00
%define I(x) [%1+16*x]
%define J(x) [%1+16*x]
vp3: port x86 SIMD to cpuflags.
2012-07-28 01:17:27 +03:00
RowIDCT
Transpose
vp3/x86: use full transpose for all IDCTs. This way, the special IDCT permutations are no longer needed. Bfin code is disabled until someone updates it. This is similar to how H264 does it, and removes the dsputil dependency imposed by the scantable code. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2013-03-12 17:28:12 +03:00
%define I(x) [%1+16*x+8]
%define J(x) [%1+16*x+8]
vp3: port x86 SIMD to cpuflags.
2012-07-28 01:17:27 +03:00
RowIDCT
Transpose
vp3/x86: use full transpose for all IDCTs. This way, the special IDCT permutations are no longer needed. Bfin code is disabled until someone updates it. This is similar to how H264 does it, and removes the dsputil dependency imposed by the scantable code. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2013-03-12 17:28:12 +03:00
%define I(x) [%1+16* x]
%define J(x) [%1+16*(x-4)+8]
vp3: port x86 SIMD to cpuflags.
2012-07-28 01:17:27 +03:00
ColumnIDCT
vp3/x86: use full transpose for all IDCTs. This way, the special IDCT permutations are no longer needed. Bfin code is disabled until someone updates it. This is similar to how H264 does it, and removes the dsputil dependency imposed by the scantable code. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2013-03-12 17:28:12 +03:00
%define I(x) [%1+16* x +64]
%define J(x) [%1+16*(x-4)+72]
vp3: port x86 SIMD to cpuflags.
2012-07-28 01:17:27 +03:00
ColumnIDCT
%endif ; mmsize == 16/8
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
%endmacro
vp3: port x86 SIMD to cpuflags.
2012-07-28 01:17:27 +03:00
%macro vp3_idct_funcs 0
cglobal vp3_idct_put, 3, 4, 9
VP3_IDCT r2
vp3: don't use calls to inline asm in yasm code. Mixing yasm and inline asm is a bad idea, since if either yasm or inline asm is not supported by your toolchain, all of the asm stops working. Thus, better to use either one or the other alone. Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
2012-07-22 23:38:56 +03:00
movsxdifnidn r1, r1d
mova m4, [pb_80]
lea r3, [r1*3]
%assign %%i 0
%rep 16/mmsize
mova m0, [r2+mmsize*0+%%i]
mova m1, [r2+mmsize*2+%%i]
mova m2, [r2+mmsize*4+%%i]
mova m3, [r2+mmsize*6+%%i]
vp3/x86: use full transpose for all IDCTs. This way, the special IDCT permutations are no longer needed. Bfin code is disabled until someone updates it. This is similar to how H264 does it, and removes the dsputil dependency imposed by the scantable code. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2013-03-12 17:28:12 +03:00
%if mmsize == 8
packsswb m0, [r2+mmsize*8+%%i]
packsswb m1, [r2+mmsize*10+%%i]
packsswb m2, [r2+mmsize*12+%%i]
packsswb m3, [r2+mmsize*14+%%i]
%else
vp3: don't use calls to inline asm in yasm code. Mixing yasm and inline asm is a bad idea, since if either yasm or inline asm is not supported by your toolchain, all of the asm stops working. Thus, better to use either one or the other alone. Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
2012-07-22 23:38:56 +03:00
packsswb m0, [r2+mmsize*1+%%i]
packsswb m1, [r2+mmsize*3+%%i]
packsswb m2, [r2+mmsize*5+%%i]
packsswb m3, [r2+mmsize*7+%%i]
vp3/x86: use full transpose for all IDCTs. This way, the special IDCT permutations are no longer needed. Bfin code is disabled until someone updates it. This is similar to how H264 does it, and removes the dsputil dependency imposed by the scantable code. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2013-03-12 17:28:12 +03:00
%endif
vp3: don't use calls to inline asm in yasm code. Mixing yasm and inline asm is a bad idea, since if either yasm or inline asm is not supported by your toolchain, all of the asm stops working. Thus, better to use either one or the other alone. Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
2012-07-22 23:38:56 +03:00
paddb m0, m4
paddb m1, m4
paddb m2, m4
paddb m3, m4
movq [r0 ], m0
%if mmsize == 8
movq [r0+r1 ], m1
movq [r0+r1*2], m2
movq [r0+r3 ], m3
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
%else
vp3: don't use calls to inline asm in yasm code. Mixing yasm and inline asm is a bad idea, since if either yasm or inline asm is not supported by your toolchain, all of the asm stops working. Thus, better to use either one or the other alone. Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
2012-07-22 23:38:56 +03:00
movhps [r0+r1 ], m0
movq [r0+r1*2], m1
movhps [r0+r3 ], m1
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
%endif
vp3: don't use calls to inline asm in yasm code. Mixing yasm and inline asm is a bad idea, since if either yasm or inline asm is not supported by your toolchain, all of the asm stops working. Thus, better to use either one or the other alone. Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
2012-07-22 23:38:56 +03:00
%if %%i == 0
lea r0, [r0+r1*4]
%endif
%if mmsize == 16
movq [r0 ], m2
movhps [r0+r1 ], m2
movq [r0+r1*2], m3
movhps [r0+r3 ], m3
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
%endif
vp3/x86: use full transpose for all IDCTs. This way, the special IDCT permutations are no longer needed. Bfin code is disabled until someone updates it. This is similar to how H264 does it, and removes the dsputil dependency imposed by the scantable code. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2013-03-12 17:28:12 +03:00
%assign %%i %%i+8
vp3: integrate clear_blocks with idct of previous block. This is identical to what e.g. vp8 does, and prevents the function call overhead (plus dependency on dsputil for this particular function). Arm asm updated by Janne Grunau <janne-libav@jannau.net>. Signed-off-by: Janne Grunau <janne-libav@jannau.net>
2013-01-18 18:43:04 +03:00
%endrep
pxor m0, m0
%assign %%offset 0
%rep 128/mmsize
mova [r2+%%offset], m0
%assign %%offset %%offset+mmsize
vp3: don't use calls to inline asm in yasm code. Mixing yasm and inline asm is a bad idea, since if either yasm or inline asm is not supported by your toolchain, all of the asm stops working. Thus, better to use either one or the other alone. Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
2012-07-22 23:38:56 +03:00
%endrep
RET
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
vp3: port x86 SIMD to cpuflags.
2012-07-28 01:17:27 +03:00
cglobal vp3_idct_add, 3, 4, 9
VP3_IDCT r2
vp3: don't use calls to inline asm in yasm code. Mixing yasm and inline asm is a bad idea, since if either yasm or inline asm is not supported by your toolchain, all of the asm stops working. Thus, better to use either one or the other alone. Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
2012-07-22 23:38:56 +03:00
movsxdifnidn r1, r1d
vp3/x86: use full transpose for all IDCTs. This way, the special IDCT permutations are no longer needed. Bfin code is disabled until someone updates it. This is similar to how H264 does it, and removes the dsputil dependency imposed by the scantable code. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2013-03-12 17:28:12 +03:00
lea r3, [r1*3]
pxor m4, m4
%if mmsize == 16
%assign %%i 0
%rep 2
vp3: don't use calls to inline asm in yasm code. Mixing yasm and inline asm is a bad idea, since if either yasm or inline asm is not supported by your toolchain, all of the asm stops working. Thus, better to use either one or the other alone. Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
2012-07-22 23:38:56 +03:00
movq m0, [r0]
movq m1, [r0+r1]
vp3/x86: use full transpose for all IDCTs. This way, the special IDCT permutations are no longer needed. Bfin code is disabled until someone updates it. This is similar to how H264 does it, and removes the dsputil dependency imposed by the scantable code. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2013-03-12 17:28:12 +03:00
movq m2, [r0+r1*2]
movq m3, [r0+r3]
vp3: don't use calls to inline asm in yasm code. Mixing yasm and inline asm is a bad idea, since if either yasm or inline asm is not supported by your toolchain, all of the asm stops working. Thus, better to use either one or the other alone. Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
2012-07-22 23:38:56 +03:00
punpcklbw m0, m4
punpcklbw m1, m4
vp3/x86: use full transpose for all IDCTs. This way, the special IDCT permutations are no longer needed. Bfin code is disabled until someone updates it. This is similar to how H264 does it, and removes the dsputil dependency imposed by the scantable code. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2013-03-12 17:28:12 +03:00
punpcklbw m2, m4
punpcklbw m3, m4
paddsw m0, [r2+ 0+%%i]
paddsw m1, [r2+16+%%i]
paddsw m2, [r2+32+%%i]
paddsw m3, [r2+48+%%i]
vp3: don't use calls to inline asm in yasm code. Mixing yasm and inline asm is a bad idea, since if either yasm or inline asm is not supported by your toolchain, all of the asm stops working. Thus, better to use either one or the other alone. Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
2012-07-22 23:38:56 +03:00
packuswb m0, m1
vp3/x86: use full transpose for all IDCTs. This way, the special IDCT permutations are no longer needed. Bfin code is disabled until someone updates it. This is similar to how H264 does it, and removes the dsputil dependency imposed by the scantable code. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2013-03-12 17:28:12 +03:00
packuswb m2, m3
movq [r0 ], m0
movhps [r0+r1 ], m0
movq [r0+r1*2], m2
movhps [r0+r3 ], m2
%if %%i == 0
lea r0, [r0+r1*4]
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
%endif
vp3/x86: use full transpose for all IDCTs. This way, the special IDCT permutations are no longer needed. Bfin code is disabled until someone updates it. This is similar to how H264 does it, and removes the dsputil dependency imposed by the scantable code. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2013-03-12 17:28:12 +03:00
%assign %%i %%i+64
%endrep
%else
%assign %%i 0
%rep 2
movq m0, [r0]
movq m1, [r0+r1]
movq m2, [r0+r1*2]
movq m3, [r0+r3]
movq m5, m0
movq m6, m1
movq m7, m2
punpcklbw m0, m4
punpcklbw m1, m4
punpcklbw m2, m4
punpckhbw m5, m4
punpckhbw m6, m4
punpckhbw m7, m4
paddsw m0, [r2+ 0+%%i]
paddsw m1, [r2+16+%%i]
paddsw m2, [r2+32+%%i]
paddsw m5, [r2+64+%%i]
paddsw m6, [r2+80+%%i]
paddsw m7, [r2+96+%%i]
packuswb m0, m5
movq m5, m3
punpcklbw m3, m4
punpckhbw m5, m4
packuswb m1, m6
paddsw m3, [r2+48+%%i]
paddsw m5, [r2+112+%%i]
packuswb m2, m7
packuswb m3, m5
movq [r0 ], m0
movq [r0+r1 ], m1
movq [r0+r1*2], m2
movq [r0+r3 ], m3
%if %%i == 0
lea r0, [r0+r1*4]
vp3: don't use calls to inline asm in yasm code. Mixing yasm and inline asm is a bad idea, since if either yasm or inline asm is not supported by your toolchain, all of the asm stops working. Thus, better to use either one or the other alone. Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
2012-07-22 23:38:56 +03:00
%endif
vp3/x86: use full transpose for all IDCTs. This way, the special IDCT permutations are no longer needed. Bfin code is disabled until someone updates it. This is similar to how H264 does it, and removes the dsputil dependency imposed by the scantable code. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2013-03-12 17:28:12 +03:00
%assign %%i %%i+8
%endrep
%endif
%assign %%i 0
%rep 128/mmsize
mova [r2+%%i], m4
%assign %%i %%i+mmsize
vp3: integrate clear_blocks with idct of previous block. This is identical to what e.g. vp8 does, and prevents the function call overhead (plus dependency on dsputil for this particular function). Arm asm updated by Janne Grunau <janne-libav@jannau.net>. Signed-off-by: Janne Grunau <janne-libav@jannau.net>
2013-01-18 18:43:04 +03:00
%endrep
vp3: don't use calls to inline asm in yasm code. Mixing yasm and inline asm is a bad idea, since if either yasm or inline asm is not supported by your toolchain, all of the asm stops working. Thus, better to use either one or the other alone. Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
2012-07-22 23:38:56 +03:00
RET
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
%endmacro
vp3: don't compile mmx IDCT functions on x86-64. 64-bit CPUs always have SSE2, and a SSE2 version exists, thus the MMX version will never be used.
2012-07-27 08:16:37 +03:00
%if ARCH_X86_32
vp3: port x86 SIMD to cpuflags.
2012-07-28 01:17:27 +03:00
INIT_MMX mmx
vp3_idct_funcs
vp3: don't compile mmx IDCT functions on x86-64. 64-bit CPUs always have SSE2, and a SSE2 version exists, thus the MMX version will never be used.
2012-07-27 08:16:37 +03:00
%endif
vp3: port x86 SIMD to cpuflags.
2012-07-28 01:17:27 +03:00
INIT_XMM sse2
vp3_idct_funcs
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
%macro DC_ADD 0
movq m2, [r0 ]
movq m3, [r0+r1 ]
paddusb m2, m0
movq m4, [r0+r1*2]
paddusb m3, m0
vp3: integrate clear_blocks with idct of previous block. This is identical to what e.g. vp8 does, and prevents the function call overhead (plus dependency on dsputil for this particular function). Arm asm updated by Janne Grunau <janne-libav@jannau.net>. Signed-off-by: Janne Grunau <janne-libav@jannau.net>
2013-01-18 18:43:04 +03:00
movq m5, [r0+r2 ]
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
paddusb m4, m0
paddusb m5, m0
psubusb m2, m1
psubusb m3, m1
movq [r0 ], m2
psubusb m4, m1
movq [r0+r1 ], m3
psubusb m5, m1
movq [r0+r1*2], m4
vp3: integrate clear_blocks with idct of previous block. This is identical to what e.g. vp8 does, and prevents the function call overhead (plus dependency on dsputil for this particular function). Arm asm updated by Janne Grunau <janne-libav@jannau.net>. Signed-off-by: Janne Grunau <janne-libav@jannau.net>
2013-01-18 18:43:04 +03:00
movq [r0+r2 ], m5
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
%endmacro
x86: mmx2 ---> mmxext in asm constructs
2012-07-10 01:04:18 +03:00
INIT_MMX mmxext
vp3: port x86 SIMD to cpuflags.
2012-07-28 01:17:27 +03:00
cglobal vp3_idct_dc_add, 3, 4
config.asm: change %ifdef directives to %if directives. This allows combining multiple conditionals in a single statement.
2012-01-23 12:45:58 +03:00
%if ARCH_X86_64
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
movsxd r1, r1d
%endif
vp3: integrate clear_blocks with idct of previous block. This is identical to what e.g. vp8 does, and prevents the function call overhead (plus dependency on dsputil for this particular function). Arm asm updated by Janne Grunau <janne-libav@jannau.net>. Signed-off-by: Janne Grunau <janne-libav@jannau.net>
2013-01-18 18:43:04 +03:00
movsx r3, word [r2]
mov word [r2], 0
lea r2, [r1*3]
add r3, 15
sar r3, 5
movd m0, r3d
Move VP3 IDCT functions from inline ASM to YASM. This fixes part of the VP3/5/6 issues on Win64. Originally committed as revision 24988 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-30 19:25:46 +03:00
pshufw m0, m0, 0x0
pxor m1, m1
psubw m1, m0
packuswb m0, m0
packuswb m1, m1
DC_ADD
lea r0, [r0+r1*4]
DC_ADD
RET
Reference in New Issue Copy Permalink