dct-a.asm 31.3 KB
Newer Older
1 2 3 4 5
;*****************************************************************************
;* dct-a.asm: h264 encoder library
;*****************************************************************************
;* Copyright (C) 2003-2008 x264 project
;*
6
;* Authors: Holger Lubitz <holger@lubitz.org>
7
;*          Loren Merritt <lorenm@u.washington.edu>
8
;*          Laurent Aimar <fenrir@via.ecp.fr>
9
;*          Min Chen <chenm001.163.com>
10 11 12 13 14 15 16 17 18 19 20 21 22
;*
;* This program is free software; you can redistribute it and/or modify
;* it under the terms of the GNU General Public License as published by
;* the Free Software Foundation; either version 2 of the License, or
;* (at your option) any later version.
;*
;* This program 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 General Public License for more details.
;*
;* You should have received a copy of the GNU General Public License
;* along with this program; if not, write to the Free Software
23
;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02111, USA.
24 25 26
;*****************************************************************************

%include "x86inc.asm"
27
%include "x86util.asm"
28

29 30 31 32 33 34 35 36
%macro SHUFFLE_16BIT 8
    %rep 8
        db %1*2
        db %1*2+1
        %rotate 1
    %endrep
%endmacro

37
SECTION_RODATA
Holger Lubitz's avatar
Holger Lubitz committed
38 39
pw_32_0: times 4 dw 32
         times 4 dw 0
40
pw_32: times 8 dw 32
41
pw_8000: times 8 dw 0x8000
42
hsub_mul: times 8 db 1, -1
43

Holger Lubitz's avatar
Holger Lubitz committed
44
pb_sub4frame:   db 0,1,4,8,5,2,3,6,9,12,13,10,7,11,14,15
45 46
pb_sub4field:   db 0,4,1,8,12,5,9,13,2,6,10,14,3,7,11,15
pb_subacmask:   dw 0,-1,-1,-1,-1,-1,-1,-1
47 48
pb_scan4framea: SHUFFLE_16BIT 6,3,7,0,4,1,2,5
pb_scan4frameb: SHUFFLE_16BIT 0,4,1,2,5,6,3,7
49
pb_idctdc_unpack: db 0,0,0,0,1,1,1,1,2,2,2,2,3,3,3,3
50
pb_idctdc_unpack2: db 4,4,4,4,5,5,5,5,6,6,6,6,7,7,7,7
Fiona Glaser's avatar
Fiona Glaser committed
51
pb_1: times 16 db 1
52
pw_1: times 8 dw 1
53 54 55

SECTION .text

56 57 58
%macro WALSH4_1D 5
    SUMSUB_BADC m%4, m%3, m%2, m%1, m%5
    SUMSUB_BADC m%4, m%2, m%3, m%1, m%5
59 60 61
    SWAP %1, %4, %3
%endmacro

62 63
%macro SUMSUB_17BIT 4 ; a, b, tmp, 0x8000
    movq  m%3, m%4
64
    pxor  m%1, m%4
65
    psubw m%3, m%2
66
    pxor  m%2, m%4
67 68
    pavgw m%3, m%1
    pavgw m%2, m%1
69 70
    pxor  m%3, m%4
    pxor  m%2, m%4
71 72 73
    SWAP %1, %2, %3
%endmacro

74
INIT_MMX
75 76 77
;-----------------------------------------------------------------------------
; void x264_dct4x4dc_mmx( int16_t d[4][4] )
;-----------------------------------------------------------------------------
Loren Merritt's avatar
Loren Merritt committed
78
cglobal x264_dct4x4dc_mmx, 1,1
79
    movq   m3, [r0+24]
80 81 82
    movq   m2, [r0+16]
    movq   m1, [r0+ 8]
    movq   m0, [r0+ 0]
83
    movq   m7, [pw_8000] ; convert to unsigned and back, so that pavgw works
84
    WALSH4_1D  0,1,2,3,4
85
    TRANSPOSE4x4W 0,1,2,3,4
86 87 88
    SUMSUB_BADC m1, m0, m3, m2, m4
    SWAP 0, 1
    SWAP 2, 3
89 90
    SUMSUB_17BIT 0,2,4,7
    SUMSUB_17BIT 1,3,5,7
91
    movq  [r0+0], m0
92 93 94
    movq  [r0+8], m2
    movq [r0+16], m3
    movq [r0+24], m1
95 96 97 98 99 100
    RET

;-----------------------------------------------------------------------------
; void x264_idct4x4dc_mmx( int16_t d[4][4] )
;-----------------------------------------------------------------------------
cglobal x264_idct4x4dc_mmx, 1,1
101 102 103 104 105
    movq   m3, [r0+24]
    movq   m2, [r0+16]
    movq   m1, [r0+ 8]
    movq   m0, [r0+ 0]
    WALSH4_1D  0,1,2,3,4
106
    TRANSPOSE4x4W 0,1,2,3,4
107
    WALSH4_1D  0,1,2,3,4
108 109 110 111 112
    movq  [r0+ 0], m0
    movq  [r0+ 8], m1
    movq  [r0+16], m2
    movq  [r0+24], m3
    RET
113

114
%macro SUB_DCT4 1
115 116 117
;-----------------------------------------------------------------------------
; void x264_sub4x4_dct_mmx( int16_t dct[4][4], uint8_t *pix1, uint8_t *pix2 )
;-----------------------------------------------------------------------------
118 119
cglobal x264_sub4x4_dct_%1, 3,3
%ifidn %1, mmx
120
.skip_prologue:
121 122 123 124 125
    LOAD_DIFF  m0, m4, m5, [r1+0*FENC_STRIDE], [r2+0*FDEC_STRIDE]
    LOAD_DIFF  m3, m4, m5, [r1+3*FENC_STRIDE], [r2+3*FDEC_STRIDE]
    LOAD_DIFF  m1, m4, m5, [r1+1*FENC_STRIDE], [r2+1*FDEC_STRIDE]
    LOAD_DIFF  m2, m4, m5, [r1+2*FENC_STRIDE], [r2+2*FDEC_STRIDE]
%else
126
    mova m5, [hsub_mul]
127 128
    LOAD_DIFF8x4_SSSE3 0, 3, 1, 2, 4, 5, r1, r2
%endif
129
    DCT4_1D 0,1,2,3,4
130
    TRANSPOSE4x4W 0,1,2,3,4
131 132 133 134 135
    DCT4_1D 0,1,2,3,4
    movq  [r0+ 0], m0
    movq  [r0+ 8], m1
    movq  [r0+16], m2
    movq  [r0+24], m3
136
    RET
137 138 139 140
%endmacro

SUB_DCT4 mmx
SUB_DCT4 ssse3
141 142 143 144

;-----------------------------------------------------------------------------
; void x264_add4x4_idct_mmx( uint8_t *p_dst, int16_t dct[4][4] )
;-----------------------------------------------------------------------------
Loren Merritt's avatar
Loren Merritt committed
145
cglobal x264_add4x4_idct_mmx, 2,2
146
    pxor m7, m7
147
.skip_prologue:
148 149
    movq  m1, [r1+ 8]
    movq  m3, [r1+24]
150 151
    movq  m2, [r1+16]
    movq  m0, [r1+ 0]
152
    IDCT4_1D 0,1,2,3,4,5
153
    TRANSPOSE4x4W 0,1,2,3,4
154
    paddw m0, [pw_32]
155
    IDCT4_1D 0,1,2,3,4,5
156 157 158 159
    STORE_DIFF  m0, m4, m7, [r0+0*FDEC_STRIDE]
    STORE_DIFF  m1, m4, m7, [r0+1*FDEC_STRIDE]
    STORE_DIFF  m2, m4, m7, [r0+2*FDEC_STRIDE]
    STORE_DIFF  m3, m4, m7, [r0+3*FDEC_STRIDE]
160 161
    RET

Holger Lubitz's avatar
Holger Lubitz committed
162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181
INIT_XMM
cglobal x264_add4x4_idct_sse4, 2,2,6
    mova      m0, [r1+0x00]     ; row1/row0
    mova      m2, [r1+0x10]     ; row3/row2
    mova      m1, m0            ; row1/row0
    psraw     m0, 1             ; row1>>1/...
    mova      m3, m2            ; row3/row2
    psraw     m2, 1             ; row3>>1/...
    movsd     m0, m1            ; row1>>1/row0
    movsd     m2, m3            ; row3>>1/row2
    psubw     m0, m3            ; row1>>1-row3/row0-2
    paddw     m2, m1            ; row3>>1+row1/row0+2
    SBUTTERFLY2 wd, 0, 2, 1
    SUMSUB_BA m2, m0, m1
    pshuflw   m1, m2, 10110001b
    pshufhw   m2, m2, 10110001b
    punpckldq m1, m0
    punpckhdq m2, m0
    SWAP 0, 1

182
    mova      m1, [pw_32_0]
Holger Lubitz's avatar
Holger Lubitz committed
183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214
    paddw     m1, m0            ; row1/row0 corrected
    psraw     m0, 1             ; row1>>1/...
    mova      m3, m2            ; row3/row2
    psraw     m2, 1             ; row3>>1/...
    movsd     m0, m1            ; row1>>1/row0
    movsd     m2, m3            ; row3>>1/row2
    psubw     m0, m3            ; row1>>1-row3/row0-2
    paddw     m2, m1            ; row3>>1+row1/row0+2
    SBUTTERFLY2 qdq, 0, 2, 1
    SUMSUB_BA m2, m0, m1

    movd      m4, [r0+FDEC_STRIDE*0]
    movd      m1, [r0+FDEC_STRIDE*1]
    movd      m3, [r0+FDEC_STRIDE*2]
    movd      m5, [r0+FDEC_STRIDE*3]
    punpckldq m1, m4            ; row0/row1
    pxor      m4, m4
    punpckldq m3, m5            ; row3/row2
    punpcklbw m1, m4
    psraw     m2, 6
    punpcklbw m3, m4
    psraw     m0, 6
    paddsw    m2, m1
    paddsw    m0, m3
    packuswb  m0, m2            ; row0/row1/row3/row2
    pextrd   [r0+FDEC_STRIDE*0], m0, 3
    pextrd   [r0+FDEC_STRIDE*1], m0, 2
    movd     [r0+FDEC_STRIDE*2], m0
    pextrd   [r0+FDEC_STRIDE*3], m0, 1
    RET

INIT_MMX
215 216 217 218
;-----------------------------------------------------------------------------
; void x264_sub8x8_dct_mmx( int16_t dct[4][4][4], uint8_t *pix1, uint8_t *pix2 )
;-----------------------------------------------------------------------------
%macro SUB_NxN_DCT 6
219 220 221 222 223
cglobal %1, 3,3,11
%if mmsize == 8
    pxor m7, m7
%else
    add r2, 4*FDEC_STRIDE
224
    mova m7, [hsub_mul]
225
%endif
226
.skip_prologue:
Anton Mitrofanov's avatar
Anton Mitrofanov committed
227 228 229
%ifdef WIN64
    sub  rsp, 8
%endif
230 231
    call %2
    add  r0, %3
Loren Merritt's avatar
Loren Merritt committed
232 233
    add  r1, %4-%5-%6*FENC_STRIDE
    add  r2, %4-%5-%6*FDEC_STRIDE
234 235
    call %2
    add  r0, %3
Loren Merritt's avatar
Loren Merritt committed
236 237
    add  r1, (%4-%6)*FENC_STRIDE-%5-%4
    add  r2, (%4-%6)*FDEC_STRIDE-%5-%4
238 239
    call %2
    add  r0, %3
Loren Merritt's avatar
Loren Merritt committed
240 241
    add  r1, %4-%5-%6*FENC_STRIDE
    add  r2, %4-%5-%6*FDEC_STRIDE
Anton Mitrofanov's avatar
Anton Mitrofanov committed
242 243
%ifdef WIN64
    add  rsp, 8
244 245 246
    call %2
    RET
%else
247
    jmp  %2
248
%endif
249 250 251 252 253
%endmacro

;-----------------------------------------------------------------------------
; void x264_add8x8_idct_mmx( uint8_t *pix, int16_t dct[4][4][4] )
;-----------------------------------------------------------------------------
254 255 256 257 258 259
%macro ADD_NxN_IDCT 6-7
cglobal %1, 2,2,11
    pxor m7, m7
%if mmsize==16
    add  r0, 4*FDEC_STRIDE
%endif
260
.skip_prologue:
Anton Mitrofanov's avatar
Anton Mitrofanov committed
261 262 263
%ifdef WIN64
    sub  rsp, 8
%endif
264
    call %2
Loren Merritt's avatar
Loren Merritt committed
265
    add  r0, %4-%5-%6*FDEC_STRIDE
266 267
    add  r1, %3
    call %2
Loren Merritt's avatar
Loren Merritt committed
268
    add  r0, (%4-%6)*FDEC_STRIDE-%5-%4
269 270
    add  r1, %3
    call %2
Loren Merritt's avatar
Loren Merritt committed
271
    add  r0, %4-%5-%6*FDEC_STRIDE
272
    add  r1, %3
Anton Mitrofanov's avatar
Anton Mitrofanov committed
273 274
%ifdef WIN64
    add  rsp, 8
275 276 277
    call %2
    RET
%else
278
    jmp  %2
279
%endif
280 281
%endmacro

282
%ifndef ARCH_X86_64
Anton Mitrofanov's avatar
Anton Mitrofanov committed
283 284 285 286
SUB_NxN_DCT  x264_sub8x8_dct_mmx,    x264_sub4x4_dct_mmx.skip_prologue,  32, 4, 0, 0
ADD_NxN_IDCT x264_add8x8_idct_mmx,   x264_add4x4_idct_mmx.skip_prologue, 32, 4, 0, 0
SUB_NxN_DCT  x264_sub16x16_dct_mmx,  x264_sub8x8_dct_mmx.skip_prologue,  32, 8, 4, 4
ADD_NxN_IDCT x264_add16x16_idct_mmx, x264_add8x8_idct_mmx.skip_prologue, 32, 8, 4, 4
287

Loren Merritt's avatar
Loren Merritt committed
288 289
cextern x264_sub8x8_dct8_mmx.skip_prologue
cextern x264_add8x8_idct8_mmx.skip_prologue
Anton Mitrofanov's avatar
Anton Mitrofanov committed
290 291
SUB_NxN_DCT  x264_sub16x16_dct8_mmx,  x264_sub8x8_dct8_mmx.skip_prologue,  128, 8, 0, 0
ADD_NxN_IDCT x264_add16x16_idct8_mmx, x264_add8x8_idct8_mmx.skip_prologue, 128, 8, 0, 0
Loren Merritt's avatar
Loren Merritt committed
292 293
%endif

294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309
INIT_XMM

cextern x264_sub8x8_dct_sse2.skip_prologue
cextern x264_sub8x8_dct_ssse3.skip_prologue
SUB_NxN_DCT  x264_sub16x16_dct_sse2, x264_sub8x8_dct_sse2.skip_prologue, 128, 8, 0, 0
SUB_NxN_DCT  x264_sub16x16_dct_ssse3, x264_sub8x8_dct_ssse3.skip_prologue, 128, 8, 0, 0
cextern x264_add8x8_idct_sse2.skip_prologue
ADD_NxN_IDCT x264_add16x16_idct_sse2, x264_add8x8_idct_sse2.skip_prologue, 2*64, 8, 0, 0

cextern x264_sub8x8_dct8_sse2.skip_prologue
cextern x264_add8x8_idct8_sse2.skip_prologue
SUB_NxN_DCT  x264_sub16x16_dct8_sse2,  x264_sub8x8_dct8_sse2.skip_prologue,  128, 8, 0, 0
ADD_NxN_IDCT x264_add16x16_idct8_sse2, x264_add8x8_idct8_sse2.skip_prologue, 128, 8, 0, 0

cextern x264_sub8x8_dct8_ssse3.skip_prologue
SUB_NxN_DCT  x264_sub16x16_dct8_ssse3,  x264_sub8x8_dct8_ssse3.skip_prologue,  128, 8, 0, 0
310

311

312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337
;-----------------------------------------------------------------------------
; void add8x8_idct_dc( uint8_t *p_dst, int16_t *dct2x2 )
;-----------------------------------------------------------------------------

%macro ADD_DC 3
    movq      mm4, [%3+FDEC_STRIDE*0]
    movq      mm5, [%3+FDEC_STRIDE*1]
    movq      mm6, [%3+FDEC_STRIDE*2]
    paddusb   mm4, %1
    paddusb   mm5, %1
    paddusb   mm6, %1
    paddusb    %1, [%3+FDEC_STRIDE*3]
    psubusb   mm4, %2
    psubusb   mm5, %2
    psubusb   mm6, %2
    psubusb    %1, %2
    movq      [%3+FDEC_STRIDE*0], mm4
    movq      [%3+FDEC_STRIDE*1], mm5
    movq      [%3+FDEC_STRIDE*2], mm6
    movq      [%3+FDEC_STRIDE*3], %1
%endmacro

cglobal x264_add8x8_idct_dc_mmx, 2,2
    movq      mm0, [r1]
    pxor      mm1, mm1
    add        r0, FDEC_STRIDE*4
338
    paddw     mm0, [pw_32]
339 340 341 342 343 344 345 346 347 348 349 350
    psraw     mm0, 6
    psubw     mm1, mm0
    packuswb  mm0, mm0
    packuswb  mm1, mm1
    punpcklbw mm0, mm0
    punpcklbw mm1, mm1
    pshufw    mm2, mm0, 0xFA
    pshufw    mm3, mm1, 0xFA
    punpcklbw mm0, mm0
    punpcklbw mm1, mm1
    ADD_DC    mm0, mm1, r0-FDEC_STRIDE*4
    ADD_DC    mm2, mm3, r0
Anton Mitrofanov's avatar
Anton Mitrofanov committed
351
    RET
352 353 354 355 356

cglobal x264_add8x8_idct_dc_ssse3, 2,2
    movq      xmm0, [r1]
    pxor      xmm1, xmm1
    add         r0, FDEC_STRIDE*4
357
    paddw     xmm0, [pw_32]
358 359
    psraw     xmm0, 6
    psubw     xmm1, xmm0
360
    movdqa    xmm5, [pb_idctdc_unpack]
361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388
    packuswb  xmm0, xmm0
    packuswb  xmm1, xmm1
    pshufb    xmm0, xmm5
    pshufb    xmm1, xmm5
    movq      xmm2, [r0+FDEC_STRIDE*-4]
    movq      xmm3, [r0+FDEC_STRIDE*-3]
    movq      xmm4, [r0+FDEC_STRIDE*-2]
    movq      xmm5, [r0+FDEC_STRIDE*-1]
    movhps    xmm2, [r0+FDEC_STRIDE* 0]
    movhps    xmm3, [r0+FDEC_STRIDE* 1]
    movhps    xmm4, [r0+FDEC_STRIDE* 2]
    movhps    xmm5, [r0+FDEC_STRIDE* 3]
    paddusb   xmm2, xmm0
    paddusb   xmm3, xmm0
    paddusb   xmm4, xmm0
    paddusb   xmm5, xmm0
    psubusb   xmm2, xmm1
    psubusb   xmm3, xmm1
    psubusb   xmm4, xmm1
    psubusb   xmm5, xmm1
    movq      [r0+FDEC_STRIDE*-4], xmm2
    movq      [r0+FDEC_STRIDE*-3], xmm3
    movq      [r0+FDEC_STRIDE*-2], xmm4
    movq      [r0+FDEC_STRIDE*-1], xmm5
    movhps    [r0+FDEC_STRIDE* 0], xmm2
    movhps    [r0+FDEC_STRIDE* 1], xmm3
    movhps    [r0+FDEC_STRIDE* 2], xmm4
    movhps    [r0+FDEC_STRIDE* 3], xmm5
Anton Mitrofanov's avatar
Anton Mitrofanov committed
389
    RET
390

391 392 393 394 395
cglobal x264_add16x16_idct_dc_mmx, 2,3
    mov       r2, 4
.loop:
    movq      mm0, [r1]
    pxor      mm1, mm1
396
    paddw     mm0, [pw_32]
397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412
    psraw     mm0, 6
    psubw     mm1, mm0
    packuswb  mm0, mm0
    packuswb  mm1, mm1
    punpcklbw mm0, mm0
    punpcklbw mm1, mm1
    pshufw    mm2, mm0, 0xFA
    pshufw    mm3, mm1, 0xFA
    punpcklbw mm0, mm0
    punpcklbw mm1, mm1
    ADD_DC    mm0, mm1, r0
    ADD_DC    mm2, mm3, r0+8
    add       r1, 8
    add       r0, FDEC_STRIDE*4
    dec       r2
    jg .loop
Anton Mitrofanov's avatar
Anton Mitrofanov committed
413
    REP_RET
414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433

%macro IDCT_DC_STORE 3
    movdqa    xmm4, [r0+%1+FDEC_STRIDE*0]
    movdqa    xmm5, [r0+%1+FDEC_STRIDE*1]
    movdqa    xmm6, [r0+%1+FDEC_STRIDE*2]
    movdqa    xmm7, [r0+%1+FDEC_STRIDE*3]
    paddusb   xmm4, %2
    paddusb   xmm5, %2
    paddusb   xmm6, %2
    paddusb   xmm7, %2
    psubusb   xmm4, %3
    psubusb   xmm5, %3
    psubusb   xmm6, %3
    psubusb   xmm7, %3
    movdqa    [r0+%1+FDEC_STRIDE*0], xmm4
    movdqa    [r0+%1+FDEC_STRIDE*1], xmm5
    movdqa    [r0+%1+FDEC_STRIDE*2], xmm6
    movdqa    [r0+%1+FDEC_STRIDE*3], xmm7
%endmacro

Anton Mitrofanov's avatar
Anton Mitrofanov committed
434
cglobal x264_add16x16_idct_dc_sse2, 2,2,8
435 436
    call .loop
    add       r0, FDEC_STRIDE*4
Anton Mitrofanov's avatar
Anton Mitrofanov committed
437 438 439 440
%ifdef WIN64
    call .loop
    RET
%endif
441 442 443 444 445 446 447 448 449
.loop:
    add       r0, FDEC_STRIDE*4
    movq      xmm0, [r1+0]
    movq      xmm2, [r1+8]
    add       r1, 16
    punpcklwd xmm0, xmm0
    punpcklwd xmm2, xmm2
    pxor      xmm1, xmm1
    pxor      xmm3, xmm3
450 451
    paddw     xmm0, [pw_32]
    paddw     xmm2, [pw_32]
452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467
    psraw     xmm0, 6
    psraw     xmm2, 6
    psubw     xmm1, xmm0
    psubw     xmm3, xmm2
    packuswb  xmm0, xmm1
    packuswb  xmm2, xmm3
    movdqa    xmm1, xmm0
    movdqa    xmm3, xmm2
    punpcklbw xmm0, xmm0
    punpcklbw xmm2, xmm2
    punpckhbw xmm1, xmm1
    punpckhbw xmm3, xmm3
    IDCT_DC_STORE FDEC_STRIDE*-4, xmm0, xmm1
    IDCT_DC_STORE 0, xmm2, xmm3
    ret

Anton Mitrofanov's avatar
Anton Mitrofanov committed
468
cglobal x264_add16x16_idct_dc_ssse3, 2,2,8
469 470
    call .loop
    add       r0, FDEC_STRIDE*4
Anton Mitrofanov's avatar
Anton Mitrofanov committed
471 472 473 474
%ifdef WIN64
    call .loop
    RET
%endif
475 476 477 478 479
.loop:
    add       r0, FDEC_STRIDE*4
    movdqa    xmm0, [r1]
    add       r1, 16
    pxor      xmm1, xmm1
480
    paddw     xmm0, [pw_32]
481 482
    psraw     xmm0, 6
    psubw     xmm1, xmm0
483 484
    movdqa    xmm5, [ pb_idctdc_unpack]
    movdqa    xmm6, [pb_idctdc_unpack2]
485 486 487 488 489 490 491 492 493 494 495 496
    packuswb  xmm0, xmm0
    packuswb  xmm1, xmm1
    movdqa    xmm2, xmm0
    movdqa    xmm3, xmm1
    pshufb    xmm0, xmm5
    pshufb    xmm2, xmm6
    pshufb    xmm1, xmm5
    pshufb    xmm3, xmm6
    IDCT_DC_STORE FDEC_STRIDE*-4, xmm0, xmm1
    IDCT_DC_STORE 0, xmm2, xmm3
    ret

497 498 499 500 501 502 503 504 505 506 507 508 509 510 511
;-----------------------------------------------------------------------------
; void sub8x8_dct_dc( int16_t dct[2][2], uint8_t *pix1, uint8_t *pix2 )
;-----------------------------------------------------------------------------

%macro DCTDC_2ROW_MMX 3
    movq      %1, [r1+FENC_STRIDE*(0+%3)]
    movq      m1, [r1+FENC_STRIDE*(1+%3)]
    movq      m2, [r2+FDEC_STRIDE*(0+%3)]
    movq      m3, [r2+FDEC_STRIDE*(1+%3)]
    movq      %2, %1
    punpckldq %1, m1
    punpckhdq %2, m1
    movq      m1, m2
    punpckldq m2, m3
    punpckhdq m1, m3
512 513 514 515 516
    pxor      m3, m3
    psadbw    %1, m3
    psadbw    %2, m3
    psadbw    m2, m3
    psadbw    m1, m3
517 518 519 520
    psubw     %1, m2
    psubw     %2, m1
%endmacro

521 522 523 524 525 526 527 528 529 530 531 532
%macro DCT2x2 2 ; reg s1/s0 (!=m1), reg s3/s2
    pshufw    mm1, %1, 10100000b  ;  s1  s1  s0  s0
    pshufw    mm0, %2, 10110001b  ;  s3  __  s2  __
    paddw     mm1, %2             ;  s1 s13  s0 s02
    psubw     mm1, mm0            ; d13 s13 d02 s02
    pshufw    mm0, mm1, 01000100b ; d02 s02 d02 s02
    psrlq     mm1, 32             ;  __  __ d13 s13
    paddw     mm0, mm1            ; d02 s02 d02+d13 s02+s13
    psllq     mm1, 32             ; d13 s13
    psubw     mm0, mm1            ; d02-d13 s02-s13 d02+d13 s02+s13
%endmacro

533 534 535 536 537 538
INIT_MMX
cglobal x264_sub8x8_dct_dc_mmxext, 3,3
    DCTDC_2ROW_MMX m0, m4, 0
    DCTDC_2ROW_MMX m5, m6, 2
    paddw     m0, m5
    paddw     m4, m6
539 540 541 542 543 544 545 546 547 548
    punpckldq m0, m4
    add       r1, FENC_STRIDE*4
    add       r2, FDEC_STRIDE*4
    DCTDC_2ROW_MMX m7, m4, 0
    DCTDC_2ROW_MMX m5, m6, 2
    paddw     m7, m5
    paddw     m4, m6
    punpckldq m7, m4
    DCT2x2    m0, m7
    movq    [r0], m0
549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
    ret

INIT_XMM
%macro DCTDC_2ROW_SSE2 3
    movq      m0, [r1+FENC_STRIDE*(0+%1)]
    movq      m1, [r1+FENC_STRIDE*(1+%1)]
    movq      m2, [r2+FDEC_STRIDE*(0+%1)]
    movq      m3, [r2+FDEC_STRIDE*(1+%1)]
    punpckldq m0, m1
    punpckldq m2, m3
    psadbw    m0, m7
    psadbw    m2, m7
%if %2
    paddw     %3, m0
    paddw     m6, m2
%else
    SWAP      %3, m0
    SWAP      m6, m2
%endif
%endmacro

cglobal x264_sub8x8_dct_dc_sse2, 3,3,8
    pxor     m7, m7
    DCTDC_2ROW_SSE2 0, 0, m4
    DCTDC_2ROW_SSE2 2, 1, m4
    add      r1, FENC_STRIDE*4
    add      r2, FDEC_STRIDE*4
576
    psubd    m4, m6
577 578
    DCTDC_2ROW_SSE2 0, 0, m5
    DCTDC_2ROW_SSE2 2, 1, m5
579
    psubd    m5, m6
580
    packssdw m4, m5
581 582 583 584 585
    movhlps  m5, m4
    movdq2q mm0, m4
    movdq2q mm7, m5
    DCT2x2  mm0, mm7
    movq   [r0], mm0
586 587
    RET

Holger Lubitz's avatar
Holger Lubitz committed
588 589 590 591
;-----------------------------------------------------------------------------
; void x264_zigzag_scan_8x8_frame_ssse3( int16_t level[64], int16_t dct[8][8] )
;-----------------------------------------------------------------------------
%macro SCAN_8x8 1
Anton Mitrofanov's avatar
Anton Mitrofanov committed
592
cglobal x264_zigzag_scan_8x8_frame_%1, 2,2,8
Holger Lubitz's avatar
Holger Lubitz committed
593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647
    movdqa    xmm0, [r1]
    movdqa    xmm1, [r1+16]
    movdq2q    mm0, xmm0
    PALIGNR   xmm1, xmm1, 14, xmm2
    movdq2q    mm1, xmm1

    movdqa    xmm2, [r1+32]
    movdqa    xmm3, [r1+48]
    PALIGNR   xmm2, xmm2, 12, xmm4
    movdq2q    mm2, xmm2
    PALIGNR   xmm3, xmm3, 10, xmm4
    movdq2q    mm3, xmm3

    punpckhwd xmm0, xmm1
    punpckhwd xmm2, xmm3

    movq       mm4, mm1
    movq       mm5, mm1
    movq       mm6, mm2
    movq       mm7, mm3
    punpckhwd  mm1, mm0
    psllq      mm0, 16
    psrlq      mm3, 16
    punpckhdq  mm1, mm1
    punpckhdq  mm2, mm0
    punpcklwd  mm0, mm4
    punpckhwd  mm4, mm3
    punpcklwd  mm4, mm2
    punpckhdq  mm0, mm2
    punpcklwd  mm6, mm3
    punpcklwd  mm5, mm7
    punpcklwd  mm5, mm6

    movdqa    xmm4, [r1+64]
    movdqa    xmm5, [r1+80]
    movdqa    xmm6, [r1+96]
    movdqa    xmm7, [r1+112]

    movq [r0+2*00], mm0
    movq [r0+2*04], mm4
    movd [r0+2*08], mm1
    movq [r0+2*36], mm5
    movq [r0+2*46], mm6

    PALIGNR   xmm4, xmm4, 14, xmm3
    movdq2q    mm4, xmm4
    PALIGNR   xmm5, xmm5, 12, xmm3
    movdq2q    mm5, xmm5
    PALIGNR   xmm6, xmm6, 10, xmm3
    movdq2q    mm6, xmm6
%ifidn %1, ssse3
    PALIGNR   xmm7, xmm7, 8, xmm3
    movdq2q    mm7, xmm7
%else
    movhlps   xmm3, xmm7
648
    punpcklqdq xmm7, xmm7
Holger Lubitz's avatar
Holger Lubitz committed
649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828
    movdq2q    mm7, xmm3
%endif

    punpckhwd xmm4, xmm5
    punpckhwd xmm6, xmm7

    movq       mm0, mm4
    movq       mm1, mm5
    movq       mm3, mm7
    punpcklwd  mm7, mm6
    psrlq      mm6, 16
    punpcklwd  mm4, mm6
    punpcklwd  mm5, mm4
    punpckhdq  mm4, mm3
    punpcklwd  mm3, mm6
    punpckhwd  mm3, mm4
    punpckhwd  mm0, mm1
    punpckldq  mm4, mm0
    punpckhdq  mm0, mm6
    pshufw     mm4, mm4, 0x6c

    movq [r0+2*14], mm4
    movq [r0+2*25], mm0
    movd [r0+2*54], mm7
    movq [r0+2*56], mm5
    movq [r0+2*60], mm3

    movdqa    xmm3, xmm0
    movdqa    xmm7, xmm4
    punpckldq xmm0, xmm2
    punpckldq xmm4, xmm6
    punpckhdq xmm3, xmm2
    punpckhdq xmm7, xmm6
    pshufhw   xmm0, xmm0, 0x1b
    pshuflw   xmm4, xmm4, 0x1b
    pshufhw   xmm3, xmm3, 0x1b
    pshuflw   xmm7, xmm7, 0x1b

    movlps [r0+2*10], xmm0
    movhps [r0+2*17], xmm0
    movlps [r0+2*21], xmm3
    movlps [r0+2*28], xmm4
    movhps [r0+2*32], xmm3
    movhps [r0+2*39], xmm4
    movlps [r0+2*43], xmm7
    movhps [r0+2*50], xmm7

    RET
%endmacro

INIT_XMM
%define PALIGNR PALIGNR_MMX
SCAN_8x8 sse2
%define PALIGNR PALIGNR_SSSE3
SCAN_8x8 ssse3

;-----------------------------------------------------------------------------
; void x264_zigzag_scan_8x8_frame_mmxext( int16_t level[64], int16_t dct[8][8] )
;-----------------------------------------------------------------------------
cglobal x264_zigzag_scan_8x8_frame_mmxext, 2,2
    movq       mm0, [r1]
    movq       mm1, [r1+2*8]
    movq       mm2, [r1+2*14]
    movq       mm3, [r1+2*21]
    movq       mm4, [r1+2*28]
    movq       mm5, mm0
    movq       mm6, mm1
    psrlq      mm0, 16
    punpckldq  mm1, mm1
    punpcklwd  mm5, mm6
    punpckhwd  mm1, mm3
    punpckhwd  mm6, mm0
    punpckldq  mm5, mm0
    movq       mm7, [r1+2*52]
    movq       mm0, [r1+2*60]
    punpckhwd  mm1, mm2
    punpcklwd  mm2, mm4
    punpckhwd  mm4, mm3
    punpckldq  mm3, mm3
    punpckhwd  mm3, mm2
    movq      [r0], mm5
    movq  [r0+2*4], mm1
    movq  [r0+2*8], mm6
    punpcklwd  mm6, mm0
    punpcklwd  mm6, mm7
    movq       mm1, [r1+2*32]
    movq       mm5, [r1+2*39]
    movq       mm2, [r1+2*46]
    movq [r0+2*35], mm3
    movq [r0+2*47], mm4
    punpckhwd  mm7, mm0
    psllq      mm0, 16
    movq       mm3, mm5
    punpcklwd  mm5, mm1
    punpckhwd  mm1, mm2
    punpckhdq  mm3, mm3
    movq [r0+2*52], mm6
    movq [r0+2*13], mm5
    movq       mm4, [r1+2*11]
    movq       mm6, [r1+2*25]
    punpcklwd  mm5, mm7
    punpcklwd  mm1, mm3
    punpckhdq  mm0, mm7
    movq       mm3, [r1+2*4]
    movq       mm7, [r1+2*18]
    punpcklwd  mm2, mm5
    movq [r0+2*25], mm1
    movq       mm1, mm4
    movq       mm5, mm6
    punpcklwd  mm4, mm3
    punpcklwd  mm6, mm7
    punpckhwd  mm1, mm3
    punpckhwd  mm5, mm7
    movq       mm3, mm6
    movq       mm7, mm5
    punpckldq  mm6, mm4
    punpckldq  mm5, mm1
    punpckhdq  mm3, mm4
    punpckhdq  mm7, mm1
    movq       mm4, [r1+2*35]
    movq       mm1, [r1+2*49]
    pshufw     mm6, mm6, 0x1b
    pshufw     mm5, mm5, 0x1b
    movq [r0+2*60], mm0
    movq [r0+2*56], mm2
    movq       mm0, [r1+2*42]
    movq       mm2, [r1+2*56]
    movq [r0+2*17], mm3
    movq [r0+2*32], mm7
    movq [r0+2*10], mm6
    movq [r0+2*21], mm5
    movq       mm3, mm0
    movq       mm7, mm2
    punpcklwd  mm0, mm4
    punpcklwd  mm2, mm1
    punpckhwd  mm3, mm4
    punpckhwd  mm7, mm1
    movq       mm4, mm2
    movq       mm1, mm7
    punpckhdq  mm2, mm0
    punpckhdq  mm7, mm3
    punpckldq  mm4, mm0
    punpckldq  mm1, mm3
    pshufw     mm2, mm2, 0x1b
    pshufw     mm7, mm7, 0x1b
    movq [r0+2*28], mm4
    movq [r0+2*43], mm1
    movq [r0+2*39], mm2
    movq [r0+2*50], mm7
    RET

;-----------------------------------------------------------------------------
; void x264_zigzag_scan_4x4_frame_mmx( int16_t level[16], int16_t dct[4][4] )
;-----------------------------------------------------------------------------
cglobal x264_zigzag_scan_4x4_frame_mmx, 2,2
    movq       mm0, [r1]
    movq       mm1, [r1+8]
    movq       mm2, [r1+16]
    movq       mm3, [r1+24]
    movq       mm4, mm0
    movq       mm5, mm1
    movq       mm6, mm2
    movq       mm7, mm3
    psllq      mm3, 16
    psrlq      mm0, 16
    punpckldq  mm2, mm2
    punpckhdq  mm1, mm1
    punpcklwd  mm4, mm5
    punpcklwd  mm5, mm3
    punpckldq  mm4, mm0
    punpckhwd  mm5, mm2
    punpckhwd  mm0, mm6
    punpckhwd  mm6, mm7
    punpcklwd  mm1, mm0
    punpckhdq  mm3, mm6
    movq      [r0], mm4
    movq    [r0+8], mm5
    movq   [r0+16], mm1
    movq   [r0+24], mm3
    RET
829

Holger Lubitz's avatar
Holger Lubitz committed
830 831 832 833 834 835
;-----------------------------------------------------------------------------
; void x264_zigzag_scan_4x4_frame_ssse3( int16_t level[16], int16_t dct[4][4] )
;-----------------------------------------------------------------------------
cglobal x264_zigzag_scan_4x4_frame_ssse3, 2,2
    movdqa    xmm1, [r1+16]
    movdqa    xmm0, [r1]
836 837
    pshufb    xmm1, [pb_scan4frameb]
    pshufb    xmm0, [pb_scan4framea]
Holger Lubitz's avatar
Holger Lubitz committed
838 839 840 841 842 843 844 845
    movdqa    xmm2, xmm1
    psrldq    xmm1, 6
    palignr   xmm2, xmm0, 6
    pslldq    xmm0, 10
    palignr   xmm1, xmm0, 10
    movdqa    [r0], xmm2
    movdqa [r0+16], xmm1
    RET
846

847
;-----------------------------------------------------------------------------
848
; void x264_zigzag_scan_4x4_field_mmxext( int16_t level[16], int16_t dct[4][4] )
849
;-----------------------------------------------------------------------------
850 851 852 853 854 855 856 857 858 859 860 861
; sse2 is only 1 cycle faster, and ssse3/pshufb is slower on core2
cglobal x264_zigzag_scan_4x4_field_mmxext, 2,3
    pshufw     mm0, [r1+4], 0xd2
    movq       mm1, [r1+16]
    movq       mm2, [r1+24]
    movq    [r0+4], mm0
    movq   [r0+16], mm1
    movq   [r0+24], mm2
    mov        r2d, [r1]
    mov       [r0], r2d
    mov        r2d, [r1+12]
    mov    [r0+12], r2d
862 863
    RET

864 865 866 867 868 869 870 871 872 873 874 875 876 877
;-----------------------------------------------------------------------------
; void x264_zigzag_scan_8x8_field_mmxext( int16_t level[64], int16_t dct[8][8] )
;-----------------------------------------------------------------------------

; Output order:
;  0  1  2  8  9  3  4 10
; 16 11  5  6  7 12 17 24
; 18 13 14 15 19 25 32 26
; 20 21 22 23 27 33 40 34
; 28 29 30 31 35 41 48 42
; 36 37 38 39 43 49 50 44
; 45 46 47 51 56 57 52 53
; 54 55 58 59 60 61 62 63

878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953
cglobal x264_zigzag_scan_8x8_field_mmxext, 2,3
    movq       mm0, [r1+2*0]        ; 03 02 01 00
    movq       mm1, [r1+2*4]        ; 07 06 05 04
    movq       mm2, [r1+2*8]        ; 11 10 09 08
    pshufw     mm3, mm0, 011111111b ; 03 03 03 03
    movd        r2, mm2             ; 09 08
    pshufw     mm2, mm2, 000111001b ; 08 11 10 09
    punpcklwd  mm3, mm1             ; 05 03 04 03
    pinsrw     mm0, r2, 3           ; 08 02 01 00
    movq       mm4, mm2
    punpcklwd  mm2, mm3             ; 04 10 03 09
    pshufw     mm2, mm2, 010110100b ; 10 04 03 09
    movq  [r0+2*0], mm0             ; 08 02 01 00
    movq  [r0+2*4], mm2             ; 10 04 03 09
    movq       mm3, [r1+2*12]       ; 15 14 13 12
    movq       mm5, [r1+2*16]       ; 19 18 17 16
    punpckldq  mm6, mm5             ; 17 16 XX XX
    psrlq      mm1, 16              ; XX 07 06 05
    punpckhwd  mm6, mm4             ; 08 17 11 16
    punpckldq  mm6, mm1             ; 06 05 11 16
    movq  [r0+2*8], mm6             ; 06 05 11 16
    psrlq      mm1, 16              ; XX XX 07 06
    punpcklwd  mm1, mm5             ; 17 07 16 06
    movq       mm0, [r1+2*20]       ; 23 22 21 20
    movq       mm2, [r1+2*24]       ; 27 26 25 24
    movq       mm6, mm3
    punpckhdq  mm1, mm1             ; 17 07 17 07
    punpcklwd  mm6, mm2             ; 25 13 24 12
    pextrw      r2, mm5, 2
    movq [r0+2*24], mm0             ; 23 22 21 20
    punpcklwd  mm1, mm6             ; 24 17 12 07
    movq [r0+2*12], mm1
    pinsrw     mm3, r2, 0           ; 15 14 13 18
    movq [r0+2*16], mm3             ; 15 14 13 18
    movq       mm7, [r1+2*28]
    movq       mm0, [r1+2*32]       ; 35 34 33 32
    psrlq      mm5, 48              ; XX XX XX 19
    pshufw     mm1, mm2, 011111001b ; 27 27 26 25
    punpcklwd  mm5, mm0             ; 33 XX 32 19
    psrlq      mm2, 48              ; XX XX XX 27
    punpcklwd  mm5, mm1             ; 26 32 25 19
    movq [r0+2*32], mm7
    movq [r0+2*20], mm5             ; 26 32 25 19
    movq       mm7, [r1+2*36]
    movq       mm1, [r1+2*40]       ; 43 42 41 40
    pshufw     mm3, mm0, 011111001b ; 35 35 34 33
    punpcklwd  mm2, mm1             ; 41 XX 40 27
    movq [r0+2*40], mm7
    punpcklwd  mm2, mm3             ; 34 40 33 27
    movq [r0+2*28], mm2
    movq       mm7, [r1+2*44]       ; 47 46 45 44
    movq       mm2, [r1+2*48]       ; 51 50 49 48
    psrlq      mm0, 48              ; XX XX XX 35
    punpcklwd  mm0, mm2             ; 49 XX 48 35
    pshufw     mm3, mm1, 011111001b ; 43 43 42 41
    punpcklwd  mm0, mm3             ; 42 48 41 35
    movq [r0+2*36], mm0
    pextrw      r2, mm2, 3          ; 51
    psrlq      mm1, 48              ; XX XX XX 43
    punpcklwd  mm1, mm7             ; 45 XX 44 43
    psrlq      mm2, 16              ; XX 51 50 49
    punpcklwd  mm1, mm2             ; 50 44 49 43
    pshufw     mm1, mm1, 010110100b ; 44 50 49 43
    movq [r0+2*44], mm1
    psrlq      mm7, 16              ; XX 47 46 45
    pinsrw     mm7, r2, 3           ; 51 47 46 45
    movq [r0+2*48], mm7
    movq       mm0, [r1+2*56]       ; 59 58 57 56
    movq       mm1, [r1+2*52]       ; 55 54 53 52
    movq       mm2, mm0
    movq       mm7, [r1+2*60]
    punpckldq  mm2, mm1             ; 53 52 57 56
    punpckhdq  mm1, mm0             ; 59 58 55 54
    movq [r0+2*52], mm2
    movq [r0+2*56], mm1
    movq [r0+2*60], mm7
954 955
    RET

Loren Merritt's avatar
Loren Merritt committed
956 957 958
;-----------------------------------------------------------------------------
; void x264_zigzag_sub_4x4_frame_ssse3( int16_t level[16], const uint8_t *src, uint8_t *dst )
;-----------------------------------------------------------------------------
959 960 961 962 963 964
%macro ZIGZAG_SUB_4x4 2
%ifidn %1, ac
cglobal x264_zigzag_sub_4x4%1_%2_ssse3, 4,4,8
%else
cglobal x264_zigzag_sub_4x4%1_%2_ssse3, 3,3,8
%endif
Loren Merritt's avatar
Loren Merritt committed
965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980
    movd      xmm0, [r1+0*FENC_STRIDE]
    movd      xmm1, [r1+1*FENC_STRIDE]
    movd      xmm2, [r1+2*FENC_STRIDE]
    movd      xmm3, [r1+3*FENC_STRIDE]
    movd      xmm4, [r2+0*FDEC_STRIDE]
    movd      xmm5, [r2+1*FDEC_STRIDE]
    movd      xmm6, [r2+2*FDEC_STRIDE]
    movd      xmm7, [r2+3*FDEC_STRIDE]
    movd      [r2+0*FDEC_STRIDE], xmm0
    movd      [r2+1*FDEC_STRIDE], xmm1
    movd      [r2+2*FDEC_STRIDE], xmm2
    movd      [r2+3*FDEC_STRIDE], xmm3
    punpckldq xmm0, xmm1
    punpckldq xmm2, xmm3
    punpckldq xmm4, xmm5
    punpckldq xmm6, xmm7
981 982
    punpcklqdq xmm0, xmm2
    punpcklqdq xmm4, xmm6
983
%ifidn %2, frame
984
    movdqa    xmm7, [pb_sub4frame]
985
%else
986
    movdqa    xmm7, [pb_sub4field]
987
%endif
Loren Merritt's avatar
Loren Merritt committed
988 989 990 991 992 993 994 995 996 997 998
    pshufb    xmm0, xmm7
    pshufb    xmm4, xmm7
    pxor      xmm6, xmm6
    movdqa    xmm1, xmm0
    movdqa    xmm5, xmm4
    punpcklbw xmm0, xmm6
    punpckhbw xmm1, xmm6
    punpcklbw xmm4, xmm6
    punpckhbw xmm5, xmm6
    psubw     xmm0, xmm4
    psubw     xmm1, xmm5
999 1000
%ifidn %1, ac
    movd       r2d, xmm0
1001
    pand      xmm0, [pb_subacmask]
1002
%endif
Loren Merritt's avatar
Loren Merritt committed
1003
    movdqa    [r0], xmm0
1004
    pxor      xmm2, xmm2
Loren Merritt's avatar
Loren Merritt committed
1005
    movdqa [r0+16], xmm1
1006 1007 1008 1009 1010 1011 1012 1013
    por       xmm0, xmm1
    pcmpeqb   xmm0, xmm2
    pmovmskb   eax, xmm0
%ifidn %1, ac
    mov       [r3], r2w
%endif
    sub        eax, 0xffff
    shr        eax, 31
Loren Merritt's avatar
Loren Merritt committed
1014
    RET
1015 1016 1017 1018 1019 1020
%endmacro

ZIGZAG_SUB_4x4   , frame
ZIGZAG_SUB_4x4 ac, frame
ZIGZAG_SUB_4x4   , field
ZIGZAG_SUB_4x4 ac, field
1021

1022 1023 1024 1025 1026 1027 1028 1029 1030
;-----------------------------------------------------------------------------
; void x264_zigzag_interleave_8x8_cavlc_mmx( int16_t *dst, int16_t *src, uint8_t *nnz )
;-----------------------------------------------------------------------------

%macro INTERLEAVE 1
    movq   m0, [r1+%1*4+ 0]
    movq   m1, [r1+%1*4+ 8]
    movq   m2, [r1+%1*4+16]
    movq   m3, [r1+%1*4+24]
1031
    TRANSPOSE4x4W 0,1,2,3,4
1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059
    movq   [r0+%1+ 0], m0
    movq   [r0+%1+32], m1
    movq   [r0+%1+64], m2
    movq   [r0+%1+96], m3
%if %1
    packsswb m0, m1
    por    m6, m2
    por    m7, m3
    por    m5, m0
%else
    packsswb m0, m1
    SWAP   m5, m0
    SWAP   m6, m2
    SWAP   m7, m3
%endif
%endmacro

INIT_MMX
cglobal x264_zigzag_interleave_8x8_cavlc_mmx, 3,3
    INTERLEAVE  0
    INTERLEAVE  8
    INTERLEAVE 16
    INTERLEAVE 24
    packsswb m6, m7
    packsswb m5, m6
    packsswb m5, m5
    pxor     m0, m0
    pcmpeqb  m5, m0
1060
    paddb    m5, [pb_1]
1061 1062 1063 1064 1065
    movd    r0d, m5
    mov  [r2+0], r0w
    shr     r0d, 16
    mov  [r2+8], r0w
    RET
Fiona Glaser's avatar
Fiona Glaser committed
1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105

%macro INTERLEAVE_XMM 1
    mova   m0, [r1+%1*4+ 0]
    mova   m1, [r1+%1*4+16]
    mova   m4, [r1+%1*4+32]
    mova   m5, [r1+%1*4+48]
    SBUTTERFLY wd, 0, 1, 6
    SBUTTERFLY wd, 4, 5, 7
    SBUTTERFLY wd, 0, 1, 6
    SBUTTERFLY wd, 4, 5, 7
    movq   [r0+%1+  0], m0
    movhps [r0+%1+ 32], m0
    movq   [r0+%1+ 64], m1
    movhps [r0+%1+ 96], m1
    movq   [r0+%1+  8], m4
    movhps [r0+%1+ 40], m4
    movq   [r0+%1+ 72], m5
    movhps [r0+%1+104], m5
%if %1
    por    m2, m0
    por    m3, m1
    por    m2, m4
    por    m3, m5
%else
    SWAP 0,2
    SWAP 3,1
    por    m2, m4
    por    m3, m5
%endif
%endmacro

INIT_XMM
cglobal x264_zigzag_interleave_8x8_cavlc_sse2, 3,3,8
    INTERLEAVE_XMM  0
    INTERLEAVE_XMM 16
    packsswb m2, m3
    pxor     m5, m5
    packsswb m2, m2
    packsswb m2, m2
    pcmpeqb  m5, m2
1106
    paddb    m5, [pb_1]
Fiona Glaser's avatar
Fiona Glaser committed
1107 1108 1109 1110 1111 1112
    movd    r0d, m5
    mov  [r2+0], r0w
    shr     r0d, 16
    mov  [r2+8], r0w
    RET