diff options
Diffstat (limited to 'libavcodec/x86/proresdsp.asm')
-rw-r--r-- | libavcodec/x86/proresdsp.asm | 148 |
1 files changed, 15 insertions, 133 deletions
diff --git a/libavcodec/x86/proresdsp.asm b/libavcodec/x86/proresdsp.asm index 9b2e11eb5d..a09d871fda 100644 --- a/libavcodec/x86/proresdsp.asm +++ b/libavcodec/x86/proresdsp.asm @@ -1,23 +1,24 @@ ;****************************************************************************** ;* x86-SIMD-optimized IDCT for prores -;* this is identical to "simple" IDCT except for the clip range +;* this is identical to "simple" IDCT written by Michael Niedermayer +;* except for the clip range ;* ;* Copyright (c) 2011 Ronald S. Bultje <rsbultje@gmail.com> ;* -;* This file is part of Libav. +;* This file is part of FFmpeg. ;* -;* Libav is free software; you can redistribute it and/or +;* 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. ;* -;* Libav is distributed in the hope that it will be useful, +;* 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 Libav; if not, write to the Free Software +;* License along with FFmpeg; if not, write to the Free Software ;* 51, Inc., Foundation Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA ;****************************************************************************** @@ -48,10 +49,10 @@ w1_plus_w5: times 4 dw W1sh2, +W5sh2 w5_min_w1: times 4 dw W5sh2, -W1sh2 w5_plus_w7: times 4 dw W5sh2, +W7sh2 w7_min_w5: times 4 dw W7sh2, -W5sh2 -row_round: times 8 dw (1<<14) +pw_88: times 8 dw 0x2008 +cextern pw_1 cextern pw_4 -cextern pw_8 cextern pw_512 cextern pw_1019 @@ -93,14 +94,12 @@ section .text align=16 ; a2 -= W6 * row[2]; ; a3 -= W2 * row[2]; %ifidn %1, col - paddw m10,[pw_8] + paddw m10,[pw_88] %endif - SBUTTERFLY3 wd, 0, 1, 10, 8 ; { row[0], row[2] }[0-3]/[4-7] %ifidn %1, row - psubw m10,[row_round] + paddw m10,[pw_1] %endif - SIGNEXTEND m8, m9, m14 ; { row[2] }[0-3] / [4-7] - SIGNEXTEND m10, m11, m14 ; { row[0] }[0-3] / [4-7] + SBUTTERFLY3 wd, 0, 1, 10, 8 ; { row[0], row[2] }[0-3]/[4-7] pmaddwd m2, m0, [w4_plus_w6] pmaddwd m3, m1, [w4_plus_w6] pmaddwd m4, m0, [w4_min_w6] @@ -109,75 +108,33 @@ section .text align=16 pmaddwd m7, m1, [w4_min_w2] pmaddwd m0, [w4_plus_w2] pmaddwd m1, [w4_plus_w2] - pslld m2, 2 - pslld m3, 2 - pslld m4, 2 - pslld m5, 2 - pslld m6, 2 - pslld m7, 2 - pslld m0, 2 - pslld m1, 2 ; a0: -1*row[0]-1*row[2] ; a1: -1*row[0] ; a2: -1*row[0] ; a3: -1*row[0]+1*row[2] - psubd m2, m10 ; a1[0-3] - psubd m3, m11 ; a1[4-7] - psubd m4, m10 ; a2[0-3] - psubd m5, m11 ; a2[4-7] - psubd m0, m10 - psubd m1, m11 - psubd m6, m10 - psubd m7, m11 - psubd m0, m8 ; a0[0-3] - psubd m1, m9 ; a0[4-7] - paddd m6, m8 ; a3[0-3] - paddd m7, m9 ; a3[4-7] ; a0 += W4*row[4] + W6*row[6]; i.e. -1*row[4] ; a1 -= W4*row[4] + W2*row[6]; i.e. -1*row[4]-1*row[6] ; a2 -= W4*row[4] - W2*row[6]; i.e. -1*row[4]+1*row[6] ; a3 += W4*row[4] - W6*row[6]; i.e. -1*row[4] SBUTTERFLY3 wd, 8, 9, 13, 12 ; { row[4], row[6] }[0-3]/[4-7] - SIGNEXTEND m13, m14, m10 ; { row[4] }[0-3] / [4-7] pmaddwd m10, m8, [w4_plus_w6] pmaddwd m11, m9, [w4_plus_w6] - pslld m10, 2 - pslld m11, 2 - psubd m10, m13 - psubd m11, m14 paddd m0, m10 ; a0[0-3] paddd m1, m11 ; a0[4-7] pmaddwd m10, m8, [w4_min_w6] pmaddwd m11, m9, [w4_min_w6] - pslld m10, 2 - pslld m11, 2 - psubd m10, m13 - psubd m11, m14 paddd m6, m10 ; a3[0-3] paddd m7, m11 ; a3[4-7] pmaddwd m10, m8, [w4_min_w2] pmaddwd m11, m9, [w4_min_w2] pmaddwd m8, [w4_plus_w2] pmaddwd m9, [w4_plus_w2] - pslld m10, 2 - pslld m11, 2 - pslld m8, 2 - pslld m9, 2 - psubd m10, m13 - psubd m11, m14 - psubd m8, m13 - psubd m9, m14 psubd m4, m10 ; a2[0-3] intermediate psubd m5, m11 ; a2[4-7] intermediate psubd m2, m8 ; a1[0-3] intermediate psubd m3, m9 ; a1[4-7] intermediate - SIGNEXTEND m12, m13, m10 ; { row[6] }[0-3] / [4-7] - psubd m4, m12 ; a2[0-3] - psubd m5, m13 ; a2[4-7] - paddd m2, m12 ; a1[0-3] - paddd m3, m13 ; a1[4-7] ; load/store mova [r2+ 0], m0 @@ -208,8 +165,6 @@ section .text align=16 ; b3 = MUL(W7, row[1]); ; MAC(b3, -W5, row[3]); SBUTTERFLY3 wd, 0, 1, 10, 8 ; { row[1], row[3] }[0-3]/[4-7] - SIGNEXTEND m10, m11, m12 ; { row[1] }[0-3] / [4-7] - SIGNEXTEND m8, m9, m12 ; { row[3] }[0-3] / [4-7] pmaddwd m2, m0, [w3_min_w7] pmaddwd m3, m1, [w3_min_w7] pmaddwd m4, m0, [w5_min_w1] @@ -218,35 +173,11 @@ section .text align=16 pmaddwd m7, m1, [w7_min_w5] pmaddwd m0, [w1_plus_w3] pmaddwd m1, [w1_plus_w3] - pslld m2, 2 - pslld m3, 2 - pslld m4, 2 - pslld m5, 2 - pslld m6, 2 - pslld m7, 2 - pslld m0, 2 - pslld m1, 2 ; b0: +1*row[1]+2*row[3] ; b1: +2*row[1]-1*row[3] ; b2: -1*row[1]-1*row[3] ; b3: +1*row[1]+1*row[3] - psubd m2, m8 - psubd m3, m9 - paddd m0, m8 - paddd m1, m9 - paddd m8, m10 ; { row[1] + row[3] }[0-3] - paddd m9, m11 ; { row[1] + row[3] }[4-7] - paddd m10, m10 - paddd m11, m11 - paddd m0, m8 ; b0[0-3] - paddd m1, m9 ; b0[4-7] - paddd m2, m10 ; b1[0-3] - paddd m3, m11 ; b2[4-7] - psubd m4, m8 ; b2[0-3] - psubd m5, m9 ; b2[4-7] - paddd m6, m8 ; b3[0-3] - paddd m7, m9 ; b3[4-7] ; MAC(b0, W5, row[5]); ; MAC(b0, W7, row[7]); @@ -257,38 +188,16 @@ section .text align=16 ; MAC(b3, W3, row[5]); ; MAC(b3, -W1, row[7]); SBUTTERFLY3 wd, 8, 9, 13, 14 ; { row[5], row[7] }[0-3]/[4-7] - SIGNEXTEND m13, m12, m11 ; { row[5] }[0-3] / [4-7] - SIGNEXTEND m14, m11, m10 ; { row[7] }[0-3] / [4-7] ; b0: -1*row[5]+1*row[7] ; b1: -1*row[5]+1*row[7] ; b2: +1*row[5]+2*row[7] ; b3: +2*row[5]-1*row[7] - paddd m4, m13 - paddd m5, m12 - paddd m6, m13 - paddd m7, m12 - psubd m13, m14 ; { row[5] - row[7] }[0-3] - psubd m12, m11 ; { row[5] - row[7] }[4-7] - paddd m14, m14 - paddd m11, m11 - psubd m0, m13 - psubd m1, m12 - psubd m2, m13 - psubd m3, m12 - paddd m4, m14 - paddd m5, m11 - paddd m6, m13 - paddd m7, m12 pmaddwd m10, m8, [w1_plus_w5] pmaddwd m11, m9, [w1_plus_w5] pmaddwd m12, m8, [w5_plus_w7] pmaddwd m13, m9, [w5_plus_w7] - pslld m10, 2 - pslld m11, 2 - pslld m12, 2 - pslld m13, 2 psubd m2, m10 ; b1[0-3] psubd m3, m11 ; b1[4-7] paddd m0, m12 ; b0[0-3] @@ -297,10 +206,6 @@ section .text align=16 pmaddwd m13, m9, [w7_plus_w3] pmaddwd m8, [w3_min_w1] pmaddwd m9, [w3_min_w1] - pslld m12, 2 - pslld m13, 2 - pslld m8, 2 - pslld m9, 2 paddd m4, m12 ; b2[0-3] paddd m5, m13 ; b2[4-7] paddd m6, m8 ; b3[0-3] @@ -347,7 +252,7 @@ cglobal prores_idct_put_10_%1, 4, 4, %2 pmullw m13,[r3+64] pmullw m12,[r3+96] - IDCT_1D row, 17, %1 + IDCT_1D row, 15, %1 ; transpose for second part of IDCT TRANSPOSE8x8W 8, 0, 1, 2, 4, 11, 9, 10, 3 @@ -362,20 +267,11 @@ cglobal prores_idct_put_10_%1, 4, 4, %2 ; for (i = 0; i < 8; i++) ; idctSparseColAdd(dest + i, line_size, block + i); - IDCT_1D col, 20, %1 + IDCT_1D col, 18, %1 ; clip/store - mova m6, [pw_512] mova m3, [pw_4] mova m5, [pw_1019] - paddw m8, m6 - paddw m0, m6 - paddw m1, m6 - paddw m2, m6 - paddw m4, m6 - paddw m11, m6 - paddw m9, m6 - paddw m10, m6 pmaxsw m8, m3 pmaxsw m0, m3 pmaxsw m1, m3 @@ -406,27 +302,13 @@ cglobal prores_idct_put_10_%1, 4, 4, %2 RET %endmacro -%macro signextend_sse2 3 ; dstlow, dsthigh, tmp - pxor %3, %3 - pcmpgtw %3, %1 - mova %2, %1 - punpcklwd %1, %3 - punpckhwd %2, %3 -%endmacro - -%macro signextend_sse4 2-3 ; dstlow, dsthigh - movhlps %2, %1 - pmovsxwd %1, %1 - pmovsxwd %2, %2 -%endmacro - INIT_XMM -%define SIGNEXTEND signextend_sse2 idct_put_fn sse2, 16 INIT_XMM -%define SIGNEXTEND signextend_sse4 idct_put_fn sse4, 16 +%if HAVE_AVX INIT_AVX idct_put_fn avx, 16 +%endif %endif |