Mercurial > libavcodec.hg
view i386/idct_sse2_xvid.c @ 8362:156137b60026 libavcodec
Split filter_mb_dir() out of filter_mb().
1% overall decoding speed up for cathedral-beta2-400extra-crop-avc.mp4
no speed change for Aladin.mpg
Benchmarks done on Pentium dual
| author | michael |
|---|---|
| date | Wed, 17 Dec 2008 02:35:14 +0000 |
| parents | eebc7209c47f |
| children | 9ea0742666b9 |
line wrap: on
line source
/* * XVID MPEG-4 VIDEO CODEC * - SSE2 inverse discrete cosine transform - * * Copyright(C) 2003 Pascal Massimino <skal@planet-d.net> * * Conversion to gcc syntax with modifications * by Alexander Strange <astrange@ithinksw.com> * * Originally from dct/x86_asm/fdct_sse2_skal.asm in Xvid. * * This file is part of FFmpeg. * * Vertical pass is an implementation of the scheme: * Loeffler C., Ligtenberg A., and Moschytz C.S.: * Practical Fast 1D DCT Algorithm with Eleven Multiplications, * Proc. ICASSP 1989, 988-991. * * Horizontal pass is a double 4x4 vector/matrix multiplication, * (see also Intel's Application Note 922: * http://developer.intel.com/vtune/cbts/strmsimd/922down.htm * Copyright (C) 1999 Intel Corporation) * * More details at http://skal.planet-d.net/coding/dct.html * * 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 Foundation, * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ #include "libavcodec/dsputil.h" #include "libavcodec/i386/idct_xvid.h" /*! * @file idct_sse2_xvid.c * @brief SSE2 idct compatible with xvidmmx */ #define X8(x) x,x,x,x,x,x,x,x #define ROW_SHIFT 11 #define COL_SHIFT 6 DECLARE_ASM_CONST(16, int16_t, tan1[]) = {X8(13036)}; // tan( pi/16) DECLARE_ASM_CONST(16, int16_t, tan2[]) = {X8(27146)}; // tan(2pi/16) = sqrt(2)-1 DECLARE_ASM_CONST(16, int16_t, tan3[]) = {X8(43790)}; // tan(3pi/16)-1 DECLARE_ASM_CONST(16, int16_t, sqrt2[])= {X8(23170)}; // 0.5/sqrt(2) DECLARE_ASM_CONST(8, uint8_t, m127[]) = {X8(127)}; DECLARE_ASM_CONST(16, int16_t, iTab1[]) = { 0x4000, 0x539f, 0xc000, 0xac61, 0x4000, 0xdd5d, 0x4000, 0xdd5d, 0x4000, 0x22a3, 0x4000, 0x22a3, 0xc000, 0x539f, 0x4000, 0xac61, 0x3249, 0x11a8, 0x4b42, 0xee58, 0x11a8, 0x4b42, 0x11a8, 0xcdb7, 0x58c5, 0x4b42, 0xa73b, 0xcdb7, 0x3249, 0xa73b, 0x4b42, 0xa73b }; DECLARE_ASM_CONST(16, int16_t, iTab2[]) = { 0x58c5, 0x73fc, 0xa73b, 0x8c04, 0x58c5, 0xcff5, 0x58c5, 0xcff5, 0x58c5, 0x300b, 0x58c5, 0x300b, 0xa73b, 0x73fc, 0x58c5, 0x8c04, 0x45bf, 0x187e, 0x6862, 0xe782, 0x187e, 0x6862, 0x187e, 0xba41, 0x7b21, 0x6862, 0x84df, 0xba41, 0x45bf, 0x84df, 0x6862, 0x84df }; DECLARE_ASM_CONST(16, int16_t, iTab3[]) = { 0x539f, 0x6d41, 0xac61, 0x92bf, 0x539f, 0xd2bf, 0x539f, 0xd2bf, 0x539f, 0x2d41, 0x539f, 0x2d41, 0xac61, 0x6d41, 0x539f, 0x92bf, 0x41b3, 0x1712, 0x6254, 0xe8ee, 0x1712, 0x6254, 0x1712, 0xbe4d, 0x73fc, 0x6254, 0x8c04, 0xbe4d, 0x41b3, 0x8c04, 0x6254, 0x8c04 }; DECLARE_ASM_CONST(16, int16_t, iTab4[]) = { 0x4b42, 0x6254, 0xb4be, 0x9dac, 0x4b42, 0xd746, 0x4b42, 0xd746, 0x4b42, 0x28ba, 0x4b42, 0x28ba, 0xb4be, 0x6254, 0x4b42, 0x9dac, 0x3b21, 0x14c3, 0x587e, 0xeb3d, 0x14c3, 0x587e, 0x14c3, 0xc4df, 0x6862, 0x587e, 0x979e, 0xc4df, 0x3b21, 0x979e, 0x587e, 0x979e }; DECLARE_ASM_CONST(16, int32_t, walkenIdctRounders[]) = { 65536, 65536, 65536, 65536, 3597, 3597, 3597, 3597, 2260, 2260, 2260, 2260, 1203, 1203, 1203, 1203, 120, 120, 120, 120, 512, 512, 512, 512 }; // Temporary storage before the column pass #define ROW1 "%%xmm6" #define ROW3 "%%xmm4" #define ROW5 "%%xmm5" #define ROW7 "%%xmm7" #define CLEAR_ODD(r) "pxor "r","r" \n\t" #define PUT_ODD(dst) "pshufhw $0x1B, %%xmm2, "dst" \n\t" #ifdef ARCH_X86_64 # define ROW0 "%%xmm8" # define REG0 ROW0 # define ROW2 "%%xmm9" # define REG2 ROW2 # define ROW4 "%%xmm10" # define REG4 ROW4 # define ROW6 "%%xmm11" # define REG6 ROW6 # define CLEAR_EVEN(r) CLEAR_ODD(r) # define PUT_EVEN(dst) PUT_ODD(dst) # define XMMS "%%xmm12" # define MOV_32_ONLY "#" # define SREG2 REG2 # define TAN3 "%%xmm13" # define TAN1 "%%xmm14" #else # define ROW0 "(%0)" # define REG0 "%%xmm4" # define ROW2 "2*16(%0)" # define REG2 "%%xmm4" # define ROW4 "4*16(%0)" # define REG4 "%%xmm6" # define ROW6 "6*16(%0)" # define REG6 "%%xmm6" # define CLEAR_EVEN(r) # define PUT_EVEN(dst) \ "pshufhw $0x1B, %%xmm2, %%xmm2 \n\t" \ "movdqa %%xmm2, "dst" \n\t" # define XMMS "%%xmm2" # define MOV_32_ONLY "movdqa " # define SREG2 "%%xmm7" # define TAN3 "%%xmm0" # define TAN1 "%%xmm2" #endif #define ROUND(x) "paddd "MANGLE(x) #define JZ(reg, to) \ "testl "reg","reg" \n\t" \ "jz "to" \n\t" #define JNZ(reg, to) \ "testl "reg","reg" \n\t" \ "jnz "to" \n\t" #define TEST_ONE_ROW(src, reg, clear) \ clear \ "movq "src", %%mm1 \n\t" \ "por 8+"src", %%mm1 \n\t" \ "paddusb %%mm0, %%mm1 \n\t" \ "pmovmskb %%mm1, "reg" \n\t" #define TEST_TWO_ROWS(row1, row2, reg1, reg2, clear1, clear2) \ clear1 \ clear2 \ "movq "row1", %%mm1 \n\t" \ "por 8+"row1", %%mm1 \n\t" \ "movq "row2", %%mm2 \n\t" \ "por 8+"row2", %%mm2 \n\t" \ "paddusb %%mm0, %%mm1 \n\t" \ "paddusb %%mm0, %%mm2 \n\t" \ "pmovmskb %%mm1, "reg1" \n\t" \ "pmovmskb %%mm2, "reg2" \n\t" ///IDCT pass on rows. #define iMTX_MULT(src, table, rounder, put) \ "movdqa "src", %%xmm3 \n\t" \ "movdqa %%xmm3, %%xmm0 \n\t" \ "pshufd $0x11, %%xmm3, %%xmm1 \n\t" /* 4602 */ \ "punpcklqdq %%xmm0, %%xmm0 \n\t" /* 0246 */ \ "pmaddwd "table", %%xmm0 \n\t" \ "pmaddwd 16+"table", %%xmm1 \n\t" \ "pshufd $0xBB, %%xmm3, %%xmm2 \n\t" /* 5713 */ \ "punpckhqdq %%xmm3, %%xmm3 \n\t" /* 1357 */ \ "pmaddwd 32+"table", %%xmm2 \n\t" \ "pmaddwd 48+"table", %%xmm3 \n\t" \ "paddd %%xmm1, %%xmm0 \n\t" \ "paddd %%xmm3, %%xmm2 \n\t" \ rounder", %%xmm0 \n\t" \ "movdqa %%xmm2, %%xmm3 \n\t" \ "paddd %%xmm0, %%xmm2 \n\t" \ "psubd %%xmm3, %%xmm0 \n\t" \ "psrad $11, %%xmm2 \n\t" \ "psrad $11, %%xmm0 \n\t" \ "packssdw %%xmm0, %%xmm2 \n\t" \ put \ "1: \n\t" #define iLLM_HEAD \ "movdqa "MANGLE(tan3)", "TAN3" \n\t" \ "movdqa "MANGLE(tan1)", "TAN1" \n\t" \ ///IDCT pass on columns. #define iLLM_PASS(dct) \ "movdqa "TAN3", %%xmm1 \n\t" \ "movdqa "TAN1", %%xmm3 \n\t" \ "pmulhw %%xmm4, "TAN3" \n\t" \ "pmulhw %%xmm5, %%xmm1 \n\t" \ "paddsw %%xmm4, "TAN3" \n\t" \ "paddsw %%xmm5, %%xmm1 \n\t" \ "psubsw %%xmm5, "TAN3" \n\t" \ "paddsw %%xmm4, %%xmm1 \n\t" \ "pmulhw %%xmm7, %%xmm3 \n\t" \ "pmulhw %%xmm6, "TAN1" \n\t" \ "paddsw %%xmm6, %%xmm3 \n\t" \ "psubsw %%xmm7, "TAN1" \n\t" \ "movdqa %%xmm3, %%xmm7 \n\t" \ "movdqa "TAN1", %%xmm6 \n\t" \ "psubsw %%xmm1, %%xmm3 \n\t" \ "psubsw "TAN3", "TAN1" \n\t" \ "paddsw %%xmm7, %%xmm1 \n\t" \ "paddsw %%xmm6, "TAN3" \n\t" \ "movdqa %%xmm3, %%xmm6 \n\t" \ "psubsw "TAN3", %%xmm3 \n\t" \ "paddsw %%xmm6, "TAN3" \n\t" \ "movdqa "MANGLE(sqrt2)", %%xmm4 \n\t" \ "pmulhw %%xmm4, %%xmm3 \n\t" \ "pmulhw %%xmm4, "TAN3" \n\t" \ "paddsw "TAN3", "TAN3" \n\t" \ "paddsw %%xmm3, %%xmm3 \n\t" \ "movdqa "MANGLE(tan2)", %%xmm7 \n\t" \ MOV_32_ONLY ROW2", "REG2" \n\t" \ MOV_32_ONLY ROW6", "REG6" \n\t" \ "movdqa %%xmm7, %%xmm5 \n\t" \ "pmulhw "REG6", %%xmm7 \n\t" \ "pmulhw "REG2", %%xmm5 \n\t" \ "paddsw "REG2", %%xmm7 \n\t" \ "psubsw "REG6", %%xmm5 \n\t" \ MOV_32_ONLY ROW0", "REG0" \n\t" \ MOV_32_ONLY ROW4", "REG4" \n\t" \ MOV_32_ONLY" "TAN1", (%0) \n\t" \ "movdqa "REG0", "XMMS" \n\t" \ "psubsw "REG4", "REG0" \n\t" \ "paddsw "XMMS", "REG4" \n\t" \ "movdqa "REG4", "XMMS" \n\t" \ "psubsw %%xmm7, "REG4" \n\t" \ "paddsw "XMMS", %%xmm7 \n\t" \ "movdqa "REG0", "XMMS" \n\t" \ "psubsw %%xmm5, "REG0" \n\t" \ "paddsw "XMMS", %%xmm5 \n\t" \ "movdqa %%xmm5, "XMMS" \n\t" \ "psubsw "TAN3", %%xmm5 \n\t" \ "paddsw "XMMS", "TAN3" \n\t" \ "movdqa "REG0", "XMMS" \n\t" \ "psubsw %%xmm3, "REG0" \n\t" \ "paddsw "XMMS", %%xmm3 \n\t" \ MOV_32_ONLY" (%0), "TAN1" \n\t" \ "psraw $6, %%xmm5 \n\t" \ "psraw $6, "REG0" \n\t" \ "psraw $6, "TAN3" \n\t" \ "psraw $6, %%xmm3 \n\t" \ "movdqa "TAN3", 1*16("dct") \n\t" \ "movdqa %%xmm3, 2*16("dct") \n\t" \ "movdqa "REG0", 5*16("dct") \n\t" \ "movdqa %%xmm5, 6*16("dct") \n\t" \ "movdqa %%xmm7, %%xmm0 \n\t" \ "movdqa "REG4", %%xmm4 \n\t" \ "psubsw %%xmm1, %%xmm7 \n\t" \ "psubsw "TAN1", "REG4" \n\t" \ "paddsw %%xmm0, %%xmm1 \n\t" \ "paddsw %%xmm4, "TAN1" \n\t" \ "psraw $6, %%xmm1 \n\t" \ "psraw $6, %%xmm7 \n\t" \ "psraw $6, "TAN1" \n\t" \ "psraw $6, "REG4" \n\t" \ "movdqa %%xmm1, ("dct") \n\t" \ "movdqa "TAN1", 3*16("dct") \n\t" \ "movdqa "REG4", 4*16("dct") \n\t" \ "movdqa %%xmm7, 7*16("dct") \n\t" ///IDCT pass on columns, assuming rows 4-7 are zero. #define iLLM_PASS_SPARSE(dct) \ "pmulhw %%xmm4, "TAN3" \n\t" \ "paddsw %%xmm4, "TAN3" \n\t" \ "movdqa %%xmm6, %%xmm3 \n\t" \ "pmulhw %%xmm6, "TAN1" \n\t" \ "movdqa %%xmm4, %%xmm1 \n\t" \ "psubsw %%xmm1, %%xmm3 \n\t" \ "paddsw %%xmm6, %%xmm1 \n\t" \ "movdqa "TAN1", %%xmm6 \n\t" \ "psubsw "TAN3", "TAN1" \n\t" \ "paddsw %%xmm6, "TAN3" \n\t" \ "movdqa %%xmm3, %%xmm6 \n\t" \ "psubsw "TAN3", %%xmm3 \n\t" \ "paddsw %%xmm6, "TAN3" \n\t" \ "movdqa "MANGLE(sqrt2)", %%xmm4 \n\t" \ "pmulhw %%xmm4, %%xmm3 \n\t" \ "pmulhw %%xmm4, "TAN3" \n\t" \ "paddsw "TAN3", "TAN3" \n\t" \ "paddsw %%xmm3, %%xmm3 \n\t" \ "movdqa "MANGLE(tan2)", %%xmm5 \n\t" \ MOV_32_ONLY ROW2", "SREG2" \n\t" \ "pmulhw "SREG2", %%xmm5 \n\t" \ MOV_32_ONLY ROW0", "REG0" \n\t" \ "movdqa "REG0", %%xmm6 \n\t" \ "psubsw "SREG2", %%xmm6 \n\t" \ "paddsw "REG0", "SREG2" \n\t" \ MOV_32_ONLY" "TAN1", (%0) \n\t" \ "movdqa "REG0", "XMMS" \n\t" \ "psubsw %%xmm5, "REG0" \n\t" \ "paddsw "XMMS", %%xmm5 \n\t" \ "movdqa %%xmm5, "XMMS" \n\t" \ "psubsw "TAN3", %%xmm5 \n\t" \ "paddsw "XMMS", "TAN3" \n\t" \ "movdqa "REG0", "XMMS" \n\t" \ "psubsw %%xmm3, "REG0" \n\t" \ "paddsw "XMMS", %%xmm3 \n\t" \ MOV_32_ONLY" (%0), "TAN1" \n\t" \ "psraw $6, %%xmm5 \n\t" \ "psraw $6, "REG0" \n\t" \ "psraw $6, "TAN3" \n\t" \ "psraw $6, %%xmm3 \n\t" \ "movdqa "TAN3", 1*16("dct") \n\t" \ "movdqa %%xmm3, 2*16("dct") \n\t" \ "movdqa "REG0", 5*16("dct") \n\t" \ "movdqa %%xmm5, 6*16("dct") \n\t" \ "movdqa "SREG2", %%xmm0 \n\t" \ "movdqa %%xmm6, %%xmm4 \n\t" \ "psubsw %%xmm1, "SREG2" \n\t" \ "psubsw "TAN1", %%xmm6 \n\t" \ "paddsw %%xmm0, %%xmm1 \n\t" \ "paddsw %%xmm4, "TAN1" \n\t" \ "psraw $6, %%xmm1 \n\t" \ "psraw $6, "SREG2" \n\t" \ "psraw $6, "TAN1" \n\t" \ "psraw $6, %%xmm6 \n\t" \ "movdqa %%xmm1, ("dct") \n\t" \ "movdqa "TAN1", 3*16("dct") \n\t" \ "movdqa %%xmm6, 4*16("dct") \n\t" \ "movdqa "SREG2", 7*16("dct") \n\t" inline void ff_idct_xvid_sse2(short *block) { __asm__ volatile( "movq "MANGLE(m127)", %%mm0 \n\t" iMTX_MULT("(%0)", MANGLE(iTab1), ROUND(walkenIdctRounders), PUT_EVEN(ROW0)) iMTX_MULT("1*16(%0)", MANGLE(iTab2), ROUND(walkenIdctRounders+1*16), PUT_ODD(ROW1)) iMTX_MULT("2*16(%0)", MANGLE(iTab3), ROUND(walkenIdctRounders+2*16), PUT_EVEN(ROW2)) TEST_TWO_ROWS("3*16(%0)", "4*16(%0)", "%%eax", "%%ecx", CLEAR_ODD(ROW3), CLEAR_EVEN(ROW4)) JZ("%%eax", "1f") iMTX_MULT("3*16(%0)", MANGLE(iTab4), ROUND(walkenIdctRounders+3*16), PUT_ODD(ROW3)) TEST_TWO_ROWS("5*16(%0)", "6*16(%0)", "%%eax", "%%edx", CLEAR_ODD(ROW5), CLEAR_EVEN(ROW6)) TEST_ONE_ROW("7*16(%0)", "%%esi", CLEAR_ODD(ROW7)) iLLM_HEAD ASMALIGN(4) JNZ("%%ecx", "2f") JNZ("%%eax", "3f") JNZ("%%edx", "4f") JNZ("%%esi", "5f") iLLM_PASS_SPARSE("%0") "jmp 6f \n\t" "2: \n\t" iMTX_MULT("4*16(%0)", MANGLE(iTab1), "#", PUT_EVEN(ROW4)) "3: \n\t" iMTX_MULT("5*16(%0)", MANGLE(iTab4), ROUND(walkenIdctRounders+4*16), PUT_ODD(ROW5)) JZ("%%edx", "1f") "4: \n\t" iMTX_MULT("6*16(%0)", MANGLE(iTab3), ROUND(walkenIdctRounders+5*16), PUT_EVEN(ROW6)) JZ("%%esi", "1f") "5: \n\t" iMTX_MULT("7*16(%0)", MANGLE(iTab2), ROUND(walkenIdctRounders+5*16), PUT_ODD(ROW7)) #ifndef ARCH_X86_64 iLLM_HEAD #endif iLLM_PASS("%0") "6: \n\t" : "+r"(block) : : "%eax", "%ecx", "%edx", "%esi", "memory"); } void ff_idct_xvid_sse2_put(uint8_t *dest, int line_size, short *block) { ff_idct_xvid_sse2(block); put_pixels_clamped_mmx(block, dest, line_size); } void ff_idct_xvid_sse2_add(uint8_t *dest, int line_size, short *block) { ff_idct_xvid_sse2(block); add_pixels_clamped_mmx(block, dest, line_size); }