From 6532a1a8286e3ecc80c30ef74ff08663f237424d Mon Sep 17 00:00:00 2001 From: Michael Niedermayer Date: Sat, 19 Jul 2014 05:09:57 +0200 Subject: sws/x86: split mmxext fast bilinear scaler out Signed-off-by: Michael Niedermayer --- libswscale/x86/Makefile | 3 +- libswscale/x86/hscale_fast_bilinear_simd.c | 372 +++++++++++++++++++++++++++++ libswscale/x86/swscale_template.c | 194 +-------------- 3 files changed, 376 insertions(+), 193 deletions(-) create mode 100644 libswscale/x86/hscale_fast_bilinear_simd.c (limited to 'libswscale/x86') diff --git a/libswscale/x86/Makefile b/libswscale/x86/Makefile index e767a5c420..c6f56d0932 100644 --- a/libswscale/x86/Makefile +++ b/libswscale/x86/Makefile @@ -1,6 +1,7 @@ $(SUBDIR)x86/swscale_mmx.o: CFLAGS += $(NOREDZONE_FLAGS) -OBJS += x86/rgb2rgb.o \ +OBJS += x86/hscale_fast_bilinear_simd.o \ + x86/rgb2rgb.o \ x86/swscale.o \ x86/yuv2rgb.o \ diff --git a/libswscale/x86/hscale_fast_bilinear_simd.c b/libswscale/x86/hscale_fast_bilinear_simd.c new file mode 100644 index 0000000000..4e547543d3 --- /dev/null +++ b/libswscale/x86/hscale_fast_bilinear_simd.c @@ -0,0 +1,372 @@ +/* + * Copyright (C) 2001-2003 Michael Niedermayer + * + * 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 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "../swscale_internal.h" +#include "libavutil/x86/asm.h" +#include "libavutil/x86/cpu.h" + +#define RET 0xC3 // near return opcode for x86 +#define PREFETCH "prefetchnta" + +av_cold int ff_init_hscaler_mmxext(int dstW, int xInc, uint8_t *filterCode, + int16_t *filter, int32_t *filterPos, + int numSplits) +{ + uint8_t *fragmentA; + x86_reg imm8OfPShufW1A; + x86_reg imm8OfPShufW2A; + x86_reg fragmentLengthA; + uint8_t *fragmentB; + x86_reg imm8OfPShufW1B; + x86_reg imm8OfPShufW2B; + x86_reg fragmentLengthB; + int fragmentPos; + + int xpos, i; + + // create an optimized horizontal scaling routine + /* This scaler is made of runtime-generated MMXEXT code using specially tuned + * pshufw instructions. For every four output pixels, if four input pixels + * are enough for the fast bilinear scaling, then a chunk of fragmentB is + * used. If five input pixels are needed, then a chunk of fragmentA is used. + */ + + // code fragment + + __asm__ volatile ( + "jmp 9f \n\t" + // Begin + "0: \n\t" + "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t" + "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t" + "movd 1(%%"REG_c", %%"REG_S"), %%mm1 \n\t" + "punpcklbw %%mm7, %%mm1 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "pshufw $0xFF, %%mm1, %%mm1 \n\t" + "1: \n\t" + "pshufw $0xFF, %%mm0, %%mm0 \n\t" + "2: \n\t" + "psubw %%mm1, %%mm0 \n\t" + "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t" + "pmullw %%mm3, %%mm0 \n\t" + "psllw $7, %%mm1 \n\t" + "paddw %%mm1, %%mm0 \n\t" + + "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t" + + "add $8, %%"REG_a" \n\t" + // End + "9: \n\t" + // "int $3 \n\t" + "lea " LOCAL_MANGLE(0b) ", %0 \n\t" + "lea " LOCAL_MANGLE(1b) ", %1 \n\t" + "lea " LOCAL_MANGLE(2b) ", %2 \n\t" + "dec %1 \n\t" + "dec %2 \n\t" + "sub %0, %1 \n\t" + "sub %0, %2 \n\t" + "lea " LOCAL_MANGLE(9b) ", %3 \n\t" + "sub %0, %3 \n\t" + + + : "=r" (fragmentA), "=r" (imm8OfPShufW1A), "=r" (imm8OfPShufW2A), + "=r" (fragmentLengthA) + ); + + __asm__ volatile ( + "jmp 9f \n\t" + // Begin + "0: \n\t" + "movq (%%"REG_d", %%"REG_a"), %%mm3 \n\t" + "movd (%%"REG_c", %%"REG_S"), %%mm0 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "pshufw $0xFF, %%mm0, %%mm1 \n\t" + "1: \n\t" + "pshufw $0xFF, %%mm0, %%mm0 \n\t" + "2: \n\t" + "psubw %%mm1, %%mm0 \n\t" + "movl 8(%%"REG_b", %%"REG_a"), %%esi \n\t" + "pmullw %%mm3, %%mm0 \n\t" + "psllw $7, %%mm1 \n\t" + "paddw %%mm1, %%mm0 \n\t" + + "movq %%mm0, (%%"REG_D", %%"REG_a") \n\t" + + "add $8, %%"REG_a" \n\t" + // End + "9: \n\t" + // "int $3 \n\t" + "lea " LOCAL_MANGLE(0b) ", %0 \n\t" + "lea " LOCAL_MANGLE(1b) ", %1 \n\t" + "lea " LOCAL_MANGLE(2b) ", %2 \n\t" + "dec %1 \n\t" + "dec %2 \n\t" + "sub %0, %1 \n\t" + "sub %0, %2 \n\t" + "lea " LOCAL_MANGLE(9b) ", %3 \n\t" + "sub %0, %3 \n\t" + + + : "=r" (fragmentB), "=r" (imm8OfPShufW1B), "=r" (imm8OfPShufW2B), + "=r" (fragmentLengthB) + ); + + xpos = 0; // lumXInc/2 - 0x8000; // difference between pixel centers + fragmentPos = 0; + + for (i = 0; i < dstW / numSplits; i++) { + int xx = xpos >> 16; + + if ((i & 3) == 0) { + int a = 0; + int b = ((xpos + xInc) >> 16) - xx; + int c = ((xpos + xInc * 2) >> 16) - xx; + int d = ((xpos + xInc * 3) >> 16) - xx; + int inc = (d + 1 < 4); + uint8_t *fragment = inc ? fragmentB : fragmentA; + x86_reg imm8OfPShufW1 = inc ? imm8OfPShufW1B : imm8OfPShufW1A; + x86_reg imm8OfPShufW2 = inc ? imm8OfPShufW2B : imm8OfPShufW2A; + x86_reg fragmentLength = inc ? fragmentLengthB : fragmentLengthA; + int maxShift = 3 - (d + inc); + int shift = 0; + + if (filterCode) { + filter[i] = ((xpos & 0xFFFF) ^ 0xFFFF) >> 9; + filter[i + 1] = (((xpos + xInc) & 0xFFFF) ^ 0xFFFF) >> 9; + filter[i + 2] = (((xpos + xInc * 2) & 0xFFFF) ^ 0xFFFF) >> 9; + filter[i + 3] = (((xpos + xInc * 3) & 0xFFFF) ^ 0xFFFF) >> 9; + filterPos[i / 2] = xx; + + memcpy(filterCode + fragmentPos, fragment, fragmentLength); + + filterCode[fragmentPos + imm8OfPShufW1] = (a + inc) | + ((b + inc) << 2) | + ((c + inc) << 4) | + ((d + inc) << 6); + filterCode[fragmentPos + imm8OfPShufW2] = a | (b << 2) | + (c << 4) | + (d << 6); + + if (i + 4 - inc >= dstW) + shift = maxShift; // avoid overread + else if ((filterPos[i / 2] & 3) <= maxShift) + shift = filterPos[i / 2] & 3; // align + + if (shift && i >= shift) { + filterCode[fragmentPos + imm8OfPShufW1] += 0x55 * shift; + filterCode[fragmentPos + imm8OfPShufW2] += 0x55 * shift; + filterPos[i / 2] -= shift; + } + } + + fragmentPos += fragmentLength; + + if (filterCode) + filterCode[fragmentPos] = RET; + } + xpos += xInc; + } + if (filterCode) + filterPos[((i / 2) + 1) & (~1)] = xpos >> 16; // needed to jump to the next part + + return fragmentPos + 1; +} + +void ff_hyscale_fast_mmxext(SwsContext *c, int16_t *dst, + int dstWidth, const uint8_t *src, + int srcW, int xInc) +{ + int32_t *filterPos = c->hLumFilterPos; + int16_t *filter = c->hLumFilter; + void *mmxextFilterCode = c->lumMmxextFilterCode; + int i; +#if defined(PIC) + uint64_t ebxsave; +#endif +#if ARCH_X86_64 + uint64_t retsave; +#endif + + __asm__ volatile( +#if defined(PIC) + "mov %%"REG_b", %5 \n\t" +#if ARCH_X86_64 + "mov -8(%%rsp), %%"REG_a" \n\t" + "mov %%"REG_a", %6 \n\t" +#endif +#else +#if ARCH_X86_64 + "mov -8(%%rsp), %%"REG_a" \n\t" + "mov %%"REG_a", %5 \n\t" +#endif +#endif + "pxor %%mm7, %%mm7 \n\t" + "mov %0, %%"REG_c" \n\t" + "mov %1, %%"REG_D" \n\t" + "mov %2, %%"REG_d" \n\t" + "mov %3, %%"REG_b" \n\t" + "xor %%"REG_a", %%"REG_a" \n\t" // i + PREFETCH" (%%"REG_c") \n\t" + PREFETCH" 32(%%"REG_c") \n\t" + PREFETCH" 64(%%"REG_c") \n\t" + +#if ARCH_X86_64 +#define CALL_MMXEXT_FILTER_CODE \ + "movl (%%"REG_b"), %%esi \n\t"\ + "call *%4 \n\t"\ + "movl (%%"REG_b", %%"REG_a"), %%esi \n\t"\ + "add %%"REG_S", %%"REG_c" \n\t"\ + "add %%"REG_a", %%"REG_D" \n\t"\ + "xor %%"REG_a", %%"REG_a" \n\t"\ + +#else +#define CALL_MMXEXT_FILTER_CODE \ + "movl (%%"REG_b"), %%esi \n\t"\ + "call *%4 \n\t"\ + "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\ + "add %%"REG_a", %%"REG_D" \n\t"\ + "xor %%"REG_a", %%"REG_a" \n\t"\ + +#endif /* ARCH_X86_64 */ + + CALL_MMXEXT_FILTER_CODE + CALL_MMXEXT_FILTER_CODE + CALL_MMXEXT_FILTER_CODE + CALL_MMXEXT_FILTER_CODE + CALL_MMXEXT_FILTER_CODE + CALL_MMXEXT_FILTER_CODE + CALL_MMXEXT_FILTER_CODE + CALL_MMXEXT_FILTER_CODE + +#if defined(PIC) + "mov %5, %%"REG_b" \n\t" +#if ARCH_X86_64 + "mov %6, %%"REG_a" \n\t" + "mov %%"REG_a", -8(%%rsp) \n\t" +#endif +#else +#if ARCH_X86_64 + "mov %5, %%"REG_a" \n\t" + "mov %%"REG_a", -8(%%rsp) \n\t" +#endif +#endif + :: "m" (src), "m" (dst), "m" (filter), "m" (filterPos), + "m" (mmxextFilterCode) +#if defined(PIC) + ,"m" (ebxsave) +#endif +#if ARCH_X86_64 + ,"m"(retsave) +#endif + : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D +#if !defined(PIC) + ,"%"REG_b +#endif + ); + + for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) + dst[i] = src[srcW-1]*128; +} + +void ff_hcscale_fast_mmxext(SwsContext *c, int16_t *dst1, int16_t *dst2, + int dstWidth, const uint8_t *src1, + const uint8_t *src2, int srcW, int xInc) +{ + int32_t *filterPos = c->hChrFilterPos; + int16_t *filter = c->hChrFilter; + void *mmxextFilterCode = c->chrMmxextFilterCode; + int i; +#if defined(PIC) + DECLARE_ALIGNED(8, uint64_t, ebxsave); +#endif +#if ARCH_X86_64 + DECLARE_ALIGNED(8, uint64_t, retsave); +#endif + + __asm__ volatile( +#if defined(PIC) + "mov %%"REG_b", %7 \n\t" +#if ARCH_X86_64 + "mov -8(%%rsp), %%"REG_a" \n\t" + "mov %%"REG_a", %8 \n\t" +#endif +#else +#if ARCH_X86_64 + "mov -8(%%rsp), %%"REG_a" \n\t" + "mov %%"REG_a", %7 \n\t" +#endif +#endif + "pxor %%mm7, %%mm7 \n\t" + "mov %0, %%"REG_c" \n\t" + "mov %1, %%"REG_D" \n\t" + "mov %2, %%"REG_d" \n\t" + "mov %3, %%"REG_b" \n\t" + "xor %%"REG_a", %%"REG_a" \n\t" // i + PREFETCH" (%%"REG_c") \n\t" + PREFETCH" 32(%%"REG_c") \n\t" + PREFETCH" 64(%%"REG_c") \n\t" + + CALL_MMXEXT_FILTER_CODE + CALL_MMXEXT_FILTER_CODE + CALL_MMXEXT_FILTER_CODE + CALL_MMXEXT_FILTER_CODE + "xor %%"REG_a", %%"REG_a" \n\t" // i + "mov %5, %%"REG_c" \n\t" // src + "mov %6, %%"REG_D" \n\t" // buf2 + PREFETCH" (%%"REG_c") \n\t" + PREFETCH" 32(%%"REG_c") \n\t" + PREFETCH" 64(%%"REG_c") \n\t" + + CALL_MMXEXT_FILTER_CODE + CALL_MMXEXT_FILTER_CODE + CALL_MMXEXT_FILTER_CODE + CALL_MMXEXT_FILTER_CODE + +#if defined(PIC) + "mov %7, %%"REG_b" \n\t" +#if ARCH_X86_64 + "mov %8, %%"REG_a" \n\t" + "mov %%"REG_a", -8(%%rsp) \n\t" +#endif +#else +#if ARCH_X86_64 + "mov %7, %%"REG_a" \n\t" + "mov %%"REG_a", -8(%%rsp) \n\t" +#endif +#endif + :: "m" (src1), "m" (dst1), "m" (filter), "m" (filterPos), + "m" (mmxextFilterCode), "m" (src2), "m"(dst2) +#if defined(PIC) + ,"m" (ebxsave) +#endif +#if ARCH_X86_64 + ,"m"(retsave) +#endif + : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D +#if !defined(PIC) + ,"%"REG_b +#endif + ); + + for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) { + dst1[i] = src1[srcW-1]*128; + dst2[i] = src2[srcW-1]*128; + } +} diff --git a/libswscale/x86/swscale_template.c b/libswscale/x86/swscale_template.c index 1213197347..36a606c5b2 100644 --- a/libswscale/x86/swscale_template.c +++ b/libswscale/x86/swscale_template.c @@ -28,11 +28,6 @@ #undef MOVNTQ2 #undef PREFETCH -#if COMPILE_TEMPLATE_MMXEXT -#define PREFETCH "prefetchnta" -#else -#define PREFETCH " # nop" -#endif #if COMPILE_TEMPLATE_MMXEXT #define REAL_MOVNTQ(a,b) "movntq " #a ", " #b " \n\t" @@ -1470,191 +1465,6 @@ static void RENAME(yuv2yuyv422_1)(SwsContext *c, const int16_t *buf0, ); } } - -#if COMPILE_TEMPLATE_MMXEXT -static void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst, - int dstWidth, const uint8_t *src, - int srcW, int xInc) -{ - int32_t *filterPos = c->hLumFilterPos; - int16_t *filter = c->hLumFilter; - void *mmxextFilterCode = c->lumMmxextFilterCode; - int i; -#if defined(PIC) - uint64_t ebxsave; -#endif -#if ARCH_X86_64 - uint64_t retsave; -#endif - - __asm__ volatile( -#if defined(PIC) - "mov %%"REG_b", %5 \n\t" -#if ARCH_X86_64 - "mov -8(%%rsp), %%"REG_a" \n\t" - "mov %%"REG_a", %6 \n\t" -#endif -#else -#if ARCH_X86_64 - "mov -8(%%rsp), %%"REG_a" \n\t" - "mov %%"REG_a", %5 \n\t" -#endif -#endif - "pxor %%mm7, %%mm7 \n\t" - "mov %0, %%"REG_c" \n\t" - "mov %1, %%"REG_D" \n\t" - "mov %2, %%"REG_d" \n\t" - "mov %3, %%"REG_b" \n\t" - "xor %%"REG_a", %%"REG_a" \n\t" // i - PREFETCH" (%%"REG_c") \n\t" - PREFETCH" 32(%%"REG_c") \n\t" - PREFETCH" 64(%%"REG_c") \n\t" - -#if ARCH_X86_64 -#define CALL_MMXEXT_FILTER_CODE \ - "movl (%%"REG_b"), %%esi \n\t"\ - "call *%4 \n\t"\ - "movl (%%"REG_b", %%"REG_a"), %%esi \n\t"\ - "add %%"REG_S", %%"REG_c" \n\t"\ - "add %%"REG_a", %%"REG_D" \n\t"\ - "xor %%"REG_a", %%"REG_a" \n\t"\ - -#else -#define CALL_MMXEXT_FILTER_CODE \ - "movl (%%"REG_b"), %%esi \n\t"\ - "call *%4 \n\t"\ - "addl (%%"REG_b", %%"REG_a"), %%"REG_c" \n\t"\ - "add %%"REG_a", %%"REG_D" \n\t"\ - "xor %%"REG_a", %%"REG_a" \n\t"\ - -#endif /* ARCH_X86_64 */ - - CALL_MMXEXT_FILTER_CODE - CALL_MMXEXT_FILTER_CODE - CALL_MMXEXT_FILTER_CODE - CALL_MMXEXT_FILTER_CODE - CALL_MMXEXT_FILTER_CODE - CALL_MMXEXT_FILTER_CODE - CALL_MMXEXT_FILTER_CODE - CALL_MMXEXT_FILTER_CODE - -#if defined(PIC) - "mov %5, %%"REG_b" \n\t" -#if ARCH_X86_64 - "mov %6, %%"REG_a" \n\t" - "mov %%"REG_a", -8(%%rsp) \n\t" -#endif -#else -#if ARCH_X86_64 - "mov %5, %%"REG_a" \n\t" - "mov %%"REG_a", -8(%%rsp) \n\t" -#endif -#endif - :: "m" (src), "m" (dst), "m" (filter), "m" (filterPos), - "m" (mmxextFilterCode) -#if defined(PIC) - ,"m" (ebxsave) -#endif -#if ARCH_X86_64 - ,"m"(retsave) -#endif - : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D -#if !defined(PIC) - ,"%"REG_b -#endif - ); - - for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) - dst[i] = src[srcW-1]*128; -} - -static void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst1, int16_t *dst2, - int dstWidth, const uint8_t *src1, - const uint8_t *src2, int srcW, int xInc) -{ - int32_t *filterPos = c->hChrFilterPos; - int16_t *filter = c->hChrFilter; - void *mmxextFilterCode = c->chrMmxextFilterCode; - int i; -#if defined(PIC) - DECLARE_ALIGNED(8, uint64_t, ebxsave); -#endif -#if ARCH_X86_64 - DECLARE_ALIGNED(8, uint64_t, retsave); -#endif - - __asm__ volatile( -#if defined(PIC) - "mov %%"REG_b", %7 \n\t" -#if ARCH_X86_64 - "mov -8(%%rsp), %%"REG_a" \n\t" - "mov %%"REG_a", %8 \n\t" -#endif -#else -#if ARCH_X86_64 - "mov -8(%%rsp), %%"REG_a" \n\t" - "mov %%"REG_a", %7 \n\t" -#endif -#endif - "pxor %%mm7, %%mm7 \n\t" - "mov %0, %%"REG_c" \n\t" - "mov %1, %%"REG_D" \n\t" - "mov %2, %%"REG_d" \n\t" - "mov %3, %%"REG_b" \n\t" - "xor %%"REG_a", %%"REG_a" \n\t" // i - PREFETCH" (%%"REG_c") \n\t" - PREFETCH" 32(%%"REG_c") \n\t" - PREFETCH" 64(%%"REG_c") \n\t" - - CALL_MMXEXT_FILTER_CODE - CALL_MMXEXT_FILTER_CODE - CALL_MMXEXT_FILTER_CODE - CALL_MMXEXT_FILTER_CODE - "xor %%"REG_a", %%"REG_a" \n\t" // i - "mov %5, %%"REG_c" \n\t" // src - "mov %6, %%"REG_D" \n\t" // buf2 - PREFETCH" (%%"REG_c") \n\t" - PREFETCH" 32(%%"REG_c") \n\t" - PREFETCH" 64(%%"REG_c") \n\t" - - CALL_MMXEXT_FILTER_CODE - CALL_MMXEXT_FILTER_CODE - CALL_MMXEXT_FILTER_CODE - CALL_MMXEXT_FILTER_CODE - -#if defined(PIC) - "mov %7, %%"REG_b" \n\t" -#if ARCH_X86_64 - "mov %8, %%"REG_a" \n\t" - "mov %%"REG_a", -8(%%rsp) \n\t" -#endif -#else -#if ARCH_X86_64 - "mov %7, %%"REG_a" \n\t" - "mov %%"REG_a", -8(%%rsp) \n\t" -#endif -#endif - :: "m" (src1), "m" (dst1), "m" (filter), "m" (filterPos), - "m" (mmxextFilterCode), "m" (src2), "m"(dst2) -#if defined(PIC) - ,"m" (ebxsave) -#endif -#if ARCH_X86_64 - ,"m"(retsave) -#endif - : "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S, "%"REG_D -#if !defined(PIC) - ,"%"REG_b -#endif - ); - - for (i=dstWidth-1; (i*xInc)>>16 >=srcW-1; i--) { - dst1[i] = src1[srcW-1]*128; - dst2[i] = src2[srcW-1]*128; - } -} -#endif /* COMPILE_TEMPLATE_MMXEXT */ - static av_cold void RENAME(sws_init_swscale)(SwsContext *c) { enum AVPixelFormat dstFormat = c->dstFormat; @@ -1723,8 +1533,8 @@ static av_cold void RENAME(sws_init_swscale)(SwsContext *c) // Use the new MMX scaler if the MMXEXT one can't be used (it is faster than the x86 ASM one). #if COMPILE_TEMPLATE_MMXEXT if (c->flags & SWS_FAST_BILINEAR && c->canMMXEXTBeUsed) { - c->hyscale_fast = RENAME(hyscale_fast); - c->hcscale_fast = RENAME(hcscale_fast); + c->hyscale_fast = ff_hyscale_fast_mmxext; + c->hcscale_fast = ff_hcscale_fast_mmxext; } else { #endif /* COMPILE_TEMPLATE_MMXEXT */ c->hyscale_fast = NULL; -- cgit v1.2.3