diff options
author | Justin Ruggles <justin.ruggles@gmail.com> | 2012-05-21 12:58:41 -0400 |
---|---|---|
committer | Justin Ruggles <justin.ruggles@gmail.com> | 2012-06-08 13:14:38 -0400 |
commit | d5a7229ba4aabc2b6407c731d9175879ae54c5ea (patch) | |
tree | 0596aaae9622f0f7b7e5d689b20ab3dc82f59897 /libavutil | |
parent | 98db4e2a4e35ccc2406004216270ceaa1c6a7d00 (diff) | |
download | ffmpeg-d5a7229ba4aabc2b6407c731d9175879ae54c5ea.tar.gz |
Add a float DSP framework to libavutil
Move vector_fmul() from DSPContext to AVFloatDSPContext.
Diffstat (limited to 'libavutil')
-rw-r--r-- | libavutil/Makefile | 1 | ||||
-rw-r--r-- | libavutil/arm/Makefile | 7 | ||||
-rw-r--r-- | libavutil/arm/float_dsp_arm.h | 29 | ||||
-rw-r--r-- | libavutil/arm/float_dsp_init_arm.c | 33 | ||||
-rw-r--r-- | libavutil/arm/float_dsp_init_neon.c | 32 | ||||
-rw-r--r-- | libavutil/arm/float_dsp_init_vfp.c | 34 | ||||
-rw-r--r-- | libavutil/arm/float_dsp_neon.S | 64 | ||||
-rw-r--r-- | libavutil/arm/float_dsp_vfp.S | 68 | ||||
-rw-r--r-- | libavutil/float_dsp.c | 42 | ||||
-rw-r--r-- | libavutil/float_dsp.h | 53 | ||||
-rw-r--r-- | libavutil/ppc/Makefile | 3 | ||||
-rw-r--r-- | libavutil/ppc/float_dsp_altivec.c | 37 | ||||
-rw-r--r-- | libavutil/ppc/float_dsp_init.c | 37 | ||||
-rw-r--r-- | libavutil/x86/Makefile | 3 | ||||
-rw-r--r-- | libavutil/x86/float_dsp.asm | 55 | ||||
-rw-r--r-- | libavutil/x86/float_dsp_init.c | 41 |
16 files changed, 539 insertions, 0 deletions
diff --git a/libavutil/Makefile b/libavutil/Makefile index fb19ebf1e2..56ccd04dff 100644 --- a/libavutil/Makefile +++ b/libavutil/Makefile @@ -57,6 +57,7 @@ OBJS = adler32.o \ eval.o \ fifo.o \ file.o \ + float_dsp.o \ imgutils.o \ intfloat_readwrite.o \ inverse.o \ diff --git a/libavutil/arm/Makefile b/libavutil/arm/Makefile index e600383b9c..ac7eca6751 100644 --- a/libavutil/arm/Makefile +++ b/libavutil/arm/Makefile @@ -1 +1,8 @@ OBJS += arm/cpu.o \ + arm/float_dsp_init_arm.o \ + +ARMVFP-OBJS += arm/float_dsp_init_vfp.o \ + arm/float_dsp_vfp.o \ + +NEON-OBJS += arm/float_dsp_init_neon.o \ + arm/float_dsp_neon.o \ diff --git a/libavutil/arm/float_dsp_arm.h b/libavutil/arm/float_dsp_arm.h new file mode 100644 index 0000000000..976e5a0a24 --- /dev/null +++ b/libavutil/arm/float_dsp_arm.h @@ -0,0 +1,29 @@ +/* + * Copyright (c) 2009 Mans Rullgard <mans@mansr.com> + * + * This file is part of Libav. + * + * Libav 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, + * 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 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#ifndef AVUTIL_ARM_FLOAT_DSP_ARM_H +#define AVUTIL_ARM_FLOAT_DSP_ARM_H + +#include "libavutil/float_dsp.h" + +void ff_dsputil_init_vfp (AVFloatDSPContext *fdsp); +void ff_dsputil_init_neon(AVFloatDSPContext *fdsp); + +#endif /* AVUTIL_ARM_FLOAT_DSP_ARM_H */ diff --git a/libavutil/arm/float_dsp_init_arm.c b/libavutil/arm/float_dsp_init_arm.c new file mode 100644 index 0000000000..0fee8611c4 --- /dev/null +++ b/libavutil/arm/float_dsp_init_arm.c @@ -0,0 +1,33 @@ +/* + * ARM optimized DSP utils + * + * This file is part of Libav. + * + * Libav 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, + * 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 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/arm/cpu.h" +#include "libavutil/float_dsp.h" +#include "float_dsp_arm.h" + +void ff_float_dsp_init_arm(AVFloatDSPContext *fdsp) +{ + int cpu_flags = av_get_cpu_flags(); + + if (have_vfp(cpu_flags)) + ff_dsputil_init_vfp(fdsp); + if (have_neon(cpu_flags)) + ff_dsputil_init_neon(fdsp); +} diff --git a/libavutil/arm/float_dsp_init_neon.c b/libavutil/arm/float_dsp_init_neon.c new file mode 100644 index 0000000000..1e732db9df --- /dev/null +++ b/libavutil/arm/float_dsp_init_neon.c @@ -0,0 +1,32 @@ +/* + * ARM NEON optimised Float DSP functions + * Copyright (c) 2008 Mans Rullgard <mans@mansr.com> + * + * This file is part of Libav. + * + * Libav 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, + * 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 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include <stdint.h> + +#include "libavutil/float_dsp.h" +#include "float_dsp_arm.h" + +void ff_vector_fmul_neon(float *dst, const float *src0, const float *src1, int len); + +void ff_dsputil_init_neon(AVFloatDSPContext *fdsp) +{ + c->vector_fmul = ff_vector_fmul_neon; +} diff --git a/libavutil/arm/float_dsp_init_vfp.c b/libavutil/arm/float_dsp_init_vfp.c new file mode 100644 index 0000000000..d58dbded3a --- /dev/null +++ b/libavutil/arm/float_dsp_init_vfp.c @@ -0,0 +1,34 @@ +/* + * Copyright (c) 2008 Siarhei Siamashka <ssvb@users.sourceforge.net> + * + * This file is part of Libav. + * + * Libav 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, + * 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 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/arm/cpu.h" +#include "libavutil/float_dsp.h" +#include "float_dsp_arm.h" + +void ff_vector_fmul_vfp(float *dst, const float *src0, const float *src1, + int len); + +void ff_dsputil_init_vfp(DSPContext* c, AVCodecContext *avctx) +{ + int cpu_flags = av_get_cpu_flags(); + + if (!have_vfpv3(cpu_flags)) + c->vector_fmul = ff_vector_fmul_vfp; +} diff --git a/libavutil/arm/float_dsp_neon.S b/libavutil/arm/float_dsp_neon.S new file mode 100644 index 0000000000..d66fa09424 --- /dev/null +++ b/libavutil/arm/float_dsp_neon.S @@ -0,0 +1,64 @@ +/* + * ARM NEON optimised Float DSP functions + * Copyright (c) 2008 Mans Rullgard <mans@mansr.com> + * + * This file is part of Libav. + * + * Libav 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, + * 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 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "config.h" +#include "asm.S" + + preserve8 + +function ff_vector_fmul_neon, export=1 + subs r3, r3, #8 + vld1.32 {d0-d3}, [r1,:128]! + vld1.32 {d4-d7}, [r2,:128]! + vmul.f32 q8, q0, q2 + vmul.f32 q9, q1, q3 + beq 3f + bics ip, r3, #15 + beq 2f +1: subs ip, ip, #16 + vld1.32 {d0-d1}, [r1,:128]! + vld1.32 {d4-d5}, [r2,:128]! + vmul.f32 q10, q0, q2 + vld1.32 {d2-d3}, [r1,:128]! + vld1.32 {d6-d7}, [r2,:128]! + vmul.f32 q11, q1, q3 + vst1.32 {d16-d19},[r0,:128]! + vld1.32 {d0-d1}, [r1,:128]! + vld1.32 {d4-d5}, [r2,:128]! + vmul.f32 q8, q0, q2 + vld1.32 {d2-d3}, [r1,:128]! + vld1.32 {d6-d7}, [r2,:128]! + vmul.f32 q9, q1, q3 + vst1.32 {d20-d23},[r0,:128]! + bne 1b + ands r3, r3, #15 + beq 3f +2: vld1.32 {d0-d1}, [r1,:128]! + vld1.32 {d4-d5}, [r2,:128]! + vst1.32 {d16-d17},[r0,:128]! + vmul.f32 q8, q0, q2 + vld1.32 {d2-d3}, [r1,:128]! + vld1.32 {d6-d7}, [r2,:128]! + vst1.32 {d18-d19},[r0,:128]! + vmul.f32 q9, q1, q3 +3: vst1.32 {d16-d19},[r0,:128]! + bx lr +endfunc diff --git a/libavutil/arm/float_dsp_vfp.S b/libavutil/arm/float_dsp_vfp.S new file mode 100644 index 0000000000..3931828381 --- /dev/null +++ b/libavutil/arm/float_dsp_vfp.S @@ -0,0 +1,68 @@ +/* + * Copyright (c) 2008 Siarhei Siamashka <ssvb@users.sourceforge.net> + * + * This file is part of Libav. + * + * Libav 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, + * 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 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "config.h" +#include "asm.S" + +/** + * Assume that len is a positive number and is multiple of 8 + */ +@ void ff_vector_fmul_vfp(float *dst, const float *src0, const float *src1, int len) +function ff_vector_fmul_vfp, export=1 + vpush {d8-d15} + fmrx r12, fpscr + orr r12, r12, #(3 << 16) /* set vector size to 4 */ + fmxr fpscr, r12 + + vldmia r1!, {s0-s3} + vldmia r2!, {s8-s11} + vldmia r1!, {s4-s7} + vldmia r2!, {s12-s15} + vmul.f32 s8, s0, s8 +1: + subs r3, r3, #16 + vmul.f32 s12, s4, s12 + itttt ge + vldmiage r1!, {s16-s19} + vldmiage r2!, {s24-s27} + vldmiage r1!, {s20-s23} + vldmiage r2!, {s28-s31} + it ge + vmulge.f32 s24, s16, s24 + vstmia r0!, {s8-s11} + vstmia r0!, {s12-s15} + it ge + vmulge.f32 s28, s20, s28 + itttt gt + vldmiagt r1!, {s0-s3} + vldmiagt r2!, {s8-s11} + vldmiagt r1!, {s4-s7} + vldmiagt r2!, {s12-s15} + ittt ge + vmulge.f32 s8, s0, s8 + vstmiage r0!, {s24-s27} + vstmiage r0!, {s28-s31} + bgt 1b + + bic r12, r12, #(7 << 16) /* set vector size back to 1 */ + fmxr fpscr, r12 + vpop {d8-d15} + bx lr +endfunc diff --git a/libavutil/float_dsp.c b/libavutil/float_dsp.c new file mode 100644 index 0000000000..039dd07d36 --- /dev/null +++ b/libavutil/float_dsp.c @@ -0,0 +1,42 @@ +/* + * This file is part of Libav. + * + * Libav 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, + * 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 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "config.h" + +#include "float_dsp.h" + +static void vector_fmul_c(float *dst, const float *src0, const float *src1, + int len) +{ + int i; + for (i = 0; i < len; i++) + dst[i] = src0[i] * src1[i]; +} + +void avpriv_float_dsp_init(AVFloatDSPContext *fdsp, int bit_exact) +{ + fdsp->vector_fmul = vector_fmul_c; + +#if ARCH_ARM + ff_float_dsp_init_arm(fdsp); +#elif ARCH_PPC + ff_float_dsp_init_ppc(fdsp, bit_exact); +#elif ARCH_X86 + ff_float_dsp_init_x86(fdsp); +#endif +} diff --git a/libavutil/float_dsp.h b/libavutil/float_dsp.h new file mode 100644 index 0000000000..30161a252b --- /dev/null +++ b/libavutil/float_dsp.h @@ -0,0 +1,53 @@ +/* + * This file is part of Libav. + * + * Libav 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, + * 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 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#ifndef AVUTIL_FLOAT_DSP_H +#define AVUTIL_FLOAT_DSP_H + +typedef struct AVFloatDSPContext { + /** + * Calculate the product of two vectors of floats and store the result in + * a vector of floats. + * + * @param dst output vector + * constraints: 32-byte aligned + * @param src0 first input vector + * constraints: 32-byte aligned + * @param src1 second input vector + * constraints: 32-byte aligned + * @param len number of elements in the input + * constraints: multiple of 16 + */ + void (*vector_fmul)(float *dst, const float *src0, const float *src1, + int len); +} AVFloatDSPContext; + +/** + * Initialize a float DSP context. + * + * @param fdsp float DSP context + * @param strict setting to non-zero avoids using functions which may not be IEEE-754 compliant + */ +void avpriv_float_dsp_init(AVFloatDSPContext *fdsp, int strict); + + +void ff_float_dsp_init_arm(AVFloatDSPContext *fdsp); +void ff_float_dsp_init_ppc(AVFloatDSPContext *fdsp, int strict); +void ff_float_dsp_init_x86(AVFloatDSPContext *fdsp); + +#endif /* AVUTIL_FLOAT_DSP_H */ diff --git a/libavutil/ppc/Makefile b/libavutil/ppc/Makefile index 5b18b08957..4fd8d6d57e 100644 --- a/libavutil/ppc/Makefile +++ b/libavutil/ppc/Makefile @@ -1 +1,4 @@ OBJS += ppc/cpu.o \ + ppc/float_dsp_init.o \ + +ALTIVEC-OBJS += ppc/float_dsp_altivec.o \ diff --git a/libavutil/ppc/float_dsp_altivec.c b/libavutil/ppc/float_dsp_altivec.c new file mode 100644 index 0000000000..f08abc8d6f --- /dev/null +++ b/libavutil/ppc/float_dsp_altivec.c @@ -0,0 +1,37 @@ +/* + * Copyright (c) 2006 Luca Barbato <lu_zero@gentoo.org> + * + * This file is part of Libav. + * + * Libav 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, + * 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 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "util_altivec.h" + +void ff_vector_fmul_altivec(float *dst, const float *src0, const float *src1, + int len) +{ + int i; + vector float d0, d1, s, zero = (vector float)vec_splat_u32(0); + for (i = 0; i < len - 7; i += 8) { + d0 = vec_ld( 0, src0 + i); + s = vec_ld( 0, src1 + i); + d1 = vec_ld(16, src0 + i); + d0 = vec_madd(d0, s, zero); + d1 = vec_madd(d1, vec_ld(16, src1 + i), zero); + vec_st(d0, 0, dst + i); + vec_st(d1, 16, dst + i); + } +} diff --git a/libavutil/ppc/float_dsp_init.c b/libavutil/ppc/float_dsp_init.c new file mode 100644 index 0000000000..0477749e83 --- /dev/null +++ b/libavutil/ppc/float_dsp_init.c @@ -0,0 +1,37 @@ +/* + * Copyright (c) 2006 Luca Barbato <lu_zero@gentoo.org> + * + * This file is part of Libav. + * + * Libav 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, + * 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 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "libavutil/cpu.h" +#include "libavutil/float_dsp.h" + +extern void ff_vector_fmul_altivec(float *dst, const float *src0, + const float *src1, int len); + +void ff_float_dsp_init_ppc(AVFloatDSPContext *fdsp, int bit_exact) +{ +#if HAVE_ALTIVEC + int mm_flags = av_get_cpu_flags(); + + if (!(mm_flags & AV_CPU_FLAG_ALTIVEC)) + return; + + fdsp->vector_fmul = ff_vector_fmul_altivec; +#endif +} diff --git a/libavutil/x86/Makefile b/libavutil/x86/Makefile index de8a341442..4546353669 100644 --- a/libavutil/x86/Makefile +++ b/libavutil/x86/Makefile @@ -1 +1,4 @@ OBJS += x86/cpu.o \ + x86/float_dsp_init.o \ + +YASM-OBJS += x86/float_dsp.o \ diff --git a/libavutil/x86/float_dsp.asm b/libavutil/x86/float_dsp.asm new file mode 100644 index 0000000000..53be7ab99a --- /dev/null +++ b/libavutil/x86/float_dsp.asm @@ -0,0 +1,55 @@ +;***************************************************************************** +;* x86-optimized Float DSP functions +;* +;* This file is part of Libav. +;* +;* Libav 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, +;* 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 +;* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA +;****************************************************************************** + +%include "x86inc.asm" + +SECTION .text + +;----------------------------------------------------------------------------- +; void vector_fmul(float *dst, const float *src0, const float *src1, int len) +;----------------------------------------------------------------------------- +%macro VECTOR_FMUL 0 +cglobal vector_fmul, 4,4,2, dst, src0, src1, len + lea lenq, [lend*4 - 2*mmsize] +ALIGN 16 +.loop + mova m0, [src0q + lenq] + mova m1, [src0q + lenq + mmsize] + mulps m0, m0, [src1q + lenq] + mulps m1, m1, [src1q + lenq + mmsize] + mova [dstq + lenq], m0 + mova [dstq + lenq + mmsize], m1 + + sub lenq, 2*mmsize + jge .loop +%if mmsize == 32 + vzeroupper + RET +%else + REP_RET +%endif +%endmacro + +INIT_XMM sse +VECTOR_FMUL +%if HAVE_AVX +INIT_YMM avx +VECTOR_FMUL +%endif diff --git a/libavutil/x86/float_dsp_init.c b/libavutil/x86/float_dsp_init.c new file mode 100644 index 0000000000..10bb226f23 --- /dev/null +++ b/libavutil/x86/float_dsp_init.c @@ -0,0 +1,41 @@ +/* + * This file is part of Libav. + * + * Libav 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, + * 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 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#include "config.h" + +#include "libavutil/cpu.h" +#include "libavutil/float_dsp.h" + +extern void ff_vector_fmul_sse(float *dst, const float *src0, const float *src1, + int len); +extern void ff_vector_fmul_avx(float *dst, const float *src0, const float *src1, + int len); + +void ff_float_dsp_init_x86(AVFloatDSPContext *fdsp) +{ +#if HAVE_YASM + int mm_flags = av_get_cpu_flags(); + + if (mm_flags & AV_CPU_FLAG_SSE && HAVE_SSE) { + fdsp->vector_fmul = ff_vector_fmul_sse; + } + if (mm_flags & AV_CPU_FLAG_AVX && HAVE_AVX) { + fdsp->vector_fmul = ff_vector_fmul_avx; + } +#endif +} |