diff options
Diffstat (limited to 'ffmpeg/libavutil/arm')
| -rw-r--r-- | ffmpeg/libavutil/arm/Makefile | 8 | ||||
| -rw-r--r-- | ffmpeg/libavutil/arm/asm.S | 304 | ||||
| -rw-r--r-- | ffmpeg/libavutil/arm/bswap.h | 67 | ||||
| -rw-r--r-- | ffmpeg/libavutil/arm/cpu.c | 146 | ||||
| -rw-r--r-- | ffmpeg/libavutil/arm/cpu.h | 32 | ||||
| -rw-r--r-- | ffmpeg/libavutil/arm/float_dsp_arm.h | 29 | ||||
| -rw-r--r-- | ffmpeg/libavutil/arm/float_dsp_init_arm.c | 33 | ||||
| -rw-r--r-- | ffmpeg/libavutil/arm/float_dsp_init_neon.c | 58 | ||||
| -rw-r--r-- | ffmpeg/libavutil/arm/float_dsp_init_vfp.c | 38 | ||||
| -rw-r--r-- | ffmpeg/libavutil/arm/float_dsp_neon.S | 271 | ||||
| -rw-r--r-- | ffmpeg/libavutil/arm/float_dsp_vfp.S | 137 | ||||
| -rw-r--r-- | ffmpeg/libavutil/arm/intmath.h | 110 | ||||
| -rw-r--r-- | ffmpeg/libavutil/arm/intreadwrite.h | 91 | ||||
| -rw-r--r-- | ffmpeg/libavutil/arm/timer.h | 40 |
14 files changed, 1364 insertions, 0 deletions
diff --git a/ffmpeg/libavutil/arm/Makefile b/ffmpeg/libavutil/arm/Makefile new file mode 100644 index 0000000..5da44b0 --- /dev/null +++ b/ffmpeg/libavutil/arm/Makefile @@ -0,0 +1,8 @@ +OBJS += arm/cpu.o \ + arm/float_dsp_init_arm.o \ + +VFP-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/ffmpeg/libavutil/arm/asm.S b/ffmpeg/libavutil/arm/asm.S new file mode 100644 index 0000000..6061e47 --- /dev/null +++ b/ffmpeg/libavutil/arm/asm.S @@ -0,0 +1,304 @@ +/* + * Copyright (c) 2008 Mans Rullgard <mans@mansr.com> + * + * 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 "config.h" + +#ifdef __ELF__ +# define ELF +#else +# define ELF @ +#endif + +#if CONFIG_THUMB +# define A @ +# define T +#else +# define A +# define T @ +#endif + +#if HAVE_NEON + .arch armv7-a +#elif HAVE_ARMV6T2 + .arch armv6t2 +#elif HAVE_ARMV6 + .arch armv6 +#elif HAVE_ARMV5TE + .arch armv5te +#endif + +#if HAVE_NEON + .fpu neon +#elif HAVE_VFP + .fpu vfp +#endif + + .syntax unified +T .thumb +ELF .eabi_attribute 25, 1 @ Tag_ABI_align_preserved + +.macro function name, export=0 + .set .Lpic_idx, 0 + .set .Lpic_gp, 0 + .macro endfunc + .if .Lpic_idx + .align 2 + .altmacro + put_pic %(.Lpic_idx - 1) + .noaltmacro + .endif +ELF .size \name, . - \name + .endfunc + .purgem endfunc + .endm + .text + .align 2 + .if \export + .global EXTERN_ASM\name +EXTERN_ASM\name: + .endif +ELF .type \name, %function + .func \name +\name: +.endm + +.macro const name, align=2 + .macro endconst +ELF .size \name, . - \name + .purgem endconst + .endm + .section .rodata + .align \align +\name: +.endm + +#if !HAVE_ARMV6T2_EXTERNAL +.macro movw rd, val + mov \rd, \val & 255 + orr \rd, \val & ~255 +.endm +#endif + +.macro mov32 rd, val +#if HAVE_ARMV6T2_EXTERNAL + movw \rd, #(\val) & 0xffff + .if (\val) >> 16 + movt \rd, #(\val) >> 16 + .endif +#else + ldr \rd, =\val +#endif +.endm + +.macro put_pic num + put_pic_\num +.endm + +.macro do_def_pic num, val, label + .macro put_pic_\num + .if \num + .altmacro + put_pic %(\num - 1) + .noaltmacro + .endif +\label: .word \val + .purgem put_pic_\num + .endm +.endm + +.macro def_pic val, label + .altmacro + do_def_pic %.Lpic_idx, \val, \label + .noaltmacro + .set .Lpic_idx, .Lpic_idx + 1 +.endm + +.macro ldpic rd, val, indir=0 + ldr \rd, .Lpicoff\@ +.Lpic\@: + .if \indir +A ldr \rd, [pc, \rd] +T add \rd, pc +T ldr \rd, [\rd] + .else + add \rd, pc + .endif + def_pic \val - (.Lpic\@ + (8 >> CONFIG_THUMB)), .Lpicoff\@ +.endm + +.macro movrel rd, val +#if CONFIG_PIC + ldpic \rd, \val +#elif HAVE_ARMV6T2_EXTERNAL && !defined(__APPLE__) + movw \rd, #:lower16:\val + movt \rd, #:upper16:\val +#else + ldr \rd, =\val +#endif +.endm + +.macro movrelx rd, val, gp +#if CONFIG_PIC && defined(__ELF__) + .ifnb \gp + .if .Lpic_gp + .unreq gp + .endif + gp .req \gp + ldpic gp, _GLOBAL_OFFSET_TABLE_ + .elseif !.Lpic_gp + gp .req r12 + ldpic gp, _GLOBAL_OFFSET_TABLE_ + .endif + .set .Lpic_gp, 1 + ldr \rd, .Lpicoff\@ + ldr \rd, [gp, \rd] + def_pic \val(GOT), .Lpicoff\@ +#elif CONFIG_PIC && defined(__APPLE__) + ldpic \rd, .Lpic\@, indir=1 + .non_lazy_symbol_pointer +.Lpic\@: + .indirect_symbol \val + .word 0 + .text +#else + movrel \rd, \val +#endif +.endm + +.macro add_sh rd, rn, rm, sh:vararg +A add \rd, \rn, \rm, \sh +T mov \rm, \rm, \sh +T add \rd, \rn, \rm +.endm + +.macro ldr_pre rt, rn, rm:vararg +A ldr \rt, [\rn, \rm]! +T add \rn, \rn, \rm +T ldr \rt, [\rn] +.endm + +.macro ldr_dpre rt, rn, rm:vararg +A ldr \rt, [\rn, -\rm]! +T sub \rn, \rn, \rm +T ldr \rt, [\rn] +.endm + +.macro ldr_nreg rt, rn, rm:vararg +A ldr \rt, [\rn, -\rm] +T sub \rt, \rn, \rm +T ldr \rt, [\rt] +.endm + +.macro ldr_post rt, rn, rm:vararg +A ldr \rt, [\rn], \rm +T ldr \rt, [\rn] +T add \rn, \rn, \rm +.endm + +.macro ldrd_reg rt, rt2, rn, rm +A ldrd \rt, \rt2, [\rn, \rm] +T add \rt, \rn, \rm +T ldrd \rt, \rt2, [\rt] +.endm + +.macro ldrd_post rt, rt2, rn, rm +A ldrd \rt, \rt2, [\rn], \rm +T ldrd \rt, \rt2, [\rn] +T add \rn, \rn, \rm +.endm + +.macro ldrh_pre rt, rn, rm +A ldrh \rt, [\rn, \rm]! +T add \rn, \rn, \rm +T ldrh \rt, [\rn] +.endm + +.macro ldrh_dpre rt, rn, rm +A ldrh \rt, [\rn, -\rm]! +T sub \rn, \rn, \rm +T ldrh \rt, [\rn] +.endm + +.macro ldrh_post rt, rn, rm +A ldrh \rt, [\rn], \rm +T ldrh \rt, [\rn] +T add \rn, \rn, \rm +.endm + +.macro ldrb_post rt, rn, rm +A ldrb \rt, [\rn], \rm +T ldrb \rt, [\rn] +T add \rn, \rn, \rm +.endm + +.macro str_post rt, rn, rm:vararg +A str \rt, [\rn], \rm +T str \rt, [\rn] +T add \rn, \rn, \rm +.endm + +.macro strb_post rt, rn, rm:vararg +A strb \rt, [\rn], \rm +T strb \rt, [\rn] +T add \rn, \rn, \rm +.endm + +.macro strd_post rt, rt2, rn, rm +A strd \rt, \rt2, [\rn], \rm +T strd \rt, \rt2, [\rn] +T add \rn, \rn, \rm +.endm + +.macro strh_pre rt, rn, rm +A strh \rt, [\rn, \rm]! +T add \rn, \rn, \rm +T strh \rt, [\rn] +.endm + +.macro strh_dpre rt, rn, rm +A strh \rt, [\rn, -\rm]! +T sub \rn, \rn, \rm +T strh \rt, [\rn] +.endm + +.macro strh_post rt, rn, rm +A strh \rt, [\rn], \rm +T strh \rt, [\rn] +T add \rn, \rn, \rm +.endm + +.macro strh_dpost rt, rn, rm +A strh \rt, [\rn], -\rm +T strh \rt, [\rn] +T sub \rn, \rn, \rm +.endm + +#if HAVE_VFP_ARGS + .eabi_attribute 28, 1 +# define VFP +# define NOVFP @ +#else +# define VFP @ +# define NOVFP +#endif + +#define GLUE(a, b) a ## b +#define JOIN(a, b) GLUE(a, b) +#define X(s) JOIN(EXTERN_ASM, s) diff --git a/ffmpeg/libavutil/arm/bswap.h b/ffmpeg/libavutil/arm/bswap.h new file mode 100644 index 0000000..ae5fdb7 --- /dev/null +++ b/ffmpeg/libavutil/arm/bswap.h @@ -0,0 +1,67 @@ +/* + * 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 + */ + +#ifndef AVUTIL_ARM_BSWAP_H +#define AVUTIL_ARM_BSWAP_H + +#include <stdint.h> +#include "config.h" +#include "libavutil/attributes.h" + +#ifdef __ARMCC_VERSION + +#if HAVE_ARMV6 +#define av_bswap32 av_bswap32 +static av_always_inline av_const uint32_t av_bswap32(uint32_t x) +{ + return __rev(x); +} +#endif /* HAVE_ARMV6 */ + +#elif HAVE_INLINE_ASM + +#if HAVE_ARMV6_INLINE +#define av_bswap16 av_bswap16 +static av_always_inline av_const unsigned av_bswap16(unsigned x) +{ + __asm__("rev16 %0, %0" : "+r"(x)); + return x; +} +#endif + +#if !AV_GCC_VERSION_AT_LEAST(4,5) +#define av_bswap32 av_bswap32 +static av_always_inline av_const uint32_t av_bswap32(uint32_t x) +{ +#if HAVE_ARMV6_INLINE + __asm__("rev %0, %0" : "+r"(x)); +#else + uint32_t t; + __asm__ ("eor %1, %0, %0, ror #16 \n\t" + "bic %1, %1, #0xFF0000 \n\t" + "mov %0, %0, ror #8 \n\t" + "eor %0, %0, %1, lsr #8 \n\t" + : "+r"(x), "=&r"(t)); +#endif /* HAVE_ARMV6_INLINE */ + return x; +} +#endif /* !AV_GCC_VERSION_AT_LEAST(4,5) */ + +#endif /* __ARMCC_VERSION */ + +#endif /* AVUTIL_ARM_BSWAP_H */ diff --git a/ffmpeg/libavutil/arm/cpu.c b/ffmpeg/libavutil/arm/cpu.c new file mode 100644 index 0000000..b4aabc3 --- /dev/null +++ b/ffmpeg/libavutil/arm/cpu.c @@ -0,0 +1,146 @@ +/* + * 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 "config.h" + +#define CORE_FLAG(f) \ + (AV_CPU_FLAG_ ## f * (HAVE_ ## f ## _EXTERNAL || HAVE_ ## f ## _INLINE)) + +#define CORE_CPU_FLAGS \ + (CORE_FLAG(ARMV5TE) | \ + CORE_FLAG(ARMV6) | \ + CORE_FLAG(ARMV6T2) | \ + CORE_FLAG(VFP) | \ + CORE_FLAG(VFPV3) | \ + CORE_FLAG(NEON)) + +#if defined __linux__ || defined __ANDROID__ + +#include <stdint.h> +#include <stdio.h> +#include <string.h> +#include "libavutil/avstring.h" + +#define AT_HWCAP 16 + +/* Relevant HWCAP values from kernel headers */ +#define HWCAP_VFP (1 << 6) +#define HWCAP_EDSP (1 << 7) +#define HWCAP_THUMBEE (1 << 11) +#define HWCAP_NEON (1 << 12) +#define HWCAP_VFPv3 (1 << 13) +#define HWCAP_TLS (1 << 15) + +static int get_hwcap(uint32_t *hwcap) +{ + struct { uint32_t a_type; uint32_t a_val; } auxv; + FILE *f = fopen("/proc/self/auxv", "r"); + int err = -1; + + if (!f) + return -1; + + while (fread(&auxv, sizeof(auxv), 1, f) > 0) { + if (auxv.a_type == AT_HWCAP) { + *hwcap = auxv.a_val; + err = 0; + break; + } + } + + fclose(f); + return err; +} + +static int get_cpuinfo(uint32_t *hwcap) +{ + FILE *f = fopen("/proc/cpuinfo", "r"); + char buf[200]; + + if (!f) + return -1; + + *hwcap = 0; + while (fgets(buf, sizeof(buf), f)) { + if (av_strstart(buf, "Features", NULL)) { + if (strstr(buf, " edsp ")) + *hwcap |= HWCAP_EDSP; + if (strstr(buf, " tls ")) + *hwcap |= HWCAP_TLS; + if (strstr(buf, " thumbee ")) + *hwcap |= HWCAP_THUMBEE; + if (strstr(buf, " vfp ")) + *hwcap |= HWCAP_VFP; + if (strstr(buf, " vfpv3 ")) + *hwcap |= HWCAP_VFPv3; + if (strstr(buf, " neon ")) + *hwcap |= HWCAP_NEON; + break; + } + } + fclose(f); + return 0; +} + +int ff_get_cpu_flags_arm(void) +{ + int flags = CORE_CPU_FLAGS; + uint32_t hwcap; + + if (get_hwcap(&hwcap) < 0) + if (get_cpuinfo(&hwcap) < 0) + return flags; + +#define check_cap(cap, flag) do { \ + if (hwcap & HWCAP_ ## cap) \ + flags |= AV_CPU_FLAG_ ## flag; \ + } while (0) + + /* No flags explicitly indicate v6 or v6T2 so check others which + imply support. */ + check_cap(EDSP, ARMV5TE); + check_cap(TLS, ARMV6); + check_cap(THUMBEE, ARMV6T2); + check_cap(VFP, VFP); + check_cap(VFPv3, VFPV3); + check_cap(NEON, NEON); + + /* The v6 checks above are not reliable so let higher flags + trickle down. */ + if (flags & (AV_CPU_FLAG_VFPV3 | AV_CPU_FLAG_NEON)) + flags |= AV_CPU_FLAG_ARMV6T2; + if (flags & AV_CPU_FLAG_ARMV6T2) + flags |= AV_CPU_FLAG_ARMV6; + + return flags; +} + +#else + +int ff_get_cpu_flags_arm(void) +{ + return AV_CPU_FLAG_ARMV5TE * HAVE_ARMV5TE | + AV_CPU_FLAG_ARMV6 * HAVE_ARMV6 | + AV_CPU_FLAG_ARMV6T2 * HAVE_ARMV6T2 | + AV_CPU_FLAG_VFP * HAVE_VFP | + AV_CPU_FLAG_VFPV3 * HAVE_VFPV3 | + AV_CPU_FLAG_NEON * HAVE_NEON; +} + +#endif diff --git a/ffmpeg/libavutil/arm/cpu.h b/ffmpeg/libavutil/arm/cpu.h new file mode 100644 index 0000000..91c959a --- /dev/null +++ b/ffmpeg/libavutil/arm/cpu.h @@ -0,0 +1,32 @@ +/* + * 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_CPU_H +#define AVUTIL_ARM_CPU_H + +#include "config.h" +#include "libavutil/cpu.h" + +#define have_armv5te(flags) (HAVE_ARMV5TE && ((flags) & AV_CPU_FLAG_ARMV5TE)) +#define have_armv6(flags) (HAVE_ARMV6 && ((flags) & AV_CPU_FLAG_ARMV6)) +#define have_armv6t2(flags) (HAVE_ARMV6T2 && ((flags) & AV_CPU_FLAG_ARMV6T2)) +#define have_vfp(flags) (HAVE_VFP && ((flags) & AV_CPU_FLAG_VFP)) +#define have_vfpv3(flags) (HAVE_VFPV3 && ((flags) & AV_CPU_FLAG_VFPV3)) +#define have_neon(flags) (HAVE_NEON && ((flags) & AV_CPU_FLAG_NEON)) + +#endif diff --git a/ffmpeg/libavutil/arm/float_dsp_arm.h b/ffmpeg/libavutil/arm/float_dsp_arm.h new file mode 100644 index 0000000..f3fafe3 --- /dev/null +++ b/ffmpeg/libavutil/arm/float_dsp_arm.h @@ -0,0 +1,29 @@ +/* + * Copyright (c) 2009 Mans Rullgard <mans@mansr.com> + * + * 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 + */ + +#ifndef AVUTIL_ARM_FLOAT_DSP_ARM_H +#define AVUTIL_ARM_FLOAT_DSP_ARM_H + +#include "libavutil/float_dsp.h" + +void ff_float_dsp_init_vfp (AVFloatDSPContext *fdsp); +void ff_float_dsp_init_neon(AVFloatDSPContext *fdsp); + +#endif /* AVUTIL_ARM_FLOAT_DSP_ARM_H */ diff --git a/ffmpeg/libavutil/arm/float_dsp_init_arm.c b/ffmpeg/libavutil/arm/float_dsp_init_arm.c new file mode 100644 index 0000000..f721344 --- /dev/null +++ b/ffmpeg/libavutil/arm/float_dsp_init_arm.c @@ -0,0 +1,33 @@ +/* + * ARM optimized DSP utils + * + * 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 "libavutil/float_dsp.h" +#include "cpu.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_float_dsp_init_vfp(fdsp); + if (have_neon(cpu_flags)) + ff_float_dsp_init_neon(fdsp); +} diff --git a/ffmpeg/libavutil/arm/float_dsp_init_neon.c b/ffmpeg/libavutil/arm/float_dsp_init_neon.c new file mode 100644 index 0000000..a7245ad --- /dev/null +++ b/ffmpeg/libavutil/arm/float_dsp_init_neon.c @@ -0,0 +1,58 @@ +/* + * 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_vector_fmac_scalar_neon(float *dst, const float *src, float mul, + int len); + +void ff_vector_fmul_scalar_neon(float *dst, const float *src, float mul, + int len); + +void ff_vector_fmul_window_neon(float *dst, const float *src0, + const float *src1, const float *win, int len); + +void ff_vector_fmul_add_neon(float *dst, const float *src0, const float *src1, + const float *src2, int len); + +void ff_vector_fmul_reverse_neon(float *dst, const float *src0, + const float *src1, int len); + +void ff_butterflies_float_neon(float *v1, float *v2, int len); + +float ff_scalarproduct_float_neon(const float *v1, const float *v2, int len); + +void ff_float_dsp_init_neon(AVFloatDSPContext *fdsp) +{ + fdsp->vector_fmul = ff_vector_fmul_neon; + fdsp->vector_fmac_scalar = ff_vector_fmac_scalar_neon; + fdsp->vector_fmul_scalar = ff_vector_fmul_scalar_neon; + fdsp->vector_fmul_window = ff_vector_fmul_window_neon; + fdsp->vector_fmul_add = ff_vector_fmul_add_neon; + fdsp->vector_fmul_reverse = ff_vector_fmul_reverse_neon; + fdsp->butterflies_float = ff_butterflies_float_neon; + fdsp->scalarproduct_float = ff_scalarproduct_float_neon; +} diff --git a/ffmpeg/libavutil/arm/float_dsp_init_vfp.c b/ffmpeg/libavutil/arm/float_dsp_init_vfp.c new file mode 100644 index 0000000..f7e2f54 --- /dev/null +++ b/ffmpeg/libavutil/arm/float_dsp_init_vfp.c @@ -0,0 +1,38 @@ +/* + * Copyright (c) 2008 Siarhei Siamashka <ssvb@users.sourceforge.net> + * + * 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 "libavutil/float_dsp.h" +#include "cpu.h" +#include "float_dsp_arm.h" + +void ff_vector_fmul_vfp(float *dst, const float *src0, const float *src1, + int len); + +void ff_vector_fmul_reverse_vfp(float *dst, const float *src0, + const float *src1, int len); + +void ff_float_dsp_init_vfp(AVFloatDSPContext *fdsp) +{ + int cpu_flags = av_get_cpu_flags(); + + if (!have_vfpv3(cpu_flags)) + fdsp->vector_fmul = ff_vector_fmul_vfp; + fdsp->vector_fmul_reverse = ff_vector_fmul_reverse_vfp; +} diff --git a/ffmpeg/libavutil/arm/float_dsp_neon.S b/ffmpeg/libavutil/arm/float_dsp_neon.S new file mode 100644 index 0000000..559b565 --- /dev/null +++ b/ffmpeg/libavutil/arm/float_dsp_neon.S @@ -0,0 +1,271 @@ +/* + * 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" + +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 + +function ff_vector_fmac_scalar_neon, export=1 +VFP len .req r2 +VFP acc .req r3 +NOVFP len .req r3 +NOVFP acc .req r2 +VFP vdup.32 q15, d0[0] +NOVFP vdup.32 q15, r2 + bics r12, len, #15 + mov acc, r0 + beq 3f + vld1.32 {q0}, [r1,:128]! + vld1.32 {q8}, [acc,:128]! + vld1.32 {q1}, [r1,:128]! + vld1.32 {q9}, [acc,:128]! +1: vmla.f32 q8, q0, q15 + vld1.32 {q2}, [r1,:128]! + vld1.32 {q10}, [acc,:128]! + vmla.f32 q9, q1, q15 + vld1.32 {q3}, [r1,:128]! + vld1.32 {q11}, [acc,:128]! + vmla.f32 q10, q2, q15 + vst1.32 {q8}, [r0,:128]! + vmla.f32 q11, q3, q15 + vst1.32 {q9}, [r0,:128]! + subs r12, r12, #16 + beq 2f + vld1.32 {q0}, [r1,:128]! + vld1.32 {q8}, [acc,:128]! + vst1.32 {q10}, [r0,:128]! + vld1.32 {q1}, [r1,:128]! + vld1.32 {q9}, [acc,:128]! + vst1.32 {q11}, [r0,:128]! + b 1b +2: vst1.32 {q10}, [r0,:128]! + vst1.32 {q11}, [r0,:128]! + ands len, len, #15 + it eq + bxeq lr +3: vld1.32 {q0}, [r1,:128]! + vld1.32 {q8}, [acc,:128]! + vmla.f32 q8, q0, q15 + vst1.32 {q8}, [r0,:128]! + subs len, len, #4 + bgt 3b + bx lr + .unreq len +endfunc + +function ff_vector_fmul_scalar_neon, export=1 +VFP len .req r2 +NOVFP len .req r3 +VFP vdup.32 q8, d0[0] +NOVFP vdup.32 q8, r2 + bics r12, len, #15 + beq 3f + vld1.32 {q0},[r1,:128]! + vld1.32 {q1},[r1,:128]! +1: vmul.f32 q0, q0, q8 + vld1.32 {q2},[r1,:128]! + vmul.f32 q1, q1, q8 + vld1.32 {q3},[r1,:128]! + vmul.f32 q2, q2, q8 + vst1.32 {q0},[r0,:128]! + vmul.f32 q3, q3, q8 + vst1.32 {q1},[r0,:128]! + subs r12, r12, #16 + beq 2f + vld1.32 {q0},[r1,:128]! + vst1.32 {q2},[r0,:128]! + vld1.32 {q1},[r1,:128]! + vst1.32 {q3},[r0,:128]! + b 1b +2: vst1.32 {q2},[r0,:128]! + vst1.32 {q3},[r0,:128]! + ands len, len, #15 + it eq + bxeq lr +3: vld1.32 {q0},[r1,:128]! + vmul.f32 q0, q0, q8 + vst1.32 {q0},[r0,:128]! + subs len, len, #4 + bgt 3b + bx lr + .unreq len +endfunc + +function ff_vector_fmul_window_neon, export=1 + push {r4,r5,lr} + ldr lr, [sp, #12] + sub r2, r2, #8 + sub r5, lr, #2 + add r2, r2, r5, lsl #2 + add r4, r3, r5, lsl #3 + add ip, r0, r5, lsl #3 + mov r5, #-16 + vld1.32 {d0,d1}, [r1,:128]! + vld1.32 {d2,d3}, [r2,:128], r5 + vld1.32 {d4,d5}, [r3,:128]! + vld1.32 {d6,d7}, [r4,:128], r5 +1: subs lr, lr, #4 + vmul.f32 d22, d0, d4 + vrev64.32 q3, q3 + vmul.f32 d23, d1, d5 + vrev64.32 q1, q1 + vmul.f32 d20, d0, d7 + vmul.f32 d21, d1, d6 + beq 2f + vmla.f32 d22, d3, d7 + vld1.32 {d0,d1}, [r1,:128]! + vmla.f32 d23, d2, d6 + vld1.32 {d18,d19},[r2,:128], r5 + vmls.f32 d20, d3, d4 + vld1.32 {d24,d25},[r3,:128]! + vmls.f32 d21, d2, d5 + vld1.32 {d6,d7}, [r4,:128], r5 + vmov q1, q9 + vrev64.32 q11, q11 + vmov q2, q12 + vswp d22, d23 + vst1.32 {d20,d21},[r0,:128]! + vst1.32 {d22,d23},[ip,:128], r5 + b 1b +2: vmla.f32 d22, d3, d7 + vmla.f32 d23, d2, d6 + vmls.f32 d20, d3, d4 + vmls.f32 d21, d2, d5 + vrev64.32 q11, q11 + vswp d22, d23 + vst1.32 {d20,d21},[r0,:128]! + vst1.32 {d22,d23},[ip,:128], r5 + pop {r4,r5,pc} +endfunc + +function ff_vector_fmul_add_neon, export=1 + ldr r12, [sp] + vld1.32 {q0-q1}, [r1,:128]! + vld1.32 {q8-q9}, [r2,:128]! + vld1.32 {q2-q3}, [r3,:128]! + vmul.f32 q10, q0, q8 + vmul.f32 q11, q1, q9 +1: vadd.f32 q12, q2, q10 + vadd.f32 q13, q3, q11 + pld [r1, #16] + pld [r2, #16] + pld [r3, #16] + subs r12, r12, #8 + beq 2f + vld1.32 {q0}, [r1,:128]! + vld1.32 {q8}, [r2,:128]! + vmul.f32 q10, q0, q8 + vld1.32 {q1}, [r1,:128]! + vld1.32 {q9}, [r2,:128]! + vmul.f32 q11, q1, q9 + vld1.32 {q2-q3}, [r3,:128]! + vst1.32 {q12-q13},[r0,:128]! + b 1b +2: vst1.32 {q12-q13},[r0,:128]! + bx lr +endfunc + +function ff_vector_fmul_reverse_neon, export=1 + add r2, r2, r3, lsl #2 + sub r2, r2, #32 + mov r12, #-32 + vld1.32 {q0-q1}, [r1,:128]! + vld1.32 {q2-q3}, [r2,:128], r12 +1: pld [r1, #32] + vrev64.32 q3, q3 + vmul.f32 d16, d0, d7 + vmul.f32 d17, d1, d6 + pld [r2, #-32] + vrev64.32 q2, q2 + vmul.f32 d18, d2, d5 + vmul.f32 d19, d3, d4 + subs r3, r3, #8 + beq 2f + vld1.32 {q0-q1}, [r1,:128]! + vld1.32 {q2-q3}, [r2,:128], r12 + vst1.32 {q8-q9}, [r0,:128]! + b 1b +2: vst1.32 {q8-q9}, [r0,:128]! + bx lr +endfunc + +function ff_butterflies_float_neon, export=1 +1: vld1.32 {q0},[r0,:128] + vld1.32 {q1},[r1,:128] + vsub.f32 q2, q0, q1 + vadd.f32 q1, q0, q1 + vst1.32 {q2},[r1,:128]! + vst1.32 {q1},[r0,:128]! + subs r2, r2, #4 + bgt 1b + bx lr +endfunc + +function ff_scalarproduct_float_neon, export=1 + vmov.f32 q2, #0.0 +1: vld1.32 {q0},[r0,:128]! + vld1.32 {q1},[r1,:128]! + vmla.f32 q2, q0, q1 + subs r2, r2, #4 + bgt 1b + vadd.f32 d0, d4, d5 + vpadd.f32 d0, d0, d0 +NOVFP vmov.32 r0, d0[0] + bx lr +endfunc diff --git a/ffmpeg/libavutil/arm/float_dsp_vfp.S b/ffmpeg/libavutil/arm/float_dsp_vfp.S new file mode 100644 index 0000000..8695fbd --- /dev/null +++ b/ffmpeg/libavutil/arm/float_dsp_vfp.S @@ -0,0 +1,137 @@ +/* + * Copyright (c) 2008 Siarhei Siamashka <ssvb@users.sourceforge.net> + * + * 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 "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 + +/** + * ARM VFP optimized implementation of 'vector_fmul_reverse_c' function. + * Assume that len is a positive number and is multiple of 8 + */ +@ void ff_vector_fmul_reverse_vfp(float *dst, const float *src0, +@ const float *src1, int len) +function ff_vector_fmul_reverse_vfp, export=1 + vpush {d8-d15} + add r2, r2, r3, lsl #2 + vldmdb r2!, {s0-s3} + vldmia r1!, {s8-s11} + vldmdb r2!, {s4-s7} + vldmia r1!, {s12-s15} + vmul.f32 s8, s3, s8 + vmul.f32 s9, s2, s9 + vmul.f32 s10, s1, s10 + vmul.f32 s11, s0, s11 +1: + subs r3, r3, #16 + it ge + vldmdbge r2!, {s16-s19} + vmul.f32 s12, s7, s12 + it ge + vldmiage r1!, {s24-s27} + vmul.f32 s13, s6, s13 + it ge + vldmdbge r2!, {s20-s23} + vmul.f32 s14, s5, s14 + it ge + vldmiage r1!, {s28-s31} + vmul.f32 s15, s4, s15 + it ge + vmulge.f32 s24, s19, s24 + it gt + vldmdbgt r2!, {s0-s3} + it ge + vmulge.f32 s25, s18, s25 + vstmia r0!, {s8-s13} + it ge + vmulge.f32 s26, s17, s26 + it gt + vldmiagt r1!, {s8-s11} + itt ge + vmulge.f32 s27, s16, s27 + vmulge.f32 s28, s23, s28 + it gt + vldmdbgt r2!, {s4-s7} + it ge + vmulge.f32 s29, s22, s29 + vstmia r0!, {s14-s15} + ittt ge + vmulge.f32 s30, s21, s30 + vmulge.f32 s31, s20, s31 + vmulge.f32 s8, s3, s8 + it gt + vldmiagt r1!, {s12-s15} + itttt ge + vmulge.f32 s9, s2, s9 + vmulge.f32 s10, s1, s10 + vstmiage r0!, {s24-s27} + vmulge.f32 s11, s0, s11 + it ge + vstmiage r0!, {s28-s31} + bgt 1b + + vpop {d8-d15} + bx lr +endfunc diff --git a/ffmpeg/libavutil/arm/intmath.h b/ffmpeg/libavutil/arm/intmath.h new file mode 100644 index 0000000..fd52648 --- /dev/null +++ b/ffmpeg/libavutil/arm/intmath.h @@ -0,0 +1,110 @@ +/* + * Copyright (c) 2010 Mans Rullgard <mans@mansr.com> + * + * 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 + */ + +#ifndef AVUTIL_ARM_INTMATH_H +#define AVUTIL_ARM_INTMATH_H + +#include <stdint.h> + +#include "config.h" +#include "libavutil/attributes.h" + +#if HAVE_INLINE_ASM + +#if HAVE_ARMV6_INLINE + +#define av_clip_uint8 av_clip_uint8_arm +static av_always_inline av_const unsigned av_clip_uint8_arm(int a) +{ + unsigned x; + __asm__ ("usat %0, #8, %1" : "=r"(x) : "r"(a)); + return x; +} + +#define av_clip_int8 av_clip_int8_arm +static av_always_inline av_const int av_clip_int8_arm(int a) +{ + int x; + __asm__ ("ssat %0, #8, %1" : "=r"(x) : "r"(a)); + return x; +} + +#define av_clip_uint16 av_clip_uint16_arm +static av_always_inline av_const unsigned av_clip_uint16_arm(int a) +{ + unsigned x; + __asm__ ("usat %0, #16, %1" : "=r"(x) : "r"(a)); + return x; +} + +#define av_clip_int16 av_clip_int16_arm +static av_always_inline av_const int av_clip_int16_arm(int a) +{ + int x; + __asm__ ("ssat %0, #16, %1" : "=r"(x) : "r"(a)); + return x; +} + +#define av_clip_uintp2 av_clip_uintp2_arm +static av_always_inline av_const unsigned av_clip_uintp2_arm(int a, int p) +{ + unsigned x; + __asm__ ("usat %0, %2, %1" : "=r"(x) : "r"(a), "i"(p)); + return x; +} + +#define av_sat_add32 av_sat_add32_arm +static av_always_inline int av_sat_add32_arm(int a, int b) +{ + int r; + __asm__ ("qadd %0, %1, %2" : "=r"(r) : "r"(a), "r"(b)); + return r; +} + +#define av_sat_dadd32 av_sat_dadd32_arm +static av_always_inline int av_sat_dadd32_arm(int a, int b) +{ + int r; + __asm__ ("qdadd %0, %1, %2" : "=r"(r) : "r"(a), "r"(b)); + return r; +} + +#endif /* HAVE_ARMV6_INLINE */ + +#if HAVE_ASM_MOD_Q + +#define av_clipl_int32 av_clipl_int32_arm +static av_always_inline av_const int32_t av_clipl_int32_arm(int64_t a) +{ + int x, y; + __asm__ ("adds %1, %R2, %Q2, lsr #31 \n\t" + "itet ne \n\t" + "mvnne %1, #1<<31 \n\t" + "moveq %0, %Q2 \n\t" + "eorne %0, %1, %R2, asr #31 \n\t" + : "=r"(x), "=&r"(y) : "r"(a) : "cc"); + return x; +} + +#endif /* HAVE_ASM_MOD_Q */ + +#endif /* HAVE_INLINE_ASM */ + +#endif /* AVUTIL_ARM_INTMATH_H */ diff --git a/ffmpeg/libavutil/arm/intreadwrite.h b/ffmpeg/libavutil/arm/intreadwrite.h new file mode 100644 index 0000000..2340a9a --- /dev/null +++ b/ffmpeg/libavutil/arm/intreadwrite.h @@ -0,0 +1,91 @@ +/* + * 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 + */ + +#ifndef AVUTIL_ARM_INTREADWRITE_H +#define AVUTIL_ARM_INTREADWRITE_H + +#include <stdint.h> +#include "config.h" +#include "libavutil/attributes.h" + +#if HAVE_FAST_UNALIGNED && HAVE_INLINE_ASM && !AV_GCC_VERSION_AT_LEAST(4,7) + +#define AV_RN16 AV_RN16 +static av_always_inline unsigned AV_RN16(const void *p) +{ + const uint8_t *q = p; + unsigned v; +#if !AV_GCC_VERSION_AT_LEAST(4,6) + __asm__ ("ldrh %0, %1" : "=r"(v) : "m"(*(const uint16_t *)q)); +#elif defined __thumb__ + __asm__ ("ldrh %0, %1" : "=r"(v) : "m"(q[0]), "m"(q[1])); +#else + __asm__ ("ldrh %0, %1" : "=r"(v) : "Uq"(q[0]), "m"(q[1])); +#endif + return v; +} + +#define AV_WN16 AV_WN16 +static av_always_inline void AV_WN16(void *p, uint16_t v) +{ + __asm__ ("strh %1, %0" : "=m"(*(uint16_t *)p) : "r"(v)); +} + +#define AV_RN32 AV_RN32 +static av_always_inline uint32_t AV_RN32(const void *p) +{ + const struct __attribute__((packed)) { uint32_t v; } *q = p; + uint32_t v; + __asm__ ("ldr %0, %1" : "=r"(v) : "m"(*q)); + return v; +} + +#define AV_WN32 AV_WN32 +static av_always_inline void AV_WN32(void *p, uint32_t v) +{ + __asm__ ("str %1, %0" : "=m"(*(uint32_t *)p) : "r"(v)); +} + +#if HAVE_ASM_MOD_Q + +#define AV_RN64 AV_RN64 +static av_always_inline uint64_t AV_RN64(const void *p) +{ + const struct __attribute__((packed)) { uint32_t v; } *q = p; + uint64_t v; + __asm__ ("ldr %Q0, %1 \n\t" + "ldr %R0, %2 \n\t" + : "=&r"(v) + : "m"(q[0]), "m"(q[1])); + return v; +} + +#define AV_WN64 AV_WN64 +static av_always_inline void AV_WN64(void *p, uint64_t v) +{ + __asm__ ("str %Q2, %0 \n\t" + "str %R2, %1 \n\t" + : "=m"(*(uint32_t*)p), "=m"(*((uint32_t*)p+1)) + : "r"(v)); +} + +#endif /* HAVE_ASM_MOD_Q */ + +#endif /* HAVE_INLINE_ASM */ + +#endif /* AVUTIL_ARM_INTREADWRITE_H */ diff --git a/ffmpeg/libavutil/arm/timer.h b/ffmpeg/libavutil/arm/timer.h new file mode 100644 index 0000000..5e8bc8e --- /dev/null +++ b/ffmpeg/libavutil/arm/timer.h @@ -0,0 +1,40 @@ +/* + * Copyright (c) 2009 Mans Rullgard <mans@mansr.com> + * + * 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 + */ + +#ifndef AVUTIL_ARM_TIMER_H +#define AVUTIL_ARM_TIMER_H + +#include <stdint.h> +#include "config.h" + +#if HAVE_INLINE_ASM && defined(__ARM_ARCH_7A__) + +#define AV_READ_TIME read_time + +static inline uint64_t read_time(void) +{ + unsigned cc; + __asm__ volatile ("mrc p15, 0, %0, c9, c13, 0" : "=r"(cc)); + return cc; +} + +#endif /* HAVE_INLINE_ASM && __ARM_ARCH_7A__ */ + +#endif /* AVUTIL_ARM_TIMER_H */ |