diff --git a/libavcodec/arm/Makefile b/libavcodec/arm/Makefile index 0e5e53a16a..201f897a61 100644 --- a/libavcodec/arm/Makefile +++ b/libavcodec/arm/Makefile @@ -52,11 +52,13 @@ ARMV6-OBJS-$(CONFIG_VP8_DECODER) += arm/vp8_armv6.o \ arm/vp8dsp_init_armv6.o \ arm/vp8dsp_armv6.o +VFP-OBJS += arm/fmtconvert_vfp.o + VFP-OBJS-$(CONFIG_DCA_DECODER) += arm/dcadsp_vfp.o \ arm/synth_filter_vfp.o VFP-OBJS-$(CONFIG_FFT) += arm/fft_vfp.o VFP-OBJS-$(CONFIG_MDCT) += arm/mdct_vfp.o -VFP-OBJS-$(HAVE_ARMV6) += arm/fmtconvert_vfp.o +VFP-OBJS-$(HAVE_ARMV6) += arm/fmtconvert_vfp_armv6.o NEON-OBJS += arm/fmtconvert_neon.o diff --git a/libavcodec/arm/fmtconvert_init_arm.c b/libavcodec/arm/fmtconvert_init_arm.c index 185f1b99dd..37319ed125 100644 --- a/libavcodec/arm/fmtconvert_init_arm.c +++ b/libavcodec/arm/fmtconvert_init_arm.c @@ -43,16 +43,15 @@ av_cold void ff_fmt_convert_init_arm(FmtConvertContext *c, AVCodecContext *avctx { int cpu_flags = av_get_cpu_flags(); - if (have_vfp(cpu_flags) && have_armv6(cpu_flags)) { + if (have_vfp(cpu_flags)) { if (!have_vfpv3(cpu_flags)) { - // These functions don't use anything armv6 specific in themselves, - // but ff_float_to_int16_vfp which is in the same assembly source - // file does, thus the whole file requires armv6 to be built. c->int32_to_float_fmul_scalar = ff_int32_to_float_fmul_scalar_vfp; c->int32_to_float_fmul_array8 = ff_int32_to_float_fmul_array8_vfp; } - c->float_to_int16 = ff_float_to_int16_vfp; + if (have_armv6(cpu_flags)) { + c->float_to_int16 = ff_float_to_int16_vfp; + } } if (have_neon(cpu_flags)) { diff --git a/libavcodec/arm/fmtconvert_vfp.S b/libavcodec/arm/fmtconvert_vfp.S index a6d4ebdbbc..b14af454eb 100644 --- a/libavcodec/arm/fmtconvert_vfp.S +++ b/libavcodec/arm/fmtconvert_vfp.S @@ -1,5 +1,4 @@ /* - * Copyright (c) 2008 Siarhei Siamashka * Copyright (c) 2013 RISC OS Open Ltd * * This file is part of FFmpeg. @@ -22,62 +21,6 @@ #include "config.h" #include "libavutil/arm/asm.S" -/** - * ARM VFP optimized float to int16 conversion. - * Assume that len is a positive number and is multiple of 8, destination - * buffer is at least 4 bytes aligned (8 bytes alignment is better for - * performance), little-endian byte sex. - */ -@ void ff_float_to_int16_vfp(int16_t *dst, const float *src, int len) -function ff_float_to_int16_vfp, export=1 - push {r4-r8,lr} - vpush {d8-d11} - vldmia r1!, {s16-s23} - vcvt.s32.f32 s0, s16 - vcvt.s32.f32 s1, s17 - vcvt.s32.f32 s2, s18 - vcvt.s32.f32 s3, s19 - vcvt.s32.f32 s4, s20 - vcvt.s32.f32 s5, s21 - vcvt.s32.f32 s6, s22 - vcvt.s32.f32 s7, s23 -1: - subs r2, r2, #8 - vmov r3, r4, s0, s1 - vmov r5, r6, s2, s3 - vmov r7, r8, s4, s5 - vmov ip, lr, s6, s7 - it gt - vldmiagt r1!, {s16-s23} - ssat r4, #16, r4 - ssat r3, #16, r3 - ssat r6, #16, r6 - ssat r5, #16, r5 - pkhbt r3, r3, r4, lsl #16 - pkhbt r4, r5, r6, lsl #16 - itttt gt - vcvtgt.s32.f32 s0, s16 - vcvtgt.s32.f32 s1, s17 - vcvtgt.s32.f32 s2, s18 - vcvtgt.s32.f32 s3, s19 - itttt gt - vcvtgt.s32.f32 s4, s20 - vcvtgt.s32.f32 s5, s21 - vcvtgt.s32.f32 s6, s22 - vcvtgt.s32.f32 s7, s23 - ssat r8, #16, r8 - ssat r7, #16, r7 - ssat lr, #16, lr - ssat ip, #16, ip - pkhbt r5, r7, r8, lsl #16 - pkhbt r6, ip, lr, lsl #16 - stmia r0!, {r3-r6} - bgt 1b - - vpop {d8-d11} - pop {r4-r8,pc} -endfunc - /** * ARM VFP optimised int32 to float conversion. * Assume len is a multiple of 8, destination buffer is at least 4 bytes aligned diff --git a/libavcodec/arm/fmtconvert_vfp_armv6.S b/libavcodec/arm/fmtconvert_vfp_armv6.S new file mode 100644 index 0000000000..7b012bc8d8 --- /dev/null +++ b/libavcodec/arm/fmtconvert_vfp_armv6.S @@ -0,0 +1,78 @@ +/* + * Copyright (c) 2008 Siarhei Siamashka + * + * 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 "libavutil/arm/asm.S" + +/** + * ARM VFP optimized float to int16 conversion. + * Assume that len is a positive number and is multiple of 8, destination + * buffer is at least 4 bytes aligned (8 bytes alignment is better for + * performance), little-endian byte sex. + */ +@ void ff_float_to_int16_vfp(int16_t *dst, const float *src, int len) +function ff_float_to_int16_vfp, export=1 + push {r4-r8,lr} + vpush {d8-d11} + vldmia r1!, {s16-s23} + vcvt.s32.f32 s0, s16 + vcvt.s32.f32 s1, s17 + vcvt.s32.f32 s2, s18 + vcvt.s32.f32 s3, s19 + vcvt.s32.f32 s4, s20 + vcvt.s32.f32 s5, s21 + vcvt.s32.f32 s6, s22 + vcvt.s32.f32 s7, s23 +1: + subs r2, r2, #8 + vmov r3, r4, s0, s1 + vmov r5, r6, s2, s3 + vmov r7, r8, s4, s5 + vmov ip, lr, s6, s7 + it gt + vldmiagt r1!, {s16-s23} + ssat r4, #16, r4 + ssat r3, #16, r3 + ssat r6, #16, r6 + ssat r5, #16, r5 + pkhbt r3, r3, r4, lsl #16 + pkhbt r4, r5, r6, lsl #16 + itttt gt + vcvtgt.s32.f32 s0, s16 + vcvtgt.s32.f32 s1, s17 + vcvtgt.s32.f32 s2, s18 + vcvtgt.s32.f32 s3, s19 + itttt gt + vcvtgt.s32.f32 s4, s20 + vcvtgt.s32.f32 s5, s21 + vcvtgt.s32.f32 s6, s22 + vcvtgt.s32.f32 s7, s23 + ssat r8, #16, r8 + ssat r7, #16, r7 + ssat lr, #16, lr + ssat ip, #16, ip + pkhbt r5, r7, r8, lsl #16 + pkhbt r6, ip, lr, lsl #16 + stmia r0!, {r3-r6} + bgt 1b + + vpop {d8-d11} + pop {r4-r8,pc} +endfunc