From patchwork Sat Jan 9 19:22:24 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Lynne X-Patchwork-Id: 24873 Return-Path: X-Original-To: patchwork@ffaux-bg.ffmpeg.org Delivered-To: patchwork@ffaux-bg.ffmpeg.org Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org [79.124.17.100]) by ffaux.localdomain (Postfix) with ESMTP id 8271A44B8EF for ; Sat, 9 Jan 2021 21:22:31 +0200 (EET) Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 67B2F68A86D; Sat, 9 Jan 2021 21:22:31 +0200 (EET) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from w4.tutanota.de (w4.tutanota.de [81.3.6.165]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id 5692568A82B for ; Sat, 9 Jan 2021 21:22:25 +0200 (EET) Received: from w3.tutanota.de (unknown [192.168.1.164]) by w4.tutanota.de (Postfix) with ESMTP id 01A5510600E3 for ; Sat, 9 Jan 2021 19:22:25 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; t=1610220144; s=s1; d=lynne.ee; h=From:From:To:To:Subject:Subject:Content-Description:Content-ID:Content-Type:Content-Type:Content-Transfer-Encoding:Cc:Date:Date:In-Reply-To:MIME-Version:MIME-Version:Message-ID:Message-ID:Reply-To:References:Sender; bh=a+5bTSHR6e5K1VOrBYGl7+BTKi4l2E7haEGLXc3Zz6M=; b=CnSzO0jS+G/uTwoiGkABZ72psbASERyunib5jQHfr5kCi1h/tXT50Mm11HLmIeYC nki+0t0zeCD5S84iWCGxk0pA+RMGmYAGNRnpT/fLaIruQwek2lf+9jXVyBK9dINC0+f HPHeWXBsae17tWZ3fSch4+TklGz8XH6plX7h4d5+FWXF3Dy7pRNRt1HsI/zU0b4jRE3 I6YZEdTM9ZGiMbzgue7PbE3q/9GNy07Blcv0FwAYxKGI4tyVsW8+Z/1yfIO91NiHVBL zGgYr33RUsiEMbZO7jg6O0MZbui6ZwVwiEZ0yxvau+9E5XnUbekKokdFyr/adqNXIAP HVZNmbWFbg== Date: Sat, 9 Jan 2021 20:22:24 +0100 (CET) From: Lynne To: Ffmpeg Devel Message-ID: MIME-Version: 1.0 Subject: [FFmpeg-devel] [PATCH 6/6] fft_fixed: remove 16-bit FFT code X-BeenThere: ffmpeg-devel@ffmpeg.org X-Mailman-Version: 2.1.20 Precedence: list List-Id: FFmpeg development discussions and patches List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Reply-To: FFmpeg development discussions and patches Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" No longer used by anything. Unfortunately the old FFT_FLOAT/FFT_FIXED_32 is left as-is. It's simply too much work for code meant to be all removed anyway. Patch attached. Read patch 1/6 to see the size savings. Subject: [PATCH 6/6] fft_fixed: remove 16-bit FFT code No longer used by anything. Unfortunately the old FFT_FLOAT/FFT_FIXED_32 is left as-is. It's simply too much work for code meant to be all removed anyway. --- libavcodec/Makefile | 7 +- libavcodec/arm/Makefile | 6 +- libavcodec/arm/fft_fixed_init_arm.c | 40 ----- libavcodec/arm/fft_fixed_neon.S | 261 ---------------------------- libavcodec/fft-internal.h | 27 +-- libavcodec/fft.h | 8 - libavcodec/fft_fixed.c | 21 --- libavcodec/fft_template.c | 2 - 8 files changed, 6 insertions(+), 366 deletions(-) delete mode 100644 libavcodec/arm/fft_fixed_init_arm.c delete mode 100644 libavcodec/arm/fft_fixed_neon.S delete mode 100644 libavcodec/fft_fixed.c diff --git a/libavcodec/Makefile b/libavcodec/Makefile index bf4256dc87..4b6d9c6e02 100644 --- a/libavcodec/Makefile +++ b/libavcodec/Makefile @@ -83,10 +83,9 @@ OBJS-$(CONFIG_EXIF) += exif.o tiff_common.o OBJS-$(CONFIG_FAANDCT) += faandct.o OBJS-$(CONFIG_FAANIDCT) += faanidct.o OBJS-$(CONFIG_FDCTDSP) += fdctdsp.o jfdctfst.o jfdctint.o -FFT-OBJS-$(CONFIG_HARDCODED_TABLES) += cos_tables.o cos_fixed_tables.o -OBJS-$(CONFIG_FFT) += avfft.o fft_fixed.o fft_float.o \ - fft_fixed_32.o fft_init_table.o \ - $(FFT-OBJS-yes) +FFT-OBJS-$(CONFIG_HARDCODED_TABLES) += cos_tables.o +OBJS-$(CONFIG_FFT) += avfft.o fft_float.o fft_fixed_32.o \ + fft_init_table.o $(FFT-OBJS-yes) OBJS-$(CONFIG_FLACDSP) += flacdsp.o OBJS-$(CONFIG_FMTCONVERT) += fmtconvert.o OBJS-$(CONFIG_GOLOMB) += golomb.o diff --git a/libavcodec/arm/Makefile b/libavcodec/arm/Makefile index 23e8abbf2e..c4ab93aeeb 100644 --- a/libavcodec/arm/Makefile +++ b/libavcodec/arm/Makefile @@ -5,8 +5,7 @@ OBJS-$(CONFIG_AC3DSP) += arm/ac3dsp_init_arm.o \ arm/ac3dsp_arm.o OBJS-$(CONFIG_AUDIODSP) += arm/audiodsp_init_arm.o OBJS-$(CONFIG_BLOCKDSP) += arm/blockdsp_init_arm.o -OBJS-$(CONFIG_FFT) += arm/fft_init_arm.o \ - arm/fft_fixed_init_arm.o +OBJS-$(CONFIG_FFT) += arm/fft_init_arm.o OBJS-$(CONFIG_FLACDSP) += arm/flacdsp_init_arm.o \ arm/flacdsp_arm.o OBJS-$(CONFIG_FMTCONVERT) += arm/fmtconvert_init_arm.o @@ -108,8 +107,7 @@ NEON-OBJS-$(CONFIG_AUDIODSP) += arm/audiodsp_init_neon.o \ arm/int_neon.o NEON-OBJS-$(CONFIG_BLOCKDSP) += arm/blockdsp_init_neon.o \ arm/blockdsp_neon.o -NEON-OBJS-$(CONFIG_FFT) += arm/fft_neon.o \ - arm/fft_fixed_neon.o +NEON-OBJS-$(CONFIG_FFT) += arm/fft_neon.o NEON-OBJS-$(CONFIG_FMTCONVERT) += arm/fmtconvert_neon.o NEON-OBJS-$(CONFIG_G722DSP) += arm/g722dsp_neon.o NEON-OBJS-$(CONFIG_H264CHROMA) += arm/h264cmc_neon.o diff --git a/libavcodec/arm/fft_fixed_init_arm.c b/libavcodec/arm/fft_fixed_init_arm.c deleted file mode 100644 index 30ce48d352..0000000000 --- a/libavcodec/arm/fft_fixed_init_arm.c +++ /dev/null @@ -1,40 +0,0 @@ -/* - * Copyright (c) 2009 Mans Rullgard - * - * 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/attributes.h" -#include "libavutil/cpu.h" -#include "libavutil/arm/cpu.h" - -#define FFT_FLOAT 0 -#include "libavcodec/fft.h" - -void ff_fft_fixed_calc_neon(FFTContext *s, FFTComplex *z); - -av_cold void ff_fft_fixed_init_arm(FFTContext *s) -{ - int cpu_flags = av_get_cpu_flags(); - - if (have_neon(cpu_flags)) { - s->fft_permutation = FF_FFT_PERM_SWAP_LSBS; -#if CONFIG_FFT - s->fft_calc = ff_fft_fixed_calc_neon; -#endif - } -} diff --git a/libavcodec/arm/fft_fixed_neon.S b/libavcodec/arm/fft_fixed_neon.S deleted file mode 100644 index 2651607544..0000000000 --- a/libavcodec/arm/fft_fixed_neon.S +++ /dev/null @@ -1,261 +0,0 @@ -/* - * Copyright (c) 2011 Mans Rullgard - * - * 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/arm/asm.S" - -.macro bflies d0, d1, r0, r1 - vrev64.32 \r0, \d1 @ t5, t6, t1, t2 - vhsub.s16 \r1, \d1, \r0 @ t1-t5, t2-t6, t5-t1, t6-t2 - vhadd.s16 \r0, \d1, \r0 @ t1+t5, t2+t6, t5+t1, t6+t2 - vext.16 \r1, \r1, \r1, #1 @ t2-t6, t5-t1, t6-t2, t1-t5 - vtrn.32 \r0, \r1 @ t1+t5, t2+t6, t2-t6, t5-t1 - @ t5, t6, t4, t3 - vhsub.s16 \d1, \d0, \r0 - vhadd.s16 \d0, \d0, \r0 -.endm - -.macro transform01 q0, q1, d3, c0, c1, r0, w0, w1 - vrev32.16 \r0, \d3 - vmull.s16 \w0, \d3, \c0 - vmlal.s16 \w0, \r0, \c1 - vshrn.s32 \d3, \w0, #15 - bflies \q0, \q1, \w0, \w1 -.endm - -.macro transform2 d0, d1, d2, d3, q0, q1, c0, c1, c2, c3, \ - r0, r1, w0, w1 - vrev32.16 \r0, \d1 - vrev32.16 \r1, \d3 - vmull.s16 \w0, \d1, \c0 - vmlal.s16 \w0, \r0, \c1 - vmull.s16 \w1, \d3, \c2 - vmlal.s16 \w1, \r1, \c3 - vshrn.s32 \d1, \w0, #15 - vshrn.s32 \d3, \w1, #15 - bflies \q0, \q1, \w0, \w1 -.endm - -.macro fft4 d0, d1, r0, r1 - vhsub.s16 \r0, \d0, \d1 @ t3, t4, t8, t7 - vhsub.s16 \r1, \d1, \d0 - vhadd.s16 \d0, \d0, \d1 @ t1, t2, t6, t5 - vmov.i64 \d1, #0xffff00000000 - vbit \r0, \r1, \d1 - vrev64.16 \r1, \r0 @ t7, t8, t4, t3 - vtrn.32 \r0, \r1 @ t3, t4, t7, t8 - vtrn.32 \d0, \r0 @ t1, t2, t3, t4, t6, t5, t8, t7 - vhsub.s16 \d1, \d0, \r0 @ r2, i2, r3, i1 - vhadd.s16 \d0, \d0, \r0 @ r0, i0, r1, i3 -.endm - -.macro fft8 d0, d1, d2, d3, q0, q1, c0, c1, r0, r1, w0, w1 - fft4 \d0, \d1, \r0, \r1 - vtrn.32 \d0, \d1 @ z0, z2, z1, z3 - vhadd.s16 \r0, \d2, \d3 @ t1, t2, t3, t4 - vhsub.s16 \d3, \d2, \d3 @ z5, z7 - vmov \d2, \r0 - transform01 \q0, \q1, \d3, \c0, \c1, \r0, \w0, \w1 -.endm - -function fft4_neon - vld1.16 {d0-d1}, [r0] - fft4 d0, d1, d2, d3 - vst1.16 {d0-d1}, [r0] - bx lr -endfunc - -function fft8_neon - vld1.16 {d0-d3}, [r0,:128] - movrel r1, coefs - vld1.16 {d30}, [r1,:64] - vdup.16 d31, d30[0] - fft8 d0, d1, d2, d3, q0, q1, d31, d30, d20, d21, q8, q9 - vtrn.32 d0, d1 - vtrn.32 d2, d3 - vst1.16 {d0-d3}, [r0,:128] - bx lr -endfunc - -function fft16_neon - vld1.16 {d0-d3}, [r0,:128]! - vld1.16 {d4-d7}, [r0,:128] - movrel r1, coefs - sub r0, r0, #32 - vld1.16 {d28-d31},[r1,:128] - vdup.16 d31, d28[0] - fft8 d0, d1, d2, d3, q0, q1, d31, d28, d20, d21, q8, q9 - vswp d5, d6 - fft4 q2, q3, q8, q9 - vswp d5, d6 - vtrn.32 q0, q1 @ z0, z4, z2, z6, z1, z5, z3, z7 - vtrn.32 q2, q3 @ z8, z12,z10,z14,z9, z13,z11,z15 - vswp d1, d2 - vdup.16 d31, d28[0] - transform01 q0, q2, d5, d31, d28, d20, q8, q9 - vdup.16 d26, d29[0] - vdup.16 d27, d30[0] - transform2 d2, d6, d3, d7, q1, q3, d26, d30, d27, d29, \ - d20, d21, q8, q9 - vtrn.32 q0, q1 - vtrn.32 q2, q3 - vst1.16 {d0-d3}, [r0,:128]! - vst1.16 {d4-d7}, [r0,:128] - bx lr -endfunc - -function fft_pass_neon - push {r4,lr} - movrel lr, coefs + 24 - vld1.16 {d30}, [lr,:64] - lsl r12, r2, #3 - vmov d31, d30 - add r3, r1, r2, lsl #2 - mov lr, #-8 - sub r3, r3, #2 - mov r4, r0 - vld1.16 {d27[]}, [r3,:16] - sub r3, r3, #6 - vld1.16 {q0}, [r4,:128], r12 - vld1.16 {q1}, [r4,:128], r12 - vld1.16 {q2}, [r4,:128], r12 - vld1.16 {q3}, [r4,:128], r12 - vld1.16 {d28}, [r1,:64]! - vld1.16 {d29}, [r3,:64], lr - vswp d1, d2 - vswp d5, d6 - vtrn.32 d0, d1 - vtrn.32 d4, d5 - vdup.16 d25, d28[1] - vmul.s16 d27, d27, d31 - transform01 q0, q2, d5, d25, d27, d20, q8, q9 - b 2f -1: - mov r4, r0 - vdup.16 d26, d29[0] - vld1.16 {q0}, [r4,:128], r12 - vld1.16 {q1}, [r4,:128], r12 - vld1.16 {q2}, [r4,:128], r12 - vld1.16 {q3}, [r4,:128], r12 - vld1.16 {d28}, [r1,:64]! - vld1.16 {d29}, [r3,:64], lr - vswp d1, d2 - vswp d5, d6 - vtrn.32 d0, d1 - vtrn.32 d4, d5 - vdup.16 d24, d28[0] - vdup.16 d25, d28[1] - vdup.16 d27, d29[3] - vmul.s16 q13, q13, q15 - transform2 d0, d4, d1, d5, q0, q2, d24, d26, d25, d27, \ - d16, d17, q9, q10 -2: - vtrn.32 d2, d3 - vtrn.32 d6, d7 - vdup.16 d24, d28[2] - vdup.16 d26, d29[2] - vdup.16 d25, d28[3] - vdup.16 d27, d29[1] - vmul.s16 q13, q13, q15 - transform2 d2, d6, d3, d7, q1, q3, d24, d26, d25, d27, \ - d16, d17, q9, q10 - vtrn.32 d0, d1 - vtrn.32 d2, d3 - vtrn.32 d4, d5 - vtrn.32 d6, d7 - vswp d1, d2 - vswp d5, d6 - mov r4, r0 - vst1.16 {q0}, [r4,:128], r12 - vst1.16 {q1}, [r4,:128], r12 - vst1.16 {q2}, [r4,:128], r12 - vst1.16 {q3}, [r4,:128], r12 - add r0, r0, #16 - subs r2, r2, #2 - bgt 1b - pop {r4,pc} -endfunc - -#define F_SQRT1_2 23170 -#define F_COS_16_1 30274 -#define F_COS_16_3 12540 - -const coefs, align=4 - .short F_SQRT1_2, -F_SQRT1_2, -F_SQRT1_2, F_SQRT1_2 - .short F_COS_16_1,-F_COS_16_1,-F_COS_16_1, F_COS_16_1 - .short F_COS_16_3,-F_COS_16_3,-F_COS_16_3, F_COS_16_3 - .short 1, -1, -1, 1 -endconst - -.macro def_fft n, n2, n4 -function fft\n\()_neon - push {r4, lr} - mov r4, r0 - bl fft\n2\()_neon - add r0, r4, #\n4*2*4 - bl fft\n4\()_neon - add r0, r4, #\n4*3*4 - bl fft\n4\()_neon - mov r0, r4 - pop {r4, lr} - movrelx r1, X(ff_cos_\n\()_fixed) - mov r2, #\n4/2 - b fft_pass_neon -endfunc -.endm - - def_fft 32, 16, 8 - def_fft 64, 32, 16 - def_fft 128, 64, 32 - def_fft 256, 128, 64 - def_fft 512, 256, 128 - def_fft 1024, 512, 256 - def_fft 2048, 1024, 512 - def_fft 4096, 2048, 1024 - def_fft 8192, 4096, 2048 - def_fft 16384, 8192, 4096 - def_fft 32768, 16384, 8192 - def_fft 65536, 32768, 16384 - -function ff_fft_fixed_calc_neon, export=1 - ldr r2, [r0] - sub r2, r2, #2 - movrel r3, fft_fixed_tab_neon - ldr r3, [r3, r2, lsl #2] - mov r0, r1 - bx r3 -endfunc - -const fft_fixed_tab_neon, relocate=1 - .word fft4_neon - .word fft8_neon - .word fft16_neon - .word fft32_neon - .word fft64_neon - .word fft128_neon - .word fft256_neon - .word fft512_neon - .word fft1024_neon - .word fft2048_neon - .word fft4096_neon - .word fft8192_neon - .word fft16384_neon - .word fft32768_neon - .word fft65536_neon -endconst diff --git a/libavcodec/fft-internal.h b/libavcodec/fft-internal.h index 5566ad9b81..3bd5a1123d 100644 --- a/libavcodec/fft-internal.h +++ b/libavcodec/fft-internal.h @@ -34,7 +34,7 @@ (dim) = (are) * (bim) + (aim) * (bre); \ } while (0) -#else +#else /* FFT_FLOAT */ #define SCALE_FLOAT(a, bits) lrint((a) * (double)(1 << (bits))) @@ -52,31 +52,6 @@ #define FIX15(a) av_clip(SCALE_FLOAT(a, 31), -2147483647, 2147483647) -#else /* FFT_FIXED_32 */ - -#include "fft.h" -#include "mathops.h" - -#define FIX15(a) av_clip(SCALE_FLOAT(a, 15), -32767, 32767) - -#define sqrthalf ((int16_t)((1<<15)*M_SQRT1_2)) - -#define BF(x, y, a, b) do { \ - x = (a - b) >> 1; \ - y = (a + b) >> 1; \ - } while (0) - -#define CMULS(dre, dim, are, aim, bre, bim, sh) do { \ - (dre) = (MUL16(are, bre) - MUL16(aim, bim)) >> sh; \ - (dim) = (MUL16(are, bim) + MUL16(aim, bre)) >> sh; \ - } while (0) - -#define CMUL(dre, dim, are, aim, bre, bim) \ - CMULS(dre, dim, are, aim, bre, bim, 15) - -#define CMULL(dre, dim, are, aim, bre, bim) \ - CMULS(dre, dim, are, aim, bre, bim, 0) - #endif /* FFT_FIXED_32 */ #endif /* FFT_FLOAT */ diff --git a/libavcodec/fft.h b/libavcodec/fft.h index ee823dd9ae..5ca2d18432 100644 --- a/libavcodec/fft.h +++ b/libavcodec/fft.h @@ -52,12 +52,6 @@ typedef float FFTDouble; typedef int32_t FFTSample; -#else /* FFT_FIXED_32 */ - -#define FFT_NAME(x) x ## _fixed - -typedef int16_t FFTSample; - #endif /* FFT_FIXED_32 */ typedef struct FFTComplex { @@ -162,8 +156,6 @@ void ff_fft_init_arm(FFTContext *s); void ff_fft_init_mips(FFTContext *s); void ff_fft_init_ppc(FFTContext *s); -void ff_fft_fixed_init_arm(FFTContext *s); - void ff_fft_end(FFTContext *s); #define ff_mdct_init FFT_NAME(ff_mdct_init) diff --git a/libavcodec/fft_fixed.c b/libavcodec/fft_fixed.c deleted file mode 100644 index 3d3bd2fca6..0000000000 --- a/libavcodec/fft_fixed.c +++ /dev/null @@ -1,21 +0,0 @@ -/* - * 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 - */ - -#define FFT_FLOAT 0 -#define FFT_FIXED_32 0 -#include "fft_template.c" diff --git a/libavcodec/fft_template.c b/libavcodec/fft_template.c index 4ffb9b1d9d..2d05990ca9 100644 --- a/libavcodec/fft_template.c +++ b/libavcodec/fft_template.c @@ -237,8 +237,6 @@ av_cold int ff_fft_init(FFTContext *s, int nbits, int inverse) if (ARCH_PPC) ff_fft_init_ppc(s); if (ARCH_X86) ff_fft_init_x86(s); if (HAVE_MIPSFPU) ff_fft_init_mips(s); -#else - if (ARCH_ARM) ff_fft_fixed_init_arm(s); #endif for(j=4; j<=nbits; j++) { ff_init_ff_cos_tabs(j);