PR #21164 opened by Sean McGovern (sean_mcg) URL: https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/21164 Patch URL: https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/21164.patch
refs: #20832 >From aa5fa0212b371e8fdb00005ec8c12dd346c8e567 Mon Sep 17 00:00:00 2001 From: Sean McGovern <[email protected]> Date: Sat, 1 Feb 2025 06:24:56 -0500 Subject: [PATCH] libswscale: deprecate big-endian AltiVec acceleration refs: #20832 --- Changelog | 1 + libswscale/ppc/Makefile | 5 +- libswscale/ppc/swscale_altivec.c | 285 ---------- libswscale/ppc/yuv2rgb_altivec.c | 874 ------------------------------- libswscale/ppc/yuv2yuv_altivec.c | 204 -------- libswscale/swscale.c | 2 +- libswscale/swscale_internal.h | 2 + libswscale/utils.c | 3 +- libswscale/version.h | 2 +- 9 files changed, 7 insertions(+), 1371 deletions(-) delete mode 100644 libswscale/ppc/swscale_altivec.c delete mode 100644 libswscale/ppc/yuv2rgb_altivec.c delete mode 100644 libswscale/ppc/yuv2yuv_altivec.c diff --git a/Changelog b/Changelog index ef72af1e93..8f44c53a3d 100644 --- a/Changelog +++ b/Changelog @@ -16,6 +16,7 @@ version <next>: - Rockchip H.264/HEVC hardware encoder - Add vf_scale_d3d12 filter +- remove big-endian AltiVec acceleration version 8.0: - Whisper filter diff --git a/libswscale/ppc/Makefile b/libswscale/ppc/Makefile index 0a31a3025b..edbf1f6f81 100644 --- a/libswscale/ppc/Makefile +++ b/libswscale/ppc/Makefile @@ -1,4 +1 @@ -OBJS += ppc/swscale_altivec.o \ - ppc/yuv2rgb_altivec.o \ - ppc/yuv2yuv_altivec.o \ - ppc/swscale_vsx.o +OBJS += ppc/swscale_vsx.o diff --git a/libswscale/ppc/swscale_altivec.c b/libswscale/ppc/swscale_altivec.c deleted file mode 100644 index 3112c6cb71..0000000000 --- a/libswscale/ppc/swscale_altivec.c +++ /dev/null @@ -1,285 +0,0 @@ -/* - * AltiVec-enhanced yuv2yuvX - * - * Copyright (C) 2004 Romain Dolbeau <[email protected]> - * based on the equivalent C code in swscale.c - * - * 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 <inttypes.h> - -#include "config.h" -#include "libswscale/swscale.h" -#include "libswscale/swscale_internal.h" -#include "libavutil/attributes.h" -#include "libavutil/cpu.h" -#include "libavutil/intfloat.h" -#include "yuv2rgb_altivec.h" -#include "libavutil/ppc/util_altivec.h" - -#if HAVE_ALTIVEC -#if HAVE_BIGENDIAN -#define vzero vec_splat_s32(0) - -#define GET_LS(a,b,c,s) {\ - vector signed short l2 = vec_ld(((b) << 1) + 16, s);\ - ls = vec_perm(a, l2, c);\ - a = l2;\ - } - -#define yuv2planeX_8(d1, d2, l1, src, x, perm, filter) do {\ - vector signed short ls;\ - vector signed int vf1, vf2, i1, i2;\ - GET_LS(l1, x, perm, src);\ - i1 = vec_mule(filter, ls);\ - i2 = vec_mulo(filter, ls);\ - vf1 = vec_mergeh(i1, i2);\ - vf2 = vec_mergel(i1, i2);\ - d1 = vec_add(d1, vf1);\ - d2 = vec_add(d2, vf2);\ - } while (0) - -#define LOAD_FILTER(vf,f) {\ - vector unsigned char perm0 = vec_lvsl(joffset, f);\ - vf = vec_ld(joffset, f);\ - vf = vec_perm(vf, vf, perm0);\ -} -#define LOAD_L1(ll1,s,p){\ - p = vec_lvsl(xoffset, s);\ - ll1 = vec_ld(xoffset, s);\ -} - -// The 3 above is 2 (filterSize == 4) + 1 (sizeof(short) == 2). - -// The neat trick: We only care for half the elements, -// high or low depending on (i<<3)%16 (it's 0 or 8 here), -// and we're going to use vec_mule, so we choose -// carefully how to "unpack" the elements into the even slots. -#define GET_VF4(a, vf, f) {\ - vf = vec_ld(a<< 3, f);\ - if ((a << 3) % 16)\ - vf = vec_mergel(vf, (vector signed short)vzero);\ - else\ - vf = vec_mergeh(vf, (vector signed short)vzero);\ -} -#define FIRST_LOAD(sv, pos, s, per) {\ - sv = vec_ld(pos, s);\ - per = vec_lvsl(pos, s);\ -} -#define UPDATE_PTR(s0, d0, s1, d1) {\ - d0 = s0;\ - d1 = s1;\ -} -#define LOAD_SRCV(pos, a, s, per, v0, v1, vf) {\ - v1 = vec_ld(pos + a + 16, s);\ - vf = vec_perm(v0, v1, per);\ -} -#define LOAD_SRCV8(pos, a, s, per, v0, v1, vf) {\ - if ((((uintptr_t)s + pos) % 16) > 8) {\ - v1 = vec_ld(pos + a + 16, s);\ - }\ - vf = vec_perm(v0, src_v1, per);\ -} -#define GET_VFD(a, b, f, vf0, vf1, per, vf, off) {\ - vf1 = vec_ld((a * 2 * filterSize) + (b * 2) + 16 + off, f);\ - vf = vec_perm(vf0, vf1, per);\ -} - -#define FUNC(name) name ## _altivec -#include "swscale_ppc_template.c" -#undef FUNC - -#undef vzero - -#endif /* HAVE_BIGENDIAN */ - -#define SHIFT 3 - -#define get_pixel(val, bias, signedness) \ - (bias + av_clip_ ## signedness ## 16(val >> shift)) - -static void -yuv2plane1_float_u(const int32_t *src, float *dest, int dstW, int start) -{ - static const int shift = 3; - static const float float_mult = 1.0f / 65535.0f; - int i, val; - uint16_t val_uint; - - for (i = start; i < dstW; ++i){ - val = src[i] + (1 << (shift - 1)); - val_uint = get_pixel(val, 0, uint); - dest[i] = float_mult * (float)val_uint; - } -} - -static void -yuv2plane1_float_bswap_u(const int32_t *src, uint32_t *dest, int dstW, int start) -{ - static const int shift = 3; - static const float float_mult = 1.0f / 65535.0f; - int i, val; - uint16_t val_uint; - - for (i = start; i < dstW; ++i){ - val = src[i] + (1 << (shift - 1)); - val_uint = get_pixel(val, 0, uint); - dest[i] = av_bswap32(av_float2int(float_mult * (float)val_uint)); - } -} - -static void yuv2plane1_float_altivec(const int32_t *src, float *dest, int dstW) -{ - const int dst_u = -(uintptr_t)dest & 3; - const int add = (1 << (SHIFT - 1)); - const int clip = (1 << 16) - 1; - const float fmult = 1.0f / 65535.0f; - const vec_u32 vadd = (vec_u32) {add, add, add, add}; - const vec_u32 vshift = (vec_u32) vec_splat_u32(SHIFT); - const vec_u32 vlargest = (vec_u32) {clip, clip, clip, clip}; - const vec_f vmul = (vec_f) {fmult, fmult, fmult, fmult}; - const vec_f vzero = (vec_f) {0, 0, 0, 0}; - vec_u32 v; - vec_f vd; - int i; - - yuv2plane1_float_u(src, dest, dst_u, 0); - - for (i = dst_u; i < dstW - 3; i += 4) { - v = vec_ld(0, (const uint32_t *) &src[i]); - v = vec_add(v, vadd); - v = vec_sr(v, vshift); - v = vec_min(v, vlargest); - - vd = vec_ctf(v, 0); - vd = vec_madd(vd, vmul, vzero); - - vec_st(vd, 0, &dest[i]); - } - - yuv2plane1_float_u(src, dest, dstW, i); -} - -static void yuv2plane1_float_bswap_altivec(const int32_t *src, uint32_t *dest, int dstW) -{ - const int dst_u = -(uintptr_t)dest & 3; - const int add = (1 << (SHIFT - 1)); - const int clip = (1 << 16) - 1; - const float fmult = 1.0f / 65535.0f; - const vec_u32 vadd = (vec_u32) {add, add, add, add}; - const vec_u32 vshift = (vec_u32) vec_splat_u32(SHIFT); - const vec_u32 vlargest = (vec_u32) {clip, clip, clip, clip}; - const vec_f vmul = (vec_f) {fmult, fmult, fmult, fmult}; - const vec_f vzero = (vec_f) {0, 0, 0, 0}; - const vec_u32 vswapbig = (vec_u32) {16, 16, 16, 16}; - const vec_u16 vswapsmall = vec_splat_u16(8); - vec_u32 v; - vec_f vd; - int i; - - yuv2plane1_float_bswap_u(src, dest, dst_u, 0); - - for (i = dst_u; i < dstW - 3; i += 4) { - v = vec_ld(0, (const uint32_t *) &src[i]); - v = vec_add(v, vadd); - v = vec_sr(v, vshift); - v = vec_min(v, vlargest); - - vd = vec_ctf(v, 0); - vd = vec_madd(vd, vmul, vzero); - - vd = (vec_f) vec_rl((vec_u32) vd, vswapbig); - vd = (vec_f) vec_rl((vec_u16) vd, vswapsmall); - - vec_st(vd, 0, (float *) &dest[i]); - } - - yuv2plane1_float_bswap_u(src, dest, dstW, i); -} - -#define yuv2plane1_float(template, dest_type, BE_LE) \ -static void yuv2plane1_float ## BE_LE ## _altivec(const int16_t *src, uint8_t *dest, \ - int dstW, \ - const uint8_t *dither, int offset) \ -{ \ - template((const int32_t *)src, (dest_type *)dest, dstW); \ -} - -#if HAVE_BIGENDIAN -yuv2plane1_float(yuv2plane1_float_altivec, float, BE) -yuv2plane1_float(yuv2plane1_float_bswap_altivec, uint32_t, LE) -#else -yuv2plane1_float(yuv2plane1_float_altivec, float, LE) -yuv2plane1_float(yuv2plane1_float_bswap_altivec, uint32_t, BE) -#endif - -#endif /* HAVE_ALTIVEC */ - -av_cold void ff_sws_init_swscale_ppc(SwsInternal *c) -{ -#if HAVE_ALTIVEC - enum AVPixelFormat dstFormat = c->opts.dst_format; - - if (!(av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC)) - return; - -#if HAVE_BIGENDIAN - if (c->srcBpc == 8 && c->dstBpc <= 14) { - c->hyScale = c->hcScale = hScale_real_altivec; - } - if (!is16BPS(dstFormat) && !isNBPS(dstFormat) && !isSemiPlanarYUV(dstFormat) && !isDataInHighBits(dstFormat) && - dstFormat != AV_PIX_FMT_GRAYF32BE && dstFormat != AV_PIX_FMT_GRAYF32LE && - !c->needAlpha) { - c->yuv2planeX = yuv2planeX_altivec; - } -#endif - - if (dstFormat == AV_PIX_FMT_GRAYF32BE) { - c->yuv2plane1 = yuv2plane1_floatBE_altivec; - } else if (dstFormat == AV_PIX_FMT_GRAYF32LE) { - c->yuv2plane1 = yuv2plane1_floatLE_altivec; - } - - /* The following list of supported dstFormat values should - * match what's found in the body of ff_yuv2packedX_altivec() */ - if (!(c->opts.flags & (SWS_BITEXACT | SWS_FULL_CHR_H_INT)) && !c->needAlpha) { - switch (c->opts.dst_format) { - case AV_PIX_FMT_ABGR: - c->yuv2packedX = ff_yuv2abgr_X_altivec; - break; - case AV_PIX_FMT_BGRA: - c->yuv2packedX = ff_yuv2bgra_X_altivec; - break; - case AV_PIX_FMT_ARGB: - c->yuv2packedX = ff_yuv2argb_X_altivec; - break; - case AV_PIX_FMT_RGBA: - c->yuv2packedX = ff_yuv2rgba_X_altivec; - break; - case AV_PIX_FMT_BGR24: - c->yuv2packedX = ff_yuv2bgr24_X_altivec; - break; - case AV_PIX_FMT_RGB24: - c->yuv2packedX = ff_yuv2rgb24_X_altivec; - break; - } - } -#endif /* HAVE_ALTIVEC */ - - ff_sws_init_swscale_vsx(c); -} diff --git a/libswscale/ppc/yuv2rgb_altivec.c b/libswscale/ppc/yuv2rgb_altivec.c deleted file mode 100644 index a6b6650e02..0000000000 --- a/libswscale/ppc/yuv2rgb_altivec.c +++ /dev/null @@ -1,874 +0,0 @@ -/* - * AltiVec acceleration for colorspace conversion - * - * copyright (C) 2004 Marc Hoffman <[email protected]> - * - * 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 - */ - -/* - * Convert I420 YV12 to RGB in various formats, - * it rejects images that are not in 420 formats, - * it rejects images that don't have widths of multiples of 16, - * it rejects images that don't have heights of multiples of 2. - * Reject defers to C simulation code. - * - * Lots of optimizations to be done here. - * - * 1. Need to fix saturation code. I just couldn't get it to fly with packs - * and adds, so we currently use max/min to clip. - * - * 2. The inefficient use of chroma loading needs a bit of brushing up. - * - * 3. Analysis of pipeline stalls needs to be done. Use shark to identify - * pipeline stalls. - * - * - * MODIFIED to calculate coeffs from currently selected color space. - * MODIFIED core to be a macro where you specify the output format. - * ADDED UYVY conversion which is never called due to some thing in swscale. - * CORRECTED algorithm selection to be strict on input formats. - * ADDED runtime detection of AltiVec. - * - * ADDED altivec_yuv2packedX vertical scl + RGB converter - * - * March 27,2004 - * PERFORMANCE ANALYSIS - * - * The C version uses 25% of the processor or ~250Mips for D1 video rawvideo - * used as test. - * The AltiVec version uses 10% of the processor or ~100Mips for D1 video - * same sequence. - * - * 720 * 480 * 30 ~10MPS - * - * so we have roughly 10 clocks per pixel. This is too high, something has - * to be wrong. - * - * OPTIMIZED clip codes to utilize vec_max and vec_packs removing the - * need for vec_min. - * - * OPTIMIZED DST OUTPUT cache/DMA controls. We are pretty much guaranteed to - * have the input video frame, it was just decompressed so it probably resides - * in L1 caches. However, we are creating the output video stream. This needs - * to use the DSTST instruction to optimize for the cache. We couple this with - * the fact that we are not going to be visiting the input buffer again so we - * mark it Least Recently Used. This shaves 25% of the processor cycles off. - * - * Now memcpy is the largest mips consumer in the system, probably due - * to the inefficient X11 stuff. - * - * GL libraries seem to be very slow on this machine 1.33Ghz PB running - * Jaguar, this is not the case for my 1Ghz PB. I thought it might be - * a versioning issue, however I have libGL.1.2.dylib for both - * machines. (We need to figure this out now.) - * - * GL2 libraries work now with patch for RGB32. - * - * NOTE: quartz vo driver ARGB32_to_RGB24 consumes 30% of the processor. - * - * Integrated luma prescaling adjustment for saturation/contrast/brightness - * adjustment. - */ - -#include <stdio.h> -#include <stdlib.h> -#include <string.h> -#include <inttypes.h> - -#include "config.h" -#include "libswscale/rgb2rgb.h" -#include "libswscale/swscale.h" -#include "libswscale/swscale_internal.h" -#include "libavutil/attributes.h" -#include "libavutil/cpu.h" -#include "libavutil/mem_internal.h" -#include "libavutil/pixdesc.h" -#include "yuv2rgb_altivec.h" - -#if HAVE_ALTIVEC - -#undef PROFILE_THE_BEAST -#undef INC_SCALING - -typedef unsigned char ubyte; -typedef signed char sbyte; - -/* RGB interleaver, 16 planar pels 8-bit samples per channel in - * homogeneous vector registers x0,x1,x2 are interleaved with the - * following technique: - * - * o0 = vec_mergeh(x0, x1); - * o1 = vec_perm(o0, x2, perm_rgb_0); - * o2 = vec_perm(o0, x2, perm_rgb_1); - * o3 = vec_mergel(x0, x1); - * o4 = vec_perm(o3, o2, perm_rgb_2); - * o5 = vec_perm(o3, o2, perm_rgb_3); - * - * perm_rgb_0: o0(RG).h v1(B) --> o1* - * 0 1 2 3 4 - * rgbr|gbrg|brgb|rgbr - * 0010 0100 1001 0010 - * 0102 3145 2673 894A - * - * perm_rgb_1: o0(RG).h v1(B) --> o2 - * 0 1 2 3 4 - * gbrg|brgb|bbbb|bbbb - * 0100 1001 1111 1111 - * B5CD 6EF7 89AB CDEF - * - * perm_rgb_2: o3(RG).l o2(rgbB.l) --> o4* - * 0 1 2 3 4 - * gbrg|brgb|rgbr|gbrg - * 1111 1111 0010 0100 - * 89AB CDEF 0182 3945 - * - * perm_rgb_2: o3(RG).l o2(rgbB.l) ---> o5* - * 0 1 2 3 4 - * brgb|rgbr|gbrg|brgb - * 1001 0010 0100 1001 - * a67b 89cA BdCD eEFf - */ -static const vector unsigned char - perm_rgb_0 = { 0x00, 0x01, 0x10, 0x02, 0x03, 0x11, 0x04, 0x05, - 0x12, 0x06, 0x07, 0x13, 0x08, 0x09, 0x14, 0x0a }, - perm_rgb_1 = { 0x0b, 0x15, 0x0c, 0x0d, 0x16, 0x0e, 0x0f, 0x17, - 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f }, - perm_rgb_2 = { 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, - 0x00, 0x01, 0x18, 0x02, 0x03, 0x19, 0x04, 0x05 }, - perm_rgb_3 = { 0x1a, 0x06, 0x07, 0x1b, 0x08, 0x09, 0x1c, 0x0a, - 0x0b, 0x1d, 0x0c, 0x0d, 0x1e, 0x0e, 0x0f, 0x1f }; - -#define vec_merge3(x2, x1, x0, y0, y1, y2) \ - do { \ - __typeof__(x0) o0, o2, o3; \ - o0 = vec_mergeh(x0, x1); \ - y0 = vec_perm(o0, x2, perm_rgb_0); \ - o2 = vec_perm(o0, x2, perm_rgb_1); \ - o3 = vec_mergel(x0, x1); \ - y1 = vec_perm(o3, o2, perm_rgb_2); \ - y2 = vec_perm(o3, o2, perm_rgb_3); \ - } while (0) - -#define vec_mstbgr24(x0, x1, x2, ptr) \ - do { \ - __typeof__(x0) _0, _1, _2; \ - vec_merge3(x0, x1, x2, _0, _1, _2); \ - vec_st(_0, 0, ptr++); \ - vec_st(_1, 0, ptr++); \ - vec_st(_2, 0, ptr++); \ - } while (0) - -#define vec_mstrgb24(x0, x1, x2, ptr) \ - do { \ - __typeof__(x0) _0, _1, _2; \ - vec_merge3(x2, x1, x0, _0, _1, _2); \ - vec_st(_0, 0, ptr++); \ - vec_st(_1, 0, ptr++); \ - vec_st(_2, 0, ptr++); \ - } while (0) - -/* pack the pixels in rgb0 format - * msb R - * lsb 0 - */ -#define vec_mstrgb32(T, x0, x1, x2, x3, ptr) \ - do { \ - T _0, _1, _2, _3; \ - _0 = vec_mergeh(x0, x1); \ - _1 = vec_mergeh(x2, x3); \ - _2 = (T) vec_mergeh((vector unsigned short) _0, \ - (vector unsigned short) _1); \ - _3 = (T) vec_mergel((vector unsigned short) _0, \ - (vector unsigned short) _1); \ - vec_st(_2, 0 * 16, (T *) ptr); \ - vec_st(_3, 1 * 16, (T *) ptr); \ - _0 = vec_mergel(x0, x1); \ - _1 = vec_mergel(x2, x3); \ - _2 = (T) vec_mergeh((vector unsigned short) _0, \ - (vector unsigned short) _1); \ - _3 = (T) vec_mergel((vector unsigned short) _0, \ - (vector unsigned short) _1); \ - vec_st(_2, 2 * 16, (T *) ptr); \ - vec_st(_3, 3 * 16, (T *) ptr); \ - ptr += 4; \ - } while (0) - -/* - * 1 0 1.4021 | | Y | - * 1 -0.3441 -0.7142 |x| Cb| - * 1 1.7718 0 | | Cr| - * - * - * Y: [-128 127] - * Cb/Cr : [-128 127] - * - * typical YUV conversion works on Y: 0-255 this version has been - * optimized for JPEG decoding. - */ - -#if HAVE_BIGENDIAN -#define vec_unh(x) \ - (vector signed short) \ - vec_perm(x, (__typeof__(x)) { 0 }, \ - ((vector unsigned char) { \ - 0x10, 0x00, 0x10, 0x01, 0x10, 0x02, 0x10, 0x03, \ - 0x10, 0x04, 0x10, 0x05, 0x10, 0x06, 0x10, 0x07 })) - -#define vec_unl(x) \ - (vector signed short) \ - vec_perm(x, (__typeof__(x)) { 0 }, \ - ((vector unsigned char) { \ - 0x10, 0x08, 0x10, 0x09, 0x10, 0x0A, 0x10, 0x0B, \ - 0x10, 0x0C, 0x10, 0x0D, 0x10, 0x0E, 0x10, 0x0F })) -#else -#define vec_unh(x)(vector signed short) vec_mergeh(x,(__typeof__(x)) { 0 }) -#define vec_unl(x)(vector signed short) vec_mergel(x,(__typeof__(x)) { 0 }) -#endif - -#define vec_clip_s16(x) \ - vec_max(vec_min(x, ((vector signed short) { \ - 235, 235, 235, 235, 235, 235, 235, 235 })), \ - ((vector signed short) { 16, 16, 16, 16, 16, 16, 16, 16 })) - -#define vec_packclp(x, y) \ - (vector unsigned char) \ - vec_packs((vector unsigned short) \ - vec_max(x, ((vector signed short) { 0 })), \ - (vector unsigned short) \ - vec_max(y, ((vector signed short) { 0 }))) - -static inline void cvtyuvtoRGB(SwsInternal *c, vector signed short Y, - vector signed short U, vector signed short V, - vector signed short *R, vector signed short *G, - vector signed short *B) -{ - vector signed short vx, ux, uvx; - - Y = vec_mradds(Y, c->CY, c->OY); - U = vec_sub(U, (vector signed short) - vec_splat((vector signed short) { 128 }, 0)); - V = vec_sub(V, (vector signed short) - vec_splat((vector signed short) { 128 }, 0)); - - // ux = (CBU * (u << c->CSHIFT) + 0x4000) >> 15; - ux = vec_sl(U, c->CSHIFT); - *B = vec_mradds(ux, c->CBU, Y); - - // vx = (CRV * (v << c->CSHIFT) + 0x4000) >> 15; - vx = vec_sl(V, c->CSHIFT); - *R = vec_mradds(vx, c->CRV, Y); - - // uvx = ((CGU * u) + (CGV * v)) >> 15; - uvx = vec_mradds(U, c->CGU, Y); - *G = vec_mradds(V, c->CGV, uvx); -} - -/* - * ------------------------------------------------------------------------------ - * CS converters - * ------------------------------------------------------------------------------ - */ - -#if !HAVE_VEC_XL -static inline vector unsigned char vec_xl(signed long long offset, const ubyte *addr) -{ - const vector unsigned char *v_addr = (const vector unsigned char *) (addr + offset); - vector unsigned char align_perm = vec_lvsl(offset, addr); - - return (vector unsigned char) vec_perm(v_addr[0], v_addr[1], align_perm); -} -#endif /* !HAVE_VEC_XL */ - -#define DEFCSP420_CVT(name, out_pixels) \ -static int altivec_ ## name(SwsInternal *c, const unsigned char *const *in, \ - const int *instrides, int srcSliceY, int srcSliceH, \ - unsigned char *const *oplanes, const int *outstrides) \ -{ \ - int w = c->opts.src_w; \ - int h = srcSliceH; \ - int i, j; \ - int instrides_scl[3]; \ - vector unsigned char y0, y1; \ - \ - vector signed char u, v; \ - \ - vector signed short Y0, Y1, Y2, Y3; \ - vector signed short U, V; \ - vector signed short vx, ux, uvx; \ - vector signed short vx0, ux0, uvx0; \ - vector signed short vx1, ux1, uvx1; \ - vector signed short R0, G0, B0; \ - vector signed short R1, G1, B1; \ - vector unsigned char R, G, B; \ - \ - vector signed short lCY = c->CY; \ - vector signed short lOY = c->OY; \ - vector signed short lCRV = c->CRV; \ - vector signed short lCBU = c->CBU; \ - vector signed short lCGU = c->CGU; \ - vector signed short lCGV = c->CGV; \ - vector unsigned short lCSHIFT = c->CSHIFT; \ - \ - const ubyte *y1i = in[0]; \ - const ubyte *y2i = in[0] + instrides[0]; \ - const ubyte *ui = in[1]; \ - const ubyte *vi = in[2]; \ - \ - vector unsigned char *oute, *outo; \ - \ - /* loop moves y{1, 2}i by w */ \ - instrides_scl[0] = instrides[0] * 2 - w; \ - /* loop moves ui by w / 2 */ \ - instrides_scl[1] = instrides[1] - w / 2; \ - /* loop moves vi by w / 2 */ \ - instrides_scl[2] = instrides[2] - w / 2; \ - \ - for (i = 0; i < h / 2; i++) { \ - oute = (vector unsigned char *)(oplanes[0] + outstrides[0] * \ - (srcSliceY + i * 2)); \ - outo = oute + (outstrides[0] >> 4); \ - vec_dstst(outo, (0x02000002 | (((w * 3 + 32) / 32) << 16)), 0); \ - vec_dstst(oute, (0x02000002 | (((w * 3 + 32) / 32) << 16)), 1); \ - \ - for (j = 0; j < w / 16; j++) { \ - y0 = vec_xl(0, y1i); \ - \ - y1 = vec_xl(0, y2i); \ - \ - u = (vector signed char) vec_xl(0, ui); \ - \ - v = (vector signed char) vec_xl(0, vi); \ - \ - u = (vector signed char) \ - vec_sub(u, \ - (vector signed char) \ - vec_splat((vector signed char) { 128 }, 0)); \ - v = (vector signed char) \ - vec_sub(v, \ - (vector signed char) \ - vec_splat((vector signed char) { 128 }, 0)); \ - \ - U = vec_unpackh(u); \ - V = vec_unpackh(v); \ - \ - Y0 = vec_unh(y0); \ - Y1 = vec_unl(y0); \ - Y2 = vec_unh(y1); \ - Y3 = vec_unl(y1); \ - \ - Y0 = vec_mradds(Y0, lCY, lOY); \ - Y1 = vec_mradds(Y1, lCY, lOY); \ - Y2 = vec_mradds(Y2, lCY, lOY); \ - Y3 = vec_mradds(Y3, lCY, lOY); \ - \ - /* ux = (CBU * (u << CSHIFT) + 0x4000) >> 15 */ \ - ux = vec_sl(U, lCSHIFT); \ - ux = vec_mradds(ux, lCBU, (vector signed short) { 0 }); \ - ux0 = vec_mergeh(ux, ux); \ - ux1 = vec_mergel(ux, ux); \ - \ - /* vx = (CRV * (v << CSHIFT) + 0x4000) >> 15; */ \ - vx = vec_sl(V, lCSHIFT); \ - vx = vec_mradds(vx, lCRV, (vector signed short) { 0 }); \ - vx0 = vec_mergeh(vx, vx); \ - vx1 = vec_mergel(vx, vx); \ - \ - /* uvx = ((CGU * u) + (CGV * v)) >> 15 */ \ - uvx = vec_mradds(U, lCGU, (vector signed short) { 0 }); \ - uvx = vec_mradds(V, lCGV, uvx); \ - uvx0 = vec_mergeh(uvx, uvx); \ - uvx1 = vec_mergel(uvx, uvx); \ - \ - R0 = vec_add(Y0, vx0); \ - G0 = vec_add(Y0, uvx0); \ - B0 = vec_add(Y0, ux0); \ - R1 = vec_add(Y1, vx1); \ - G1 = vec_add(Y1, uvx1); \ - B1 = vec_add(Y1, ux1); \ - \ - R = vec_packclp(R0, R1); \ - G = vec_packclp(G0, G1); \ - B = vec_packclp(B0, B1); \ - \ - out_pixels(R, G, B, oute); \ - \ - R0 = vec_add(Y2, vx0); \ - G0 = vec_add(Y2, uvx0); \ - B0 = vec_add(Y2, ux0); \ - R1 = vec_add(Y3, vx1); \ - G1 = vec_add(Y3, uvx1); \ - B1 = vec_add(Y3, ux1); \ - R = vec_packclp(R0, R1); \ - G = vec_packclp(G0, G1); \ - B = vec_packclp(B0, B1); \ - \ - \ - out_pixels(R, G, B, outo); \ - \ - y1i += 16; \ - y2i += 16; \ - ui += 8; \ - vi += 8; \ - } \ - \ - ui += instrides_scl[1]; \ - vi += instrides_scl[2]; \ - y1i += instrides_scl[0]; \ - y2i += instrides_scl[0]; \ - } \ - return srcSliceH; \ -} - -#define out_abgr(a, b, c, ptr) \ - vec_mstrgb32(__typeof__(a), ((__typeof__(a)) vec_splat((__typeof__(a)){ 255 }, 0)), c, b, a, ptr) -#define out_bgra(a, b, c, ptr) \ - vec_mstrgb32(__typeof__(a), c, b, a, ((__typeof__(a)) vec_splat((__typeof__(a)){ 255 }, 0)), ptr) -#define out_rgba(a, b, c, ptr) \ - vec_mstrgb32(__typeof__(a), a, b, c, ((__typeof__(a)) vec_splat((__typeof__(a)){ 255 }, 0)), ptr) -#define out_argb(a, b, c, ptr) \ - vec_mstrgb32(__typeof__(a), ((__typeof__(a)) vec_splat((__typeof__(a)){ 255 }, 0)), a, b, c, ptr) -#define out_rgb24(a, b, c, ptr) vec_mstrgb24(a, b, c, ptr) -#define out_bgr24(a, b, c, ptr) vec_mstbgr24(a, b, c, ptr) - -DEFCSP420_CVT(yuv2_abgr, out_abgr) -DEFCSP420_CVT(yuv2_bgra, out_bgra) -DEFCSP420_CVT(yuv2_rgba, out_rgba) -DEFCSP420_CVT(yuv2_argb, out_argb) -DEFCSP420_CVT(yuv2_rgb24, out_rgb24) -DEFCSP420_CVT(yuv2_bgr24, out_bgr24) - -// uyvy|uyvy|uyvy|uyvy -// 0123 4567 89ab cdef -static const vector unsigned char - demux_u = { 0x10, 0x00, 0x10, 0x00, - 0x10, 0x04, 0x10, 0x04, - 0x10, 0x08, 0x10, 0x08, - 0x10, 0x0c, 0x10, 0x0c }, - demux_v = { 0x10, 0x02, 0x10, 0x02, - 0x10, 0x06, 0x10, 0x06, - 0x10, 0x0A, 0x10, 0x0A, - 0x10, 0x0E, 0x10, 0x0E }, - demux_y = { 0x10, 0x01, 0x10, 0x03, - 0x10, 0x05, 0x10, 0x07, - 0x10, 0x09, 0x10, 0x0B, - 0x10, 0x0D, 0x10, 0x0F }; - -/* - * this is so I can play live CCIR raw video - */ -static int altivec_uyvy_rgb32(SwsInternal *c, const unsigned char *const *in, - const int *instrides, int srcSliceY, int srcSliceH, - unsigned char *const *oplanes, const int *outstrides) -{ - int w = c->opts.src_w; - int h = srcSliceH; - int i, j; - vector unsigned char uyvy; - vector signed short Y, U, V; - vector signed short R0, G0, B0, R1, G1, B1; - vector unsigned char R, G, B; - vector unsigned char *out; - const ubyte *img; - - img = in[0]; - out = (vector unsigned char *) (oplanes[0] + srcSliceY * outstrides[0]); - - for (i = 0; i < h; i++) - for (j = 0; j < w / 16; j++) { - uyvy = vec_ld(0, img); - - U = (vector signed short) - vec_perm(uyvy, (vector unsigned char) { 0 }, demux_u); - V = (vector signed short) - vec_perm(uyvy, (vector unsigned char) { 0 }, demux_v); - Y = (vector signed short) - vec_perm(uyvy, (vector unsigned char) { 0 }, demux_y); - - cvtyuvtoRGB(c, Y, U, V, &R0, &G0, &B0); - - uyvy = vec_ld(16, img); - - U = (vector signed short) - vec_perm(uyvy, (vector unsigned char) { 0 }, demux_u); - V = (vector signed short) - vec_perm(uyvy, (vector unsigned char) { 0 }, demux_v); - Y = (vector signed short) - vec_perm(uyvy, (vector unsigned char) { 0 }, demux_y); - - cvtyuvtoRGB(c, Y, U, V, &R1, &G1, &B1); - - R = vec_packclp(R0, R1); - G = vec_packclp(G0, G1); - B = vec_packclp(B0, B1); - - // vec_mstbgr24 (R,G,B, out); - out_rgba(R, G, B, out); - - img += 32; - } - return srcSliceH; -} - -#endif /* HAVE_ALTIVEC */ - -/* Ok currently the acceleration routine only supports - * inputs of widths a multiple of 16 - * and heights a multiple 2 - * - * So we just fall back to the C codes for this. - */ -av_cold SwsFunc ff_yuv2rgb_init_ppc(SwsInternal *c) -{ -#if HAVE_ALTIVEC - if (!(av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC)) - return NULL; - - /* - * and this seems not to matter too much I tried a bunch of - * videos with abnormal widths and MPlayer crashes elsewhere. - * mplayer -vo x11 -rawvideo on:w=350:h=240 raw-350x240.eyuv - * boom with X11 bad match. - * - */ - if ((c->opts.src_w & 0xf) != 0) - return NULL; - - switch (c->opts.src_format) { - case AV_PIX_FMT_YUV410P: - case AV_PIX_FMT_YUV420P: - /*case IMGFMT_CLPL: ??? */ - case AV_PIX_FMT_GRAY8: - case AV_PIX_FMT_NV12: - case AV_PIX_FMT_NV21: - if ((c->opts.src_h & 0x1) != 0) - return NULL; - -/* - * The below accelerations for YUV2RGB are known broken. - * See: 'fate-checkasm-sw_yuv2rgb' with --enable-altivec - * They are disabled for the moment, until such time as - * they can be repaired. - */ -#if 0 - switch (c->opts.dst_format) { - case AV_PIX_FMT_RGB24: - av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space RGB24\n"); - return altivec_yuv2_rgb24; - case AV_PIX_FMT_BGR24: - av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space BGR24\n"); - return altivec_yuv2_bgr24; - case AV_PIX_FMT_ARGB: - av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space ARGB\n"); - return altivec_yuv2_argb; - case AV_PIX_FMT_ABGR: - av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space ABGR\n"); - return altivec_yuv2_abgr; - case AV_PIX_FMT_RGBA: - av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space RGBA\n"); - return altivec_yuv2_rgba; - case AV_PIX_FMT_BGRA: - av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space BGRA\n"); - return altivec_yuv2_bgra; - default: return NULL; - } -#endif /* disabled YUV2RGB acceleration */ - break; - - case AV_PIX_FMT_UYVY422: - switch (c->opts.dst_format) { - case AV_PIX_FMT_BGR32: - av_log(c, AV_LOG_WARNING, "ALTIVEC: Color Space UYVY -> RGB32\n"); - return altivec_uyvy_rgb32; - default: return NULL; - } - break; - } -#endif /* HAVE_ALTIVEC */ - - return NULL; -} - -av_cold void ff_yuv2rgb_init_tables_ppc(SwsInternal *c, - const int inv_table[4], - int brightness, - int contrast, - int saturation) -{ -#if HAVE_ALTIVEC - union { - DECLARE_ALIGNED(16, signed short, tmp)[8]; - vector signed short vec; - } buf; - - if (!(av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC)) - return; - - buf.tmp[0] = ((0xffffLL) * contrast >> 8) >> 9; // cy - buf.tmp[1] = -256 * brightness; // oy - buf.tmp[2] = (inv_table[0] >> 3) * (contrast >> 16) * (saturation >> 16); // crv - buf.tmp[3] = (inv_table[1] >> 3) * (contrast >> 16) * (saturation >> 16); // cbu - buf.tmp[4] = -((inv_table[2] >> 1) * (contrast >> 16) * (saturation >> 16)); // cgu - buf.tmp[5] = -((inv_table[3] >> 1) * (contrast >> 16) * (saturation >> 16)); // cgv - - c->CSHIFT = (vector unsigned short) vec_splat_u16(2); - c->CY = vec_splat((vector signed short) buf.vec, 0); - c->OY = vec_splat((vector signed short) buf.vec, 1); - c->CRV = vec_splat((vector signed short) buf.vec, 2); - c->CBU = vec_splat((vector signed short) buf.vec, 3); - c->CGU = vec_splat((vector signed short) buf.vec, 4); - c->CGV = vec_splat((vector signed short) buf.vec, 5); - return; -#endif /* HAVE_ALTIVEC */ -} - -#if HAVE_ALTIVEC - -static av_always_inline void yuv2packedX_altivec(SwsInternal *c, - const int16_t *lumFilter, - const int16_t **lumSrc, - int lumFilterSize, - const int16_t *chrFilter, - const int16_t **chrUSrc, - const int16_t **chrVSrc, - int chrFilterSize, - const int16_t **alpSrc, - uint8_t *dest, - int dstW, int dstY, - enum AVPixelFormat target) -{ - int i, j; - vector signed short X, X0, X1, Y0, U0, V0, Y1, U1, V1, U, V; - vector signed short R0, G0, B0, R1, G1, B1; - - vector unsigned char R, G, B; - vector unsigned char *out, *nout; - - vector signed short RND = vec_splat_s16(1 << 3); - vector unsigned short SCL = vec_splat_u16(4); - DECLARE_ALIGNED(16, unsigned int, scratch)[16]; - - vector signed short *YCoeffs, *CCoeffs; - - YCoeffs = c->vYCoeffsBank + dstY * lumFilterSize; - CCoeffs = c->vCCoeffsBank + dstY * chrFilterSize; - - out = (vector unsigned char *) dest; - - for (i = 0; i < dstW; i += 16) { - Y0 = RND; - Y1 = RND; - /* extract 16 coeffs from lumSrc */ - for (j = 0; j < lumFilterSize; j++) { - X0 = vec_ld(0, &lumSrc[j][i]); - X1 = vec_ld(16, &lumSrc[j][i]); - Y0 = vec_mradds(X0, YCoeffs[j], Y0); - Y1 = vec_mradds(X1, YCoeffs[j], Y1); - } - - U = RND; - V = RND; - /* extract 8 coeffs from U,V */ - for (j = 0; j < chrFilterSize; j++) { - X = vec_ld(0, &chrUSrc[j][i / 2]); - U = vec_mradds(X, CCoeffs[j], U); - X = vec_ld(0, &chrVSrc[j][i / 2]); - V = vec_mradds(X, CCoeffs[j], V); - } - - /* scale and clip signals */ - Y0 = vec_sra(Y0, SCL); - Y1 = vec_sra(Y1, SCL); - U = vec_sra(U, SCL); - V = vec_sra(V, SCL); - - Y0 = vec_clip_s16(Y0); - Y1 = vec_clip_s16(Y1); - U = vec_clip_s16(U); - V = vec_clip_s16(V); - - /* now we have - * Y0 = y0 y1 y2 y3 y4 y5 y6 y7 Y1 = y8 y9 y10 y11 y12 y13 y14 y15 - * U = u0 u1 u2 u3 u4 u5 u6 u7 V = v0 v1 v2 v3 v4 v5 v6 v7 - * - * Y0 = y0 y1 y2 y3 y4 y5 y6 y7 Y1 = y8 y9 y10 y11 y12 y13 y14 y15 - * U0 = u0 u0 u1 u1 u2 u2 u3 u3 U1 = u4 u4 u5 u5 u6 u6 u7 u7 - * V0 = v0 v0 v1 v1 v2 v2 v3 v3 V1 = v4 v4 v5 v5 v6 v6 v7 v7 - */ - - U0 = vec_mergeh(U, U); - V0 = vec_mergeh(V, V); - - U1 = vec_mergel(U, U); - V1 = vec_mergel(V, V); - - cvtyuvtoRGB(c, Y0, U0, V0, &R0, &G0, &B0); - cvtyuvtoRGB(c, Y1, U1, V1, &R1, &G1, &B1); - - R = vec_packclp(R0, R1); - G = vec_packclp(G0, G1); - B = vec_packclp(B0, B1); - - switch (target) { - case AV_PIX_FMT_ABGR: - out_abgr(R, G, B, out); - break; - case AV_PIX_FMT_BGRA: - out_bgra(R, G, B, out); - break; - case AV_PIX_FMT_RGBA: - out_rgba(R, G, B, out); - break; - case AV_PIX_FMT_ARGB: - out_argb(R, G, B, out); - break; - case AV_PIX_FMT_RGB24: - out_rgb24(R, G, B, out); - break; - case AV_PIX_FMT_BGR24: - out_bgr24(R, G, B, out); - break; - default: - { - /* If this is reached, the caller should have called yuv2packedXinC - * instead. */ - static int printed_error_message; - if (!printed_error_message) { - av_log(c, AV_LOG_ERROR, - "altivec_yuv2packedX doesn't support %s output\n", - av_get_pix_fmt_name(c->opts.dst_format)); - printed_error_message = 1; - } - return; - } - } - } - - if (i < dstW) { - i -= 16; - - Y0 = RND; - Y1 = RND; - /* extract 16 coeffs from lumSrc */ - for (j = 0; j < lumFilterSize; j++) { - X0 = vec_ld(0, &lumSrc[j][i]); - X1 = vec_ld(16, &lumSrc[j][i]); - Y0 = vec_mradds(X0, YCoeffs[j], Y0); - Y1 = vec_mradds(X1, YCoeffs[j], Y1); - } - - U = RND; - V = RND; - /* extract 8 coeffs from U,V */ - for (j = 0; j < chrFilterSize; j++) { - X = vec_ld(0, &chrUSrc[j][i / 2]); - U = vec_mradds(X, CCoeffs[j], U); - X = vec_ld(0, &chrVSrc[j][i / 2]); - V = vec_mradds(X, CCoeffs[j], V); - } - - /* scale and clip signals */ - Y0 = vec_sra(Y0, SCL); - Y1 = vec_sra(Y1, SCL); - U = vec_sra(U, SCL); - V = vec_sra(V, SCL); - - Y0 = vec_clip_s16(Y0); - Y1 = vec_clip_s16(Y1); - U = vec_clip_s16(U); - V = vec_clip_s16(V); - - /* now we have - * Y0 = y0 y1 y2 y3 y4 y5 y6 y7 Y1 = y8 y9 y10 y11 y12 y13 y14 y15 - * U = u0 u1 u2 u3 u4 u5 u6 u7 V = v0 v1 v2 v3 v4 v5 v6 v7 - * - * Y0 = y0 y1 y2 y3 y4 y5 y6 y7 Y1 = y8 y9 y10 y11 y12 y13 y14 y15 - * U0 = u0 u0 u1 u1 u2 u2 u3 u3 U1 = u4 u4 u5 u5 u6 u6 u7 u7 - * V0 = v0 v0 v1 v1 v2 v2 v3 v3 V1 = v4 v4 v5 v5 v6 v6 v7 v7 - */ - - U0 = vec_mergeh(U, U); - V0 = vec_mergeh(V, V); - - U1 = vec_mergel(U, U); - V1 = vec_mergel(V, V); - - cvtyuvtoRGB(c, Y0, U0, V0, &R0, &G0, &B0); - cvtyuvtoRGB(c, Y1, U1, V1, &R1, &G1, &B1); - - R = vec_packclp(R0, R1); - G = vec_packclp(G0, G1); - B = vec_packclp(B0, B1); - - nout = (vector unsigned char *) scratch; - switch (target) { - case AV_PIX_FMT_ABGR: - out_abgr(R, G, B, nout); - break; - case AV_PIX_FMT_BGRA: - out_bgra(R, G, B, nout); - break; - case AV_PIX_FMT_RGBA: - out_rgba(R, G, B, nout); - break; - case AV_PIX_FMT_ARGB: - out_argb(R, G, B, nout); - break; - case AV_PIX_FMT_RGB24: - out_rgb24(R, G, B, nout); - break; - case AV_PIX_FMT_BGR24: - out_bgr24(R, G, B, nout); - break; - default: - /* Unreachable, I think. */ - av_log(c, AV_LOG_ERROR, - "altivec_yuv2packedX doesn't support %s output\n", - av_get_pix_fmt_name(c->opts.dst_format)); - return; - } - - memcpy(&((uint32_t *) dest)[i], scratch, (dstW - i) / 4); - } -} - -#define YUV2PACKEDX_WRAPPER(suffix, pixfmt) \ -void ff_yuv2 ## suffix ## _X_altivec(SwsInternal *c, \ - const int16_t *lumFilter, \ - const int16_t **lumSrc, \ - int lumFilterSize, \ - const int16_t *chrFilter, \ - const int16_t **chrUSrc, \ - const int16_t **chrVSrc, \ - int chrFilterSize, \ - const int16_t **alpSrc, \ - uint8_t *dest, int dstW, int dstY) \ -{ \ - yuv2packedX_altivec(c, lumFilter, lumSrc, lumFilterSize, \ - chrFilter, chrUSrc, chrVSrc, \ - chrFilterSize, alpSrc, \ - dest, dstW, dstY, pixfmt); \ -} - -YUV2PACKEDX_WRAPPER(abgr, AV_PIX_FMT_ABGR); -YUV2PACKEDX_WRAPPER(bgra, AV_PIX_FMT_BGRA); -YUV2PACKEDX_WRAPPER(argb, AV_PIX_FMT_ARGB); -YUV2PACKEDX_WRAPPER(rgba, AV_PIX_FMT_RGBA); -YUV2PACKEDX_WRAPPER(rgb24, AV_PIX_FMT_RGB24); -YUV2PACKEDX_WRAPPER(bgr24, AV_PIX_FMT_BGR24); - -#endif /* HAVE_ALTIVEC */ diff --git a/libswscale/ppc/yuv2yuv_altivec.c b/libswscale/ppc/yuv2yuv_altivec.c deleted file mode 100644 index 7884c38d69..0000000000 --- a/libswscale/ppc/yuv2yuv_altivec.c +++ /dev/null @@ -1,204 +0,0 @@ -/* - * AltiVec-enhanced yuv-to-yuv conversion routines. - * - * Copyright (C) 2004 Romain Dolbeau <[email protected]> - * based on the equivalent C code in swscale.c - * - * 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 <inttypes.h> - -#include "config.h" -#include "libavutil/attributes.h" -#include "libavutil/cpu.h" -#include "libswscale/swscale.h" -#include "libswscale/swscale_internal.h" - -#if HAVE_ALTIVEC - -static int yv12toyuy2_unscaled_altivec(SwsInternal *c, const uint8_t *const src[], - const int srcStride[], int srcSliceY, - int srcSliceH, uint8_t *const dstParam[], - const int dstStride_a[]) -{ - uint8_t *dst = dstParam[0] + dstStride_a[0] * srcSliceY; - // yv12toyuy2(src[0], src[1], src[2], dst, c->opts.src_w, srcSliceH, - // srcStride[0], srcStride[1], dstStride[0]); - const uint8_t *ysrc = src[0]; - const uint8_t *usrc = src[1]; - const uint8_t *vsrc = src[2]; - const int width = c->opts.src_w; - const int height = srcSliceH; - const int lumStride = srcStride[0]; - const int chromStride = srcStride[1]; - const int dstStride = dstStride_a[0]; - const vector unsigned char yperm = vec_lvsl(0, ysrc); - const int vertLumPerChroma = 2; - register unsigned int y; - - /* This code assumes: - * - * 1) dst is 16 bytes-aligned - * 2) dstStride is a multiple of 16 - * 3) width is a multiple of 16 - * 4) lum & chrom stride are multiples of 8 - */ - - for (y = 0; y < height; y++) { - int i; - for (i = 0; i < width - 31; i += 32) { - const unsigned int j = i >> 1; - vector unsigned char v_yA = vec_ld(i, ysrc); - vector unsigned char v_yB = vec_ld(i + 16, ysrc); - vector unsigned char v_yC = vec_ld(i + 32, ysrc); - vector unsigned char v_y1 = vec_perm(v_yA, v_yB, yperm); - vector unsigned char v_y2 = vec_perm(v_yB, v_yC, yperm); - vector unsigned char v_uA = vec_ld(j, usrc); - vector unsigned char v_uB = vec_ld(j + 16, usrc); - vector unsigned char v_u = vec_perm(v_uA, v_uB, vec_lvsl(j, usrc)); - vector unsigned char v_vA = vec_ld(j, vsrc); - vector unsigned char v_vB = vec_ld(j + 16, vsrc); - vector unsigned char v_v = vec_perm(v_vA, v_vB, vec_lvsl(j, vsrc)); - vector unsigned char v_uv_a = vec_mergeh(v_u, v_v); - vector unsigned char v_uv_b = vec_mergel(v_u, v_v); - vector unsigned char v_yuy2_0 = vec_mergeh(v_y1, v_uv_a); - vector unsigned char v_yuy2_1 = vec_mergel(v_y1, v_uv_a); - vector unsigned char v_yuy2_2 = vec_mergeh(v_y2, v_uv_b); - vector unsigned char v_yuy2_3 = vec_mergel(v_y2, v_uv_b); - vec_st(v_yuy2_0, (i << 1), dst); - vec_st(v_yuy2_1, (i << 1) + 16, dst); - vec_st(v_yuy2_2, (i << 1) + 32, dst); - vec_st(v_yuy2_3, (i << 1) + 48, dst); - } - if (i < width) { - const unsigned int j = i >> 1; - vector unsigned char v_y1 = vec_ld(i, ysrc); - vector unsigned char v_u = vec_ld(j, usrc); - vector unsigned char v_v = vec_ld(j, vsrc); - vector unsigned char v_uv_a = vec_mergeh(v_u, v_v); - vector unsigned char v_yuy2_0 = vec_mergeh(v_y1, v_uv_a); - vector unsigned char v_yuy2_1 = vec_mergel(v_y1, v_uv_a); - vec_st(v_yuy2_0, (i << 1), dst); - vec_st(v_yuy2_1, (i << 1) + 16, dst); - } - if ((y & (vertLumPerChroma - 1)) == vertLumPerChroma - 1) { - usrc += chromStride; - vsrc += chromStride; - } - ysrc += lumStride; - dst += dstStride; - } - - return srcSliceH; -} - -static int yv12touyvy_unscaled_altivec(SwsInternal *c, const uint8_t *const src[], - const int srcStride[], int srcSliceY, - int srcSliceH, uint8_t *const dstParam[], - const int dstStride_a[]) -{ - uint8_t *dst = dstParam[0] + dstStride_a[0] * srcSliceY; - // yv12toyuy2(src[0], src[1], src[2], dst, c->opts.src_w, srcSliceH, - // srcStride[0], srcStride[1], dstStride[0]); - const uint8_t *ysrc = src[0]; - const uint8_t *usrc = src[1]; - const uint8_t *vsrc = src[2]; - const int width = c->opts.src_w; - const int height = srcSliceH; - const int lumStride = srcStride[0]; - const int chromStride = srcStride[1]; - const int dstStride = dstStride_a[0]; - const int vertLumPerChroma = 2; - const vector unsigned char yperm = vec_lvsl(0, ysrc); - register unsigned int y; - - /* This code assumes: - * - * 1) dst is 16 bytes-aligned - * 2) dstStride is a multiple of 16 - * 3) width is a multiple of 16 - * 4) lum & chrom stride are multiples of 8 - */ - - for (y = 0; y < height; y++) { - int i; - for (i = 0; i < width - 31; i += 32) { - const unsigned int j = i >> 1; - vector unsigned char v_yA = vec_ld(i, ysrc); - vector unsigned char v_yB = vec_ld(i + 16, ysrc); - vector unsigned char v_yC = vec_ld(i + 32, ysrc); - vector unsigned char v_y1 = vec_perm(v_yA, v_yB, yperm); - vector unsigned char v_y2 = vec_perm(v_yB, v_yC, yperm); - vector unsigned char v_uA = vec_ld(j, usrc); - vector unsigned char v_uB = vec_ld(j + 16, usrc); - vector unsigned char v_u = vec_perm(v_uA, v_uB, vec_lvsl(j, usrc)); - vector unsigned char v_vA = vec_ld(j, vsrc); - vector unsigned char v_vB = vec_ld(j + 16, vsrc); - vector unsigned char v_v = vec_perm(v_vA, v_vB, vec_lvsl(j, vsrc)); - vector unsigned char v_uv_a = vec_mergeh(v_u, v_v); - vector unsigned char v_uv_b = vec_mergel(v_u, v_v); - vector unsigned char v_uyvy_0 = vec_mergeh(v_uv_a, v_y1); - vector unsigned char v_uyvy_1 = vec_mergel(v_uv_a, v_y1); - vector unsigned char v_uyvy_2 = vec_mergeh(v_uv_b, v_y2); - vector unsigned char v_uyvy_3 = vec_mergel(v_uv_b, v_y2); - vec_st(v_uyvy_0, (i << 1), dst); - vec_st(v_uyvy_1, (i << 1) + 16, dst); - vec_st(v_uyvy_2, (i << 1) + 32, dst); - vec_st(v_uyvy_3, (i << 1) + 48, dst); - } - if (i < width) { - const unsigned int j = i >> 1; - vector unsigned char v_y1 = vec_ld(i, ysrc); - vector unsigned char v_u = vec_ld(j, usrc); - vector unsigned char v_v = vec_ld(j, vsrc); - vector unsigned char v_uv_a = vec_mergeh(v_u, v_v); - vector unsigned char v_uyvy_0 = vec_mergeh(v_uv_a, v_y1); - vector unsigned char v_uyvy_1 = vec_mergel(v_uv_a, v_y1); - vec_st(v_uyvy_0, (i << 1), dst); - vec_st(v_uyvy_1, (i << 1) + 16, dst); - } - if ((y & (vertLumPerChroma - 1)) == vertLumPerChroma - 1) { - usrc += chromStride; - vsrc += chromStride; - } - ysrc += lumStride; - dst += dstStride; - } - return srcSliceH; -} - -#endif /* HAVE_ALTIVEC */ - -av_cold void ff_get_unscaled_swscale_ppc(SwsInternal *c) -{ -#if HAVE_ALTIVEC - if (!(av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC)) - return; - - if (!(c->opts.src_w & 15) && !(c->opts.flags & SWS_BITEXACT) && - c->opts.src_format == AV_PIX_FMT_YUV420P) { - enum AVPixelFormat dstFormat = c->opts.dst_format; - - // unscaled YV12 -> packed YUV, we want speed - if (dstFormat == AV_PIX_FMT_YUYV422) - c->convert_unscaled = yv12toyuy2_unscaled_altivec; - else if (dstFormat == AV_PIX_FMT_UYVY422) - c->convert_unscaled = yv12touyvy_unscaled_altivec; - } -#endif /* HAVE_ALTIVEC */ -} diff --git a/libswscale/swscale.c b/libswscale/swscale.c index 96df4ed3f4..69c3795cce 100644 --- a/libswscale/swscale.c +++ b/libswscale/swscale.c @@ -695,7 +695,7 @@ void ff_sws_init_scale(SwsInternal *c) { sws_init_swscale(c); -#if ARCH_PPC +#if ARCH_PPC && !HAVE_BIGENDIAN ff_sws_init_swscale_ppc(c); #elif ARCH_X86 ff_sws_init_swscale_x86(c); diff --git a/libswscale/swscale_internal.h b/libswscale/swscale_internal.h index 5c58272664..3d85233cb4 100644 --- a/libswscale/swscale_internal.h +++ b/libswscale/swscale_internal.h @@ -69,6 +69,8 @@ #define RETCODE_USE_CASCADE -12345 +#define SWS_ACCEL_BE_ALTIVEC (AV_HAVE_BIGENDIAN && LIBSWSCALE_VERSION_MAJOR < 10) + typedef struct SwsInternal SwsInternal; static inline SwsInternal *sws_internal(const SwsContext *sws) diff --git a/libswscale/utils.c b/libswscale/utils.c index 8f8789b24d..8c2d4d7d29 100644 --- a/libswscale/utils.c +++ b/libswscale/utils.c @@ -548,7 +548,6 @@ static av_cold int initFilter(int16_t **outFilter, int32_t **filterPos, } // Note the +1 is for the MMX scaler which reads over the end - /* align at 16 for AltiVec (needed by hScale_altivec_real) */ if (!FF_ALLOCZ_TYPED_ARRAY(*outFilter, *outFilterSize * (dstW + 3))) goto nomem; @@ -980,7 +979,7 @@ int sws_setColorspaceDetails(SwsContext *sws, const int inv_table[4], contrast, saturation); // FIXME factorize -#if ARCH_PPC +#if ARCH_PPC && !HAVE_BIGENDIAN ff_yuv2rgb_init_tables_ppc(c, inv_table, brightness, contrast, saturation); #endif diff --git a/libswscale/version.h b/libswscale/version.h index 51eb013a29..9de77f8777 100644 --- a/libswscale/version.h +++ b/libswscale/version.h @@ -29,7 +29,7 @@ #include "version_major.h" #define LIBSWSCALE_VERSION_MINOR 3 -#define LIBSWSCALE_VERSION_MICRO 100 +#define LIBSWSCALE_VERSION_MICRO 101 #define LIBSWSCALE_VERSION_INT AV_VERSION_INT(LIBSWSCALE_VERSION_MAJOR, \ LIBSWSCALE_VERSION_MINOR, \ -- 2.49.1 _______________________________________________ ffmpeg-devel mailing list -- [email protected] To unsubscribe send an email to [email protected]
