liyepan/openloongarch-buildroot-2021.02
1
0
Fork 0
mirror of https://gitee.com/open-loongarch/buildroot.git synced 2026-10-28 22:17:14 +08:00
Code Issues Packages Projects Releases Wiki Activity
Files
master
openloongarch-buildroot-202…/package/ffmpeg/0017-swscale-mips-Add-following-msa-optimizations.patch
T
michael 75e64c9ee9 init src 1d072180a
2024-06-17 20:39:36 +08:00

524 lines
25 KiB
Diff
Raw Permalink Blame History

From 95a87430a309ea304f819c60bbbbf0e620d6c147 Mon Sep 17 00:00:00 2001
From: jinbo <jinbo@loongson.cn>
Date: Wed, 29 Mar 2023 14:21:23 +0800
Subject: [PATCH 07/10] swscale [mips]: Add following msa optimizations
yuv420_rgb24_msa
yuv420_bgr24_msa
yuv420_rgba32_msa
yuv420_argb32_msa
yuv420_bgra32_msa
yuv420_abgr32_msa
Change-Id: I2a04879d5cf59a89aa551a1f8ca611452b9c108c
---
libswscale/mips/Makefile | 1 +
libswscale/mips/swscale_init_mips.c | 36 +++
libswscale/mips/swscale_mips.h | 19 ++
libswscale/mips/yuv2rgb_msa.c | 385 ++++++++++++++++++++++++++++
libswscale/swscale_internal.h | 1 +
libswscale/yuv2rgb.c | 2 +
6 files changed, 444 insertions(+)
create mode 100644 libswscale/mips/yuv2rgb_msa.c
diff --git a/libswscale/mips/Makefile b/libswscale/mips/Makefile
index 48c2425eff..8721561930 100644
--- a/libswscale/mips/Makefile
+++ b/libswscale/mips/Makefile
@@ -2,3 +2,4 @@ OBJS += mips/swscale_init_mips.o
OBJS += mips/rgb2rgb_init_mips.o
MSA-OBJS += mips/swscale_msa.o
MSA-OBJS += mips/rgb2rgb_msa.o
+MSA-OBJS += mips/yuv2rgb_msa.o
diff --git a/libswscale/mips/swscale_init_mips.c b/libswscale/mips/swscale_init_mips.c
index 8e26486f1e..4ed8b1b172 100644
--- a/libswscale/mips/swscale_init_mips.c
+++ b/libswscale/mips/swscale_init_mips.c
@@ -209,3 +209,39 @@ av_cold void ff_sws_init_swscale_mips(SwsContext *c)
}
#endif /* #if HAVE_MSA */
}
+
+av_cold SwsFunc ff_yuv2rgb_init_mips(SwsContext *c)
+{
+ int cpu_flags = av_get_cpu_flags();
+#if HAVE_MSA
+ if (have_msa(cpu_flags)) {
+ switch (c->dstFormat) {
+ case AV_PIX_FMT_RGB24:
+ return yuv420_rgb24_msa;
+ case AV_PIX_FMT_BGR24:
+ return yuv420_bgr24_msa;
+ case AV_PIX_FMT_RGBA:
+ if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat)) {
+ break;
+ } else
+ return yuv420_rgba32_msa;
+ case AV_PIX_FMT_ARGB:
+ if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat)) {
+ break;
+ } else
+ return yuv420_argb32_msa;
+ case AV_PIX_FMT_BGRA:
+ if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat)) {
+ break;
+ } else
+ return yuv420_bgra32_msa;
+ case AV_PIX_FMT_ABGR:
+ if (CONFIG_SWSCALE_ALPHA && isALPHA(c->srcFormat)) {
+ break;
+ } else
+ return yuv420_abgr32_msa;
+ }
+ }
+ return NULL;
+#endif
+}
diff --git a/libswscale/mips/swscale_mips.h b/libswscale/mips/swscale_mips.h
index ff30593e0d..683b3f5c72 100644
--- a/libswscale/mips/swscale_mips.h
+++ b/libswscale/mips/swscale_mips.h
@@ -401,4 +401,23 @@ void planar_rgb_to_uv_msa(uint8_t *_dstU, uint8_t *_dstV, const uint8_t *src[4],
void planar_rgb_to_y_msa(uint8_t *_dst, const uint8_t *src[4], int width,
int32_t *rgb2yuv);
+int yuv420_rgb24_msa(SwsContext *c, const uint8_t *src[], int srcStride[],
+ int srcSliceY, int srcSliceH, uint8_t *dst[], int dstStride[]);
+
+int yuv420_bgr24_msa(SwsContext *c, const uint8_t *src[], int srcStride[],
+ int srcSliceY, int srcSliceH, uint8_t *dst[], int dstStride[]);
+
+int yuv420_rgba32_msa(SwsContext *c, const uint8_t *src[], int srcStride[],
+ int srcSliceY, int srcSliceH, uint8_t *dst[], int dstStride[]);
+
+int yuv420_bgra32_msa(SwsContext *c, const uint8_t *src[], int srcStride[],
+ int srcSliceY, int srcSliceH, uint8_t *dst[], int dstStride[]);
+
+int yuv420_argb32_msa(SwsContext *c, const uint8_t *src[], int srcStride[],
+ int srcSliceY, int srcSliceH, uint8_t *dst[], int dstStride[]);
+
+int yuv420_abgr32_msa(SwsContext *c, const uint8_t *src[], int srcStride[],
+ int srcSliceY, int srcSliceH, uint8_t *dst[], int dstStride[]);
+
+
#endif /* SWSCALE_MIPS_SWSCALE_MIPS_H */
diff --git a/libswscale/mips/yuv2rgb_msa.c b/libswscale/mips/yuv2rgb_msa.c
new file mode 100644
index 0000000000..33e951f6ba
--- /dev/null
+++ b/libswscale/mips/yuv2rgb_msa.c
@@ -0,0 +1,385 @@
+/*
+ * Copyright (C) 2023 Loongson Technology Co. Ltd.
+ * Contributed by jinbo(jinbo@loongson.cn)
+ * All rights reserved.
+ *
+ * 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 "swscale_mips.h"
+#include "libavutil/mips/generic_macros_msa.h"
+
+//Using ST_D1 will result in great regression, reason is not clear.
+#define SD_NEW(val, pdst) \
+{ \
+ uint64_t *pdst_sd_m = (uint64_t *) (pdst); \
+ *pdst_sd_m = (val); \
+}
+
+#define YUV2RGB_LOAD_COE \
+ /* Load x_offset */ \
+ v8i16 y_offset = (v8i16)__msa_fill_d(c->yOffset); \
+ v8i16 u_offset = (v8i16)__msa_fill_d(c->uOffset); \
+ v8i16 v_offset = (v8i16)__msa_fill_d(c->vOffset); \
+ /*Load x_coeff */ \
+ v8i16 ug_coeff = (v8i16)__msa_fill_d(c->ugCoeff); \
+ v8i16 vg_coeff = (v8i16)__msa_fill_d(c->vgCoeff); \
+ v8i16 y_coeff = (v8i16)__msa_fill_d(c->yCoeff); \
+ v8i16 ub_coeff = (v8i16)__msa_fill_d(c->ubCoeff); \
+ v8i16 vr_coeff = (v8i16)__msa_fill_d(c->vrCoeff); \
+
+#define LOAD_YUV_16 \
+ m_y1 = LD_SB(py_1); \
+ m_y2 = LD_SB(py_2); \
+ m_u = LD_SB(pu); \
+ m_v = LD_SB(pv); \
+ m_u = __msa_ilvr_b((v16i8)m_u, (v16i8)m_u); \
+ m_v = __msa_ilvr_b((v16i8)m_v, (v16i8)m_v); \
+ m_u_l = __msa_ilvl_b(zero, (v16i8)m_u); \
+ m_u = __msa_ilvr_b(zero, (v16i8)m_u); \
+ m_v_l = __msa_ilvl_b(zero, (v16i8)m_v); \
+ m_v = __msa_ilvr_b(zero, (v16i8)m_v); \
+ m_y1_l = __msa_ilvl_b(zero, (v16i8)m_y1); \
+ m_y1 = __msa_ilvr_b(zero, (v16i8)m_y1); \
+ m_y2_l = __msa_ilvl_b(zero, (v16i8)m_y2); \
+ m_y2 = __msa_ilvr_b(zero, (v16i8)m_y2); \
+
+#define MUH_H(in0, in1, in2, in3, out0) \
+{ \
+ v4i32 tmp0, tmp1; \
+ tmp0 = in0 * in1; \
+ tmp1 = in2 * in3; \
+ out0 = __msa_pckod_h((v8i16)tmp1, (v8i16)tmp0); \
+}
+
+#define YUV2RGB(y1, y2, u, v, r1, g1, b1, r2, g2, b2) \
+{ \
+ y1 = (v16i8)__msa_slli_h((v8i16)y1, 3); \
+ y2 = (v16i8)__msa_slli_h((v8i16)y2, 3); \
+ u = (v16i8)__msa_slli_h((v8i16)u, 3); \
+ v = (v16i8)__msa_slli_h((v8i16)v, 3); \
+ y1 = (v16i8)__msa_subv_h((v8i16)y1, y_offset); \
+ y2 = (v16i8)__msa_subv_h((v8i16)y2, y_offset); \
+ u = (v16i8)__msa_subv_h((v8i16)u, u_offset); \
+ v = (v16i8)__msa_subv_h((v8i16)v, v_offset); \
+ UNPCK_SH_SW(y_coeff, y_coeff_r, y_coeff_l); \
+ UNPCK_SH_SW(y1, tmp_r, tmp_l); \
+ MUH_H(tmp_r, y_coeff_r, tmp_l, y_coeff_l, y_1_r); \
+ UNPCK_SH_SW(y2, tmp_r, tmp_l); \
+ MUH_H(tmp_r, y_coeff_r, tmp_l, y_coeff_l, y_2_r); \
+ UNPCK_SH_SW(ug_coeff, ug_coeff_r, ug_coeff_l); \
+ UNPCK_SH_SW(ub_coeff, ub_coeff_r, ub_coeff_l); \
+ UNPCK_SH_SW(u, tmp_r, tmp_l); \
+ MUH_H(tmp_r, ug_coeff_r, tmp_l, ug_coeff_l, u2g_r); \
+ MUH_H(tmp_r, ub_coeff_r, tmp_l, ub_coeff_l, u2b_r); \
+ UNPCK_SH_SW(vr_coeff, vr_coeff_r, vr_coeff_l); \
+ UNPCK_SH_SW(vg_coeff, vg_coeff_r, vg_coeff_l); \
+ UNPCK_SH_SW(v, tmp_r, tmp_l); \
+ MUH_H(tmp_r, vr_coeff_r, tmp_l, vr_coeff_l, v2r_r); \
+ MUH_H(tmp_r, vg_coeff_r, tmp_l, vg_coeff_l, v2g_r); \
+ r1 = __msa_adds_s_h(y_1_r, v2r_r); \
+ v2g_r = __msa_adds_s_h(v2g_r, u2g_r); \
+ g1 = __msa_adds_s_h(y_1_r, v2g_r); \
+ b1 = __msa_adds_s_h(y_1_r, u2b_r); \
+ r2 = __msa_adds_s_h(y_2_r, v2r_r); \
+ g2 = __msa_adds_s_h(y_2_r, v2g_r); \
+ b2 = __msa_adds_s_h(y_2_r, u2b_r); \
+ CLIP_SH4_0_255(r1, g1, b1, r2); \
+ CLIP_SH2_0_255(g2, b2); \
+}
+
+#define RGB_PACK(r, g, b, rgb_l, rgb_h) \
+{ \
+ v16i8 rg; \
+ rg = __msa_ilvev_b((v16i8)g, (v16i8)r); \
+ rgb_l = (v8i16)__msa_vshf_b((v16i8)shuf2, (v16i8)b, rg); \
+ rgb_h = (v8i16)__msa_vshf_b((v16i8)shuf3, (v16i8)b, rg); \
+}
+
+#define RGB32_PACK(a, r, g, b, rgb_l, rgb_h) \
+{ \
+ v16i8 ra, bg; \
+ ra = __msa_ilvev_b((v16i8)r, (v16i8)a); \
+ bg = __msa_ilvev_b((v16i8)b, (v16i8)g); \
+ rgb_l = __msa_ilvr_h((v8i16)bg, (v8i16)ra); \
+ rgb_h = __msa_ilvl_h((v8i16)bg, (v8i16)ra); \
+}
+
+#define RGB_STORE(rgb_l, rgb_h, image) \
+{ \
+ ST_SH(rgb_l, image); \
+ SD_NEW(__msa_copy_u_d((v2i64)rgb_h, 0), image); \
+}
+
+#define RGB32_STORE(rgb_l, rgb_h, image) \
+{ \
+ ST_SH2(rgb_l, rgb_h, image, 4); \
+}
+
+#define YUV2RGBFUNC(func_name, dst_type, alpha) \
+ int func_name(SwsContext *c, const uint8_t *src[], \
+ int srcStride[], int srcSliceY, int srcSliceH, \
+ uint8_t *dst[], int dstStride[]) \
+{ \
+ int x, y, h_size, vshift, res; \
+ v16i8 m_y1, m_y2, m_u, m_v; \
+ v16i8 m_u_l, m_v_l, m_y1_l, m_y2_l; \
+ v4i32 y_coeff_r, y_coeff_l, ug_coeff_r, ug_coeff_l; \
+ v4i32 ub_coeff_r, ub_coeff_l, vr_coeff_r, vr_coeff_l; \
+ v4i32 vg_coeff_r, vg_coeff_l, tmp_r, tmp_l; \
+ v8i16 y_1_r, y_2_r, u2g_r, v2g_r, u2b_r, v2r_r, rgb1_l, rgb1_h; \
+ v8i16 rgb2_l, rgb2_h, r1, g1, b1, r2, g2, b2; \
+ v2i64 shuf2 = {0x0504120302100100, 0x0A18090816070614}; \
+ v2i64 shuf3 = {0x1E0F0E1C0D0C1A0B, 0x0101010101010101}; \
+ v16i8 zero = __msa_fill_b(0); \
+ \
+ YUV2RGB_LOAD_COE \
+ \
+ h_size = c->dstW >> 4; \
+ res = (c->dstW & 15) >> 1; \
+ vshift = c->srcFormat != AV_PIX_FMT_YUV422P; \
+ for (y = 0; y < srcSliceH; y += 2) { \
+ dst_type av_unused *r, *g, *b; \
+ dst_type *image1 = (dst_type *)(dst[0] + (y + srcSliceY) * dstStride[0]);\
+ dst_type *image2 = (dst_type *)(image1 + dstStride[0]);\
+ const uint8_t *py_1 = src[0] + y * srcStride[0]; \
+ const uint8_t *py_2 = py_1 + srcStride[0]; \
+ const uint8_t *pu = src[1] + (y >> vshift) * srcStride[1]; \
+ const uint8_t *pv = src[2] + (y >> vshift) * srcStride[2]; \
+ for(x = 0; x < h_size; x++) { \
+
+#define YUV2RGBFUNC32(func_name, dst_type, alpha) \
+ int func_name(SwsContext *c, const uint8_t *src[], \
+ int srcStride[], int srcSliceY, int srcSliceH, \
+ uint8_t *dst[], int dstStride[]) \
+{ \
+ int x, y, h_size, vshift, res; \
+ v16i8 m_y1, m_y2, m_u, m_v; \
+ v16i8 m_u_l, m_v_l, m_y1_l, m_y2_l; \
+ v4i32 y_coeff_r, y_coeff_l, ug_coeff_r, ug_coeff_l; \
+ v4i32 ub_coeff_r, ub_coeff_l, vr_coeff_r, vr_coeff_l; \
+ v4i32 vg_coeff_r, vg_coeff_l, tmp_r, tmp_l; \
+ v8i16 y_1_r, y_2_r, u2g_r, v2g_r, u2b_r, v2r_r, rgb1_l, rgb1_h; \
+ v8i16 rgb2_l, rgb2_h, r1, g1, b1, r2, g2, b2; \
+ v16u8 a = (v16u8)__msa_fill_b(0xFF); \
+ v16i8 zero = __msa_fill_b(0); \
+ \
+ YUV2RGB_LOAD_COE \
+ \
+ h_size = c->dstW >> 4; \
+ res = (c->dstW & 15) >> 1; \
+ vshift = c->srcFormat != AV_PIX_FMT_YUV422P; \
+ for (y = 0; y < srcSliceH; y += 2) { \
+ int yd = y + srcSliceY; \
+ dst_type av_unused *r, *g, *b; \
+ dst_type *image1 = (dst_type *)(dst[0] + (yd) * dstStride[0]); \
+ dst_type *image2 = (dst_type *)(dst[0] + (yd + 1) * dstStride[0]); \
+ const uint8_t *py_1 = src[0] + y * srcStride[0]; \
+ const uint8_t *py_2 = py_1 + srcStride[0]; \
+ const uint8_t *pu = src[1] + (y >> vshift) * srcStride[1]; \
+ const uint8_t *pv = src[2] + (y >> vshift) * srcStride[2]; \
+ for(x = 0; x < h_size; x++) { \
+
+#define DEALYUV2RGBREMAIN \
+ py_1 += 16; \
+ py_2 += 16; \
+ pu += 8; \
+ pv += 8; \
+ image1 += 48; \
+ image2 += 48; \
+ } \
+ for (x = 0; x < res; x++) { \
+ int av_unused U, V, Y; \
+ U = pu[0]; \
+ V = pv[0]; \
+ r = (void *)c->table_rV[V+YUVRGB_TABLE_HEADROOM]; \
+ g = (void *)(c->table_gU[U+YUVRGB_TABLE_HEADROOM] \
+ + c->table_gV[V+YUVRGB_TABLE_HEADROOM]); \
+ b = (void *)c->table_bU[U+YUVRGB_TABLE_HEADROOM];
+
+#define DEALYUV2RGBREMAIN32 \
+ py_1 += 16; \
+ py_2 += 16; \
+ pu += 8; \
+ pv += 8; \
+ image1 += 16; \
+ image2 += 16; \
+ } \
+ for (x = 0; x < res; x++) { \
+ int av_unused U, V, Y; \
+ U = pu[0]; \
+ V = pv[0]; \
+ r = (void *)c->table_rV[V+YUVRGB_TABLE_HEADROOM]; \
+ g = (void *)(c->table_gU[U+YUVRGB_TABLE_HEADROOM] \
+ + c->table_gV[V+YUVRGB_TABLE_HEADROOM]); \
+ b = (void *)c->table_bU[U+YUVRGB_TABLE_HEADROOM]; \
+
+#define PUTRGB24(dst, src) \
+ Y = src[0]; \
+ dst[0] = r[Y]; \
+ dst[1] = g[Y]; \
+ dst[2] = b[Y]; \
+ Y = src[1]; \
+ dst[3] = r[Y]; \
+ dst[4] = g[Y]; \
+ dst[5] = b[Y];
+
+#define PUTBGR24(dst, src) \
+ Y = src[0]; \
+ dst[0] = b[Y]; \
+ dst[1] = g[Y]; \
+ dst[2] = r[Y]; \
+ Y = src[1]; \
+ dst[3] = b[Y]; \
+ dst[4] = g[Y]; \
+ dst[5] = r[Y];
+
+#define PUTRGB(dst, src) \
+ Y = src[0]; \
+ dst[0] = r[Y] + g[Y] + b[Y]; \
+ Y = src[1]; \
+ dst[1] = r[Y] + g[Y] + b[Y]; \
+
+#define ENDRES \
+ pu += 1; \
+ pv += 1; \
+ py_1 += 2; \
+ py_2 += 2; \
+ image1 += 6; \
+ image2 += 6; \
+
+#define ENDRES32 \
+ pu += 1; \
+ pv += 1; \
+ py_1 += 2; \
+ py_2 += 2; \
+ image1 += 2; \
+ image2 += 2; \
+
+#define END_FUNC() \
+ } \
+ } \
+ return srcSliceH; \
+}
+
+YUV2RGBFUNC(yuv420_rgb24_msa, uint8_t, 0)
+ LOAD_YUV_16
+ YUV2RGB(m_y1, m_y2, m_u, m_v, r1, g1, b1, r2, g2, b2);
+ RGB_PACK(r1, g1, b1, rgb1_l, rgb1_h);
+ RGB_PACK(r2, g2, b2, rgb2_l, rgb2_h);
+ RGB_STORE(rgb1_l, rgb1_h, image1);
+ RGB_STORE(rgb2_l, rgb2_h, image2);
+ YUV2RGB(m_y1_l, m_y2_l, m_u_l, m_v_l, r1, g1, b1, r2, g2, b2);
+ RGB_PACK(r1, g1, b1, rgb1_l, rgb1_h);
+ RGB_PACK(r2, g2, b2, rgb2_l, rgb2_h);
+ RGB_STORE(rgb1_l, rgb1_h, image1 + 24);
+ RGB_STORE(rgb2_l, rgb2_h, image2 + 24);
+ DEALYUV2RGBREMAIN
+ PUTRGB24(image1, py_1);
+ PUTRGB24(image2, py_2);
+ ENDRES
+ END_FUNC()
+
+YUV2RGBFUNC(yuv420_bgr24_msa, uint8_t, 0)
+ LOAD_YUV_16
+ YUV2RGB(m_y1, m_y2, m_u, m_v, r1, g1, b1, r2, g2, b2);
+ RGB_PACK(b1, g1, r1, rgb1_l, rgb1_h);
+ RGB_PACK(b2, g2, r2, rgb2_l, rgb2_h);
+ RGB_STORE(rgb1_l, rgb1_h, image1);
+ RGB_STORE(rgb2_l, rgb2_h, image2);
+ YUV2RGB(m_y1_l, m_y2_l, m_u_l, m_v_l, r1, g1, b1, r2, g2, b2);
+ RGB_PACK(b1, g1, r1, rgb1_l, rgb1_h);
+ RGB_PACK(b2, g2, r2, rgb2_l, rgb2_h);
+ RGB_STORE(rgb1_l, rgb1_h, image1 + 24);
+ RGB_STORE(rgb2_l, rgb2_h, image2 + 24);
+ DEALYUV2RGBREMAIN
+ PUTBGR24(image1, py_1);
+ PUTBGR24(image2, py_2);
+ ENDRES
+ END_FUNC()
+
+YUV2RGBFUNC32(yuv420_rgba32_msa, uint32_t, 0)
+ LOAD_YUV_16
+ YUV2RGB(m_y1, m_y2, m_u, m_v, r1, g1, b1, r2, g2, b2);
+ RGB32_PACK(r1, g1, b1, a, rgb1_l, rgb1_h);
+ RGB32_PACK(r2, g2, b2, a, rgb2_l, rgb2_h);
+ RGB32_STORE(rgb1_l, rgb1_h, image1);
+ RGB32_STORE(rgb2_l, rgb2_h, image2);
+ YUV2RGB(m_y1_l, m_y2_l, m_u_l, m_v_l, r1, g1, b1, r2, g2, b2);
+ RGB32_PACK(r1, g1, b1, a, rgb1_l, rgb1_h);
+ RGB32_PACK(r2, g2, b2, a, rgb2_l, rgb2_h);
+ RGB32_STORE(rgb1_l, rgb1_h, image1 + 8);
+ RGB32_STORE(rgb2_l, rgb2_h, image2 + 8);
+ DEALYUV2RGBREMAIN32
+ PUTRGB(image1, py_1);
+ PUTRGB(image2, py_2);
+ ENDRES32
+ END_FUNC()
+
+YUV2RGBFUNC32(yuv420_bgra32_msa, uint32_t, 0)
+ LOAD_YUV_16
+ YUV2RGB(m_y1, m_y2, m_u, m_v, r1, g1, b1, r2, g2, b2);
+ RGB32_PACK(b1, g1, r1, a, rgb1_l, rgb1_h);
+ RGB32_PACK(b2, g2, r2, a, rgb2_l, rgb2_h);
+ RGB32_STORE(rgb1_l, rgb1_h, image1);
+ RGB32_STORE(rgb2_l, rgb2_h, image2);
+ YUV2RGB(m_y1_l, m_y2_l, m_u_l, m_v_l, r1, g1, b1, r2, g2, b2);
+ RGB32_PACK(b1, g1, r1, a, rgb1_l, rgb1_h);
+ RGB32_PACK(b2, g2, r2, a, rgb2_l, rgb2_h);
+ RGB32_STORE(rgb1_l, rgb1_h, image1 + 8);
+ RGB32_STORE(rgb2_l, rgb2_h, image2 + 8);
+ DEALYUV2RGBREMAIN32
+ PUTRGB(image1, py_1);
+ PUTRGB(image2, py_2);
+ ENDRES32
+ END_FUNC()
+
+YUV2RGBFUNC32(yuv420_argb32_msa, uint32_t, 0)
+ LOAD_YUV_16
+ YUV2RGB(m_y1, m_y2, m_u, m_v, r1, g1, b1, r2, g2, b2);
+ RGB32_PACK(a, r1, g1, b1, rgb1_l, rgb1_h);
+ RGB32_PACK(a, r2, g2, b2, rgb2_l, rgb2_h);
+ RGB32_STORE(rgb1_l, rgb1_h, image1);
+ RGB32_STORE(rgb2_l, rgb2_h, image2);
+ YUV2RGB(m_y1_l, m_y2_l, m_u_l, m_v_l, r1, g1, b1, r2, g2, b2);
+ RGB32_PACK(a, r1, g1, b1, rgb1_l, rgb1_h);
+ RGB32_PACK(a, r2, g2, b2, rgb2_l, rgb2_h);
+ RGB32_STORE(rgb1_l, rgb1_h, image1 + 8);
+ RGB32_STORE(rgb2_l, rgb2_h, image2 + 8);
+ DEALYUV2RGBREMAIN32
+ PUTRGB(image1, py_1);
+ PUTRGB(image2, py_2);
+ ENDRES32
+ END_FUNC()
+
+YUV2RGBFUNC32(yuv420_abgr32_msa, uint32_t, 0)
+ LOAD_YUV_16
+ YUV2RGB(m_y1, m_y2, m_u, m_v, r1, g1, b1, r2, g2, b2);
+ RGB32_PACK(a, b1, g1, r1, rgb1_l, rgb1_h);
+ RGB32_PACK(a, b2, g2, r2, rgb2_l, rgb2_h);
+ RGB32_STORE(rgb1_l, rgb1_h, image1);
+ RGB32_STORE(rgb2_l, rgb2_h, image2);
+ YUV2RGB(m_y1_l, m_y2_l, m_u_l, m_v_l, r1, g1, b1, r2, g2, b2);
+ RGB32_PACK(a, b1, g1, r1, rgb1_l, rgb1_h);
+ RGB32_PACK(a, b2, g2, r2, rgb2_l, rgb2_h);
+ RGB32_STORE(rgb1_l, rgb1_h, image1 + 8);
+ RGB32_STORE(rgb2_l, rgb2_h, image2 + 8);
+ DEALYUV2RGBREMAIN32
+ PUTRGB(image1, py_1);
+ PUTRGB(image2, py_2);
+ ENDRES32
+ END_FUNC()
diff --git a/libswscale/swscale_internal.h b/libswscale/swscale_internal.h
index 1d2932dc43..4d1aeb1cf2 100644
--- a/libswscale/swscale_internal.h
+++ b/libswscale/swscale_internal.h
@@ -643,6 +643,7 @@ av_cold void ff_sws_init_range_convert(SwsContext *c);
SwsFunc ff_yuv2rgb_init_x86(SwsContext *c);
SwsFunc ff_yuv2rgb_init_ppc(SwsContext *c);
SwsFunc ff_yuv2rgb_init_loongarch(SwsContext *c);
+SwsFunc ff_yuv2rgb_init_mips(SwsContext *c);
static av_always_inline int is16BPS(enum AVPixelFormat pix_fmt)
{
diff --git a/libswscale/yuv2rgb.c b/libswscale/yuv2rgb.c
index f7edfd3d0c..1b36dd319e 100644
--- a/libswscale/yuv2rgb.c
+++ b/libswscale/yuv2rgb.c
@@ -686,6 +686,8 @@ SwsFunc ff_yuv2rgb_get_func_ptr(SwsContext *c)
t = ff_yuv2rgb_init_x86(c);
if (ARCH_LOONGARCH)
t = ff_yuv2rgb_init_loongarch(c);
+ if (ARCH_MIPS)
+ t = ff_yuv2rgb_init_mips(c);
if (t)
return t;
--
2.20.1
Reference in New Issue View Git Blame Copy Permalink