mirror of
https://chromium.googlesource.com/libyuv/libyuv
synced 2025-12-06 16:56:55 +08:00
44396e6e9a
* Run on SiFive internal FPGA: ARGBToRAW_Opt (~1.55x vs scalar) ARGBToRGB24_Opt (~1.44x vs scalar) RGB24ToARGB_Opt (~1.77x vs scalar) LIBYUV_WIDTH=1280 LIBYUV_HEIGHT=720 LIBYUV_REPEAT=10 Bug: libyuv:956 Change-Id: I26722f6848cd68684d95d9a7ee06ce0416e7985d Signed-off-by: Darren Hsieh <darren.hsieh@sifive.com> Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/4413083 Reviewed-by: Frank Barchard <fbarchard@chromium.org> Commit-Queue: Frank Barchard <fbarchard@chromium.org>
118 lines
3.4 KiB
C++
118 lines
3.4 KiB
C++
/*
|
|
* Copyright 2023 The LibYuv Project Authors. All rights reserved.
|
|
*
|
|
* Use of this source code is governed by a BSD-style license
|
|
* that can be found in the LICENSE file in the root of the source
|
|
* tree. An additional intellectual property rights grant can be found
|
|
* in the file PATENTS. All contributing project authors may
|
|
* be found in the AUTHORS file in the root of the source tree.
|
|
*/
|
|
|
|
/*
|
|
* Copyright (c) 2023 SiFive, Inc. All rights reserved.
|
|
*
|
|
* Contributed by Darren Hsieh <darren.hsieh@sifive.com>
|
|
*
|
|
*/
|
|
|
|
#include <assert.h>
|
|
|
|
#include "libyuv/row.h"
|
|
|
|
#if !defined(LIBYUV_DISABLE_RVV) && defined(__riscv)
|
|
#include <riscv_vector.h>
|
|
|
|
#ifdef __cplusplus
|
|
namespace libyuv {
|
|
extern "C" {
|
|
#endif
|
|
|
|
void RAWToARGBRow_RVV(const uint8_t* src_raw, uint8_t* dst_argb, int width) {
|
|
size_t vl = __riscv_vsetvl_e8m2(width);
|
|
vuint8m2_t v_a = __riscv_vmv_v_x_u8m2(255u, vl);
|
|
do {
|
|
vuint8m2_t v_b, v_g, v_r;
|
|
__riscv_vlseg3e8_v_u8m2(&v_r, &v_g, &v_b, src_raw, vl);
|
|
__riscv_vsseg4e8_v_u8m2(dst_argb, v_b, v_g, v_r, v_a, vl);
|
|
width -= vl;
|
|
src_raw += (3 * vl);
|
|
dst_argb += (4 * vl);
|
|
vl = __riscv_vsetvl_e8m2(width);
|
|
} while (width > 0);
|
|
}
|
|
|
|
void RAWToRGBARow_RVV(const uint8_t* src_raw, uint8_t* dst_rgba, int width) {
|
|
size_t vl = __riscv_vsetvl_e8m2(width);
|
|
vuint8m2_t v_a = __riscv_vmv_v_x_u8m2(255u, vl);
|
|
do {
|
|
vuint8m2_t v_b, v_g, v_r;
|
|
__riscv_vlseg3e8_v_u8m2(&v_r, &v_g, &v_b, src_raw, vl);
|
|
__riscv_vsseg4e8_v_u8m2(dst_rgba, v_a, v_b, v_g, v_r, vl);
|
|
width -= vl;
|
|
src_raw += (3 * vl);
|
|
dst_rgba += (4 * vl);
|
|
vl = __riscv_vsetvl_e8m2(width);
|
|
} while (width > 0);
|
|
}
|
|
|
|
void RAWToRGB24Row_RVV(const uint8_t* src_raw, uint8_t* dst_rgb24, int width) {
|
|
do {
|
|
vuint8m2_t v_b, v_g, v_r;
|
|
size_t vl = __riscv_vsetvl_e8m2(width);
|
|
__riscv_vlseg3e8_v_u8m2(&v_b, &v_g, &v_r, src_raw, vl);
|
|
__riscv_vsseg3e8_v_u8m2(dst_rgb24, v_r, v_g, v_b, vl);
|
|
width -= vl;
|
|
src_raw += (3 * vl);
|
|
dst_rgb24 += (3 * vl);
|
|
} while (width > 0);
|
|
}
|
|
|
|
void ARGBToRAWRow_RVV(const uint8_t* src_argb, uint8_t* dst_raw, int width) {
|
|
do {
|
|
vuint8m2_t v_b, v_g, v_r, v_a;
|
|
size_t vl = __riscv_vsetvl_e8m2(width);
|
|
__riscv_vlseg4e8_v_u8m2(&v_b, &v_g, &v_r, &v_a, src_argb, vl);
|
|
__riscv_vsseg3e8_v_u8m2(dst_raw, v_r, v_g, v_b, vl);
|
|
width -= vl;
|
|
src_argb += (4 * vl);
|
|
dst_raw += (3 * vl);
|
|
} while (width > 0);
|
|
}
|
|
|
|
void ARGBToRGB24Row_RVV(const uint8_t* src_argb,
|
|
uint8_t* dst_rgb24,
|
|
int width) {
|
|
do {
|
|
vuint8m2_t v_b, v_g, v_r, v_a;
|
|
size_t vl = __riscv_vsetvl_e8m2(width);
|
|
__riscv_vlseg4e8_v_u8m2(&v_b, &v_g, &v_r, &v_a, src_argb, vl);
|
|
__riscv_vsseg3e8_v_u8m2(dst_rgb24, v_b, v_g, v_r, vl);
|
|
width -= vl;
|
|
src_argb += (4 * vl);
|
|
dst_rgb24 += (3 * vl);
|
|
} while (width > 0);
|
|
}
|
|
|
|
void RGB24ToARGBRow_RVV(const uint8_t* src_rgb24,
|
|
uint8_t* dst_argb,
|
|
int width) {
|
|
size_t vl = __riscv_vsetvl_e8m2(width);
|
|
vuint8m2_t v_a = __riscv_vmv_v_x_u8m2(255u, vl);
|
|
do {
|
|
vuint8m2_t v_b, v_g, v_r;
|
|
__riscv_vlseg3e8_v_u8m2(&v_b, &v_g, &v_r, src_rgb24, vl);
|
|
__riscv_vsseg4e8_v_u8m2(dst_argb, v_b, v_g, v_r, v_a, vl);
|
|
width -= vl;
|
|
src_rgb24 += (3 * vl);
|
|
dst_argb += (4 * vl);
|
|
vl = __riscv_vsetvl_e8m2(width);
|
|
} while (width > 0);
|
|
}
|
|
|
|
#ifdef __cplusplus
|
|
} // extern "C"
|
|
} // namespace libyuv
|
|
#endif
|
|
|
|
#endif // !defined(LIBYUV_DISABLE_RVV) && defined(__riscv)
|