coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2025-12-07 09:16:48 +08:00
Code Issues Packages Projects Releases Wiki Activity

RVV check __riscv_v_intrinsic version

Browse Source 
Bug: libyuv:965
Change-Id: I9b02abd13ab3345288655fa7a16383f59cf66bb8
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/4750230
Commit-Queue: Frank Barchard <fbarchard@chromium.org>
Reviewed-by: Mirko Bonadei <mbonadei@chromium.org>
This commit is contained in:
Frank Barchard 2023-08-03 19:03:55 -07:00 committed by libyuv LUCI CQ
parent 506f93e9db
commit 696e619571
4 changed files with 198 additions and 24 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

23
include/libyuv/row.h
View File

@ -791,7 +791,11 @@ extern "C" {
#endif #endif
#if !defined(LIBYUV_DISABLE_RVV) && defined(__riscv_vector) #if !defined(LIBYUV_DISABLE_RVV) && defined(__riscv_vector)
#define HAS_COPYROW_RVV
#if __riscv_v_intrinsic == 11000
#define HAS_AB64TOARGBROW_RVV #define HAS_AB64TOARGBROW_RVV
#define HAS_ABGRTOYJROW_RVV
#define HAS_ABGRTOYROW_RVV
#define HAS_AR64TOARGBROW_RVV #define HAS_AR64TOARGBROW_RVV
#define HAS_ARGBATTENUATEROW_RVV #define HAS_ARGBATTENUATEROW_RVV
#define HAS_ARGBBLENDROW_RVV #define HAS_ARGBBLENDROW_RVV
@ -801,13 +805,11 @@ extern "C" {
#define HAS_ARGBTOAR64ROW_RVV #define HAS_ARGBTOAR64ROW_RVV
#define HAS_ARGBTORAWROW_RVV #define HAS_ARGBTORAWROW_RVV
#define HAS_ARGBTORGB24ROW_RVV #define HAS_ARGBTORGB24ROW_RVV
#define HAS_ARGBTOYROW_RVV
#define HAS_ARGBTOYJROW_RVV #define HAS_ARGBTOYJROW_RVV
#define HAS_ABGRTOYROW_RVV #define HAS_ARGBTOYMATRIXROW_RVV
#define HAS_ABGRTOYJROW_RVV #define HAS_ARGBTOYROW_RVV
#define HAS_BGRATOYROW_RVV #define HAS_BGRATOYROW_RVV
#define HAS_BLENDPLANEROW_RVV #define HAS_BLENDPLANEROW_RVV
#define HAS_COPYROW_RVV
#define HAS_I400TOARGBROW_RVV #define HAS_I400TOARGBROW_RVV
#define HAS_I422ALPHATOARGBROW_RVV #define HAS_I422ALPHATOARGBROW_RVV
#define HAS_I422TOARGBROW_RVV #define HAS_I422TOARGBROW_RVV
@ -826,10 +828,6 @@ extern "C" {
#define HAS_NV12TORGB24ROW_RVV #define HAS_NV12TORGB24ROW_RVV
#define HAS_NV21TOARGBROW_RVV #define HAS_NV21TOARGBROW_RVV
#define HAS_NV21TORGB24ROW_RVV #define HAS_NV21TORGB24ROW_RVV
#define HAS_SPLITARGBROW_RVV
#define HAS_SPLITRGBROW_RVV
#define HAS_SPLITUVROW_RVV
#define HAS_SPLITXRGBROW_RVV
#define HAS_RAWTOARGBROW_RVV #define HAS_RAWTOARGBROW_RVV
#define HAS_RAWTORGB24ROW_RVV #define HAS_RAWTORGB24ROW_RVV
#define HAS_RAWTORGBAROW_RVV #define HAS_RAWTORGBAROW_RVV
@ -838,8 +836,15 @@ extern "C" {
#define HAS_RGB24TOARGBROW_RVV #define HAS_RGB24TOARGBROW_RVV
#define HAS_RGB24TOYJROW_RVV #define HAS_RGB24TOYJROW_RVV
#define HAS_RGB24TOYROW_RVV #define HAS_RGB24TOYROW_RVV
#define HAS_RGBATOYROW_RVV
#define HAS_RGBATOYJROW_RVV #define HAS_RGBATOYJROW_RVV
#define HAS_RGBATOYMATRIXROW_RVV
#define HAS_RGBATOYROW_RVV
#define HAS_RGBTOYMATRIXROW_RVV
#define HAS_SPLITARGBROW_RVV
#define HAS_SPLITRGBROW_RVV
#define HAS_SPLITUVROW_RVV
#define HAS_SPLITXRGBROW_RVV
#endif
#endif #endif
#if defined(_MSC_VER) && !defined(__CLR_VER) && !defined(__clang__) #if defined(_MSC_VER) && !defined(__CLR_VER) && !defined(__clang__)

22
include/libyuv/scale_row.h
View File

@ -177,20 +177,32 @@ extern "C" {
#if !defined(LIBYUV_DISABLE_RVV) && defined(__riscv_vector) #if !defined(LIBYUV_DISABLE_RVV) && defined(__riscv_vector)
#define HAS_SCALEADDROW_RVV #define HAS_SCALEADDROW_RVV
#define HAS_SCALEUVROWDOWN4_RVV
#define HAS_SCALEUVROWDOWNEVEN_RVV
#if __riscv_v_intrinsic == 11000
#define HAS_SCALEARGBROWDOWN2_RVV #define HAS_SCALEARGBROWDOWN2_RVV
#define HAS_SCALEARGBROWDOWNEVEN_RVV #define HAS_SCALEARGBROWDOWN2BOX_RVV
#define HAS_SCALEARGBROWDOWN2LINEAR_RVV
#define HAS_SCALEARGBROWDOWNEVENBOX_RVV
#define HAS_SCALEROWDOWN2_RVV #define HAS_SCALEROWDOWN2_RVV
#define HAS_SCALEROWDOWN2BOX_RVV
#define HAS_SCALEROWDOWN2LINEAR_RVV
#define HAS_SCALEROWDOWN34_0_BOX_RVV
#define HAS_SCALEROWDOWN34_1_BOX_RVV
#define HAS_SCALEROWDOWN34_RVV #define HAS_SCALEROWDOWN34_RVV
#define HAS_SCALEROWDOWN38_2_BOX_RVV
#define HAS_SCALEROWDOWN38_3_BOX_RVV
#define HAS_SCALEROWDOWN38_RVV #define HAS_SCALEROWDOWN38_RVV
#define HAS_SCALEROWDOWN4_RVV #define HAS_SCALEROWDOWN4_RVV
#define HAS_SCALEROWUP2_LINEAR_RVV #define HAS_SCALEROWDOWN4BOX_RVV
#define HAS_SCALEROWUP2_BILINEAR_RVV #define HAS_SCALEROWUP2_BILINEAR_RVV
#define HAS_SCALEROWUP2_LINEAR_RVV
#define HAS_SCALEUVROWDOWN2_RVV #define HAS_SCALEUVROWDOWN2_RVV
#define HAS_SCALEUVROWDOWN2LINEAR_RVV
#define HAS_SCALEUVROWDOWN2BOX_RVV #define HAS_SCALEUVROWDOWN2BOX_RVV
#define HAS_SCALEUVROWDOWNEVEN_RVV #define HAS_SCALEUVROWDOWN2LINEAR_RVV
#define HAS_SCALEUVROWUP2_LINEAR_RVV
#define HAS_SCALEUVROWUP2_BILINEAR_RVV #define HAS_SCALEUVROWUP2_BILINEAR_RVV
#define HAS_SCALEUVROWUP2_LINEAR_RVV
#endif
#endif #endif
// Scale ARGB vertically with bilinear interpolation. // Scale ARGB vertically with bilinear interpolation.

107
source/row_rvv.cc
View File

@ -141,6 +141,7 @@ extern "C" {
v_y_16 = __riscv_vwaddu_vx_u16m4(v_y, 0, vl); \ v_y_16 = __riscv_vwaddu_vx_u16m4(v_y, 0, vl); \
} }
#ifdef HAS_ARGBTOAR64ROW_RVV
void ARGBToAR64Row_RVV(const uint8_t* src_argb, uint16_t* dst_ar64, int width) { void ARGBToAR64Row_RVV(const uint8_t* src_argb, uint16_t* dst_ar64, int width) {
size_t avl = (size_t)4 * width; size_t avl = (size_t)4 * width;
do { do {
@ -156,7 +157,9 @@ void ARGBToAR64Row_RVV(const uint8_t* src_argb, uint16_t* dst_ar64, int width) {
dst_ar64 += vl; dst_ar64 += vl;
} while (avl > 0); } while (avl > 0);
} }
#endif
#ifdef HAS_ARGBTOAB64ROW_RVV
void ARGBToAB64Row_RVV(const uint8_t* src_argb, uint16_t* dst_ab64, int width) { void ARGBToAB64Row_RVV(const uint8_t* src_argb, uint16_t* dst_ab64, int width) {
size_t avl = (size_t)width; size_t avl = (size_t)width;
do { do {
@ -178,7 +181,9 @@ void ARGBToAB64Row_RVV(const uint8_t* src_argb, uint16_t* dst_ab64, int width) {
dst_ab64 += 4 * vl; dst_ab64 += 4 * vl;
} while (avl > 0); } while (avl > 0);
} }
#endif
#ifdef HAS_AR64TOARGBROW_RVV
void AR64ToARGBRow_RVV(const uint16_t* src_ar64, uint8_t* dst_argb, int width) { void AR64ToARGBRow_RVV(const uint16_t* src_ar64, uint8_t* dst_argb, int width) {
size_t avl = (size_t)4 * width; size_t avl = (size_t)4 * width;
do { do {
@ -193,7 +198,9 @@ void AR64ToARGBRow_RVV(const uint16_t* src_ar64, uint8_t* dst_argb, int width) {
dst_argb += vl; dst_argb += vl;
} while (avl > 0); } while (avl > 0);
} }
#endif
#ifdef HAS_AB64TOARGBROW_RVV
void AB64ToARGBRow_RVV(const uint16_t* src_ab64, uint8_t* dst_argb, int width) { void AB64ToARGBRow_RVV(const uint16_t* src_ab64, uint8_t* dst_argb, int width) {
size_t avl = (size_t)width; size_t avl = (size_t)width;
do { do {
@ -211,7 +218,9 @@ void AB64ToARGBRow_RVV(const uint16_t* src_ab64, uint8_t* dst_argb, int width) {
dst_argb += 4 * vl; dst_argb += 4 * vl;
} while (avl > 0); } while (avl > 0);
} }
#endif
#ifdef HAS_RAWTOARGBROW_RVV
void RAWToARGBRow_RVV(const uint8_t* src_raw, uint8_t* dst_argb, int width) { void RAWToARGBRow_RVV(const uint8_t* src_raw, uint8_t* dst_argb, int width) {
size_t w = (size_t)width; size_t w = (size_t)width;
size_t vl = __riscv_vsetvl_e8m2(w); size_t vl = __riscv_vsetvl_e8m2(w);
@ -226,7 +235,9 @@ void RAWToARGBRow_RVV(const uint8_t* src_raw, uint8_t* dst_argb, int width) {
vl = __riscv_vsetvl_e8m2(w); vl = __riscv_vsetvl_e8m2(w);
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_RAWTORGBAROW_RVV
void RAWToRGBARow_RVV(const uint8_t* src_raw, uint8_t* dst_rgba, int width) { void RAWToRGBARow_RVV(const uint8_t* src_raw, uint8_t* dst_rgba, int width) {
size_t w = (size_t)width; size_t w = (size_t)width;
size_t vl = __riscv_vsetvl_e8m2(w); size_t vl = __riscv_vsetvl_e8m2(w);
@ -241,7 +252,9 @@ void RAWToRGBARow_RVV(const uint8_t* src_raw, uint8_t* dst_rgba, int width) {
vl = __riscv_vsetvl_e8m2(w); vl = __riscv_vsetvl_e8m2(w);
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_RAWTORGB24ROW_RVV
void RAWToRGB24Row_RVV(const uint8_t* src_raw, uint8_t* dst_rgb24, int width) { void RAWToRGB24Row_RVV(const uint8_t* src_raw, uint8_t* dst_rgb24, int width) {
size_t w = (size_t)width; size_t w = (size_t)width;
do { do {
@ -254,7 +267,9 @@ void RAWToRGB24Row_RVV(const uint8_t* src_raw, uint8_t* dst_rgb24, int width) {
dst_rgb24 += vl * 3; dst_rgb24 += vl * 3;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_ARGBTORAWROW_RVV
void ARGBToRAWRow_RVV(const uint8_t* src_argb, uint8_t* dst_raw, int width) { void ARGBToRAWRow_RVV(const uint8_t* src_argb, uint8_t* dst_raw, int width) {
size_t w = (size_t)width; size_t w = (size_t)width;
do { do {
@ -267,7 +282,9 @@ void ARGBToRAWRow_RVV(const uint8_t* src_argb, uint8_t* dst_raw, int width) {
dst_raw += vl * 3; dst_raw += vl * 3;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_ARGBTORGB24ROW_RVV
void ARGBToRGB24Row_RVV(const uint8_t* src_argb, void ARGBToRGB24Row_RVV(const uint8_t* src_argb,
uint8_t* dst_rgb24, uint8_t* dst_rgb24,
int width) { int width) {
@ -282,7 +299,9 @@ void ARGBToRGB24Row_RVV(const uint8_t* src_argb,
dst_rgb24 += vl * 3; dst_rgb24 += vl * 3;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_RGB24TOARGBROW_RVV
void RGB24ToARGBRow_RVV(const uint8_t* src_rgb24, void RGB24ToARGBRow_RVV(const uint8_t* src_rgb24,
uint8_t* dst_argb, uint8_t* dst_argb,
int width) { int width) {
@ -299,7 +318,9 @@ void RGB24ToARGBRow_RVV(const uint8_t* src_rgb24,
vl = __riscv_vsetvl_e8m2(w); vl = __riscv_vsetvl_e8m2(w);
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_I444TOARGBROW_RVV
void I444ToARGBRow_RVV(const uint8_t* src_y, void I444ToARGBRow_RVV(const uint8_t* src_y,
const uint8_t* src_u, const uint8_t* src_u,
const uint8_t* src_v, const uint8_t* src_v,
@ -328,7 +349,9 @@ void I444ToARGBRow_RVV(const uint8_t* src_y,
dst_argb += vl * 4; dst_argb += vl * 4;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_I444ALPHATOARGBROW_RVV
void I444AlphaToARGBRow_RVV(const uint8_t* src_y, void I444AlphaToARGBRow_RVV(const uint8_t* src_y,
const uint8_t* src_u, const uint8_t* src_u,
const uint8_t* src_v, const uint8_t* src_v,
@ -359,7 +382,9 @@ void I444AlphaToARGBRow_RVV(const uint8_t* src_y,
dst_argb += vl * 4; dst_argb += vl * 4;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_I444TORGB24ROW_RVV
void I444ToRGB24Row_RVV(const uint8_t* src_y, void I444ToRGB24Row_RVV(const uint8_t* src_y,
const uint8_t* src_u, const uint8_t* src_u,
const uint8_t* src_v, const uint8_t* src_v,
@ -387,7 +412,9 @@ void I444ToRGB24Row_RVV(const uint8_t* src_y,
dst_rgb24 += vl * 3; dst_rgb24 += vl * 3;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_I422TOARGBROW_RVV
void I422ToARGBRow_RVV(const uint8_t* src_y, void I422ToARGBRow_RVV(const uint8_t* src_y,
const uint8_t* src_u, const uint8_t* src_u,
const uint8_t* src_v, const uint8_t* src_v,
@ -416,7 +443,9 @@ void I422ToARGBRow_RVV(const uint8_t* src_y,
dst_argb += vl * 4; dst_argb += vl * 4;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_I422ALPHATOARGBROW_RVV
void I422AlphaToARGBRow_RVV(const uint8_t* src_y, void I422AlphaToARGBRow_RVV(const uint8_t* src_y,
const uint8_t* src_u, const uint8_t* src_u,
const uint8_t* src_v, const uint8_t* src_v,
@ -447,7 +476,9 @@ void I422AlphaToARGBRow_RVV(const uint8_t* src_y,
dst_argb += vl * 4; dst_argb += vl * 4;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_I422TORGBAROW_RVV
void I422ToRGBARow_RVV(const uint8_t* src_y, void I422ToRGBARow_RVV(const uint8_t* src_y,
const uint8_t* src_u, const uint8_t* src_u,
const uint8_t* src_v, const uint8_t* src_v,
@ -476,7 +507,9 @@ void I422ToRGBARow_RVV(const uint8_t* src_y,
dst_rgba += vl * 4; dst_rgba += vl * 4;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_I422TORGB24ROW_RVV
void I422ToRGB24Row_RVV(const uint8_t* src_y, void I422ToRGB24Row_RVV(const uint8_t* src_y,
const uint8_t* src_u, const uint8_t* src_u,
const uint8_t* src_v, const uint8_t* src_v,
@ -504,7 +537,9 @@ void I422ToRGB24Row_RVV(const uint8_t* src_y,
dst_rgb24 += vl * 3; dst_rgb24 += vl * 3;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_I400TOARGBROW_RVV
void I400ToARGBRow_RVV(const uint8_t* src_y, void I400ToARGBRow_RVV(const uint8_t* src_y,
uint8_t* dst_argb, uint8_t* dst_argb,
const struct YuvConstants* yuvconstants, const struct YuvConstants* yuvconstants,
@ -543,7 +578,9 @@ void I400ToARGBRow_RVV(const uint8_t* src_y,
dst_argb += vl * 4; dst_argb += vl * 4;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_J400TOARGBROW_RVV
void J400ToARGBRow_RVV(const uint8_t* src_y, uint8_t* dst_argb, int width) { void J400ToARGBRow_RVV(const uint8_t* src_y, uint8_t* dst_argb, int width) {
size_t w = (size_t)width; size_t w = (size_t)width;
size_t vl = __riscv_vsetvl_e8m2(w); size_t vl = __riscv_vsetvl_e8m2(w);
@ -558,7 +595,9 @@ void J400ToARGBRow_RVV(const uint8_t* src_y, uint8_t* dst_argb, int width) {
vl = __riscv_vsetvl_e8m2(w); vl = __riscv_vsetvl_e8m2(w);
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_COPYROW_RVV
void CopyRow_RVV(const uint8_t* src, uint8_t* dst, int width) { void CopyRow_RVV(const uint8_t* src, uint8_t* dst, int width) {
size_t w = (size_t)width; size_t w = (size_t)width;
do { do {
@ -570,7 +609,9 @@ void CopyRow_RVV(const uint8_t* src, uint8_t* dst, int width) {
dst += vl; dst += vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_NV12TOARGBROW_RVV
void NV12ToARGBRow_RVV(const uint8_t* src_y, void NV12ToARGBRow_RVV(const uint8_t* src_y,
const uint8_t* src_uv, const uint8_t* src_uv,
uint8_t* dst_argb, uint8_t* dst_argb,
@ -597,7 +638,9 @@ void NV12ToARGBRow_RVV(const uint8_t* src_y,
dst_argb += vl * 4; dst_argb += vl * 4;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_NV12TORGB24ROW_RVV
void NV12ToRGB24Row_RVV(const uint8_t* src_y, void NV12ToRGB24Row_RVV(const uint8_t* src_y,
const uint8_t* src_uv, const uint8_t* src_uv,
uint8_t* dst_rgb24, uint8_t* dst_rgb24,
@ -623,7 +666,9 @@ void NV12ToRGB24Row_RVV(const uint8_t* src_y,
dst_rgb24 += vl * 3; dst_rgb24 += vl * 3;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_NV21TOARGBROW_RVV
void NV21ToARGBRow_RVV(const uint8_t* src_y, void NV21ToARGBRow_RVV(const uint8_t* src_y,
const uint8_t* src_vu, const uint8_t* src_vu,
uint8_t* dst_argb, uint8_t* dst_argb,
@ -650,7 +695,9 @@ void NV21ToARGBRow_RVV(const uint8_t* src_y,
dst_argb += vl * 4; dst_argb += vl * 4;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_NV21TORGB24ROW_RVV
void NV21ToRGB24Row_RVV(const uint8_t* src_y, void NV21ToRGB24Row_RVV(const uint8_t* src_y,
const uint8_t* src_vu, const uint8_t* src_vu,
uint8_t* dst_rgb24, uint8_t* dst_rgb24,
@ -676,8 +723,11 @@ void NV21ToRGB24Row_RVV(const uint8_t* src_y,
dst_rgb24 += vl * 3; dst_rgb24 += vl * 3;
} while (w > 0); } while (w > 0);
} }
#endif
// Bilinear filter [VLEN/8]x2 -> [VLEN/8]x1 // Bilinear filter [VLEN/8]x2 -> [VLEN/8]x1
#ifdef HAS_INTERPOLATEROW_RVV
void InterpolateRow_RVV(uint8_t* dst_ptr, void InterpolateRow_RVV(uint8_t* dst_ptr,
const uint8_t* src_ptr, const uint8_t* src_ptr,
ptrdiff_t src_stride, ptrdiff_t src_stride,
@ -734,7 +784,9 @@ void InterpolateRow_RVV(uint8_t* dst_ptr,
dst_ptr += vl; dst_ptr += vl;
} while (dst_w > 0); } while (dst_w > 0);
} }
#endif
#ifdef HAS_SPLITRGBROW_RVV
void SplitRGBRow_RVV(const uint8_t* src_rgb, void SplitRGBRow_RVV(const uint8_t* src_rgb,
uint8_t* dst_r, uint8_t* dst_r,
uint8_t* dst_g, uint8_t* dst_g,
@ -755,7 +807,9 @@ void SplitRGBRow_RVV(const uint8_t* src_rgb,
src_rgb += vl * 3; src_rgb += vl * 3;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_MERGERGBROW_RVV
void MergeRGBRow_RVV(const uint8_t* src_r, void MergeRGBRow_RVV(const uint8_t* src_r,
const uint8_t* src_g, const uint8_t* src_g,
const uint8_t* src_b, const uint8_t* src_b,
@ -775,7 +829,9 @@ void MergeRGBRow_RVV(const uint8_t* src_r,
dst_rgb += vl * 3; dst_rgb += vl * 3;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SPLITARGBROW_RVV
void SplitARGBRow_RVV(const uint8_t* src_argb, void SplitARGBRow_RVV(const uint8_t* src_argb,
uint8_t* dst_r, uint8_t* dst_r,
uint8_t* dst_g, uint8_t* dst_g,
@ -799,7 +855,9 @@ void SplitARGBRow_RVV(const uint8_t* src_argb,
src_argb += vl * 4; src_argb += vl * 4;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_MERGEARGBROW_RVV
void MergeARGBRow_RVV(const uint8_t* src_r, void MergeARGBRow_RVV(const uint8_t* src_r,
const uint8_t* src_g, const uint8_t* src_g,
const uint8_t* src_b, const uint8_t* src_b,
@ -822,7 +880,9 @@ void MergeARGBRow_RVV(const uint8_t* src_r,
dst_argb += vl * 4; dst_argb += vl * 4;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SPLITXRGBROW_RVV
void SplitXRGBRow_RVV(const uint8_t* src_argb, void SplitXRGBRow_RVV(const uint8_t* src_argb,
uint8_t* dst_r, uint8_t* dst_r,
uint8_t* dst_g, uint8_t* dst_g,
@ -843,7 +903,9 @@ void SplitXRGBRow_RVV(const uint8_t* src_argb,
src_argb += vl * 4; src_argb += vl * 4;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_MERGEXRGBROW_RVV
void MergeXRGBRow_RVV(const uint8_t* src_r, void MergeXRGBRow_RVV(const uint8_t* src_r,
const uint8_t* src_g, const uint8_t* src_g,
const uint8_t* src_b, const uint8_t* src_b,
@ -866,7 +928,9 @@ void MergeXRGBRow_RVV(const uint8_t* src_r,
vl = __riscv_vsetvl_e8m2(w); vl = __riscv_vsetvl_e8m2(w);
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SPLITUVROW_RVV
void SplitUVRow_RVV(const uint8_t* src_uv, void SplitUVRow_RVV(const uint8_t* src_uv,
uint8_t* dst_u, uint8_t* dst_u,
uint8_t* dst_v, uint8_t* dst_v,
@ -884,7 +948,9 @@ void SplitUVRow_RVV(const uint8_t* src_uv,
src_uv += 2 * vl; src_uv += 2 * vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_MERGEUVROW_RVV
void MergeUVRow_RVV(const uint8_t* src_u, void MergeUVRow_RVV(const uint8_t* src_u,
const uint8_t* src_v, const uint8_t* src_v,
uint8_t* dst_uv, uint8_t* dst_uv,
@ -902,6 +968,7 @@ void MergeUVRow_RVV(const uint8_t* src_u,
dst_uv += 2 * vl; dst_uv += 2 * vl;
} while (w > 0); } while (w > 0);
} }
#endif
struct RgbConstants { struct RgbConstants {
uint8_t kRGBToY[4]; uint8_t kRGBToY[4];
@ -934,7 +1001,8 @@ static const struct RgbConstants kRawI601Constants = {{66, 129, 25, 0},
0x1080, 0x1080,
0}; 0};
// ARGB expects first 3 values to contain RGB and 4th value is ignored. // ARGB expects first 3 values to contain RGB and 4th value is ignored
#ifdef HAS_ARGBTOYMATRIXROW_RVV
void ARGBToYMatrixRow_RVV(const uint8_t* src_argb, void ARGBToYMatrixRow_RVV(const uint8_t* src_argb,
uint8_t* dst_y, uint8_t* dst_y,
int width, int width,
@ -964,24 +1032,34 @@ void ARGBToYMatrixRow_RVV(const uint8_t* src_argb,
dst_y += vl; dst_y += vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_ARGBTOYROW_RVV
void ARGBToYRow_RVV(const uint8_t* src_argb, uint8_t* dst_y, int width) { void ARGBToYRow_RVV(const uint8_t* src_argb, uint8_t* dst_y, int width) {
ARGBToYMatrixRow_RVV(src_argb, dst_y, width, &kRgb24I601Constants); ARGBToYMatrixRow_RVV(src_argb, dst_y, width, &kRgb24I601Constants);
} }
#endif
#ifdef HAS_ARGBTOYJROW_RVV
void ARGBToYJRow_RVV(const uint8_t* src_argb, uint8_t* dst_yj, int width) { void ARGBToYJRow_RVV(const uint8_t* src_argb, uint8_t* dst_yj, int width) {
ARGBToYMatrixRow_RVV(src_argb, dst_yj, width, &kRgb24JPEGConstants); ARGBToYMatrixRow_RVV(src_argb, dst_yj, width, &kRgb24JPEGConstants);
} }
#endif
#ifdef HAS_ABGRTOYROW_RVV
void ABGRToYRow_RVV(const uint8_t* src_abgr, uint8_t* dst_y, int width) { void ABGRToYRow_RVV(const uint8_t* src_abgr, uint8_t* dst_y, int width) {
ARGBToYMatrixRow_RVV(src_abgr, dst_y, width, &kRawI601Constants); ARGBToYMatrixRow_RVV(src_abgr, dst_y, width, &kRawI601Constants);
} }
#endif
#ifdef HAS_ABGRTOYJROW_RVV
void ABGRToYJRow_RVV(const uint8_t* src_abgr, uint8_t* dst_yj, int width) { void ABGRToYJRow_RVV(const uint8_t* src_abgr, uint8_t* dst_yj, int width) {
ARGBToYMatrixRow_RVV(src_abgr, dst_yj, width, &kRawJPEGConstants); ARGBToYMatrixRow_RVV(src_abgr, dst_yj, width, &kRawJPEGConstants);
} }
#endif
// RGBA expects first value to be A and ignored, then 3 values to contain RGB. // RGBA expects first value to be A and ignored, then 3 values to contain RGB.
#ifdef HAS_RGBATOYMATRIXROW_RVV
void RGBAToYMatrixRow_RVV(const uint8_t* src_rgba, void RGBAToYMatrixRow_RVV(const uint8_t* src_rgba,
uint8_t* dst_y, uint8_t* dst_y,
int width, int width,
@ -1011,19 +1089,27 @@ void RGBAToYMatrixRow_RVV(const uint8_t* src_rgba,
dst_y += vl; dst_y += vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_RGBATOYROW_RVV
void RGBAToYRow_RVV(const uint8_t* src_rgba, uint8_t* dst_y, int width) { void RGBAToYRow_RVV(const uint8_t* src_rgba, uint8_t* dst_y, int width) {
RGBAToYMatrixRow_RVV(src_rgba, dst_y, width, &kRgb24I601Constants); RGBAToYMatrixRow_RVV(src_rgba, dst_y, width, &kRgb24I601Constants);
} }
#endif
#ifdef HAS_RGBATOYJROW_RVV
void RGBAToYJRow_RVV(const uint8_t* src_rgba, uint8_t* dst_yj, int width) { void RGBAToYJRow_RVV(const uint8_t* src_rgba, uint8_t* dst_yj, int width) {
RGBAToYMatrixRow_RVV(src_rgba, dst_yj, width, &kRgb24JPEGConstants); RGBAToYMatrixRow_RVV(src_rgba, dst_yj, width, &kRgb24JPEGConstants);
} }
#endif
#ifdef HAS_BGRATOYROW_RVV
void BGRAToYRow_RVV(const uint8_t* src_bgra, uint8_t* dst_y, int width) { void BGRAToYRow_RVV(const uint8_t* src_bgra, uint8_t* dst_y, int width) {
RGBAToYMatrixRow_RVV(src_bgra, dst_y, width, &kRawI601Constants); RGBAToYMatrixRow_RVV(src_bgra, dst_y, width, &kRawI601Constants);
} }
#endif
#ifdef HAS_RGBTOYMATRIXROW_RVV
void RGBToYMatrixRow_RVV(const uint8_t* src_rgb, void RGBToYMatrixRow_RVV(const uint8_t* src_rgb,
uint8_t* dst_y, uint8_t* dst_y,
int width, int width,
@ -1053,26 +1139,36 @@ void RGBToYMatrixRow_RVV(const uint8_t* src_rgb,
dst_y += vl; dst_y += vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_RGB24TOYJROW_RVV
void RGB24ToYJRow_RVV(const uint8_t* src_rgb24, uint8_t* dst_yj, int width) { void RGB24ToYJRow_RVV(const uint8_t* src_rgb24, uint8_t* dst_yj, int width) {
RGBToYMatrixRow_RVV(src_rgb24, dst_yj, width, &kRgb24JPEGConstants); RGBToYMatrixRow_RVV(src_rgb24, dst_yj, width, &kRgb24JPEGConstants);
} }
#endif
#ifdef HAS_RAWTOYJROW_RVV
void RAWToYJRow_RVV(const uint8_t* src_raw, uint8_t* dst_yj, int width) { void RAWToYJRow_RVV(const uint8_t* src_raw, uint8_t* dst_yj, int width) {
RGBToYMatrixRow_RVV(src_raw, dst_yj, width, &kRawJPEGConstants); RGBToYMatrixRow_RVV(src_raw, dst_yj, width, &kRawJPEGConstants);
} }
#endif
#ifdef HAS_RGB24TOYROW_RVV
void RGB24ToYRow_RVV(const uint8_t* src_rgb24, uint8_t* dst_y, int width) { void RGB24ToYRow_RVV(const uint8_t* src_rgb24, uint8_t* dst_y, int width) {
RGBToYMatrixRow_RVV(src_rgb24, dst_y, width, &kRgb24I601Constants); RGBToYMatrixRow_RVV(src_rgb24, dst_y, width, &kRgb24I601Constants);
} }
#endif
#ifdef HAS_RAWTOYROW_RVV
void RAWToYRow_RVV(const uint8_t* src_raw, uint8_t* dst_y, int width) { void RAWToYRow_RVV(const uint8_t* src_raw, uint8_t* dst_y, int width) {
RGBToYMatrixRow_RVV(src_raw, dst_y, width, &kRawI601Constants); RGBToYMatrixRow_RVV(src_raw, dst_y, width, &kRawI601Constants);
} }
#endif
// Blend src_argb over src_argb1 and store to dst_argb. // Blend src_argb over src_argb1 and store to dst_argb.
// dst_argb may be src_argb or src_argb1. // dst_argb may be src_argb or src_argb1.
// src_argb: RGB values have already been pre-multiplied by the a. // src_argb: RGB values have already been pre-multiplied by the a.
#ifdef HAS_ARGBBLENDROW_RVV
void ARGBBlendRow_RVV(const uint8_t* src_argb, void ARGBBlendRow_RVV(const uint8_t* src_argb,
const uint8_t* src_argb1, const uint8_t* src_argb1,
uint8_t* dst_argb, uint8_t* dst_argb,
@ -1113,7 +1209,9 @@ void ARGBBlendRow_RVV(const uint8_t* src_argb,
dst_argb += 4 * vl; dst_argb += 4 * vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_BLENDPLANEROW_RVV
void BlendPlaneRow_RVV(const uint8_t* src0, void BlendPlaneRow_RVV(const uint8_t* src0,
const uint8_t* src1, const uint8_t* src1,
const uint8_t* alpha, const uint8_t* alpha,
@ -1144,8 +1242,10 @@ void BlendPlaneRow_RVV(const uint8_t* src0,
dst += vl; dst += vl;
} while (w > 0); } while (w > 0);
} }
#endif
// Attenuate: (f * a + 255) >> 8 // Attenuate: (f * a + 255) >> 8
#ifdef HAS_ARGBATTENUATEROW_RVV
void ARGBAttenuateRow_RVV(const uint8_t* src_argb, void ARGBAttenuateRow_RVV(const uint8_t* src_argb,
uint8_t* dst_argb, uint8_t* dst_argb,
int width) { int width) {
@ -1173,7 +1273,9 @@ void ARGBAttenuateRow_RVV(const uint8_t* src_argb,
dst_argb += vl * 4; dst_argb += vl * 4;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_ARGBEXTRACTALPHAROW_RVV
void ARGBExtractAlphaRow_RVV(const uint8_t* src_argb, void ARGBExtractAlphaRow_RVV(const uint8_t* src_argb,
uint8_t* dst_a, uint8_t* dst_a,
int width) { int width) {
@ -1188,7 +1290,9 @@ void ARGBExtractAlphaRow_RVV(const uint8_t* src_argb,
dst_a += vl; dst_a += vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_ARGBCOPYYTOALPHAROW_RVV
void ARGBCopyYToAlphaRow_RVV(const uint8_t* src, uint8_t* dst, int width) { void ARGBCopyYToAlphaRow_RVV(const uint8_t* src, uint8_t* dst, int width) {
size_t w = (size_t)width; size_t w = (size_t)width;
const ptrdiff_t dst_stride = 4; const ptrdiff_t dst_stride = 4;
@ -1202,6 +1306,7 @@ void ARGBCopyYToAlphaRow_RVV(const uint8_t* src, uint8_t* dst, int width) {
dst += vl * dst_stride; dst += vl * dst_stride;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef __cplusplus #ifdef __cplusplus
} // extern "C" } // extern "C"

52
source/scale_rvv.cc
View File

@ -28,6 +28,7 @@ namespace libyuv {
extern "C" { extern "C" {
#endif #endif
#ifdef HAS_SCALEADDROW_RVV
void ScaleAddRow_RVV(const uint8_t* src_ptr, uint16_t* dst_ptr, int src_width) { void ScaleAddRow_RVV(const uint8_t* src_ptr, uint16_t* dst_ptr, int src_width) {
size_t w = (size_t)src_width; size_t w = (size_t)src_width;
do { do {
@ -42,7 +43,9 @@ void ScaleAddRow_RVV(const uint8_t* src_ptr, uint16_t* dst_ptr, int src_width) {
dst_ptr += vl; dst_ptr += vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEARGBROWDOWN2_RVV
void ScaleARGBRowDown2_RVV(const uint8_t* src_argb, void ScaleARGBRowDown2_RVV(const uint8_t* src_argb,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_argb, uint8_t* dst_argb,
@ -61,7 +64,9 @@ void ScaleARGBRowDown2_RVV(const uint8_t* src_argb,
dst += vl; dst += vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEARGBROWDOWN2LINEAR_RVV
void ScaleARGBRowDown2Linear_RVV(const uint8_t* src_argb, void ScaleARGBRowDown2Linear_RVV(const uint8_t* src_argb,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_argb, uint8_t* dst_argb,
@ -87,7 +92,9 @@ void ScaleARGBRowDown2Linear_RVV(const uint8_t* src_argb,
dst_argb += vl * 4; dst_argb += vl * 4;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEARGBROWDOWN2BOX_RVV
void ScaleARGBRowDown2Box_RVV(const uint8_t* src_argb, void ScaleARGBRowDown2Box_RVV(const uint8_t* src_argb,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_argb, uint8_t* dst_argb,
@ -121,6 +128,7 @@ void ScaleARGBRowDown2Box_RVV(const uint8_t* src_argb,
dst_argb += vl * 4; dst_argb += vl * 4;
} while (w > 0); } while (w > 0);
} }
#endif
void ScaleARGBRowDownEven_RVV(const uint8_t* src_argb, void ScaleARGBRowDownEven_RVV(const uint8_t* src_argb,
ptrdiff_t src_stride, ptrdiff_t src_stride,
@ -141,6 +149,7 @@ void ScaleARGBRowDownEven_RVV(const uint8_t* src_argb,
} while (w > 0); } while (w > 0);
} }
#ifdef HAS_SCALEARGBROWDOWNEVENBOX_RVV
void ScaleARGBRowDownEvenBox_RVV(const uint8_t* src_argb, void ScaleARGBRowDownEvenBox_RVV(const uint8_t* src_argb,
ptrdiff_t src_stride, ptrdiff_t src_stride,
int src_stepx, int src_stepx,
@ -178,7 +187,9 @@ void ScaleARGBRowDownEvenBox_RVV(const uint8_t* src_argb,
dst_argb += vl * 4; dst_argb += vl * 4;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEROWDOWN2_RVV
void ScaleRowDown2_RVV(const uint8_t* src_ptr, void ScaleRowDown2_RVV(const uint8_t* src_ptr,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst, uint8_t* dst,
@ -196,7 +207,9 @@ void ScaleRowDown2_RVV(const uint8_t* src_ptr,
dst += vl; dst += vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEROWDOWN2LINEAR_RVV
void ScaleRowDown2Linear_RVV(const uint8_t* src_ptr, void ScaleRowDown2Linear_RVV(const uint8_t* src_ptr,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst, uint8_t* dst,
@ -218,7 +231,9 @@ void ScaleRowDown2Linear_RVV(const uint8_t* src_ptr,
dst += vl; dst += vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEROWDOWN2BOX_RVV
void ScaleRowDown2Box_RVV(const uint8_t* src_ptr, void ScaleRowDown2Box_RVV(const uint8_t* src_ptr,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst, uint8_t* dst,
@ -248,7 +263,9 @@ void ScaleRowDown2Box_RVV(const uint8_t* src_ptr,
dst += vl; dst += vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEROWDOWN4_RVV
void ScaleRowDown4_RVV(const uint8_t* src_ptr, void ScaleRowDown4_RVV(const uint8_t* src_ptr,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_ptr, uint8_t* dst_ptr,
@ -265,7 +282,9 @@ void ScaleRowDown4_RVV(const uint8_t* src_ptr,
dst_ptr += vl; dst_ptr += vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEROWDOWN4BOX_RVV
void ScaleRowDown4Box_RVV(const uint8_t* src_ptr, void ScaleRowDown4Box_RVV(const uint8_t* src_ptr,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_ptr, uint8_t* dst_ptr,
@ -326,7 +345,9 @@ void ScaleRowDown4Box_RVV(const uint8_t* src_ptr,
dst_ptr += vl; dst_ptr += vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEROWDOWN34_RVV
void ScaleRowDown34_RVV(const uint8_t* src_ptr, void ScaleRowDown34_RVV(const uint8_t* src_ptr,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_ptr, uint8_t* dst_ptr,
@ -342,7 +363,9 @@ void ScaleRowDown34_RVV(const uint8_t* src_ptr,
dst_ptr += 3 * vl; dst_ptr += 3 * vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEROWDOWN34_0_BOX_RVV
void ScaleRowDown34_0_Box_RVV(const uint8_t* src_ptr, void ScaleRowDown34_0_Box_RVV(const uint8_t* src_ptr,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_ptr, uint8_t* dst_ptr,
@ -408,7 +431,9 @@ void ScaleRowDown34_0_Box_RVV(const uint8_t* src_ptr,
dst_ptr += 3 * vl; dst_ptr += 3 * vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEROWDOWN34_1_BOX_RVV
void ScaleRowDown34_1_Box_RVV(const uint8_t* src_ptr, void ScaleRowDown34_1_Box_RVV(const uint8_t* src_ptr,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_ptr, uint8_t* dst_ptr,
@ -462,7 +487,9 @@ void ScaleRowDown34_1_Box_RVV(const uint8_t* src_ptr,
dst_ptr += 3 * vl; dst_ptr += 3 * vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEROWDOWN38_RVV
void ScaleRowDown38_RVV(const uint8_t* src_ptr, void ScaleRowDown38_RVV(const uint8_t* src_ptr,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_ptr, uint8_t* dst_ptr,
@ -481,7 +508,9 @@ void ScaleRowDown38_RVV(const uint8_t* src_ptr,
dst_ptr += 3 * vl; dst_ptr += 3 * vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEROWDOWN38_2_BOX_RVV
void ScaleRowDown38_2_Box_RVV(const uint8_t* src_ptr, void ScaleRowDown38_2_Box_RVV(const uint8_t* src_ptr,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_ptr, uint8_t* dst_ptr,
@ -537,7 +566,9 @@ void ScaleRowDown38_2_Box_RVV(const uint8_t* src_ptr,
dst_ptr += 3 * vl; dst_ptr += 3 * vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEROWDOWN38_3_BOX_RVV
void ScaleRowDown38_3_Box_RVV(const uint8_t* src_ptr, void ScaleRowDown38_3_Box_RVV(const uint8_t* src_ptr,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_ptr, uint8_t* dst_ptr,
@ -608,10 +639,13 @@ void ScaleRowDown38_3_Box_RVV(const uint8_t* src_ptr,
dst_ptr += 3 * vl; dst_ptr += 3 * vl;
} while (w > 0); } while (w > 0);
} }
#endif
// ScaleUVRowUp2_(Bi)linear_RVV function is equal to other platforms' // ScaleUVRowUp2_(Bi)linear_RVV function is equal to other platforms'
// ScaleRowUp2_(Bi)linear_Any_XXX. We process entire row in this function. Other // ScaleRowUp2_(Bi)linear_Any_XXX. We process entire row in this function. Other
// platforms only implement non-edge part of image and process edge with scalar. // platforms only implement non-edge part of image and process edge with scalar.
#ifdef HAS_SCALEROWUP2_LINEAR_RVV
void ScaleRowUp2_Linear_RVV(const uint8_t* src_ptr, void ScaleRowUp2_Linear_RVV(const uint8_t* src_ptr,
uint8_t* dst_ptr, uint8_t* dst_ptr,
int dst_width) { int dst_width) {
@ -645,7 +679,9 @@ void ScaleRowUp2_Linear_RVV(const uint8_t* src_ptr,
} }
dst_ptr[dst_width - 1] = src_ptr[(dst_width - 1) / 2]; dst_ptr[dst_width - 1] = src_ptr[(dst_width - 1) / 2];
} }
#endif
#ifdef HAS_SCALEROWUP2_BILINEAR_RVV
void ScaleRowUp2_Bilinear_RVV(const uint8_t* src_ptr, void ScaleRowUp2_Bilinear_RVV(const uint8_t* src_ptr,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_ptr, uint8_t* dst_ptr,
@ -715,7 +751,9 @@ void ScaleRowUp2_Bilinear_RVV(const uint8_t* src_ptr,
e[dst_width - 1] = e[dst_width - 1] =
(s[(dst_width - 1) / 2] + 3 * t[(dst_width - 1) / 2] + 2) >> 2; (s[(dst_width - 1) / 2] + 3 * t[(dst_width - 1) / 2] + 2) >> 2;
} }
#endif
#ifdef HAS_SCALEUVROWDOWN2_RVV
void ScaleUVRowDown2_RVV(const uint8_t* src_uv, void ScaleUVRowDown2_RVV(const uint8_t* src_uv,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_uv, uint8_t* dst_uv,
@ -734,7 +772,9 @@ void ScaleUVRowDown2_RVV(const uint8_t* src_uv,
dst += vl; dst += vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEUVROWDOWN2LINEAR_RVV
void ScaleUVRowDown2Linear_RVV(const uint8_t* src_uv, void ScaleUVRowDown2Linear_RVV(const uint8_t* src_uv,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_uv, uint8_t* dst_uv,
@ -760,7 +800,9 @@ void ScaleUVRowDown2Linear_RVV(const uint8_t* src_uv,
dst_uv += vl * 2; dst_uv += vl * 2;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEUVROWDOWN2BOX_RVV
void ScaleUVRowDown2Box_RVV(const uint8_t* src_uv, void ScaleUVRowDown2Box_RVV(const uint8_t* src_uv,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_uv, uint8_t* dst_uv,
@ -802,7 +844,9 @@ void ScaleUVRowDown2Box_RVV(const uint8_t* src_uv,
src_uv_row1 += 4 * vl; src_uv_row1 += 4 * vl;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEUVROWDOWN4_RVV
void ScaleUVRowDown4_RVV(const uint8_t* src_uv, void ScaleUVRowDown4_RVV(const uint8_t* src_uv,
ptrdiff_t src_stride, ptrdiff_t src_stride,
int src_stepx, int src_stepx,
@ -827,7 +871,9 @@ void ScaleUVRowDown4_RVV(const uint8_t* src_uv,
dst_uv += vl / 4; dst_uv += vl / 4;
} while (w > 0); } while (w > 0);
} }
#endif
#ifdef HAS_SCALEUVROWDOWNEVEN_RVV
void ScaleUVRowDownEven_RVV(const uint8_t* src_uv, void ScaleUVRowDownEven_RVV(const uint8_t* src_uv,
ptrdiff_t src_stride, ptrdiff_t src_stride,
int src_stepx, int src_stepx,
@ -847,11 +893,14 @@ void ScaleUVRowDownEven_RVV(const uint8_t* src_uv,
dst += vl; dst += vl;
} while (w > 0); } while (w > 0);
} }
#endif
// ScaleUVRowUp2_(Bi)linear_RVV function is equal to other platforms' // ScaleUVRowUp2_(Bi)linear_RVV function is equal to other platforms'
// ScaleUVRowUp2_(Bi)linear_Any_XXX. We process entire row in this function. // ScaleUVRowUp2_(Bi)linear_Any_XXX. We process entire row in this function.
// Other platforms only implement non-edge part of image and process edge with // Other platforms only implement non-edge part of image and process edge with
// scalar. // scalar.
#ifdef HAS_SCALEUVROWUP2_LINEAR_RVV
void ScaleUVRowUp2_Linear_RVV(const uint8_t* src_ptr, void ScaleUVRowUp2_Linear_RVV(const uint8_t* src_ptr,
uint8_t* dst_ptr, uint8_t* dst_ptr,
int dst_width) { int dst_width) {
@ -891,7 +940,9 @@ void ScaleUVRowUp2_Linear_RVV(const uint8_t* src_ptr,
dst_ptr[2 * dst_width - 2] = src_ptr[((dst_width + 1) & ~1) - 2]; dst_ptr[2 * dst_width - 2] = src_ptr[((dst_width + 1) & ~1) - 2];
dst_ptr[2 * dst_width - 1] = src_ptr[((dst_width + 1) & ~1) - 1]; dst_ptr[2 * dst_width - 1] = src_ptr[((dst_width + 1) & ~1) - 1];
} }
#endif
#ifdef HAS_SCALEUVROWUP2_BILINEAR_RVV
void ScaleUVRowUp2_Bilinear_RVV(const uint8_t* src_ptr, void ScaleUVRowUp2_Bilinear_RVV(const uint8_t* src_ptr,
ptrdiff_t src_stride, ptrdiff_t src_stride,
uint8_t* dst_ptr, uint8_t* dst_ptr,
@ -976,6 +1027,7 @@ void ScaleUVRowUp2_Bilinear_RVV(const uint8_t* src_ptr,
(s[((dst_width + 1) & ~1) - 1] + 3 * t[((dst_width + 1) & ~1) - 1] + 2) >> (s[((dst_width + 1) & ~1) - 1] + 3 * t[((dst_width + 1) & ~1) - 1] + 2) >>
2; 2;
} }
#endif
#ifdef __cplusplus #ifdef __cplusplus
} // extern "C" } // extern "C"