coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2025-12-06 08:46:47 +08:00
Code Issues Packages Projects Releases Wiki Activity

Deprecate MIPS and MSA support.

Browse Source 
- Remove *_msa.cc source files
- Update build files
- Update header references, planar ifdefs for row functions
- Update documentation on supported platforms
- Version bumped to 1921
- clang-format applied

Bug: 434383432
Change-Id: I072d6aac4956f0ed668e64614ac8557612171f76
Reviewed-on: https://chromium-review.googlesource.com/c/libyuv/libyuv/+/7045953
Reviewed-by: Justin Green <greenjustin@google.com>
This commit is contained in:
Frank Barchard 2025-10-16 11:24:52 -07:00
parent 28ebe5d0b0
commit 2b4453d46f
42 changed files with 41 additions and 8131 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
Android.bp
View File

@ -34,7 +34,6 @@ cc_library {
"source/compare.cc",
"source/compare_common.cc",
"source/compare_gcc.cc",
"source/compare_msa.cc",
"source/compare_neon.cc",
"source/compare_neon64.cc",
"source/convert.cc",
@ -53,13 +52,11 @@ cc_library {
"source/rotate_argb.cc",
"source/rotate_common.cc",
"source/rotate_gcc.cc",
"source/rotate_msa.cc",
"source/rotate_neon.cc",
"source/rotate_neon64.cc",
"source/row_any.cc",
"source/row_common.cc",
"source/row_gcc.cc",
"source/row_msa.cc",
"source/row_neon.cc",
"source/row_neon64.cc",
"source/row_rvv.cc",
@ -68,7 +65,6 @@ cc_library {
"source/scale_argb.cc",
"source/scale_common.cc",
"source/scale_gcc.cc",
"source/scale_msa.cc",
"source/scale_neon.cc",
"source/scale_neon64.cc",
"source/scale_rgb.cc",

4
Android.mk
View File

@ -9,7 +9,6 @@ LOCAL_SRC_FILES := \
source/compare.cc \
source/compare_common.cc \
source/compare_gcc.cc \
source/compare_msa.cc \
source/compare_neon.cc \
source/compare_neon64.cc \
source/compare_win.cc \
@ -26,14 +25,12 @@ LOCAL_SRC_FILES := \
source/rotate_argb.cc \
source/rotate_common.cc \
source/rotate_gcc.cc \
source/rotate_msa.cc \
source/rotate_neon.cc \
source/rotate_neon64.cc \
source/rotate_win.cc \
source/row_any.cc \
source/row_common.cc \
source/row_gcc.cc \
source/row_msa.cc \
source/row_neon.cc \
source/row_neon64.cc \
source/row_win.cc \
@ -42,7 +39,6 @@ LOCAL_SRC_FILES := \
source/scale_argb.cc \
source/scale_common.cc \
source/scale_gcc.cc \
source/scale_msa.cc \
source/scale_neon.cc \
source/scale_neon64.cc \
source/scale_rgb.cc \

25
BUILD.gn
View File

@ -23,7 +23,7 @@ declare_args() {
config("libyuv_config") {
include_dirs = [ "include" ]
if (is_android) {
if (target_cpu == "arm" || target_cpu == "x86" || target_cpu == "mipsel") {
if (target_cpu == "arm" || target_cpu == "x86") {
ldflags = [ "-Wl,--dynamic-linker,/system/bin/linker" ]
} else {
ldflags = [ "-Wl,--dynamic-linker,/system/bin/linker64" ]
@ -92,10 +92,6 @@ group("libyuv") {
deps += [ ":libyuv_sme" ]
}
if (libyuv_use_msa) {
deps += [ ":libyuv_msa" ]
}
if (libyuv_use_lsx) {
deps += [ ":libyuv_lsx" ]
}
@ -127,7 +123,6 @@ static_library("libyuv_internal") {
"include/libyuv/convert_from_argb.h",
"include/libyuv/cpu_id.h",
"include/libyuv/loongson_intrinsics.h",
"include/libyuv/macros_msa.h",
"include/libyuv/mjpeg_decoder.h",
"include/libyuv/planar_functions.h",
"include/libyuv/rotate.h",
@ -255,6 +250,7 @@ if (libyuv_use_sve) {
# The -march arguments below are incompatible with libc++ modules.
use_libcxx_modules = false
# SVE2 is an Armv9-A feature.
cflags = [ "-march=armv9-a+i8mm+sve2" ]
}
@ -274,27 +270,12 @@ if (libyuv_use_sme) {
# The -march arguments below are incompatible with libc++ modules.
use_libcxx_modules = false
# SME is an Armv9-A feature.
cflags = [ "-march=armv9-a+i8mm+sme" ]
}
}
if (libyuv_use_msa) {
static_library("libyuv_msa") {
sources = [
# MSA Source Files
"source/compare_msa.cc",
"source/rotate_msa.cc",
"source/row_msa.cc",
"source/scale_msa.cc",
]
deps = [ ":libyuv_internal" ]
public_configs = [ ":libyuv_config" ]
}
}
if (libyuv_use_lsx) {
static_library("libyuv_lsx") {
sources = [

4
CMakeLists.txt
View File

@ -23,7 +23,6 @@ set ( ly_common_source_files
${ly_src_dir}/compare.cc
${ly_src_dir}/compare_common.cc
${ly_src_dir}/compare_gcc.cc
${ly_src_dir}/compare_msa.cc
${ly_src_dir}/compare_win.cc
${ly_src_dir}/convert_argb.cc
${ly_src_dir}/convert.cc
@ -42,14 +41,12 @@ set ( ly_common_source_files
${ly_src_dir}/rotate_common.cc
${ly_src_dir}/rotate_gcc.cc
${ly_src_dir}/rotate_lsx.cc
${ly_src_dir}/rotate_msa.cc
${ly_src_dir}/rotate_win.cc
${ly_src_dir}/row_any.cc
${ly_src_dir}/row_common.cc
${ly_src_dir}/row_gcc.cc
${ly_src_dir}/row_lasx.cc
${ly_src_dir}/row_lsx.cc
${ly_src_dir}/row_msa.cc
${ly_src_dir}/row_rvv.cc
${ly_src_dir}/row_win.cc
${ly_src_dir}/scale_any.cc
@ -58,7 +55,6 @@ set ( ly_common_source_files
${ly_src_dir}/scale_common.cc
${ly_src_dir}/scale_gcc.cc
${ly_src_dir}/scale_lsx.cc
${ly_src_dir}/scale_msa.cc
${ly_src_dir}/scale_rgb.cc
${ly_src_dir}/scale_rvv.cc
${ly_src_dir}/scale_uv.cc

7
DEPS
View File

@ -1532,13 +1532,6 @@ hooks = [
'action': ['python3', 'src/build/linux/sysroot_scripts/install-sysroot.py',
'--arch=x86'],
},
{
'name': 'sysroot_mips',
'pattern': '.',
'condition': 'checkout_linux and checkout_mips',
'action': ['python3', 'src/build/linux/sysroot_scripts/install-sysroot.py',
'--arch=mips'],
},
{
'name': 'sysroot_x64',
'pattern': '.',

2
README.chromium
View File

@ -1,6 +1,6 @@
Name: libyuv
URL: https://chromium.googlesource.com/libyuv/libyuv/
Version: 1920
Version: 1921
License: BSD-3-Clause
License File: LICENSE
Shipped: yes

16
docs/deprecated_builds.md
View File

@ -141,12 +141,6 @@ ia32
GYP_DEFINES="OS=android target_arch=ia32 android_full_debug=1" GYP_CROSSCOMPILE=1 ./gyp_libyuv
ninja -j7 -C out/Debug yuv_unittest_apk
mipsel
GYP_DEFINES="OS=android target_arch=mipsel" GYP_CROSSCOMPILE=1 ./gyp_libyuv
ninja -j7 -C out/Debug yuv_unittest_apk
ninja -j7 -C out/Release yuv_unittest_apk
arm32 disassembly:
llvm-objdump -d out/Release/obj/source/libyuv.row_neon.o
@ -179,16 +173,6 @@ Running test with C code:
gn gen out/Official "--args=is_debug=false is_official_build=true is_chrome_branded=true"
ninja -C out/Official
#### Building mips with GN
mipsel
gn gen out/Default "--args=is_debug=false target_cpu=\"mipsel\" target_os = \"android\" mips_arch_variant = \"r6\" mips_use_msa = true is_component_build = true is_clang = false"
ninja -C out/Default
mips64el
gn gen out/Default "--args=is_debug=false target_cpu=\"mips64el\" target_os = \"android\" mips_arch_variant = \"r6\" mips_use_msa = true is_component_build = true is_clang = false"
ninja -C out/Default
### Linux
GYP_DEFINES="target_arch=x64" ./gyp_libyuv

4
docs/environment_variables.md
View File

@ -43,10 +43,6 @@ By default the cpu is detected and the most advanced form of SIMD is used. But
LIBYUV_DISABLE_SVE2
LIBYUV_DISABLE_SME
## MIPS CPUs
LIBYUV_DISABLE_MSA
## LOONGARCH CPUs
LIBYUV_DISABLE_LSX

14
docs/getting_started.md
View File

@ -116,13 +116,6 @@ ia32
ninja -v -C out/Debug libyuv_unittest
ninja -v -C out/Release libyuv_unittest
mips
gn gen out/Release "--args=is_debug=false target_os=\"android\" target_cpu=\"mips64el\" mips_arch_variant=\"r6\" mips_use_msa=true is_component_build=true"
gn gen out/Debug "--args=is_debug=true target_os=\"android\" target_cpu=\"mips64el\" mips_arch_variant=\"r6\" mips_use_msa=true is_component_build=true"
ninja -v -C out/Debug libyuv_unittest
ninja -v -C out/Release libyuv_unittest
arm disassembly:
llvm-objdump -d ./out/Release/obj/libyuv/row_common.o >row_common.txt
@ -162,13 +155,6 @@ Running test with C code:
ninja -v -C out/Debug libyuv_unittest
ninja -v -C out/Release libyuv_unittest
### MIPS Linux
gn gen out/Release "--args=is_debug=false target_os=\"linux\" target_cpu=\"mips64el\" mips_arch_variant=\"loongson3\" is_component_build=false use_sysroot=false use_gold=false"
gn gen out/Debug "--args=is_debug=true target_os=\"linux\" target_cpu=\"mips64el\" mips_arch_variant=\"loongson3\" is_component_build=false use_sysroot=false use_gold=false"
ninja -v -C out/Debug libyuv_unittest
ninja -v -C out/Release libyuv_unittest
## Building the Library with make
### Linux

11
include/libyuv/compare_row.h
View File

@ -65,11 +65,6 @@ extern "C" {
#define HAS_SUMSQUAREERROR_NEON_DOTPROD
#endif
#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)
#define HAS_HAMMINGDISTANCE_MSA
#define HAS_SUMSQUAREERROR_MSA
#endif
uint32_t HammingDistance_C(const uint8_t* src_a,
const uint8_t* src_b,
int count);
@ -88,9 +83,6 @@ uint32_t HammingDistance_NEON(const uint8_t* src_a,
uint32_t HammingDistance_NEON_DotProd(const uint8_t* src_a,
const uint8_t* src_b,
int count);
uint32_t HammingDistance_MSA(const uint8_t* src_a,
const uint8_t* src_b,
int count);
uint32_t SumSquareError_C(const uint8_t* src_a,
const uint8_t* src_b,
int count);
@ -106,9 +98,6 @@ uint32_t SumSquareError_NEON(const uint8_t* src_a,
uint32_t SumSquareError_NEON_DotProd(const uint8_t* src_a,
const uint8_t* src_b,
int count);
uint32_t SumSquareError_MSA(const uint8_t* src_a,
const uint8_t* src_b,
int count);
uint32_t HashDjb2_C(const uint8_t* src, int count, uint32_t seed);
uint32_t HashDjb2_SSE41(const uint8_t* src, int count, uint32_t seed);

6
include/libyuv/cpu_id.h
View File

@ -60,10 +60,6 @@ static const int kCpuHasAVXVNNI = 0x4000000;
static const int kCpuHasAVXVNNIINT8 = 0x8000000;
static const int kCpuHasAMXINT8 = 0x10000000;
// These flags are only valid on MIPS processors.
static const int kCpuHasMIPS = 0x10;
static const int kCpuHasMSA = 0x100;
// These flags are only valid on LOONGARCH processors.
static const int kCpuHasLOONGARCH = 0x20;
static const int kCpuHasLSX = 0x100;
@ -91,8 +87,6 @@ static __inline int TestCpuFlag(int test_flag) {
LIBYUV_API
int ArmCpuCaps(const char* cpuinfo_name);
LIBYUV_API
int MipsCpuCaps(const char* cpuinfo_name);
LIBYUV_API
int RiscvCpuCaps(const char* cpuinfo_name);
#ifdef __linux__

244
include/libyuv/macros_msa.h
View File

@ -1,244 +0,0 @@
/*
* Copyright 2016 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.
*/
#ifndef INCLUDE_LIBYUV_MACROS_MSA_H_
#define INCLUDE_LIBYUV_MACROS_MSA_H_
#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)
#include <msa.h>
#include <stdint.h>
#if (__mips_isa_rev >= 6)
#define LW(psrc) \
({ \
const uint8_t* psrc_lw_m = (const uint8_t*)(psrc); \
uint32_t val_m; \
asm("lw %[val_m], %[psrc_lw_m] \n" \
: [val_m] "=r"(val_m) \
: [psrc_lw_m] "m"(*psrc_lw_m)); \
val_m; \
})
#if (__mips == 64)
#define LD(psrc) \
({ \
const uint8_t* psrc_ld_m = (const uint8_t*)(psrc); \
uint64_t val_m = 0; \
asm("ld %[val_m], %[psrc_ld_m] \n" \
: [val_m] "=r"(val_m) \
: [psrc_ld_m] "m"(*psrc_ld_m)); \
val_m; \
})
#else // !(__mips == 64)
#define LD(psrc) \
({ \
const uint8_t* psrc_ld_m = (const uint8_t*)(psrc); \
uint32_t val0_m, val1_m; \
uint64_t val_m = 0; \
val0_m = LW(psrc_ld_m); \
val1_m = LW(psrc_ld_m + 4); \
val_m = (uint64_t)(val1_m); /* NOLINT */ \
val_m = (uint64_t)((val_m << 32) & 0xFFFFFFFF00000000); /* NOLINT */ \
val_m = (uint64_t)(val_m | (uint64_t)val0_m); /* NOLINT */ \
val_m; \
})
#endif // (__mips == 64)
#define SW(val, pdst) \
({ \
uint8_t* pdst_sw_m = (uint8_t*)(pdst); /* NOLINT */ \
uint32_t val_m = (val); \
asm("sw %[val_m], %[pdst_sw_m] \n" \
: [pdst_sw_m] "=m"(*pdst_sw_m) \
: [val_m] "r"(val_m)); \
})
#if (__mips == 64)
#define SD(val, pdst) \
({ \
uint8_t* pdst_sd_m = (uint8_t*)(pdst); /* NOLINT */ \
uint64_t val_m = (val); \
asm("sd %[val_m], %[pdst_sd_m] \n" \
: [pdst_sd_m] "=m"(*pdst_sd_m) \
: [val_m] "r"(val_m)); \
})
#else // !(__mips == 64)
#define SD(val, pdst) \
({ \
uint8_t* pdst_sd_m = (uint8_t*)(pdst); /* NOLINT */ \
uint32_t val0_m, val1_m; \
val0_m = (uint32_t)((val)&0x00000000FFFFFFFF); \
val1_m = (uint32_t)(((val) >> 32) & 0x00000000FFFFFFFF); \
SW(val0_m, pdst_sd_m); \
SW(val1_m, pdst_sd_m + 4); \
})
#endif // !(__mips == 64)
#else // !(__mips_isa_rev >= 6)
#define LW(psrc) \
({ \
uint8_t* psrc_lw_m = (uint8_t*)(psrc); \
uint32_t val_lw_m; \
\
asm("lwr %[val_lw_m], 0(%[psrc_lw_m]) \n\t" \
"lwl %[val_lw_m], 3(%[psrc_lw_m]) \n\t" \
\
: [val_lw_m] "=&r"(val_lw_m) \
: [psrc_lw_m] "r"(psrc_lw_m)); \
\
val_lw_m; \
})
#if (__mips == 64)
#define LD(psrc) \
({ \
uint8_t* psrc_ld_m = (uint8_t*)(psrc); \
uint64_t val_ld_m = 0; \
\
asm("ldr %[val_ld_m], 0(%[psrc_ld_m]) \n\t" \
"ldl %[val_ld_m], 7(%[psrc_ld_m]) \n\t" \
\
: [val_ld_m] "=&r"(val_ld_m) \
: [psrc_ld_m] "r"(psrc_ld_m)); \
\
val_ld_m; \
})
#else // !(__mips == 64)
#define LD(psrc) \
({ \
const uint8_t* psrc_ld_m = (const uint8_t*)(psrc); \
uint32_t val0_m, val1_m; \
uint64_t val_m = 0; \
val0_m = LW(psrc_ld_m); \
val1_m = LW(psrc_ld_m + 4); \
val_m = (uint64_t)(val1_m); /* NOLINT */ \
val_m = (uint64_t)((val_m << 32) & 0xFFFFFFFF00000000); /* NOLINT */ \
val_m = (uint64_t)(val_m | (uint64_t)val0_m); /* NOLINT */ \
val_m; \
})
#endif // (__mips == 64)
#define SW(val, pdst) \
({ \
uint8_t* pdst_sw_m = (uint8_t*)(pdst); /* NOLINT */ \
uint32_t val_m = (val); \
asm("usw %[val_m], %[pdst_sw_m] \n" \
: [pdst_sw_m] "=m"(*pdst_sw_m) \
: [val_m] "r"(val_m)); \
})
#define SD(val, pdst) \
({ \
uint8_t* pdst_sd_m = (uint8_t*)(pdst); /* NOLINT */ \
uint32_t val0_m, val1_m; \
val0_m = (uint32_t)((val)&0x00000000FFFFFFFF); \
val1_m = (uint32_t)(((val) >> 32) & 0x00000000FFFFFFFF); \
SW(val0_m, pdst_sd_m); \
SW(val1_m, pdst_sd_m + 4); \
})
#endif // (__mips_isa_rev >= 6)
// TODO(fbarchard): Consider removing __VAR_ARGS versions.
#define LD_B(RTYPE, psrc) *((RTYPE*)(psrc)) /* NOLINT */
#define LD_UB(...) LD_B(const v16u8, __VA_ARGS__)
#define LD_H(RTYPE, psrc) *((RTYPE*)(psrc)) /* NOLINT */
#define LD_UH(...) LD_H(const v8u16, __VA_ARGS__)
#define ST_B(RTYPE, in, pdst) *((RTYPE*)(pdst)) = (in) /* NOLINT */
#define ST_UB(...) ST_B(v16u8, __VA_ARGS__)
#define ST_H(RTYPE, in, pdst) *((RTYPE*)(pdst)) = (in) /* NOLINT */
#define ST_UH(...) ST_H(v8u16, __VA_ARGS__)
/* Description : Load two vectors with 16 'byte' sized elements
Arguments : Inputs - psrc, stride
Outputs - out0, out1
Return Type - as per RTYPE
Details : Load 16 byte elements in 'out0' from (psrc)
Load 16 byte elements in 'out1' from (psrc + stride)
*/
#define LD_B2(RTYPE, psrc, stride, out0, out1) \
{ \
out0 = LD_B(RTYPE, (psrc)); \
out1 = LD_B(RTYPE, (psrc) + stride); \
}
#define LD_UB2(...) LD_B2(const v16u8, __VA_ARGS__)
#define LD_B4(RTYPE, psrc, stride, out0, out1, out2, out3) \
{ \
LD_B2(RTYPE, (psrc), stride, out0, out1); \
LD_B2(RTYPE, (psrc) + 2 * stride, stride, out2, out3); \
}
#define LD_UB4(...) LD_B4(const v16u8, __VA_ARGS__)
/* Description : Store two vectors with stride each having 16 'byte' sized
elements
Arguments : Inputs - in0, in1, pdst, stride
Details : Store 16 byte elements from 'in0' to (pdst)
Store 16 byte elements from 'in1' to (pdst + stride)
*/
#define ST_B2(RTYPE, in0, in1, pdst, stride) \
{ \
ST_B(RTYPE, in0, (pdst)); \
ST_B(RTYPE, in1, (pdst) + stride); \
}
#define ST_UB2(...) ST_B2(v16u8, __VA_ARGS__)
#define ST_B4(RTYPE, in0, in1, in2, in3, pdst, stride) \
{ \
ST_B2(RTYPE, in0, in1, (pdst), stride); \
ST_B2(RTYPE, in2, in3, (pdst) + 2 * stride, stride); \
}
#define ST_UB4(...) ST_B4(v16u8, __VA_ARGS__)
/* Description : Store vectors of 8 halfword elements with stride
Arguments : Inputs - in0, in1, pdst, stride
Details : Store 8 halfword elements from 'in0' to (pdst)
Store 8 halfword elements from 'in1' to (pdst + stride)
*/
#define ST_H2(RTYPE, in0, in1, pdst, stride) \
{ \
ST_H(RTYPE, in0, (pdst)); \
ST_H(RTYPE, in1, (pdst) + stride); \
}
#define ST_UH2(...) ST_H2(v8u16, __VA_ARGS__)
// TODO(fbarchard): Consider using __msa_vshf_b and __msa_ilvr_b directly.
/* Description : Shuffle byte vector elements as per mask vector
Arguments : Inputs - in0, in1, in2, in3, mask0, mask1
Outputs - out0, out1
Return Type - as per RTYPE
Details : Byte elements from 'in0' & 'in1' are copied selectively to
'out0' as per control vector 'mask0'
*/
#define VSHF_B2(RTYPE, in0, in1, in2, in3, mask0, mask1, out0, out1) \
{ \
out0 = (RTYPE)__msa_vshf_b((v16i8)mask0, (v16i8)in1, (v16i8)in0); \
out1 = (RTYPE)__msa_vshf_b((v16i8)mask1, (v16i8)in3, (v16i8)in2); \
}
#define VSHF_B2_UB(...) VSHF_B2(v16u8, __VA_ARGS__)
/* Description : Interleave both left and right half of input vectors
Arguments : Inputs - in0, in1
Outputs - out0, out1
Return Type - as per RTYPE
Details : Right half of byte elements from 'in0' and 'in1' are
interleaved and written to 'out0'
*/
#define ILVRL_B2(RTYPE, in0, in1, out0, out1) \
{ \
out0 = (RTYPE)__msa_ilvr_b((v16i8)in0, (v16i8)in1); \
out1 = (RTYPE)__msa_ilvl_b((v16i8)in0, (v16i8)in1); \
}
#define ILVRL_B2_UB(...) ILVRL_B2(v16u8, __VA_ARGS__)
#endif /* !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa) */
#endif // INCLUDE_LIBYUV_MACROS_MSA_H_

29
include/libyuv/rotate_row.h
View File

@ -60,11 +60,6 @@ extern "C" {
#define HAS_TRANSPOSEUVWXH_SME
#endif
#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)
#define HAS_TRANSPOSEWX16_MSA
#define HAS_TRANSPOSEUVWX16_MSA
#endif
#if !defined(LIBYUV_DISABLE_LSX) && defined(__loongarch_sx)
#define HAS_TRANSPOSEWX16_LSX
#define HAS_TRANSPOSEUVWX16_LSX
@ -113,11 +108,6 @@ void TransposeWx8_Fast_SSSE3(const uint8_t* src,
uint8_t* dst,
int dst_stride,
int width);
void TransposeWx16_MSA(const uint8_t* src,
int src_stride,
uint8_t* dst,
int dst_stride,
int width);
void TransposeWx16_LSX(const uint8_t* src,
int src_stride,
uint8_t* dst,
@ -144,11 +134,6 @@ void TransposeWx8_Fast_Any_SSSE3(const uint8_t* src,
uint8_t* dst,
int dst_stride,
int width);
void TransposeWx16_Any_MSA(const uint8_t* src,
int src_stride,
uint8_t* dst,
int dst_stride,
int width);
void TransposeWx16_Any_LSX(const uint8_t* src,
int src_stride,
uint8_t* dst,
@ -200,13 +185,6 @@ void TransposeUVWxH_SME(const uint8_t* src,
int dst_stride_b,
int width,
int height);
void TransposeUVWx16_MSA(const uint8_t* src,
int src_stride,
uint8_t* dst_a,
int dst_stride_a,
uint8_t* dst_b,
int dst_stride_b,
int width);
void TransposeUVWx16_LSX(const uint8_t* src,
int src_stride,
uint8_t* dst_a,
@ -229,13 +207,6 @@ void TransposeUVWx8_Any_NEON(const uint8_t* src,
uint8_t* dst_b,
int dst_stride_b,
int width);
void TransposeUVWx16_Any_MSA(const uint8_t* src,
int src_stride,
uint8_t* dst_a,
int dst_stride_a,
uint8_t* dst_b,
int dst_stride_b,
int width);
void TransposeUVWx16_Any_LSX(const uint8_t* src,
int src_stride,
uint8_t* dst_a,

622
include/libyuv/row.h
View File

File diff suppressed because it is too large Load Diff

230
include/libyuv/scale_row.h
View File

@ -136,19 +136,6 @@ extern "C" {
#define HAS_SCALEUVROWDOWN2_SME
#endif
#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)
#define HAS_SCALEADDROW_MSA
#define HAS_SCALEARGBCOLS_MSA
#define HAS_SCALEARGBFILTERCOLS_MSA
#define HAS_SCALEARGBROWDOWN2_MSA
#define HAS_SCALEARGBROWDOWNEVEN_MSA
#define HAS_SCALEFILTERCOLS_MSA
#define HAS_SCALEROWDOWN2_MSA
#define HAS_SCALEROWDOWN34_MSA
#define HAS_SCALEROWDOWN38_MSA
#define HAS_SCALEROWDOWN4_MSA
#endif
#if !defined(LIBYUV_DISABLE_LSX) && defined(__loongarch_sx)
#define HAS_SCALEARGBROWDOWN2_LSX
#define HAS_SCALEARGBROWDOWNEVEN_LSX
@ -269,17 +256,12 @@ enum FilterMode ScaleFilterReduce(int src_width,
// Divide num by div and return as 16.16 fixed point result.
int FixedDiv_C(int num, int div);
int FixedDiv_X86(int num, int div);
int FixedDiv_MIPS(int num, int div);
// Divide num - 1 by div - 1 and return as 16.16 fixed point result.
int FixedDiv1_C(int num, int div);
int FixedDiv1_X86(int num, int div);
int FixedDiv1_MIPS(int num, int div);
#ifdef HAS_FIXEDDIV_X86
#define FixedDiv FixedDiv_X86
#define FixedDiv1 FixedDiv1_X86
#elif defined HAS_FIXEDDIV_MIPS
#define FixedDiv FixedDiv_MIPS
#define FixedDiv1 FixedDiv1_MIPS
#else
#define FixedDiv FixedDiv_C
#define FixedDiv1 FixedDiv1_C
@ -935,26 +917,6 @@ void ScaleARGBCols_Any_NEON(uint8_t* dst_ptr,
int dst_width,
int x,
int dx);
void ScaleARGBFilterCols_MSA(uint8_t* dst_argb,
const uint8_t* src_argb,
int dst_width,
int x,
int dx);
void ScaleARGBCols_MSA(uint8_t* dst_argb,
const uint8_t* src_argb,
int dst_width,
int x,
int dx);
void ScaleARGBFilterCols_Any_MSA(uint8_t* dst_ptr,
const uint8_t* src_ptr,
int dst_width,
int x,
int dx);
void ScaleARGBCols_Any_MSA(uint8_t* dst_ptr,
const uint8_t* src_ptr,
int dst_width,
int x,
int dx);
void ScaleARGBFilterCols_RVV(uint8_t* dst_argb,
const uint8_t* src_argb,
int dst_width,
@ -1010,18 +972,6 @@ void ScaleARGBRowDown2Box_RVV(const uint8_t* src_argb,
ptrdiff_t src_stride,
uint8_t* dst_argb,
int dst_width);
void ScaleARGBRowDown2_MSA(const uint8_t* src_argb,
ptrdiff_t src_stride,
uint8_t* dst_argb,
int dst_width);
void ScaleARGBRowDown2Linear_MSA(const uint8_t* src_argb,
ptrdiff_t src_stride,
uint8_t* dst_argb,
int dst_width);
void ScaleARGBRowDown2Box_MSA(const uint8_t* src_argb,
ptrdiff_t src_stride,
uint8_t* dst_argb,
int dst_width);
void ScaleARGBRowDown2_LSX(const uint8_t* src_argb,
ptrdiff_t src_stride,
uint8_t* dst_argb,
@ -1058,18 +1008,6 @@ void ScaleARGBRowDown2Box_Any_NEON(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleARGBRowDown2_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleARGBRowDown2Linear_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleARGBRowDown2Box_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleARGBRowDown2_Any_LSX(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
@ -1102,16 +1040,6 @@ void ScaleARGBRowDownEvenBox_NEON(const uint8_t* src_argb,
int src_stepx,
uint8_t* dst_argb,
int dst_width);
void ScaleARGBRowDownEven_MSA(const uint8_t* src_argb,
ptrdiff_t src_stride,
int32_t src_stepx,
uint8_t* dst_argb,
int dst_width);
void ScaleARGBRowDownEvenBox_MSA(const uint8_t* src_argb,
ptrdiff_t src_stride,
int src_stepx,
uint8_t* dst_argb,
int dst_width);
void ScaleARGBRowDownEven_LSX(const uint8_t* src_argb,
ptrdiff_t src_stride,
int32_t src_stepx,
@ -1152,16 +1080,6 @@ void ScaleARGBRowDownEvenBox_Any_NEON(const uint8_t* src_ptr,
int src_stepx,
uint8_t* dst_ptr,
int dst_width);
void ScaleARGBRowDownEven_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
int32_t src_stepx,
uint8_t* dst_ptr,
int dst_width);
void ScaleARGBRowDownEvenBox_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
int src_stepx,
uint8_t* dst_ptr,
int dst_width);
void ScaleARGBRowDownEven_Any_LSX(const uint8_t* src_ptr,
ptrdiff_t src_stride,
int32_t src_stepx,
@ -1214,18 +1132,6 @@ void ScaleUVRowDown2Box_SME(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,
int dst_width);
void ScaleUVRowDown2_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_uv,
int dst_width);
void ScaleUVRowDown2Linear_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_uv,
int dst_width);
void ScaleUVRowDown2Box_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_uv,
int dst_width);
void ScaleUVRowDown2_RVV(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_uv,
@ -1266,18 +1172,6 @@ void ScaleUVRowDown2Box_Any_NEON(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleUVRowDown2_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleUVRowDown2Linear_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleUVRowDown2Box_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleUVRowDownEven_SSSE3(const uint8_t* src_ptr,
ptrdiff_t src_stride,
int src_stepx,
@ -1308,16 +1202,6 @@ void ScaleUVRowDownEven_RVV(const uint8_t* src_ptr,
int32_t src_stepx,
uint8_t* dst_uv,
int dst_width);
void ScaleUVRowDownEven_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
int32_t src_stepx,
uint8_t* dst_uv,
int dst_width);
void ScaleUVRowDownEvenBox_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
int src_stepx,
uint8_t* dst_uv,
int dst_width);
void ScaleUVRowDownEven_Any_SSSE3(const uint8_t* src_ptr,
ptrdiff_t src_stride,
int src_stepx,
@ -1338,16 +1222,6 @@ void ScaleUVRowDownEvenBox_Any_NEON(const uint8_t* src_ptr,
int src_stepx,
uint8_t* dst_ptr,
int dst_width);
void ScaleUVRowDownEven_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
int32_t src_stepx,
uint8_t* dst_ptr,
int dst_width);
void ScaleUVRowDownEvenBox_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
int src_stepx,
uint8_t* dst_ptr,
int dst_width);
void ScaleUVRowUp2_Linear_SSSE3(const uint8_t* src_ptr,
uint8_t* dst_ptr,
@ -1664,110 +1538,6 @@ void ScaleFilterCols_Any_NEON(uint8_t* dst_ptr,
int x,
int dx);
void ScaleRowDown2_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,
int dst_width);
void ScaleRowDown2Linear_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,
int dst_width);
void ScaleRowDown2Box_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,
int dst_width);
void ScaleRowDown4_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,
int dst_width);
void ScaleRowDown4Box_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,
int dst_width);
void ScaleRowDown38_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,
int dst_width);
void ScaleRowDown38_2_Box_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleRowDown38_3_Box_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleAddRow_MSA(const uint8_t* src_ptr, uint16_t* dst_ptr, int src_width);
void ScaleFilterCols_MSA(uint8_t* dst_ptr,
const uint8_t* src_ptr,
int dst_width,
int x,
int dx);
void ScaleRowDown34_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,
int dst_width);
void ScaleRowDown34_0_Box_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* d,
int dst_width);
void ScaleRowDown34_1_Box_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* d,
int dst_width);
void ScaleRowDown2_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleRowDown2Linear_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleRowDown2Box_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleRowDown4_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleRowDown4Box_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleRowDown38_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleRowDown38_2_Box_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleRowDown38_3_Box_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleAddRow_Any_MSA(const uint8_t* src_ptr,
uint16_t* dst_ptr,
int src_width);
void ScaleFilterCols_Any_MSA(uint8_t* dst_ptr,
const uint8_t* src_ptr,
int dst_width,
int x,
int dx);
void ScaleRowDown34_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleRowDown34_0_Box_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleRowDown34_1_Box_Any_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width);
void ScaleRowDown2_LSX(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,

12
libyuv.gyp
View File

@ -26,20 +26,13 @@
'clang%': 0,
# Link-Time Optimizations.
'use_lto%': 0,
'mips_msa%': 0, # Default to msa off.
'build_neon': 0,
'build_msa': 0,
'conditions': [
['(target_arch == "armv7" or target_arch == "armv7s" or \
(target_arch == "arm" and arm_version >= 7) or target_arch == "arm64")\
and (arm_neon == 1 or arm_neon_optional == 1)', {
'build_neon': 1,
}],
['(target_arch == "mipsel" or target_arch == "mips64el")\
and (mips_msa == 1)',
{
'build_msa': 1,
}],
],
},
@ -87,11 +80,6 @@
}],
],
}],
['build_msa != 0', {
'defines': [
'LIBYUV_MSA',
],
}],
['OS != "ios" and libyuv_disable_jpeg != 1', {
'defines': [
'HAVE_JPEG'

5
libyuv.gypi
View File

@ -21,7 +21,6 @@
'include/libyuv/cpu_id.h',
'include/libyuv/cpu_support.h',
'include/libyuv/loongson_intrinsics.h',
'include/libyuv/macros_msa.h',
'include/libyuv/mjpeg_decoder.h',
'include/libyuv/planar_functions.h',
'include/libyuv/rotate.h',
@ -40,7 +39,6 @@
'source/compare.cc',
'source/compare_common.cc',
'source/compare_gcc.cc',
'source/compare_msa.cc',
'source/compare_neon.cc',
'source/compare_neon64.cc',
'source/compare_win.cc',
@ -61,14 +59,12 @@
'source/rotate_common.cc',
'source/rotate_gcc.cc',
'source/rotate_lsx.cc',
'source/rotate_msa.cc',
'source/rotate_neon.cc',
'source/rotate_neon64.cc',
'source/rotate_win.cc',
'source/row_any.cc',
'source/row_common.cc',
'source/row_gcc.cc',
'source/row_msa.cc',
'source/row_lasx.cc',
'source/row_lsx.cc',
'source/row_neon.cc',
@ -80,7 +76,6 @@
'source/scale_common.cc',
'source/scale_gcc.cc',
'source/scale_lsx.cc',
'source/scale_msa.cc',
'source/scale_neon.cc',
'source/scale_neon64.cc',
'source/scale_rgb.cc',

4
linux.mk
View File

@ -13,7 +13,6 @@ LOCAL_OBJ_FILES := \
source/compare.o \
source/compare_common.o \
source/compare_gcc.o \
source/compare_msa.o \
source/compare_neon.o \
source/compare_neon64.o \
source/compare_win.o \
@ -34,7 +33,6 @@ LOCAL_OBJ_FILES := \
source/rotate_common.o \
source/rotate_gcc.o \
source/rotate_lsx.o \
source/rotate_msa.o \
source/rotate_neon.o \
source/rotate_neon64.o \
source/rotate_win.o \
@ -43,7 +41,6 @@ LOCAL_OBJ_FILES := \
source/row_gcc.o \
source/row_lasx.o \
source/row_lsx.o \
source/row_msa.o \
source/row_neon.o \
source/row_neon64.o \
source/row_rvv.o \
@ -54,7 +51,6 @@ LOCAL_OBJ_FILES := \
source/scale_common.o \
source/scale_gcc.o \
source/scale_lsx.o \
source/scale_msa.o \
source/scale_neon.o \
source/scale_neon64.o \
source/scale_rgb.o \

10
source/compare.cc
View File

@ -159,11 +159,6 @@ uint64_t ComputeHammingDistance(const uint8_t* src_a,
HammingDistance = HammingDistance_AVX2;
}
#endif
#if defined(HAS_HAMMINGDISTANCE_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
HammingDistance = HammingDistance_MSA;
}
#endif
#ifdef _OPENMP
#pragma omp parallel for reduction(+ : diff)
@ -221,11 +216,6 @@ uint64_t ComputeSumSquareError(const uint8_t* src_a,
SumSquareError = SumSquareError_AVX2;
}
#endif
#if defined(HAS_SUMSQUAREERROR_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
SumSquareError = SumSquareError_MSA;
}
#endif
#ifdef _OPENMP
#pragma omp parallel for reduction(+ : sse)
#endif

97
source/compare_msa.cc
View File

@ -1,97 +0,0 @@
/*
* Copyright 2017 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.
*/
#include "libyuv/basic_types.h"
#include "libyuv/compare_row.h"
#include "libyuv/row.h"
// This module is for GCC MSA
#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)
#include "libyuv/macros_msa.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
uint32_t HammingDistance_MSA(const uint8_t* src_a,
const uint8_t* src_b,
int count) {
uint32_t diff = 0u;
int i;
v16u8 src0, src1, src2, src3;
v2i64 vec0 = {0}, vec1 = {0};
for (i = 0; i < count; i += 32) {
src0 = (v16u8)__msa_ld_b((v16i8*)src_a, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)src_a, 16);
src2 = (v16u8)__msa_ld_b((v16i8*)src_b, 0);
src3 = (v16u8)__msa_ld_b((v16i8*)src_b, 16);
src0 ^= src2;
src1 ^= src3;
vec0 += __msa_pcnt_d((v2i64)src0);
vec1 += __msa_pcnt_d((v2i64)src1);
src_a += 32;
src_b += 32;
}
vec0 += vec1;
diff = (uint32_t)__msa_copy_u_w((v4i32)vec0, 0);
diff += (uint32_t)__msa_copy_u_w((v4i32)vec0, 2);
return diff;
}
uint32_t SumSquareError_MSA(const uint8_t* src_a,
const uint8_t* src_b,
int count) {
uint32_t sse = 0u;
int i;
v16u8 src0, src1, src2, src3;
v8i16 vec0, vec1, vec2, vec3;
v4i32 reg0 = {0}, reg1 = {0}, reg2 = {0}, reg3 = {0};
v2i64 tmp0;
for (i = 0; i < count; i += 32) {
src0 = (v16u8)__msa_ld_b((v16i8*)src_a, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)src_a, 16);
src2 = (v16u8)__msa_ld_b((v16i8*)src_b, 0);
src3 = (v16u8)__msa_ld_b((v16i8*)src_b, 16);
vec0 = (v8i16)__msa_ilvr_b((v16i8)src2, (v16i8)src0);
vec1 = (v8i16)__msa_ilvl_b((v16i8)src2, (v16i8)src0);
vec2 = (v8i16)__msa_ilvr_b((v16i8)src3, (v16i8)src1);
vec3 = (v8i16)__msa_ilvl_b((v16i8)src3, (v16i8)src1);
vec0 = __msa_hsub_u_h((v16u8)vec0, (v16u8)vec0);
vec1 = __msa_hsub_u_h((v16u8)vec1, (v16u8)vec1);
vec2 = __msa_hsub_u_h((v16u8)vec2, (v16u8)vec2);
vec3 = __msa_hsub_u_h((v16u8)vec3, (v16u8)vec3);
reg0 = __msa_dpadd_s_w(reg0, vec0, vec0);
reg1 = __msa_dpadd_s_w(reg1, vec1, vec1);
reg2 = __msa_dpadd_s_w(reg2, vec2, vec2);
reg3 = __msa_dpadd_s_w(reg3, vec3, vec3);
src_a += 32;
src_b += 32;
}
reg0 += reg1;
reg2 += reg3;
reg0 += reg2;
tmp0 = __msa_hadd_s_d(reg0, reg0);
sse = (uint32_t)__msa_copy_u_w((v4i32)tmp0, 0);
sse += (uint32_t)__msa_copy_u_w((v4i32)tmp0, 2);
return sse;
}
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
#endif // !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)

308
source/convert.cc
View File

@ -748,14 +748,6 @@ int I010ToNV12(const uint16_t* src_y,
MergeUVRow = MergeUVRow_SME;
}
#endif
#if defined(HAS_MERGEUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
MergeUVRow = MergeUVRow_Any_MSA;
if (IS_ALIGNED(halfwidth, 16)) {
MergeUVRow = MergeUVRow_MSA;
}
}
#endif
#if defined(HAS_MERGEUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MergeUVRow = MergeUVRow_Any_LSX;
@ -1196,14 +1188,6 @@ int I422ToNV21(const uint8_t* src_y,
MergeUVRow = MergeUVRow_SME;
}
#endif
#if defined(HAS_MERGEUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
MergeUVRow = MergeUVRow_Any_MSA;
if (IS_ALIGNED(halfwidth, 16)) {
MergeUVRow = MergeUVRow_MSA;
}
}
#endif
#if defined(HAS_MERGEUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MergeUVRow = MergeUVRow_Any_LSX;
@ -1246,14 +1230,6 @@ int I422ToNV21(const uint8_t* src_y,
InterpolateRow = InterpolateRow_SME;
}
#endif
#if defined(HAS_INTERPOLATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
InterpolateRow = InterpolateRow_Any_MSA;
if (IS_ALIGNED(width, 32)) {
InterpolateRow = InterpolateRow_MSA;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
@ -1724,16 +1700,6 @@ int YUY2ToI420(const uint8_t* src_yuy2,
}
}
#endif
#if defined(HAS_YUY2TOYROW_MSA) && defined(HAS_YUY2TOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
YUY2ToYRow = YUY2ToYRow_Any_MSA;
YUY2ToUVRow = YUY2ToUVRow_Any_MSA;
if (IS_ALIGNED(width, 32)) {
YUY2ToYRow = YUY2ToYRow_MSA;
YUY2ToUVRow = YUY2ToUVRow_MSA;
}
}
#endif
#if defined(HAS_YUY2TOYROW_LSX) && defined(HAS_YUY2TOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
YUY2ToYRow = YUY2ToYRow_Any_LSX;
@ -1825,16 +1791,6 @@ int UYVYToI420(const uint8_t* src_uyvy,
}
}
#endif
#if defined(HAS_UYVYTOYROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
UYVYToYRow = UYVYToYRow_Any_MSA;
UYVYToUVRow = UYVYToUVRow_Any_MSA;
if (IS_ALIGNED(width, 32)) {
UYVYToYRow = UYVYToYRow_MSA;
UYVYToUVRow = UYVYToUVRow_MSA;
}
}
#endif
#if defined(HAS_UYVYTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
UYVYToYRow = UYVYToYRow_Any_LSX;
@ -2143,18 +2099,6 @@ int ARGBToI420(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYRow = ARGBToYRow_Any_MSA;
ARGBToUVRow = ARGBToUVRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_MSA;
}
if (IS_ALIGNED(width, 32)) {
ARGBToUVRow = ARGBToUVRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
@ -2337,18 +2281,6 @@ int ARGBToI420Alpha(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYRow = ARGBToYRow_Any_MSA;
ARGBToUVRow = ARGBToUVRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_MSA;
}
if (IS_ALIGNED(width, 32)) {
ARGBToUVRow = ARGBToUVRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
@ -2385,12 +2317,6 @@ int ARGBToI420Alpha(const uint8_t* src_argb,
: ARGBExtractAlphaRow_Any_NEON;
}
#endif
#if defined(HAS_ARGBEXTRACTALPHAROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBExtractAlphaRow = IS_ALIGNED(width, 16) ? ARGBExtractAlphaRow_MSA
: ARGBExtractAlphaRow_Any_MSA;
}
#endif
#if defined(HAS_ARGBEXTRACTALPHAROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBExtractAlphaRow = IS_ALIGNED(width, 16) ? ARGBExtractAlphaRow_LSX
@ -2532,18 +2458,6 @@ int BGRAToI420(const uint8_t* src_bgra,
}
}
#endif
#if defined(HAS_BGRATOYROW_MSA) && defined(HAS_BGRATOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
BGRAToYRow = BGRAToYRow_Any_MSA;
BGRAToUVRow = BGRAToUVRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
BGRAToYRow = BGRAToYRow_MSA;
}
if (IS_ALIGNED(width, 32)) {
BGRAToUVRow = BGRAToUVRow_MSA;
}
}
#endif
#if defined(HAS_BGRATOYROW_LSX) && defined(HAS_BGRATOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
BGRAToYRow = BGRAToYRow_Any_LSX;
@ -2691,16 +2605,6 @@ int ABGRToI420(const uint8_t* src_abgr,
}
}
#endif
#if defined(HAS_ABGRTOYROW_MSA) && defined(HAS_ABGRTOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ABGRToYRow = ABGRToYRow_Any_MSA;
ABGRToUVRow = ABGRToUVRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ABGRToYRow = ABGRToYRow_MSA;
ABGRToUVRow = ABGRToUVRow_MSA;
}
}
#endif
#if defined(HAS_ABGRTOYROW_LSX) && defined(HAS_ABGRTOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ABGRToYRow = ABGRToYRow_Any_LSX;
@ -2832,16 +2736,6 @@ int RGBAToI420(const uint8_t* src_rgba,
}
}
#endif
#if defined(HAS_RGBATOYROW_MSA) && defined(HAS_RGBATOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RGBAToYRow = RGBAToYRow_Any_MSA;
RGBAToUVRow = RGBAToUVRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RGBAToYRow = RGBAToYRow_MSA;
RGBAToUVRow = RGBAToUVRow_MSA;
}
}
#endif
#if defined(HAS_RGBATOYROW_LSX) && defined(HAS_RGBATOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RGBAToYRow = RGBAToYRow_Any_LSX;
@ -2883,8 +2777,8 @@ int RGBAToI420(const uint8_t* src_rgba,
}
// Enabled if 1 pass is available
#if (defined(HAS_RGB24TOYROW_NEON) || defined(HAS_RGB24TOYROW_MSA) || \
defined(HAS_RGB24TOYROW_LSX) || defined(HAS_RGB24TOYROW_RVV))
#if (defined(HAS_RGB24TOYROW_NEON) || defined(HAS_RGB24TOYROW_LSX) || \
defined(HAS_RGB24TOYROW_RVV))
#define HAS_RGB24TOYROW
#endif
@ -2939,16 +2833,6 @@ int RGB24ToI420(const uint8_t* src_rgb24,
}
}
#endif
#if defined(HAS_RGB24TOYROW_MSA) && defined(HAS_RGB24TOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RGB24ToUVRow = RGB24ToUVRow_Any_MSA;
RGB24ToYRow = RGB24ToYRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RGB24ToYRow = RGB24ToYRow_MSA;
RGB24ToUVRow = RGB24ToUVRow_MSA;
}
}
#endif
#if defined(HAS_RGB24TOYROW_LSX) && defined(HAS_RGB24TOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RGB24ToUVRow = RGB24ToUVRow_Any_LSX;
@ -3065,8 +2949,7 @@ int RGB24ToI420(const uint8_t* src_rgb24,
#undef HAS_RGB24TOYROW
// Enabled if 1 pass is available
#if defined(HAS_RGB24TOYJROW_NEON) || defined(HAS_RGB24TOYJROW_MSA) || \
defined(HAS_RGB24TOYJROW_RVV)
#if defined(HAS_RGB24TOYJROW_NEON) || defined(HAS_RGB24TOYJROW_RVV)
#define HAS_RGB24TOYJROW
#endif
@ -3121,16 +3004,6 @@ int RGB24ToJ420(const uint8_t* src_rgb24,
}
}
#endif
#if defined(HAS_RGB24TOYJROW_MSA) && defined(HAS_RGB24TOUVJROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RGB24ToUVJRow = RGB24ToUVJRow_Any_MSA;
RGB24ToYJRow = RGB24ToYJRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RGB24ToYJRow = RGB24ToYJRow_MSA;
RGB24ToUVJRow = RGB24ToUVJRow_MSA;
}
}
#endif
#if defined(HAS_RGB24TOYJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RGB24ToYJRow = RGB24ToYJRow_Any_LSX;
@ -3243,8 +3116,8 @@ int RGB24ToJ420(const uint8_t* src_rgb24,
#undef HAS_RGB24TOYJROW
// Enabled if 1 pass is available
#if (defined(HAS_RAWTOYROW_NEON) || defined(HAS_RAWTOYROW_MSA) || \
defined(HAS_RAWTOYROW_LSX) || defined(HAS_RAWTOYROW_RVV))
#if (defined(HAS_RAWTOYROW_NEON) || defined(HAS_RAWTOYROW_LSX) || \
defined(HAS_RAWTOYROW_RVV))
#define HAS_RAWTOYROW
#endif
@ -3298,16 +3171,6 @@ int RAWToI420(const uint8_t* src_raw,
}
}
#endif
#if defined(HAS_RAWTOYROW_MSA) && defined(HAS_RAWTOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RAWToUVRow = RAWToUVRow_Any_MSA;
RAWToYRow = RAWToYRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RAWToYRow = RAWToYRow_MSA;
RAWToUVRow = RAWToUVRow_MSA;
}
}
#endif
#if defined(HAS_RAWTOYROW_LSX) && defined(HAS_RAWTOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RAWToUVRow = RAWToUVRow_Any_LSX;
@ -3432,8 +3295,7 @@ int RAWToI420(const uint8_t* src_raw,
#undef HAS_RAWTOYROW
// Enabled if 1 pass is available
#if defined(HAS_RAWTOYJROW_NEON) || defined(HAS_RAWTOYJROW_MSA) || \
defined(HAS_RAWTOYJROW_RVV)
#if defined(HAS_RAWTOYJROW_NEON) || defined(HAS_RAWTOYJROW_RVV)
#define HAS_RAWTOYJROW
#endif
@ -3488,16 +3350,6 @@ int RAWToJ420(const uint8_t* src_raw,
}
}
#endif
#if defined(HAS_RAWTOYJROW_MSA) && defined(HAS_RAWTOUVJROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RAWToUVJRow = RAWToUVJRow_Any_MSA;
RAWToYJRow = RAWToYJRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RAWToYJRow = RAWToYJRow_MSA;
RAWToUVJRow = RAWToUVJRow_MSA;
}
}
#endif
#if defined(HAS_RAWTOYJROW_LSX) && defined(HAS_RAWTOUVJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RAWToUVJRow = RAWToUVJRow_Any_LSX;
@ -3683,14 +3535,6 @@ int RAWToI444(const uint8_t* src_raw,
}
}
#endif
#if defined(HAS_ARGBTOUV444ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToUV444Row = ARGBToUV444Row_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToUV444Row = ARGBToUV444Row_MSA;
}
}
#endif
#if defined(HAS_ARGBTOUV444ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToUV444Row = ARGBToUV444Row_Any_LSX;
@ -3739,14 +3583,6 @@ int RAWToI444(const uint8_t* src_raw,
}
}
#endif
#if defined(HAS_ARGBTOYROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYRow = ARGBToYRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
@ -3798,14 +3634,6 @@ int RAWToI444(const uint8_t* src_raw,
RAWToARGBRow = RAWToARGBRow_SVE2;
}
#endif
#if defined(HAS_RAWTOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RAWToARGBRow = RAWToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RAWToARGBRow = RAWToARGBRow_MSA;
}
}
#endif
#if defined(HAS_RAWTOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RAWToARGBRow = RAWToARGBRow_Any_LSX;
@ -3910,14 +3738,6 @@ int RAWToJ444(const uint8_t* src_raw,
}
}
#endif
#if defined(HAS_ARGBTOUVJ444ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToUVJ444Row = ARGBToUVJ444Row_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToUVJ444Row = ARGBToUVJ444Row_MSA;
}
}
#endif
#if defined(HAS_ARGBTOUVJ444ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToUVJ444Row = ARGBToUVJ444Row_Any_LSX;
@ -3966,14 +3786,6 @@ int RAWToJ444(const uint8_t* src_raw,
}
}
#endif
#if defined(HAS_ARGBTOYJROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYJRow = ARGBToYJRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYJRow = ARGBToYJRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYJRow = ARGBToYJRow_Any_LSX;
@ -4025,14 +3837,6 @@ int RAWToJ444(const uint8_t* src_raw,
RAWToARGBRow = RAWToARGBRow_SVE2;
}
#endif
#if defined(HAS_RAWTOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RAWToARGBRow = RAWToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RAWToARGBRow = RAWToARGBRow_MSA;
}
}
#endif
#if defined(HAS_RAWTOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RAWToARGBRow = RAWToARGBRow_Any_LSX;
@ -4089,8 +3893,8 @@ int RGB565ToI420(const uint8_t* src_rgb565,
int width,
int height) {
int y;
#if (defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_MSA) || \
defined(HAS_RGB565TOYROW_LSX) || defined(HAS_RGB565TOYROW_LASX))
#if (defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_LSX) || \
defined(HAS_RGB565TOYROW_LASX))
void (*RGB565ToUVRow)(const uint8_t* src_rgb565, int src_stride_rgb565,
uint8_t* dst_u, uint8_t* dst_v, int width) =
RGB565ToUVRow_C;
@ -4133,17 +3937,7 @@ int RGB565ToI420(const uint8_t* src_rgb565,
}
}
#endif
// MSA version does direct RGB565 to YUV.
#if defined(HAS_RGB565TOYROW_MSA) && defined(HAS_RGB565TOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RGB565ToUVRow = RGB565ToUVRow_Any_MSA;
RGB565ToYRow = RGB565ToYRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RGB565ToYRow = RGB565ToYRow_MSA;
RGB565ToUVRow = RGB565ToUVRow_MSA;
}
}
#endif
// LSX version does direct RGB565 to YUV.
#if defined(HAS_RGB565TOYROW_LSX) && defined(HAS_RGB565TOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RGB565ToUVRow = RGB565ToUVRow_Any_LSX;
@ -4214,8 +4008,8 @@ int RGB565ToI420(const uint8_t* src_rgb565,
}
#endif
{
#if !(defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_MSA) || \
defined(HAS_RGB565TOYROW_LSX) || defined(HAS_RGB565TOYROW_LASX))
#if !(defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_LSX) || \
defined(HAS_RGB565TOYROW_LASX))
// Allocate 2 rows of ARGB.
const int row_size = (width * 4 + 31) & ~31;
align_buffer_64(row, row_size * 2);
@ -4223,8 +4017,8 @@ int RGB565ToI420(const uint8_t* src_rgb565,
return 1;
#endif
for (y = 0; y < height - 1; y += 2) {
#if (defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_MSA) || \
defined(HAS_RGB565TOYROW_LSX) || defined(HAS_RGB565TOYROW_LASX))
#if (defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_LSX) || \
defined(HAS_RGB565TOYROW_LASX))
RGB565ToUVRow(src_rgb565, src_stride_rgb565, dst_u, dst_v, width);
RGB565ToYRow(src_rgb565, dst_y, width);
RGB565ToYRow(src_rgb565 + src_stride_rgb565, dst_y + dst_stride_y, width);
@ -4241,8 +4035,8 @@ int RGB565ToI420(const uint8_t* src_rgb565,
dst_v += dst_stride_v;
}
if (height & 1) {
#if (defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_MSA) || \
defined(HAS_RGB565TOYROW_LSX) || defined(HAS_RGB565TOYROW_LASX))
#if (defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_LSX) || \
defined(HAS_RGB565TOYROW_LASX))
RGB565ToUVRow(src_rgb565, 0, dst_u, dst_v, width);
RGB565ToYRow(src_rgb565, dst_y, width);
#else
@ -4251,8 +4045,8 @@ int RGB565ToI420(const uint8_t* src_rgb565,
ARGBToYRow(row, dst_y, width);
#endif
}
#if !(defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_MSA) || \
defined(HAS_RGB565TOYROW_LSX) || defined(HAS_RGB565TOYROW_LASX))
#if !(defined(HAS_RGB565TOYROW_NEON) || defined(HAS_RGB565TOYROW_LSX) || \
defined(HAS_RGB565TOYROW_LASX))
free_aligned_buffer_64(row);
#endif
}
@ -4272,8 +4066,8 @@ int ARGB1555ToI420(const uint8_t* src_argb1555,
int width,
int height) {
int y;
#if (defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_MSA) || \
defined(HAS_ARGB1555TOYROW_LSX) || defined(HAS_ARGB1555TOYROW_LASX))
#if (defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_LSX) || \
defined(HAS_ARGB1555TOYROW_LASX))
void (*ARGB1555ToUVRow)(const uint8_t* src_argb1555, int src_stride_argb1555,
uint8_t* dst_u, uint8_t* dst_v, int width) =
ARGB1555ToUVRow_C;
@ -4317,17 +4111,7 @@ int ARGB1555ToI420(const uint8_t* src_argb1555,
}
#endif
// MSA version does direct ARGB1555 to YUV.
#if defined(HAS_ARGB1555TOYROW_MSA) && defined(HAS_ARGB1555TOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGB1555ToUVRow = ARGB1555ToUVRow_Any_MSA;
ARGB1555ToYRow = ARGB1555ToYRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGB1555ToYRow = ARGB1555ToYRow_MSA;
ARGB1555ToUVRow = ARGB1555ToUVRow_MSA;
}
}
#endif
// LSX version does direct ARGB1555 to YUV.
#if defined(HAS_ARGB1555TOYROW_LSX) && defined(HAS_ARGB1555TOUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGB1555ToUVRow = ARGB1555ToUVRow_Any_LSX;
@ -4399,8 +4183,8 @@ int ARGB1555ToI420(const uint8_t* src_argb1555,
}
#endif
{
#if !(defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_MSA) || \
defined(HAS_ARGB1555TOYROW_LSX) || defined(HAS_ARGB1555TOYROW_LASX))
#if !(defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_LSX) || \
defined(HAS_ARGB1555TOYROW_LASX))
// Allocate 2 rows of ARGB.
const int row_size = (width * 4 + 31) & ~31;
align_buffer_64(row, row_size * 2);
@ -4409,8 +4193,8 @@ int ARGB1555ToI420(const uint8_t* src_argb1555,
#endif
for (y = 0; y < height - 1; y += 2) {
#if (defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_MSA) || \
defined(HAS_ARGB1555TOYROW_LSX) || defined(HAS_ARGB1555TOYROW_LASX))
#if (defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_LSX) || \
defined(HAS_ARGB1555TOYROW_LASX))
ARGB1555ToUVRow(src_argb1555, src_stride_argb1555, dst_u, dst_v, width);
ARGB1555ToYRow(src_argb1555, dst_y, width);
ARGB1555ToYRow(src_argb1555 + src_stride_argb1555, dst_y + dst_stride_y,
@ -4429,8 +4213,8 @@ int ARGB1555ToI420(const uint8_t* src_argb1555,
dst_v += dst_stride_v;
}
if (height & 1) {
#if (defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_MSA) || \
defined(HAS_ARGB1555TOYROW_LSX) || defined(HAS_ARGB1555TOYROW_LASX))
#if (defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_LSX) || \
defined(HAS_ARGB1555TOYROW_LASX))
ARGB1555ToUVRow(src_argb1555, 0, dst_u, dst_v, width);
ARGB1555ToYRow(src_argb1555, dst_y, width);
#else
@ -4439,8 +4223,8 @@ int ARGB1555ToI420(const uint8_t* src_argb1555,
ARGBToYRow(row, dst_y, width);
#endif
}
#if !(defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_MSA) || \
defined(HAS_ARGB1555TOYROW_LSX) || defined(HAS_ARGB1555TOYROW_LASX))
#if !(defined(HAS_ARGB1555TOYROW_NEON) || defined(HAS_ARGB1555TOYROW_LSX) || \
defined(HAS_ARGB1555TOYROW_LASX))
free_aligned_buffer_64(row);
#endif
}
@ -4519,14 +4303,6 @@ int ARGB4444ToI420(const uint8_t* src_argb4444,
}
}
#endif
#if defined(HAS_ARGB4444TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGB4444ToARGBRow = ARGB4444ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_ARGB4444TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_LSX;
@ -4575,18 +4351,6 @@ int ARGB4444ToI420(const uint8_t* src_argb4444,
}
}
#endif
#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToUVRow = ARGBToUVRow_Any_MSA;
ARGBToYRow = ARGBToYRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_MSA;
if (IS_ALIGNED(width, 32)) {
ARGBToUVRow = ARGBToUVRow_MSA;
}
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
@ -4710,14 +4474,6 @@ int RGB24ToJ400(const uint8_t* src_rgb24,
}
}
#endif
#if defined(HAS_RGB24TOYJROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RGB24ToYJRow = RGB24ToYJRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RGB24ToYJRow = RGB24ToYJRow_MSA;
}
}
#endif
#if defined(HAS_RGB24TOYJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RGB24ToYJRow = RGB24ToYJRow_Any_LSX;
@ -4799,14 +4555,6 @@ int RAWToJ400(const uint8_t* src_raw,
}
}
#endif
#if defined(HAS_RAWTOYJROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RAWToYJRow = RAWToYJRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RAWToYJRow = RAWToYJRow_MSA;
}
}
#endif
#if defined(HAS_RAWTOYJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RAWToYJRow = RAWToYJRow_Any_LSX;

352
source/convert_argb.cc
View File

@ -119,14 +119,6 @@ int I420ToARGBMatrix(const uint8_t* src_y,
I422ToARGBRow = I422ToARGBRow_SME;
}
#endif
#if defined(HAS_I422TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToARGBRow = I422ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I422ToARGBRow = I422ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_I422TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToARGBRow = I422ToARGBRow_Any_LSX;
@ -383,14 +375,6 @@ int I422ToARGBMatrix(const uint8_t* src_y,
I422ToARGBRow = I422ToARGBRow_SME;
}
#endif
#if defined(HAS_I422TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToARGBRow = I422ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I422ToARGBRow = I422ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_I422TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToARGBRow = I422ToARGBRow_Any_LSX;
@ -636,14 +620,6 @@ int I444ToARGBMatrix(const uint8_t* src_y,
I444ToARGBRow = I444ToARGBRow_SME;
}
#endif
#if defined(HAS_I444TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I444ToARGBRow = I444ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I444ToARGBRow = I444ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_I444TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I444ToARGBRow = I444ToARGBRow_Any_LSX;
@ -881,14 +857,6 @@ int I444ToRGB24Matrix(const uint8_t* src_y,
I444ToRGB24Row = I444ToRGB24Row_SME;
}
#endif
#if defined(HAS_I444TORGB24ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I444ToRGB24Row = I444ToRGB24Row_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I444ToRGB24Row = I444ToRGB24Row_MSA;
}
}
#endif
#if defined(HAS_I444TORGB24ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I444ToRGB24Row = I444ToRGB24Row_Any_LSX;
@ -2298,14 +2266,6 @@ int I420AlphaToARGBMatrix(const uint8_t* src_y,
I422AlphaToARGBRow = I422AlphaToARGBRow_SME;
}
#endif
#if defined(HAS_I422ALPHATOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_MSA;
}
}
#endif
#if defined(HAS_I422ALPHATOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_Any_LSX;
@ -2351,14 +2311,6 @@ int I420AlphaToARGBMatrix(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_MSA;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_RVV)
if (TestCpuFlag(kCpuHasRVV)) {
ARGBAttenuateRow = ARGBAttenuateRow_RVV;
@ -2467,14 +2419,6 @@ int I422AlphaToARGBMatrix(const uint8_t* src_y,
I422AlphaToARGBRow = I422AlphaToARGBRow_SME;
}
#endif
#if defined(HAS_I422ALPHATOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_MSA;
}
}
#endif
#if defined(HAS_I422ALPHATOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422AlphaToARGBRow = I422AlphaToARGBRow_Any_LSX;
@ -2520,14 +2464,6 @@ int I422AlphaToARGBMatrix(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_MSA;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_RVV)
if (TestCpuFlag(kCpuHasRVV)) {
ARGBAttenuateRow = ARGBAttenuateRow_RVV;
@ -2634,14 +2570,6 @@ int I444AlphaToARGBMatrix(const uint8_t* src_y,
I444AlphaToARGBRow = I444AlphaToARGBRow_SME;
}
#endif
#if defined(HAS_I444ALPHATOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_MSA;
}
}
#endif
#if defined(HAS_I444ALPHATOARGBROW_RVV)
if (TestCpuFlag(kCpuHasRVV)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_RVV;
@ -2671,14 +2599,6 @@ int I444AlphaToARGBMatrix(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_MSA;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_RVV)
if (TestCpuFlag(kCpuHasRVV)) {
ARGBAttenuateRow = ARGBAttenuateRow_RVV;
@ -2938,14 +2858,6 @@ int I010AlphaToARGBMatrix(const uint16_t* src_y,
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_MSA;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_RVV)
if (TestCpuFlag(kCpuHasRVV)) {
ARGBAttenuateRow = ARGBAttenuateRow_RVV;
@ -3078,14 +2990,6 @@ int I210AlphaToARGBMatrix(const uint16_t* src_y,
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_MSA;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_RVV)
if (TestCpuFlag(kCpuHasRVV)) {
ARGBAttenuateRow = ARGBAttenuateRow_RVV;
@ -3216,14 +3120,6 @@ int I410AlphaToARGBMatrix(const uint16_t* src_y,
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_MSA;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_RVV)
if (TestCpuFlag(kCpuHasRVV)) {
ARGBAttenuateRow = ARGBAttenuateRow_RVV;
@ -3324,14 +3220,6 @@ int I400ToARGBMatrix(const uint8_t* src_y,
I400ToARGBRow = I400ToARGBRow_SME;
}
#endif
#if defined(HAS_I400TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I400ToARGBRow = I400ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
I400ToARGBRow = I400ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_I400TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I400ToARGBRow = I400ToARGBRow_Any_LSX;
@ -3416,14 +3304,6 @@ int J400ToARGB(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_J400TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
J400ToARGBRow = J400ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
J400ToARGBRow = J400ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_J400TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
J400ToARGBRow = J400ToARGBRow_Any_LSX;
@ -3771,14 +3651,6 @@ int RGB24ToARGB(const uint8_t* src_rgb24,
RGB24ToARGBRow = RGB24ToARGBRow_SVE2;
}
#endif
#if defined(HAS_RGB24TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RGB24ToARGBRow = RGB24ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RGB24ToARGBRow = RGB24ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_RGB24TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RGB24ToARGBRow = RGB24ToARGBRow_Any_LSX;
@ -3864,14 +3736,6 @@ int RAWToARGB(const uint8_t* src_raw,
RAWToARGBRow = RAWToARGBRow_SVE2;
}
#endif
#if defined(HAS_RAWTOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RAWToARGBRow = RAWToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RAWToARGBRow = RAWToARGBRow_MSA;
}
}
#endif
#if defined(HAS_RAWTOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RAWToARGBRow = RAWToARGBRow_Any_LSX;
@ -4013,14 +3877,6 @@ int RGB565ToARGB(const uint8_t* src_rgb565,
}
}
#endif
#if defined(HAS_RGB565TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RGB565ToARGBRow = RGB565ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RGB565ToARGBRow = RGB565ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_RGB565TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RGB565ToARGBRow = RGB565ToARGBRow_Any_LSX;
@ -4101,14 +3957,6 @@ int ARGB1555ToARGB(const uint8_t* src_argb1555,
ARGB1555ToARGBRow = ARGB1555ToARGBRow_SVE2;
}
#endif
#if defined(HAS_ARGB1555TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGB1555ToARGBRow = ARGB1555ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGB1555ToARGBRow = ARGB1555ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_ARGB1555TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGB1555ToARGBRow = ARGB1555ToARGBRow_Any_LSX;
@ -4184,14 +4032,6 @@ int ARGB4444ToARGB(const uint8_t* src_argb4444,
}
}
#endif
#if defined(HAS_ARGB4444TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGB4444ToARGBRow = ARGB4444ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_ARGB4444TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGB4444ToARGBRow = ARGB4444ToARGBRow_Any_LSX;
@ -4500,14 +4340,6 @@ int NV12ToARGBMatrix(const uint8_t* src_y,
NV12ToARGBRow = NV12ToARGBRow_SME;
}
#endif
#if defined(HAS_NV12TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
NV12ToARGBRow = NV12ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
NV12ToARGBRow = NV12ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_NV12TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
NV12ToARGBRow = NV12ToARGBRow_Any_LSX;
@ -4600,14 +4432,6 @@ int NV21ToARGBMatrix(const uint8_t* src_y,
NV21ToARGBRow = NV21ToARGBRow_SME;
}
#endif
#if defined(HAS_NV21TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
NV21ToARGBRow = NV21ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
NV21ToARGBRow = NV21ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_NV21TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
NV21ToARGBRow = NV21ToARGBRow_Any_LSX;
@ -5028,14 +4852,6 @@ int YUY2ToARGBMatrix(const uint8_t* src_yuy2,
YUY2ToARGBRow = YUY2ToARGBRow_SME;
}
#endif
#if defined(HAS_YUY2TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
YUY2ToARGBRow = YUY2ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
YUY2ToARGBRow = YUY2ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_YUY2TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
YUY2ToARGBRow = YUY2ToARGBRow_Any_LSX;
@ -5126,14 +4942,6 @@ int UYVYToARGBMatrix(const uint8_t* src_uyvy,
UYVYToARGBRow = UYVYToARGBRow_SME;
}
#endif
#if defined(HAS_UYVYTOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
UYVYToARGBRow = UYVYToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
UYVYToARGBRow = UYVYToARGBRow_MSA;
}
}
#endif
#if defined(HAS_UYVYTOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
UYVYToARGBRow = UYVYToARGBRow_Any_LSX;
@ -5343,14 +5151,6 @@ int I422ToRGBAMatrix(const uint8_t* src_y,
I422ToRGBARow = I422ToRGBARow_SME;
}
#endif
#if defined(HAS_I422TORGBAROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToRGBARow = I422ToRGBARow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I422ToRGBARow = I422ToRGBARow_MSA;
}
}
#endif
#if defined(HAS_I422TORGBAROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToRGBARow = I422ToRGBARow_Any_LSX;
@ -5468,14 +5268,6 @@ int NV12ToRGB565Matrix(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_NV12TORGB565ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
NV12ToRGB565Row = NV12ToRGB565Row_Any_MSA;
if (IS_ALIGNED(width, 8)) {
NV12ToRGB565Row = NV12ToRGB565Row_MSA;
}
}
#endif
#if defined(HAS_NV12TORGB565ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
NV12ToRGB565Row = NV12ToRGB565Row_Any_LSX;
@ -5581,14 +5373,6 @@ int I420ToRGBAMatrix(const uint8_t* src_y,
I422ToRGBARow = I422ToRGBARow_SME;
}
#endif
#if defined(HAS_I422TORGBAROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToRGBARow = I422ToRGBARow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I422ToRGBARow = I422ToRGBARow_MSA;
}
}
#endif
#if defined(HAS_I422TORGBAROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToRGBARow = I422ToRGBARow_Any_LSX;
@ -5721,14 +5505,6 @@ int I420ToRGB24Matrix(const uint8_t* src_y,
I422ToRGB24Row = I422ToRGB24Row_SME;
}
#endif
#if defined(HAS_I422TORGB24ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToRGB24Row = I422ToRGB24Row_Any_MSA;
if (IS_ALIGNED(width, 16)) {
I422ToRGB24Row = I422ToRGB24Row_MSA;
}
}
#endif
#if defined(HAS_I422TORGB24ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToRGB24Row = I422ToRGB24Row_Any_LSX;
@ -5933,14 +5709,6 @@ int I422ToRGB24Matrix(const uint8_t* src_y,
I422ToRGB24Row = I422ToRGB24Row_SME;
}
#endif
#if defined(HAS_I422TORGB24ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToRGB24Row = I422ToRGB24Row_Any_MSA;
if (IS_ALIGNED(width, 16)) {
I422ToRGB24Row = I422ToRGB24Row_MSA;
}
}
#endif
#if defined(HAS_I422TORGB24ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToRGB24Row = I422ToRGB24Row_Any_LSX;
@ -6070,14 +5838,6 @@ int I420ToARGB1555(const uint8_t* src_y,
I422ToARGB1555Row = I422ToARGB1555Row_SME;
}
#endif
#if defined(HAS_I422TOARGB1555ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToARGB1555Row = I422ToARGB1555Row_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I422ToARGB1555Row = I422ToARGB1555Row_MSA;
}
}
#endif
#if defined(HAS_I422TOARGB1555ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToARGB1555Row = I422ToARGB1555Row_Any_LSX;
@ -6169,14 +5929,6 @@ int I420ToARGB4444(const uint8_t* src_y,
I422ToARGB4444Row = I422ToARGB4444Row_SME;
}
#endif
#if defined(HAS_I422TOARGB4444ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToARGB4444Row = I422ToARGB4444Row_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I422ToARGB4444Row = I422ToARGB4444Row_MSA;
}
}
#endif
#if defined(HAS_I422TOARGB4444ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToARGB4444Row = I422ToARGB4444Row_Any_LSX;
@ -6269,14 +6021,6 @@ int I420ToRGB565Matrix(const uint8_t* src_y,
I422ToRGB565Row = I422ToRGB565Row_SME;
}
#endif
#if defined(HAS_I422TORGB565ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToRGB565Row = I422ToRGB565Row_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I422ToRGB565Row = I422ToRGB565Row_MSA;
}
}
#endif
#if defined(HAS_I422TORGB565ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToRGB565Row = I422ToRGB565Row_Any_LSX;
@ -6419,14 +6163,6 @@ int I422ToRGB565Matrix(const uint8_t* src_y,
I422ToRGB565Row = I422ToRGB565Row_SME;
}
#endif
#if defined(HAS_I422TORGB565ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToRGB565Row = I422ToRGB565Row_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I422ToRGB565Row = I422ToRGB565Row_MSA;
}
}
#endif
#if defined(HAS_I422TORGB565ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToRGB565Row = I422ToRGB565Row_Any_LSX;
@ -6552,14 +6288,6 @@ int I420ToRGB565Dither(const uint8_t* src_y,
I422ToARGBRow = I422ToARGBRow_SME;
}
#endif
#if defined(HAS_I422TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToARGBRow = I422ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I422ToARGBRow = I422ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_I422TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToARGBRow = I422ToARGBRow_Any_LSX;
@ -6610,14 +6338,6 @@ int I420ToRGB565Dither(const uint8_t* src_y,
ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_SVE2;
}
#endif
#if defined(HAS_ARGBTORGB565DITHERROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTORGB565DITHERROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_LSX;
@ -6866,14 +6586,6 @@ static int I420ToARGBMatrixBilinear(const uint8_t* src_y,
I444ToARGBRow = I444ToARGBRow_SME;
}
#endif
#if defined(HAS_I444TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I444ToARGBRow = I444ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I444ToARGBRow = I444ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_I444TOARGBROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
I444ToARGBRow = I444ToARGBRow_Any_LASX;
@ -7023,14 +6735,6 @@ static int I422ToARGBMatrixLinear(const uint8_t* src_y,
I444ToARGBRow = I444ToARGBRow_SME;
}
#endif
#if defined(HAS_I444TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I444ToARGBRow = I444ToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I444ToARGBRow = I444ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_I444TOARGBROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
I444ToARGBRow = I444ToARGBRow_Any_LASX;
@ -7157,14 +6861,6 @@ static int I420ToRGB24MatrixBilinear(const uint8_t* src_y,
I444ToRGB24Row = I444ToRGB24Row_SME;
}
#endif
#if defined(HAS_I444TORGB24ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I444ToRGB24Row = I444ToRGB24Row_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I444ToRGB24Row = I444ToRGB24Row_MSA;
}
}
#endif
#if defined(HAS_I444TORGB24ROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
I444ToRGB24Row = I444ToRGB24Row_Any_LASX;
@ -7779,14 +7475,6 @@ static int I420AlphaToARGBMatrixBilinear(
I444AlphaToARGBRow = I444AlphaToARGBRow_SME;
}
#endif
#if defined(HAS_I444ALPHATOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_MSA;
}
}
#endif
#if defined(HAS_I444ALPHATOARGBROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_Any_LASX;
@ -7824,14 +7512,6 @@ static int I420AlphaToARGBMatrixBilinear(
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_MSA;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_RVV)
if (TestCpuFlag(kCpuHasRVV)) {
ARGBAttenuateRow = ARGBAttenuateRow_RVV;
@ -8015,14 +7695,6 @@ static int I422AlphaToARGBMatrixLinear(const uint8_t* src_y,
I444AlphaToARGBRow = I444AlphaToARGBRow_SME;
}
#endif
#if defined(HAS_I444ALPHATOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_MSA;
}
}
#endif
#if defined(HAS_I444ALPHATOARGBROW_LASX)
if (TestCpuFlag(kCpuHasLASX)) {
I444AlphaToARGBRow = I444AlphaToARGBRow_Any_LASX;
@ -8060,14 +7732,6 @@ static int I422AlphaToARGBMatrixLinear(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_MSA;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_RVV)
if (TestCpuFlag(kCpuHasRVV)) {
ARGBAttenuateRow = ARGBAttenuateRow_RVV;
@ -8240,14 +7904,6 @@ static int I010AlphaToARGBMatrixBilinear(
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_MSA;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_RVV)
if (TestCpuFlag(kCpuHasRVV)) {
ARGBAttenuateRow = ARGBAttenuateRow_RVV;
@ -8442,14 +8098,6 @@ static int I210AlphaToARGBMatrixLinear(const uint16_t* src_y,
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_MSA;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_RVV)
if (TestCpuFlag(kCpuHasRVV)) {
ARGBAttenuateRow = ARGBAttenuateRow_RVV;

24
source/convert_from.cc
View File

@ -445,14 +445,6 @@ int I420ToYUY2(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_I422TOYUY2ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToYUY2Row = I422ToYUY2Row_Any_MSA;
if (IS_ALIGNED(width, 32)) {
I422ToYUY2Row = I422ToYUY2Row_MSA;
}
}
#endif
#if defined(HAS_I422TOYUY2ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToYUY2Row = I422ToYUY2Row_Any_LSX;
@ -540,14 +532,6 @@ int I422ToUYVY(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_I422TOUYVYROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToUYVYRow = I422ToUYVYRow_Any_MSA;
if (IS_ALIGNED(width, 32)) {
I422ToUYVYRow = I422ToUYVYRow_MSA;
}
}
#endif
#if defined(HAS_I422TOUYVYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToUYVYRow = I422ToUYVYRow_Any_LSX;
@ -623,14 +607,6 @@ int I420ToUYVY(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_I422TOUYVYROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToUYVYRow = I422ToUYVYRow_Any_MSA;
if (IS_ALIGNED(width, 32)) {
I422ToUYVYRow = I422ToUYVYRow_MSA;
}
}
#endif
#if defined(HAS_I422TOUYVYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToUYVYRow = I422ToUYVYRow_Any_LSX;

311
source/convert_from_argb.cc
View File

@ -84,14 +84,6 @@ int ARGBToI444(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOUV444ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToUV444Row = ARGBToUV444Row_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToUV444Row = ARGBToUV444Row_MSA;
}
}
#endif
#if defined(HAS_ARGBTOUV444ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToUV444Row = ARGBToUV444Row_Any_LSX;
@ -140,14 +132,6 @@ int ARGBToI444(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYRow = ARGBToYRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
@ -295,18 +279,6 @@ int ARGBToI422(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYRow = ARGBToYRow_Any_MSA;
ARGBToUVRow = ARGBToUVRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_MSA;
}
if (IS_ALIGNED(width, 32)) {
ARGBToUVRow = ARGBToUVRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
@ -459,18 +431,6 @@ int ARGBToNV12(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYRow = ARGBToYRow_Any_MSA;
ARGBToUVRow = ARGBToUVRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_MSA;
}
if (IS_ALIGNED(width, 32)) {
ARGBToUVRow = ARGBToUVRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
@ -531,14 +491,6 @@ int ARGBToNV12(const uint8_t* src_argb,
MergeUVRow_ = MergeUVRow_SME;
}
#endif
#if defined(HAS_MERGEUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
MergeUVRow_ = MergeUVRow_Any_MSA;
if (IS_ALIGNED(halfwidth, 16)) {
MergeUVRow_ = MergeUVRow_MSA;
}
}
#endif
#if defined(HAS_MERGEUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MergeUVRow_ = MergeUVRow_Any_LSX;
@ -686,18 +638,6 @@ int ARGBToNV21(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYRow = ARGBToYRow_Any_MSA;
ARGBToUVRow = ARGBToUVRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_MSA;
}
if (IS_ALIGNED(width, 32)) {
ARGBToUVRow = ARGBToUVRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
@ -768,14 +708,6 @@ int ARGBToNV21(const uint8_t* src_argb,
MergeUVRow_ = MergeUVRow_SME;
}
#endif
#if defined(HAS_MERGEUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
MergeUVRow_ = MergeUVRow_Any_MSA;
if (IS_ALIGNED(halfwidth, 16)) {
MergeUVRow_ = MergeUVRow_MSA;
}
}
#endif
#if defined(HAS_MERGEUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MergeUVRow_ = MergeUVRow_Any_LSX;
@ -922,18 +854,6 @@ int ABGRToNV12(const uint8_t* src_abgr,
}
}
#endif
#if defined(HAS_ABGRTOYROW_MSA) && defined(HAS_ABGRTOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ABGRToYRow = ABGRToYRow_Any_MSA;
ABGRToUVRow = ABGRToUVRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ABGRToYRow = ABGRToYRow_MSA;
}
if (IS_ALIGNED(width, 32)) {
ABGRToUVRow = ABGRToUVRow_MSA;
}
}
#endif
#if defined(HAS_ABGRTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ABGRToYRow = ABGRToYRow_Any_LSX;
@ -992,14 +912,6 @@ int ABGRToNV12(const uint8_t* src_abgr,
MergeUVRow_ = MergeUVRow_SME;
}
#endif
#if defined(HAS_MERGEUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
MergeUVRow_ = MergeUVRow_Any_MSA;
if (IS_ALIGNED(halfwidth, 16)) {
MergeUVRow_ = MergeUVRow_MSA;
}
}
#endif
#if defined(HAS_MERGEUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MergeUVRow_ = MergeUVRow_Any_LSX;
@ -1147,18 +1059,6 @@ int ABGRToNV21(const uint8_t* src_abgr,
}
}
#endif
#if defined(HAS_ABGRTOYROW_MSA) && defined(HAS_ABGRTOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ABGRToYRow = ABGRToYRow_Any_MSA;
ABGRToUVRow = ABGRToUVRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ABGRToYRow = ABGRToYRow_MSA;
}
if (IS_ALIGNED(width, 32)) {
ABGRToUVRow = ABGRToUVRow_MSA;
}
}
#endif
#if defined(HAS_ABGRTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ABGRToYRow = ABGRToYRow_Any_LSX;
@ -1217,14 +1117,6 @@ int ABGRToNV21(const uint8_t* src_abgr,
MergeUVRow_ = MergeUVRow_SME;
}
#endif
#if defined(HAS_MERGEUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
MergeUVRow_ = MergeUVRow_Any_MSA;
if (IS_ALIGNED(halfwidth, 16)) {
MergeUVRow_ = MergeUVRow_MSA;
}
}
#endif
#if defined(HAS_MERGEUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MergeUVRow_ = MergeUVRow_Any_LSX;
@ -1377,18 +1269,6 @@ int ARGBToYUY2(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYRow = ARGBToYRow_Any_MSA;
ARGBToUVRow = ARGBToUVRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_MSA;
}
if (IS_ALIGNED(width, 32)) {
ARGBToUVRow = ARGBToUVRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
@ -1446,14 +1326,6 @@ int ARGBToYUY2(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_I422TOYUY2ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToYUY2Row = I422ToYUY2Row_Any_MSA;
if (IS_ALIGNED(width, 32)) {
I422ToYUY2Row = I422ToYUY2Row_MSA;
}
}
#endif
#if defined(HAS_I422TOYUY2ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToYUY2Row = I422ToYUY2Row_Any_LSX;
@ -1605,18 +1477,6 @@ int ARGBToUYVY(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYROW_MSA) && defined(HAS_ARGBTOUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYRow = ARGBToYRow_Any_MSA;
ARGBToUVRow = ARGBToUVRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_MSA;
}
if (IS_ALIGNED(width, 32)) {
ARGBToUVRow = ARGBToUVRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
@ -1674,14 +1534,6 @@ int ARGBToUYVY(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_I422TOUYVYROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToUYVYRow = I422ToUYVYRow_Any_MSA;
if (IS_ALIGNED(width, 32)) {
I422ToUYVYRow = I422ToUYVYRow_MSA;
}
}
#endif
#if defined(HAS_I422TOUYVYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToUYVYRow = I422ToUYVYRow_Any_LSX;
@ -1777,14 +1629,6 @@ int ARGBToI400(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYRow = ARGBToYRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYRow = ARGBToYRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYRow = ARGBToYRow_Any_LSX;
@ -1935,14 +1779,6 @@ int ARGBToRGB24(const uint8_t* src_argb,
ARGBToRGB24Row = ARGBToRGB24Row_SVE2;
}
#endif
#if defined(HAS_ARGBTORGB24ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToRGB24Row = ARGBToRGB24Row_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToRGB24Row = ARGBToRGB24Row_MSA;
}
}
#endif
#if defined(HAS_ARGBTORGB24ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToRGB24Row = ARGBToRGB24Row_Any_LSX;
@ -2027,14 +1863,6 @@ int ARGBToRAW(const uint8_t* src_argb,
ARGBToRAWRow = ARGBToRAWRow_SVE2;
}
#endif
#if defined(HAS_ARGBTORAWROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToRAWRow = ARGBToRAWRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToRAWRow = ARGBToRAWRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTORAWROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToRAWRow = ARGBToRAWRow_Any_LSX;
@ -2123,14 +1951,6 @@ int ARGBToRGB565Dither(const uint8_t* src_argb,
ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_SVE2;
}
#endif
#if defined(HAS_ARGBTORGB565DITHERROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTORGB565DITHERROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToRGB565DitherRow = ARGBToRGB565DitherRow_Any_LSX;
@ -2213,14 +2033,6 @@ int ARGBToRGB565(const uint8_t* src_argb,
ARGBToRGB565Row = ARGBToRGB565Row_SVE2;
}
#endif
#if defined(HAS_ARGBTORGB565ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToRGB565Row = ARGBToRGB565Row_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBToRGB565Row = ARGBToRGB565Row_MSA;
}
}
#endif
#if defined(HAS_ARGBTORGB565ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToRGB565Row = ARGBToRGB565Row_Any_LSX;
@ -2296,14 +2108,6 @@ int ARGBToARGB1555(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOARGB1555ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToARGB1555Row = ARGBToARGB1555Row_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBToARGB1555Row = ARGBToARGB1555Row_MSA;
}
}
#endif
#if defined(HAS_ARGBTOARGB1555ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToARGB1555Row = ARGBToARGB1555Row_Any_LSX;
@ -2378,14 +2182,6 @@ int ARGBToARGB4444(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOARGB4444ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToARGB4444Row = ARGBToARGB4444Row_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBToARGB4444Row = ARGBToARGB4444Row_MSA;
}
}
#endif
#if defined(HAS_ARGBTOARGB4444ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToARGB4444Row = ARGBToARGB4444Row_Any_LSX;
@ -2589,14 +2385,6 @@ int ARGBToJ444(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOUVJ444ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToUVJ444Row = ARGBToUVJ444Row_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToUVJ444Row = ARGBToUVJ444Row_MSA;
}
}
#endif
#if defined(HAS_ARGBTOUVJ444ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToUVJ444Row = ARGBToUVJ444Row_Any_LSX;
@ -2645,14 +2433,6 @@ int ARGBToJ444(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYJROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYJRow = ARGBToYJRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYJRow = ARGBToYJRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYJRow = ARGBToYJRow_Any_LSX;
@ -2793,18 +2573,6 @@ int ARGBToJ420(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYJROW_MSA) && defined(HAS_ARGBTOUVJROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYJRow = ARGBToYJRow_Any_MSA;
ARGBToUVJRow = ARGBToUVJRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYJRow = ARGBToYJRow_MSA;
}
if (IS_ALIGNED(width, 32)) {
ARGBToUVJRow = ARGBToUVJRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYJROW_LSX) && defined(HAS_ARGBTOUVJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYJRow = ARGBToYJRow_Any_LSX;
@ -2961,18 +2729,6 @@ int ARGBToJ422(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYJROW_MSA) && defined(HAS_ARGBTOUVJROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYJRow = ARGBToYJRow_Any_MSA;
ARGBToUVJRow = ARGBToUVJRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYJRow = ARGBToYJRow_MSA;
}
if (IS_ALIGNED(width, 32)) {
ARGBToUVJRow = ARGBToUVJRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYJROW_LSX) && defined(HAS_ARGBTOUVJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYJRow = ARGBToYJRow_Any_LSX;
@ -3067,14 +2823,6 @@ int ARGBToJ400(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYJROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYJRow = ARGBToYJRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYJRow = ARGBToYJRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYJROW_RVV)
if (TestCpuFlag(kCpuHasRVV)) {
ARGBToYJRow = ARGBToYJRow_RVV;
@ -3146,14 +2894,6 @@ int RGBAToJ400(const uint8_t* src_rgba,
}
}
#endif
#if defined(HAS_RGBATOYJROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RGBAToYJRow = RGBAToYJRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RGBAToYJRow = RGBAToYJRow_MSA;
}
}
#endif
#if defined(HAS_RGBATOYJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RGBAToYJRow = RGBAToYJRow_Any_LSX;
@ -3291,16 +3031,6 @@ int ABGRToJ420(const uint8_t* src_abgr,
}
}
#endif
#if defined(HAS_ABGRTOYJROW_MSA) && defined(HAS_ABGRTOUVJROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ABGRToYJRow = ABGRToYJRow_Any_MSA;
ABGRToUVJRow = ABGRToUVJRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ABGRToYJRow = ABGRToYJRow_MSA;
ABGRToUVJRow = ABGRToUVJRow_MSA;
}
}
#endif
#if defined(HAS_ABGRTOYJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ABGRToYJRow = ABGRToYJRow_Any_LSX;
@ -3453,18 +3183,6 @@ int ABGRToJ422(const uint8_t* src_abgr,
}
}
#endif
#if defined(HAS_ABGRTOYJROW_MSA) && defined(HAS_ABGRTOUVJROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ABGRToYJRow = ABGRToYJRow_Any_MSA;
ABGRToUVJRow = ABGRToUVJRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ABGRToYJRow = ABGRToYJRow_MSA;
}
if (IS_ALIGNED(width, 32)) {
ABGRToUVJRow = ABGRToUVJRow_MSA;
}
}
#endif
#if defined(HAS_ABGRTOYJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ABGRToYJRow = ABGRToYJRow_Any_LSX;
@ -3555,14 +3273,6 @@ int ABGRToJ400(const uint8_t* src_abgr,
}
}
#endif
#if defined(HAS_ABGRTOYJROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ABGRToYJRow = ABGRToYJRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ABGRToYJRow = ABGRToYJRow_MSA;
}
}
#endif
#if defined(HAS_ABGRTOYJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ABGRToYJRow = ABGRToYJRow_Any_LSX;
@ -3722,8 +3432,7 @@ int ARGBToAB64(const uint8_t* src_argb,
}
// Enabled if 1 pass is available
#if defined(HAS_RAWTOYJROW_NEON) || defined(HAS_RAWTOYJROW_MSA) || \
defined(HAS_RAWTOYJROW_RVV)
#if defined(HAS_RAWTOYJROW_NEON) || defined(HAS_RAWTOYJROW_RVV)
#define HAS_RAWTOYJROW
#endif
@ -3779,16 +3488,6 @@ int RAWToJNV21(const uint8_t* src_raw,
}
}
#endif
#if defined(HAS_RAWTOYJROW_MSA) && defined(HAS_RAWTOUVJROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RAWToUVJRow = RAWToUVJRow_Any_MSA;
RAWToYJRow = RAWToYJRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RAWToYJRow = RAWToYJRow_MSA;
RAWToUVJRow = RAWToUVJRow_MSA;
}
}
#endif
#if defined(HAS_RAWTOYJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RAWToYJRow = RAWToYJRow_Any_LSX;
@ -3900,14 +3599,6 @@ int RAWToJNV21(const uint8_t* src_raw,
MergeUVRow_ = MergeUVRow_SME;
}
#endif
#if defined(HAS_MERGEUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
MergeUVRow_ = MergeUVRow_Any_MSA;
if (IS_ALIGNED(halfwidth, 16)) {
MergeUVRow_ = MergeUVRow_MSA;
}
}
#endif
#if defined(HAS_MERGEUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MergeUVRow_ = MergeUVRow_Any_LSX;

37
source/cpu_id.cc
View File

@ -359,42 +359,13 @@ LIBYUV_API SAFEBUFFERS int RiscvCpuCaps(const char* cpuinfo_name) {
return flag;
}
LIBYUV_API SAFEBUFFERS int MipsCpuCaps(const char* cpuinfo_name) {
char cpuinfo_line[512];
int flag = 0;
FILE* f = fopen(cpuinfo_name, "re");
if (!f) {
// Assume nothing if /proc/cpuinfo is unavailable.
// This will occur for Chrome sandbox for Pepper or Render process.
return 0;
}
memset(cpuinfo_line, 0, sizeof(cpuinfo_line));
while (fgets(cpuinfo_line, sizeof(cpuinfo_line), f)) {
if (memcmp(cpuinfo_line, "cpu model", 9) == 0) {
// Workaround early kernel without MSA in ASEs line.
if (strstr(cpuinfo_line, "Loongson-2K")) {
flag |= kCpuHasMSA;
}
}
if (memcmp(cpuinfo_line, "ASEs implemented", 16) == 0) {
if (strstr(cpuinfo_line, "msa")) {
flag |= kCpuHasMSA;
}
// ASEs is the last line, so we can break here.
break;
}
}
fclose(f);
return flag;
}
#if defined(__loongarch__) && defined(__linux__)
// Define hwcap values ourselves: building with an old auxv header where these
// hwcap values are not defined should not prevent features from being enabled.
#define YUV_LOONGARCH_HWCAP_LSX (1 << 4)
#define YUV_LOONGARCH_HWCAP_LASX (1 << 5)
LIBYUV_API SAFEBUFFERS int LoongarchCpuCaps(void) {
LIBYUV_API SAFEBUFFERS int LoongArchCpuCaps(void) {
int flag = 0;
unsigned long hwcap = getauxval(AT_HWCAP);
@ -462,12 +433,8 @@ static SAFEBUFFERS int GetCpuFlags(void) {
}
}
#endif
#if defined(__mips__) && defined(__linux__)
cpu_info = MipsCpuCaps("/proc/cpuinfo");
cpu_info |= kCpuHasMIPS;
#endif
#if defined(__loongarch__) && defined(__linux__)
cpu_info = LoongarchCpuCaps();
cpu_info = LoongArchCpuCaps();
cpu_info |= kCpuHasLOONGARCH;
#endif
#if defined(__aarch64__)

268
source/planar_functions.cc
View File

@ -631,14 +631,6 @@ void SplitUVPlane(const uint8_t* src_uv,
}
}
#endif
#if defined(HAS_SPLITUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
SplitUVRow = SplitUVRow_Any_MSA;
if (IS_ALIGNED(width, 32)) {
SplitUVRow = SplitUVRow_MSA;
}
}
#endif
#if defined(HAS_SPLITUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
SplitUVRow = SplitUVRow_Any_LSX;
@ -727,14 +719,6 @@ void MergeUVPlane(const uint8_t* src_u,
MergeUVRow = MergeUVRow_SME;
}
#endif
#if defined(HAS_MERGEUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
MergeUVRow = MergeUVRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
MergeUVRow = MergeUVRow_MSA;
}
}
#endif
#if defined(HAS_MERGEUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MergeUVRow = MergeUVRow_Any_LSX;
@ -2251,16 +2235,6 @@ int YUY2ToI422(const uint8_t* src_yuy2,
}
}
#endif
#if defined(HAS_YUY2TOYROW_MSA) && defined(HAS_YUY2TOUV422ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
YUY2ToYRow = YUY2ToYRow_Any_MSA;
YUY2ToUV422Row = YUY2ToUV422Row_Any_MSA;
if (IS_ALIGNED(width, 32)) {
YUY2ToYRow = YUY2ToYRow_MSA;
YUY2ToUV422Row = YUY2ToUV422Row_MSA;
}
}
#endif
#if defined(HAS_YUY2TOYROW_LSX) && defined(HAS_YUY2TOUV422ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
YUY2ToYRow = YUY2ToYRow_Any_LSX;
@ -2357,16 +2331,6 @@ int UYVYToI422(const uint8_t* src_uyvy,
}
}
#endif
#if defined(HAS_UYVYTOYROW_MSA) && defined(HAS_UYVYTOUV422ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
UYVYToYRow = UYVYToYRow_Any_MSA;
UYVYToUV422Row = UYVYToUV422Row_Any_MSA;
if (IS_ALIGNED(width, 32)) {
UYVYToYRow = UYVYToYRow_MSA;
UYVYToUV422Row = UYVYToUV422Row_MSA;
}
}
#endif
#if defined(HAS_UYVYTOYROW_LSX) && defined(HAS_UYVYTOUV422ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
UYVYToYRow = UYVYToYRow_Any_LSX;
@ -2449,14 +2413,6 @@ int YUY2ToY(const uint8_t* src_yuy2,
}
}
#endif
#if defined(HAS_YUY2TOYROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
YUY2ToYRow = YUY2ToYRow_Any_MSA;
if (IS_ALIGNED(width, 32)) {
YUY2ToYRow = YUY2ToYRow_MSA;
}
}
#endif
for (y = 0; y < height; ++y) {
YUY2ToYRow(src_yuy2, dst_y, width);
@ -2516,14 +2472,6 @@ int UYVYToY(const uint8_t* src_uyvy,
}
}
#endif
#if defined(HAS_UYVYTOYROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
UYVYToYRow = UYVYToYRow_Any_MSA;
if (IS_ALIGNED(width, 32)) {
UYVYToYRow = UYVYToYRow_MSA;
}
}
#endif
#if defined(HAS_UYVYTOYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
UYVYToYRow = UYVYToYRow_Any_LSX;
@ -2582,14 +2530,6 @@ void MirrorPlane(const uint8_t* src_y,
}
}
#endif
#if defined(HAS_MIRRORROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
MirrorRow = MirrorRow_Any_MSA;
if (IS_ALIGNED(width, 64)) {
MirrorRow = MirrorRow_MSA;
}
}
#endif
#if defined(HAS_MIRRORROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MirrorRow = MirrorRow_Any_LSX;
@ -2656,14 +2596,6 @@ void MirrorUVPlane(const uint8_t* src_uv,
}
}
#endif
#if defined(HAS_MIRRORUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
MirrorUVRow = MirrorUVRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
MirrorUVRow = MirrorUVRow_MSA;
}
}
#endif
#if defined(HAS_MIRRORUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MirrorUVRow = MirrorUVRow_Any_LSX;
@ -2836,14 +2768,6 @@ int ARGBMirror(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBMIRRORROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBMirrorRow = ARGBMirrorRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBMirrorRow = ARGBMirrorRow_MSA;
}
}
#endif
#if defined(HAS_ARGBMIRRORROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBMirrorRow = ARGBMirrorRow_Any_LSX;
@ -2955,11 +2879,6 @@ int ARGBBlend(const uint8_t* src_argb0,
ARGBBlendRow = ARGBBlendRow_NEON;
}
#endif
#if defined(HAS_ARGBBLENDROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBBlendRow = ARGBBlendRow_MSA;
}
#endif
#if defined(HAS_ARGBBLENDROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBBlendRow = ARGBBlendRow_LSX;
@ -3241,14 +3160,6 @@ int ARGBMultiply(const uint8_t* src_argb0,
ARGBMultiplyRow = ARGBMultiplyRow_SME;
}
#endif
#if defined(HAS_ARGBMULTIPLYROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBMultiplyRow = ARGBMultiplyRow_Any_MSA;
if (IS_ALIGNED(width, 4)) {
ARGBMultiplyRow = ARGBMultiplyRow_MSA;
}
}
#endif
#if defined(HAS_ARGBMULTIPLYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBMultiplyRow = ARGBMultiplyRow_Any_LSX;
@ -3334,14 +3245,6 @@ int ARGBAdd(const uint8_t* src_argb0,
}
}
#endif
#if defined(HAS_ARGBADDROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBAddRow = ARGBAddRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBAddRow = ARGBAddRow_MSA;
}
}
#endif
#if defined(HAS_ARGBADDROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBAddRow = ARGBAddRow_Any_LSX;
@ -3422,14 +3325,6 @@ int ARGBSubtract(const uint8_t* src_argb0,
}
}
#endif
#if defined(HAS_ARGBSUBTRACTROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBSubtractRow = ARGBSubtractRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBSubtractRow = ARGBSubtractRow_MSA;
}
}
#endif
#if defined(HAS_ARGBSUBTRACTROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBSubtractRow = ARGBSubtractRow_Any_LSX;
@ -3504,14 +3399,6 @@ int RAWToRGB24(const uint8_t* src_raw,
RAWToRGB24Row = RAWToRGB24Row_SVE2;
}
#endif
#if defined(HAS_RAWTORGB24ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
RAWToRGB24Row = RAWToRGB24Row_Any_MSA;
if (IS_ALIGNED(width, 16)) {
RAWToRGB24Row = RAWToRGB24Row_MSA;
}
}
#endif
#if defined(HAS_RAWTORGB24ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
RAWToRGB24Row = RAWToRGB24Row_Any_LSX;
@ -3579,11 +3466,6 @@ void SetPlane(uint8_t* dst_y,
SetRow = SetRow_ERMS;
}
#endif
#if defined(HAS_SETROW_MSA)
if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 16)) {
SetRow = SetRow_MSA;
}
#endif
#if defined(HAS_SETROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
SetRow = SetRow_Any_LSX;
@ -3674,14 +3556,6 @@ int ARGBRect(uint8_t* dst_argb,
ARGBSetRow = ARGBSetRow_X86;
}
#endif
#if defined(HAS_ARGBSETROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBSetRow = ARGBSetRow_Any_MSA;
if (IS_ALIGNED(width, 4)) {
ARGBSetRow = ARGBSetRow_MSA;
}
}
#endif
#if defined(HAS_ARGBSETROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBSetRow = ARGBSetRow_Any_LSX;
@ -3760,14 +3634,6 @@ int ARGBAttenuate(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBAttenuateRow = ARGBAttenuateRow_MSA;
}
}
#endif
#if defined(HAS_ARGBATTENUATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBAttenuateRow = ARGBAttenuateRow_Any_LSX;
@ -3889,11 +3755,6 @@ int ARGBGrayTo(const uint8_t* src_argb,
ARGBGrayRow = ARGBGrayRow_NEON_DotProd;
}
#endif
#if defined(HAS_ARGBGRAYROW_MSA)
if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 8)) {
ARGBGrayRow = ARGBGrayRow_MSA;
}
#endif
#if defined(HAS_ARGBGRAYROW_LSX)
if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 8)) {
ARGBGrayRow = ARGBGrayRow_LSX;
@ -3949,11 +3810,6 @@ int ARGBGray(uint8_t* dst_argb,
ARGBGrayRow = ARGBGrayRow_NEON_DotProd;
}
#endif
#if defined(HAS_ARGBGRAYROW_MSA)
if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 8)) {
ARGBGrayRow = ARGBGrayRow_MSA;
}
#endif
#if defined(HAS_ARGBGRAYROW_LSX)
if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 8)) {
ARGBGrayRow = ARGBGrayRow_LSX;
@ -4007,11 +3863,6 @@ int ARGBSepia(uint8_t* dst_argb,
ARGBSepiaRow = ARGBSepiaRow_NEON_DotProd;
}
#endif
#if defined(HAS_ARGBSEPIAROW_MSA)
if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 8)) {
ARGBSepiaRow = ARGBSepiaRow_MSA;
}
#endif
#if defined(HAS_ARGBSEPIAROW_LSX)
if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 8)) {
ARGBSepiaRow = ARGBSepiaRow_LSX;
@ -4073,11 +3924,6 @@ int ARGBColorMatrix(const uint8_t* src_argb,
ARGBColorMatrixRow = ARGBColorMatrixRow_NEON_I8MM;
}
#endif
#if defined(HAS_ARGBCOLORMATRIXROW_MSA)
if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 8)) {
ARGBColorMatrixRow = ARGBColorMatrixRow_MSA;
}
#endif
#if defined(HAS_ARGBCOLORMATRIXROW_LSX)
if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 8)) {
ARGBColorMatrixRow = ARGBColorMatrixRow_LSX;
@ -4243,11 +4089,6 @@ int ARGBQuantize(uint8_t* dst_argb,
ARGBQuantizeRow = ARGBQuantizeRow_NEON;
}
#endif
#if defined(HAS_ARGBQUANTIZEROW_MSA)
if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 8)) {
ARGBQuantizeRow = ARGBQuantizeRow_MSA;
}
#endif
#if defined(HAS_ARGBQUANTIZEROW_LSX)
if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 8)) {
ARGBQuantizeRow = ARGBQuantizeRow_LSX;
@ -4442,11 +4283,6 @@ int ARGBShade(const uint8_t* src_argb,
ARGBShadeRow = ARGBShadeRow_NEON;
}
#endif
#if defined(HAS_ARGBSHADEROW_MSA)
if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 4)) {
ARGBShadeRow = ARGBShadeRow_MSA;
}
#endif
#if defined(HAS_ARGBSHADEROW_LSX)
if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 4)) {
ARGBShadeRow = ARGBShadeRow_LSX;
@ -4525,14 +4361,6 @@ int InterpolatePlane(const uint8_t* src0,
InterpolateRow = InterpolateRow_SME;
}
#endif
#if defined(HAS_INTERPOLATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
InterpolateRow = InterpolateRow_Any_MSA;
if (IS_ALIGNED(width, 32)) {
InterpolateRow = InterpolateRow_MSA;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
@ -4615,14 +4443,6 @@ int InterpolatePlane_16(const uint16_t* src0,
InterpolateRow_16 = InterpolateRow_16_SME;
}
#endif
#if defined(HAS_INTERPOLATEROW_16_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
InterpolateRow_16 = InterpolateRow_16_Any_MSA;
if (IS_ALIGNED(width, 32)) {
InterpolateRow_16 = InterpolateRow_16_MSA;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_16_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow_16 = InterpolateRow_16_Any_LSX;
@ -4748,14 +4568,6 @@ int ARGBShuffle(const uint8_t* src_bgra,
}
}
#endif
#if defined(HAS_ARGBSHUFFLEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBShuffleRow = ARGBShuffleRow_Any_MSA;
if (IS_ALIGNED(width, 8)) {
ARGBShuffleRow = ARGBShuffleRow_MSA;
}
}
#endif
#if defined(HAS_ARGBSHUFFLEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBShuffleRow = ARGBShuffleRow_Any_LSX;
@ -4973,14 +4785,6 @@ static int ARGBSobelize(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBTOYJROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBToYJRow = ARGBToYJRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBToYJRow = ARGBToYJRow_MSA;
}
}
#endif
#if defined(HAS_ARGBTOYJROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBToYJRow = ARGBToYJRow_Any_LSX;
@ -5013,11 +4817,6 @@ static int ARGBSobelize(const uint8_t* src_argb,
SobelYRow = SobelYRow_NEON;
}
#endif
#if defined(HAS_SOBELYROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
SobelYRow = SobelYRow_MSA;
}
#endif
#if defined(HAS_SOBELXROW_SSE2)
if (TestCpuFlag(kCpuHasSSE2)) {
SobelXRow = SobelXRow_SSE2;
@ -5027,11 +4826,6 @@ static int ARGBSobelize(const uint8_t* src_argb,
if (TestCpuFlag(kCpuHasNEON)) {
SobelXRow = SobelXRow_NEON;
}
#endif
#if defined(HAS_SOBELXROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
SobelXRow = SobelXRow_MSA;
}
#endif
{
// 3 rows with edges before/after.
@ -5109,14 +4903,6 @@ int ARGBSobel(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_SOBELROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
SobelRow = SobelRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
SobelRow = SobelRow_MSA;
}
}
#endif
#if defined(HAS_SOBELROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
SobelRow = SobelRow_Any_LSX;
@ -5155,14 +4941,6 @@ int ARGBSobelToPlane(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_SOBELTOPLANEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
SobelToPlaneRow = SobelToPlaneRow_Any_MSA;
if (IS_ALIGNED(width, 32)) {
SobelToPlaneRow = SobelToPlaneRow_MSA;
}
}
#endif
#if defined(HAS_SOBELTOPLANEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
SobelToPlaneRow = SobelToPlaneRow_Any_LSX;
@ -5202,14 +4980,6 @@ int ARGBSobelXY(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_SOBELXYROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
SobelXYRow = SobelXYRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
SobelXYRow = SobelXYRow_MSA;
}
}
#endif
#if defined(HAS_SOBELXYROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
SobelXYRow = SobelXYRow_Any_LSX;
@ -5345,14 +5115,6 @@ int HalfFloatPlane(const uint16_t* src_y,
HalfFloatRow = scale == 1.0f ? HalfFloat1Row_SVE2 : HalfFloatRow_SVE2;
}
#endif
#if defined(HAS_HALFFLOATROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
HalfFloatRow = HalfFloatRow_Any_MSA;
if (IS_ALIGNED(width, 32)) {
HalfFloatRow = HalfFloatRow_MSA;
}
}
#endif
#if defined(HAS_HALFFLOATROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
HalfFloatRow = HalfFloatRow_Any_LSX;
@ -5527,12 +5289,6 @@ int ARGBExtractAlpha(const uint8_t* src_argb,
: ARGBExtractAlphaRow_Any_NEON;
}
#endif
#if defined(HAS_ARGBEXTRACTALPHAROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBExtractAlphaRow = IS_ALIGNED(width, 16) ? ARGBExtractAlphaRow_MSA
: ARGBExtractAlphaRow_Any_MSA;
}
#endif
#if defined(HAS_ARGBEXTRACTALPHAROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBExtractAlphaRow = IS_ALIGNED(width, 16) ? ARGBExtractAlphaRow_LSX
@ -5657,14 +5413,6 @@ int YUY2ToNV12(const uint8_t* src_yuy2,
}
}
#endif
#if defined(HAS_YUY2TOYROW_MSA) && defined(HAS_YUY2TOUV422ROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
YUY2ToYRow = YUY2ToYRow_Any_MSA;
if (IS_ALIGNED(width, 32)) {
YUY2ToYRow = YUY2ToYRow_MSA;
}
}
#endif
#if defined(HAS_YUY2TOYROW_LSX) && defined(HAS_YUY2TOUV422ROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
YUY2ToYRow = YUY2ToYRow_Any_LSX;
@ -5773,14 +5521,6 @@ int UYVYToNV12(const uint8_t* src_uyvy,
}
}
#endif
#if defined(HAS_SPLITUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
SplitUVRow = SplitUVRow_Any_MSA;
if (IS_ALIGNED(width, 32)) {
SplitUVRow = SplitUVRow_MSA;
}
}
#endif
#if defined(HAS_SPLITUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
SplitUVRow = SplitUVRow_Any_LSX;
@ -5824,14 +5564,6 @@ int UYVYToNV12(const uint8_t* src_uyvy,
InterpolateRow = InterpolateRow_SME;
}
#endif
#if defined(HAS_INTERPOLATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
InterpolateRow = InterpolateRow_Any_MSA;
if (IS_ALIGNED(width, 32)) {
InterpolateRow = InterpolateRow_MSA;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;

57
source/rotate.cc
View File

@ -35,8 +35,7 @@ void TransposePlane(const uint8_t* src,
void (*TransposeWxH)(const uint8_t* src, int src_stride, uint8_t* dst,
int dst_stride, int width, int height) = NULL;
#endif
#if defined(HAS_TRANSPOSEWX16_MSA) || defined(HAS_TRANSPOSEWX16_LSX) || \
defined(HAS_TRANSPOSEWX16_NEON)
#if defined(HAS_TRANSPOSEWX16_LSX) || defined(HAS_TRANSPOSEWX16_NEON)
void (*TransposeWx16)(const uint8_t* src, int src_stride, uint8_t* dst,
int dst_stride, int width) = TransposeWx16_C;
#else
@ -81,14 +80,6 @@ void TransposePlane(const uint8_t* src,
}
}
#endif
#if defined(HAS_TRANSPOSEWX16_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
TransposeWx16 = TransposeWx16_Any_MSA;
if (IS_ALIGNED(width, 16)) {
TransposeWx16 = TransposeWx16_MSA;
}
}
#endif
#if defined(HAS_TRANSPOSEWX16_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
TransposeWx16 = TransposeWx16_Any_LSX;
@ -104,8 +95,7 @@ void TransposePlane(const uint8_t* src,
return;
}
#endif
#if defined(HAS_TRANSPOSEWX16_MSA) || defined(HAS_TRANSPOSEWX16_LSX) || \
defined(HAS_TRANSPOSEWX16_NEON)
#if defined(HAS_TRANSPOSEWX16_LSX) || defined(HAS_TRANSPOSEWX16_NEON)
// Work across the source in 16x16 tiles
while (i >= 16) {
TransposeWx16(src, src_stride, dst, dst_stride, width);
@ -200,14 +190,6 @@ void RotatePlane180(const uint8_t* src,
}
}
#endif
#if defined(HAS_MIRRORROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
MirrorRow = MirrorRow_Any_MSA;
if (IS_ALIGNED(width, 64)) {
MirrorRow = MirrorRow_MSA;
}
}
#endif
#if defined(HAS_MIRRORROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
MirrorRow = MirrorRow_Any_LSX;
@ -288,11 +270,7 @@ void SplitTransposeUV(const uint8_t* src,
int dst_stride_a, uint8_t* dst_b, int dst_stride_b,
int width, int height) = TransposeUVWxH_C;
#endif
#if defined(HAS_TRANSPOSEUVWX16_MSA)
void (*TransposeUVWx16)(const uint8_t* src, int src_stride, uint8_t* dst_a,
int dst_stride_a, uint8_t* dst_b, int dst_stride_b,
int width) = TransposeUVWx16_C;
#elif defined(HAS_TRANSPOSEUVWX16_LSX)
#if defined(HAS_TRANSPOSEUVWX16_LSX)
void (*TransposeUVWx16)(const uint8_t* src, int src_stride, uint8_t* dst_a,
int dst_stride_a, uint8_t* dst_b, int dst_stride_b,
int width) = TransposeUVWx16_C;
@ -302,21 +280,14 @@ void SplitTransposeUV(const uint8_t* src,
int width) = TransposeUVWx8_C;
#endif
#if defined(HAS_TRANSPOSEUVWX16_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
TransposeUVWx16 = TransposeUVWx16_Any_MSA;
if (IS_ALIGNED(width, 8)) {
TransposeUVWx16 = TransposeUVWx16_MSA;
}
}
#elif defined(HAS_TRANSPOSEUVWX16_LSX)
#if defined(HAS_TRANSPOSEUVWX16_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
TransposeUVWx16 = TransposeUVWx16_Any_LSX;
if (IS_ALIGNED(width, 8)) {
TransposeUVWx16 = TransposeUVWx16_LSX;
}
}
#else
#endif
#if defined(HAS_TRANSPOSEUVWX8_NEON)
if (TestCpuFlag(kCpuHasNEON)) {
TransposeUVWx8 = TransposeUVWx8_Any_NEON;
@ -338,7 +309,6 @@ void SplitTransposeUV(const uint8_t* src,
}
}
#endif
#endif /* defined(HAS_TRANSPOSEUVWX16_MSA) */
#if defined(HAS_TRANSPOSEUVWXH_SME)
if (TestCpuFlag(kCpuHasSME)) {
@ -347,17 +317,7 @@ void SplitTransposeUV(const uint8_t* src,
return;
}
#endif
#if defined(HAS_TRANSPOSEUVWX16_MSA)
// Work through the source in 8x8 tiles.
while (i >= 16) {
TransposeUVWx16(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b,
width);
src += 16 * src_stride; // Go down 16 rows.
dst_a += 16; // Move over 8 columns.
dst_b += 16; // Move over 8 columns.
i -= 16;
}
#elif defined(HAS_TRANSPOSEUVWX16_LSX)
#if defined(HAS_TRANSPOSEUVWX16_LSX)
// Work through the source in 8x8 tiles.
while (i >= 16) {
TransposeUVWx16(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b,
@ -442,11 +402,6 @@ void SplitRotateUV180(const uint8_t* src,
MirrorSplitUVRow = MirrorSplitUVRow_SSSE3;
}
#endif
#if defined(HAS_MIRRORSPLITUVROW_MSA)
if (TestCpuFlag(kCpuHasMSA) && IS_ALIGNED(width, 32)) {
MirrorSplitUVRow = MirrorSplitUVRow_MSA;
}
#endif
#if defined(HAS_MIRRORSPLITUVROW_LSX)
if (TestCpuFlag(kCpuHasLSX) && IS_ALIGNED(width, 32)) {
MirrorSplitUVRow = MirrorSplitUVRow_LSX;

6
source/rotate_any.cc
View File

@ -41,9 +41,6 @@ TANY(TransposeWx8_Any_SSSE3, TransposeWx8_SSSE3, TransposeWx8_C, 7)
#ifdef HAS_TRANSPOSEWX8_FAST_SSSE3
TANY(TransposeWx8_Fast_Any_SSSE3, TransposeWx8_Fast_SSSE3, TransposeWx8_C, 15)
#endif
#ifdef HAS_TRANSPOSEWX16_MSA
TANY(TransposeWx16_Any_MSA, TransposeWx16_MSA, TransposeWx16_C, 15)
#endif
#ifdef HAS_TRANSPOSEWX16_LSX
TANY(TransposeWx16_Any_LSX, TransposeWx16_LSX, TransposeWx16_C, 15)
#endif
@ -68,9 +65,6 @@ TUVANY(TransposeUVWx8_Any_NEON, TransposeUVWx8_NEON, 7)
#ifdef HAS_TRANSPOSEUVWX8_SSE2
TUVANY(TransposeUVWx8_Any_SSE2, TransposeUVWx8_SSE2, 7)
#endif
#ifdef HAS_TRANSPOSEUVWX16_MSA
TUVANY(TransposeUVWx16_Any_MSA, TransposeUVWx16_MSA, 7)
#endif
#ifdef HAS_TRANSPOSEUVWX16_LSX
TUVANY(TransposeUVWx16_Any_LSX, TransposeUVWx16_LSX, 7)
#endif

16
source/rotate_argb.cc
View File

@ -53,14 +53,6 @@ static int ARGBTranspose(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_SCALEARGBROWDOWNEVEN_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ScaleARGBRowDownEven = ScaleARGBRowDownEven_Any_MSA;
if (IS_ALIGNED(height, 4)) { // Width of dest.
ScaleARGBRowDownEven = ScaleARGBRowDownEven_MSA;
}
}
#endif
#if defined(HAS_SCALEARGBROWDOWNEVEN_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ScaleARGBRowDownEven = ScaleARGBRowDownEven_Any_LSX;
@ -155,14 +147,6 @@ static int ARGBRotate180(const uint8_t* src_argb,
}
}
#endif
#if defined(HAS_ARGBMIRRORROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ARGBMirrorRow = ARGBMirrorRow_Any_MSA;
if (IS_ALIGNED(width, 16)) {
ARGBMirrorRow = ARGBMirrorRow_MSA;
}
}
#endif
#if defined(HAS_ARGBMIRRORROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ARGBMirrorRow = ARGBMirrorRow_Any_LSX;

240
source/rotate_msa.cc
View File

@ -1,240 +0,0 @@
/*
* Copyright 2016 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.
*/
#include "libyuv/rotate_row.h"
// This module is for GCC MSA
#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)
#include "libyuv/macros_msa.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
#define ILVRL_B(in0, in1, in2, in3, out0, out1, out2, out3) \
{ \
out0 = (v16u8)__msa_ilvr_b((v16i8)in1, (v16i8)in0); \
out1 = (v16u8)__msa_ilvl_b((v16i8)in1, (v16i8)in0); \
out2 = (v16u8)__msa_ilvr_b((v16i8)in3, (v16i8)in2); \
out3 = (v16u8)__msa_ilvl_b((v16i8)in3, (v16i8)in2); \
}
#define ILVRL_H(in0, in1, in2, in3, out0, out1, out2, out3) \
{ \
out0 = (v16u8)__msa_ilvr_h((v8i16)in1, (v8i16)in0); \
out1 = (v16u8)__msa_ilvl_h((v8i16)in1, (v8i16)in0); \
out2 = (v16u8)__msa_ilvr_h((v8i16)in3, (v8i16)in2); \
out3 = (v16u8)__msa_ilvl_h((v8i16)in3, (v8i16)in2); \
}
#define ILVRL_W(in0, in1, in2, in3, out0, out1, out2, out3) \
{ \
out0 = (v16u8)__msa_ilvr_w((v4i32)in1, (v4i32)in0); \
out1 = (v16u8)__msa_ilvl_w((v4i32)in1, (v4i32)in0); \
out2 = (v16u8)__msa_ilvr_w((v4i32)in3, (v4i32)in2); \
out3 = (v16u8)__msa_ilvl_w((v4i32)in3, (v4i32)in2); \
}
#define ILVRL_D(in0, in1, in2, in3, out0, out1, out2, out3) \
{ \
out0 = (v16u8)__msa_ilvr_d((v2i64)in1, (v2i64)in0); \
out1 = (v16u8)__msa_ilvl_d((v2i64)in1, (v2i64)in0); \
out2 = (v16u8)__msa_ilvr_d((v2i64)in3, (v2i64)in2); \
out3 = (v16u8)__msa_ilvl_d((v2i64)in3, (v2i64)in2); \
}
void TransposeUVWx16_C(const uint8_t* src,
int src_stride,
uint8_t* dst_a,
int dst_stride_a,
uint8_t* dst_b,
int dst_stride_b,
int width) {
TransposeUVWx8_C(src, src_stride, dst_a, dst_stride_a, dst_b, dst_stride_b,
width);
TransposeUVWx8_C((src + 8 * src_stride), src_stride, (dst_a + 8),
dst_stride_a, (dst_b + 8), dst_stride_b, width);
}
void TransposeWx16_MSA(const uint8_t* src,
int src_stride,
uint8_t* dst,
int dst_stride,
int width) {
int x;
const uint8_t* s;
v16u8 src0, src1, src2, src3, dst0, dst1, dst2, dst3, vec0, vec1, vec2, vec3;
v16u8 reg0, reg1, reg2, reg3, reg4, reg5, reg6, reg7;
v16u8 res0, res1, res2, res3, res4, res5, res6, res7, res8, res9;
for (x = 0; x < width; x += 16) {
s = src;
src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
ILVRL_H(vec0, vec2, vec1, vec3, reg0, reg1, reg2, reg3);
src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
ILVRL_H(vec0, vec2, vec1, vec3, reg4, reg5, reg6, reg7);
ILVRL_W(reg0, reg4, reg1, reg5, res0, res1, res2, res3);
ILVRL_W(reg2, reg6, reg3, reg7, res4, res5, res6, res7);
src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
ILVRL_H(vec0, vec2, vec1, vec3, reg0, reg1, reg2, reg3);
src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
ILVRL_H(vec0, vec2, vec1, vec3, reg4, reg5, reg6, reg7);
res8 = (v16u8)__msa_ilvr_w((v4i32)reg4, (v4i32)reg0);
res9 = (v16u8)__msa_ilvl_w((v4i32)reg4, (v4i32)reg0);
ILVRL_D(res0, res8, res1, res9, dst0, dst1, dst2, dst3);
ST_UB4(dst0, dst1, dst2, dst3, dst, dst_stride);
dst += dst_stride * 4;
res8 = (v16u8)__msa_ilvr_w((v4i32)reg5, (v4i32)reg1);
res9 = (v16u8)__msa_ilvl_w((v4i32)reg5, (v4i32)reg1);
ILVRL_D(res2, res8, res3, res9, dst0, dst1, dst2, dst3);
ST_UB4(dst0, dst1, dst2, dst3, dst, dst_stride);
dst += dst_stride * 4;
res8 = (v16u8)__msa_ilvr_w((v4i32)reg6, (v4i32)reg2);
res9 = (v16u8)__msa_ilvl_w((v4i32)reg6, (v4i32)reg2);
ILVRL_D(res4, res8, res5, res9, dst0, dst1, dst2, dst3);
ST_UB4(dst0, dst1, dst2, dst3, dst, dst_stride);
dst += dst_stride * 4;
res8 = (v16u8)__msa_ilvr_w((v4i32)reg7, (v4i32)reg3);
res9 = (v16u8)__msa_ilvl_w((v4i32)reg7, (v4i32)reg3);
ILVRL_D(res6, res8, res7, res9, dst0, dst1, dst2, dst3);
ST_UB4(dst0, dst1, dst2, dst3, dst, dst_stride);
src += 16;
dst += dst_stride * 4;
}
}
void TransposeUVWx16_MSA(const uint8_t* src,
int src_stride,
uint8_t* dst_a,
int dst_stride_a,
uint8_t* dst_b,
int dst_stride_b,
int width) {
int x;
const uint8_t* s;
v16u8 src0, src1, src2, src3, dst0, dst1, dst2, dst3, vec0, vec1, vec2, vec3;
v16u8 reg0, reg1, reg2, reg3, reg4, reg5, reg6, reg7;
v16u8 res0, res1, res2, res3, res4, res5, res6, res7, res8, res9;
for (x = 0; x < width; x += 8) {
s = src;
src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
ILVRL_H(vec0, vec2, vec1, vec3, reg0, reg1, reg2, reg3);
src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
ILVRL_H(vec0, vec2, vec1, vec3, reg4, reg5, reg6, reg7);
ILVRL_W(reg0, reg4, reg1, reg5, res0, res1, res2, res3);
ILVRL_W(reg2, reg6, reg3, reg7, res4, res5, res6, res7);
src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
ILVRL_H(vec0, vec2, vec1, vec3, reg0, reg1, reg2, reg3);
src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src1 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src2 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
src3 = (v16u8)__msa_ld_b((v16i8*)s, 0);
s += src_stride;
ILVRL_B(src0, src1, src2, src3, vec0, vec1, vec2, vec3);
ILVRL_H(vec0, vec2, vec1, vec3, reg4, reg5, reg6, reg7);
res8 = (v16u8)__msa_ilvr_w((v4i32)reg4, (v4i32)reg0);
res9 = (v16u8)__msa_ilvl_w((v4i32)reg4, (v4i32)reg0);
ILVRL_D(res0, res8, res1, res9, dst0, dst1, dst2, dst3);
ST_UB2(dst0, dst2, dst_a, dst_stride_a);
ST_UB2(dst1, dst3, dst_b, dst_stride_b);
dst_a += dst_stride_a * 2;
dst_b += dst_stride_b * 2;
res8 = (v16u8)__msa_ilvr_w((v4i32)reg5, (v4i32)reg1);
res9 = (v16u8)__msa_ilvl_w((v4i32)reg5, (v4i32)reg1);
ILVRL_D(res2, res8, res3, res9, dst0, dst1, dst2, dst3);
ST_UB2(dst0, dst2, dst_a, dst_stride_a);
ST_UB2(dst1, dst3, dst_b, dst_stride_b);
dst_a += dst_stride_a * 2;
dst_b += dst_stride_b * 2;
res8 = (v16u8)__msa_ilvr_w((v4i32)reg6, (v4i32)reg2);
res9 = (v16u8)__msa_ilvl_w((v4i32)reg6, (v4i32)reg2);
ILVRL_D(res4, res8, res5, res9, dst0, dst1, dst2, dst3);
ST_UB2(dst0, dst2, dst_a, dst_stride_a);
ST_UB2(dst1, dst3, dst_b, dst_stride_b);
dst_a += dst_stride_a * 2;
dst_b += dst_stride_b * 2;
res8 = (v16u8)__msa_ilvr_w((v4i32)reg7, (v4i32)reg3);
res9 = (v16u8)__msa_ilvl_w((v4i32)reg7, (v4i32)reg3);
ILVRL_D(res6, res8, res7, res9, dst0, dst1, dst2, dst3);
ST_UB2(dst0, dst2, dst_a, dst_stride_a);
ST_UB2(dst1, dst3, dst_b, dst_stride_b);
src += 16;
dst_a += dst_stride_a * 2;
dst_b += dst_stride_b * 2;
}
}
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
#endif // !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)

198
source/row_any.cc
View File

@ -110,12 +110,6 @@ ANY41C(I444AlphaToARGBRow_Any_NEON, I444AlphaToARGBRow_NEON, 0, 0, 4, 7)
#ifdef HAS_I422ALPHATOARGBROW_NEON
ANY41C(I422AlphaToARGBRow_Any_NEON, I422AlphaToARGBRow_NEON, 1, 0, 4, 7)
#endif
#ifdef HAS_I444ALPHATOARGBROW_MSA
ANY41C(I444AlphaToARGBRow_Any_MSA, I444AlphaToARGBRow_MSA, 0, 0, 4, 7)
#endif
#ifdef HAS_I422ALPHATOARGBROW_MSA
ANY41C(I422AlphaToARGBRow_Any_MSA, I422AlphaToARGBRow_MSA, 1, 0, 4, 7)
#endif
#ifdef HAS_I422ALPHATOARGBROW_LSX
ANY41C(I422AlphaToARGBRow_Any_LSX, I422AlphaToARGBRow_LSX, 1, 0, 4, 15)
#endif
@ -311,9 +305,6 @@ ANY31(I422ToUYVYRow_Any_AVX2, I422ToUYVYRow_AVX2, 1, 1, 4, 31)
#ifdef HAS_I422TOYUY2ROW_NEON
ANY31(I422ToYUY2Row_Any_NEON, I422ToYUY2Row_NEON, 1, 1, 4, 15)
#endif
#ifdef HAS_I422TOYUY2ROW_MSA
ANY31(I422ToYUY2Row_Any_MSA, I422ToYUY2Row_MSA, 1, 1, 4, 31)
#endif
#ifdef HAS_I422TOYUY2ROW_LSX
ANY31(I422ToYUY2Row_Any_LSX, I422ToYUY2Row_LSX, 1, 1, 4, 15)
#endif
@ -323,9 +314,6 @@ ANY31(I422ToYUY2Row_Any_LASX, I422ToYUY2Row_LASX, 1, 1, 4, 31)
#ifdef HAS_I422TOUYVYROW_NEON
ANY31(I422ToUYVYRow_Any_NEON, I422ToUYVYRow_NEON, 1, 1, 4, 15)
#endif
#ifdef HAS_I422TOUYVYROW_MSA
ANY31(I422ToUYVYRow_Any_MSA, I422ToUYVYRow_MSA, 1, 1, 4, 31)
#endif
#ifdef HAS_I422TOUYVYROW_LSX
ANY31(I422ToUYVYRow_Any_LSX, I422ToUYVYRow_LSX, 1, 1, 4, 15)
#endif
@ -439,15 +427,6 @@ ANY31C(I422ToARGB4444Row_Any_NEON, I422ToARGB4444Row_NEON, 1, 0, 2, 7)
ANY31C(I422ToARGB1555Row_Any_NEON, I422ToARGB1555Row_NEON, 1, 0, 2, 7)
ANY31C(I422ToRGB565Row_Any_NEON, I422ToRGB565Row_NEON, 1, 0, 2, 7)
#endif
#ifdef HAS_I422TOARGBROW_MSA
ANY31C(I444ToARGBRow_Any_MSA, I444ToARGBRow_MSA, 0, 0, 4, 7)
ANY31C(I422ToARGBRow_Any_MSA, I422ToARGBRow_MSA, 1, 0, 4, 7)
ANY31C(I422ToRGBARow_Any_MSA, I422ToRGBARow_MSA, 1, 0, 4, 7)
ANY31C(I422ToRGB24Row_Any_MSA, I422ToRGB24Row_MSA, 1, 0, 3, 15)
ANY31C(I422ToARGB4444Row_Any_MSA, I422ToARGB4444Row_MSA, 1, 0, 2, 7)
ANY31C(I422ToARGB1555Row_Any_MSA, I422ToARGB1555Row_MSA, 1, 0, 2, 7)
ANY31C(I422ToRGB565Row_Any_MSA, I422ToRGB565Row_MSA, 1, 0, 2, 7)
#endif
#ifdef HAS_I422TOARGBROW_LSX
ANY31C(I422ToARGBRow_Any_LSX, I422ToARGBRow_LSX, 1, 0, 4, 15)
ANY31C(I422ToRGBARow_Any_LSX, I422ToRGBARow_LSX, 1, 0, 4, 15)
@ -645,9 +624,6 @@ ANY21(MergeUVRow_Any_AVX512BW, MergeUVRow_AVX512BW, 0, 1, 1, 2, 31)
#ifdef HAS_MERGEUVROW_NEON
ANY21(MergeUVRow_Any_NEON, MergeUVRow_NEON, 0, 1, 1, 2, 15)
#endif
#ifdef HAS_MERGEUVROW_MSA
ANY21(MergeUVRow_Any_MSA, MergeUVRow_MSA, 0, 1, 1, 2, 15)
#endif
#ifdef HAS_MERGEUVROW_LSX
ANY21(MergeUVRow_Any_LSX, MergeUVRow_LSX, 0, 1, 1, 2, 15)
#endif
@ -688,27 +664,18 @@ ANY21(ARGBAddRow_Any_NEON, ARGBAddRow_NEON, 0, 4, 4, 4, 7)
#ifdef HAS_ARGBSUBTRACTROW_NEON
ANY21(ARGBSubtractRow_Any_NEON, ARGBSubtractRow_NEON, 0, 4, 4, 4, 7)
#endif
#ifdef HAS_ARGBMULTIPLYROW_MSA
ANY21(ARGBMultiplyRow_Any_MSA, ARGBMultiplyRow_MSA, 0, 4, 4, 4, 3)
#endif
#ifdef HAS_ARGBMULTIPLYROW_LSX
ANY21(ARGBMultiplyRow_Any_LSX, ARGBMultiplyRow_LSX, 0, 4, 4, 4, 3)
#endif
#ifdef HAS_ARGBMULTIPLYROW_LASX
ANY21(ARGBMultiplyRow_Any_LASX, ARGBMultiplyRow_LASX, 0, 4, 4, 4, 7)
#endif
#ifdef HAS_ARGBADDROW_MSA
ANY21(ARGBAddRow_Any_MSA, ARGBAddRow_MSA, 0, 4, 4, 4, 7)
#endif
#ifdef HAS_ARGBADDROW_LSX
ANY21(ARGBAddRow_Any_LSX, ARGBAddRow_LSX, 0, 4, 4, 4, 3)
#endif
#ifdef HAS_ARGBADDROW_LASX
ANY21(ARGBAddRow_Any_LASX, ARGBAddRow_LASX, 0, 4, 4, 4, 7)
#endif
#ifdef HAS_ARGBSUBTRACTROW_MSA
ANY21(ARGBSubtractRow_Any_MSA, ARGBSubtractRow_MSA, 0, 4, 4, 4, 7)
#endif
#ifdef HAS_ARGBSUBTRACTROW_LSX
ANY21(ARGBSubtractRow_Any_LSX, ARGBSubtractRow_LSX, 0, 4, 4, 4, 3)
#endif
@ -721,9 +688,6 @@ ANY21(SobelRow_Any_SSE2, SobelRow_SSE2, 0, 1, 1, 4, 15)
#ifdef HAS_SOBELROW_NEON
ANY21(SobelRow_Any_NEON, SobelRow_NEON, 0, 1, 1, 4, 7)
#endif
#ifdef HAS_SOBELROW_MSA
ANY21(SobelRow_Any_MSA, SobelRow_MSA, 0, 1, 1, 4, 15)
#endif
#ifdef HAS_SOBELROW_LSX
ANY21(SobelRow_Any_LSX, SobelRow_LSX, 0, 1, 1, 4, 15)
#endif
@ -733,9 +697,6 @@ ANY21(SobelToPlaneRow_Any_SSE2, SobelToPlaneRow_SSE2, 0, 1, 1, 1, 15)
#ifdef HAS_SOBELTOPLANEROW_NEON
ANY21(SobelToPlaneRow_Any_NEON, SobelToPlaneRow_NEON, 0, 1, 1, 1, 15)
#endif
#ifdef HAS_SOBELTOPLANEROW_MSA
ANY21(SobelToPlaneRow_Any_MSA, SobelToPlaneRow_MSA, 0, 1, 1, 1, 31)
#endif
#ifdef HAS_SOBELTOPLANEROW_LSX
ANY21(SobelToPlaneRow_Any_LSX, SobelToPlaneRow_LSX, 0, 1, 1, 1, 31)
#endif
@ -745,9 +706,6 @@ ANY21(SobelXYRow_Any_SSE2, SobelXYRow_SSE2, 0, 1, 1, 4, 15)
#ifdef HAS_SOBELXYROW_NEON
ANY21(SobelXYRow_Any_NEON, SobelXYRow_NEON, 0, 1, 1, 4, 7)
#endif
#ifdef HAS_SOBELXYROW_MSA
ANY21(SobelXYRow_Any_MSA, SobelXYRow_MSA, 0, 1, 1, 4, 15)
#endif
#ifdef HAS_SOBELXYROW_LSX
ANY21(SobelXYRow_Any_LSX, SobelXYRow_LSX, 0, 1, 1, 4, 15)
#endif
@ -814,9 +772,6 @@ ANY21C(NV12ToARGBRow_Any_AVX2, NV12ToARGBRow_AVX2, 1, 1, 2, 4, 15)
#ifdef HAS_NV12TOARGBROW_NEON
ANY21C(NV12ToARGBRow_Any_NEON, NV12ToARGBRow_NEON, 1, 1, 2, 4, 7)
#endif
#ifdef HAS_NV12TOARGBROW_MSA
ANY21C(NV12ToARGBRow_Any_MSA, NV12ToARGBRow_MSA, 1, 1, 2, 4, 7)
#endif
#ifdef HAS_NV12TOARGBROW_LSX
ANY21C(NV12ToARGBRow_Any_LSX, NV12ToARGBRow_LSX, 1, 1, 2, 4, 7)
#endif
@ -832,9 +787,6 @@ ANY21C(NV21ToARGBRow_Any_AVX2, NV21ToARGBRow_AVX2, 1, 1, 2, 4, 15)
#ifdef HAS_NV21TOARGBROW_NEON
ANY21C(NV21ToARGBRow_Any_NEON, NV21ToARGBRow_NEON, 1, 1, 2, 4, 7)
#endif
#ifdef HAS_NV21TOARGBROW_MSA
ANY21C(NV21ToARGBRow_Any_MSA, NV21ToARGBRow_MSA, 1, 1, 2, 4, 7)
#endif
#ifdef HAS_NV21TOARGBROW_LSX
ANY21C(NV21ToARGBRow_Any_LSX, NV21ToARGBRow_LSX, 1, 1, 2, 4, 7)
#endif
@ -868,9 +820,6 @@ ANY21C(NV12ToRGB565Row_Any_AVX2, NV12ToRGB565Row_AVX2, 1, 1, 2, 2, 15)
#ifdef HAS_NV12TORGB565ROW_NEON
ANY21C(NV12ToRGB565Row_Any_NEON, NV12ToRGB565Row_NEON, 1, 1, 2, 2, 7)
#endif
#ifdef HAS_NV12TORGB565ROW_MSA
ANY21C(NV12ToRGB565Row_Any_MSA, NV12ToRGB565Row_MSA, 1, 1, 2, 2, 7)
#endif
#ifdef HAS_NV12TORGB565ROW_LSX
ANY21C(NV12ToRGB565Row_Any_LSX, NV12ToRGB565Row_LSX, 1, 1, 2, 2, 7)
#endif
@ -1074,14 +1023,6 @@ ANY11(ARGBToARGB1555Row_Any_NEON, ARGBToARGB1555Row_NEON, 0, 4, 2, 7)
ANY11(ARGBToARGB4444Row_Any_NEON, ARGBToARGB4444Row_NEON, 0, 4, 2, 7)
ANY11(J400ToARGBRow_Any_NEON, J400ToARGBRow_NEON, 0, 1, 4, 7)
#endif
#if defined(HAS_ARGBTORGB24ROW_MSA)
ANY11(ARGBToRGB24Row_Any_MSA, ARGBToRGB24Row_MSA, 0, 4, 3, 15)
ANY11(ARGBToRAWRow_Any_MSA, ARGBToRAWRow_MSA, 0, 4, 3, 15)
ANY11(ARGBToRGB565Row_Any_MSA, ARGBToRGB565Row_MSA, 0, 4, 2, 7)
ANY11(ARGBToARGB1555Row_Any_MSA, ARGBToARGB1555Row_MSA, 0, 4, 2, 7)
ANY11(ARGBToARGB4444Row_Any_MSA, ARGBToARGB4444Row_MSA, 0, 4, 2, 7)
ANY11(J400ToARGBRow_Any_MSA, J400ToARGBRow_MSA, 0, 1, 4, 15)
#endif
#if defined(HAS_ARGBTORGB24ROW_LSX)
ANY11(ARGBToRGB24Row_Any_LSX, ARGBToRGB24Row_LSX, 0, 4, 3, 15)
ANY11(ARGBToRAWRow_Any_LSX, ARGBToRAWRow_LSX, 0, 4, 3, 15)
@ -1102,9 +1043,6 @@ ANY11(J400ToARGBRow_Any_LSX, J400ToARGBRow_LSX, 0, 1, 4, 15)
#if defined(HAS_RAWTORGB24ROW_NEON)
ANY11(RAWToRGB24Row_Any_NEON, RAWToRGB24Row_NEON, 0, 3, 3, 7)
#endif
#if defined(HAS_RAWTORGB24ROW_MSA)
ANY11(RAWToRGB24Row_Any_MSA, RAWToRGB24Row_MSA, 0, 3, 3, 15)
#endif
#if defined(HAS_RAWTORGB24ROW_LSX)
ANY11(RAWToRGB24Row_Any_LSX, RAWToRGB24Row_LSX, 0, 3, 3, 15)
#endif
@ -1156,9 +1094,6 @@ ANY11(ARGBToYRow_Any_NEON, ARGBToYRow_NEON, 0, 4, 1, 15)
#ifdef HAS_ARGBTOYROW_NEON_DOTPROD
ANY11(ARGBToYRow_Any_NEON_DotProd, ARGBToYRow_NEON_DotProd, 0, 4, 1, 15)
#endif
#ifdef HAS_ARGBTOYROW_MSA
ANY11(ARGBToYRow_Any_MSA, ARGBToYRow_MSA, 0, 4, 1, 15)
#endif
#ifdef HAS_ARGBTOYROW_LSX
ANY11(ARGBToYRow_Any_LSX, ARGBToYRow_LSX, 0, 4, 1, 15)
#endif
@ -1183,9 +1118,6 @@ ANY11(RGBAToYJRow_Any_NEON, RGBAToYJRow_NEON, 0, 4, 1, 15)
#ifdef HAS_RGBATOYJROW_NEON_DOTPROD
ANY11(RGBAToYJRow_Any_NEON_DotProd, RGBAToYJRow_NEON_DotProd, 0, 4, 1, 15)
#endif
#ifdef HAS_ARGBTOYJROW_MSA
ANY11(ARGBToYJRow_Any_MSA, ARGBToYJRow_MSA, 0, 4, 1, 15)
#endif
#ifdef HAS_ARGBTOYJROW_LSX
ANY11(ARGBToYJRow_Any_LSX, ARGBToYJRow_LSX, 0, 4, 1, 15)
#endif
@ -1210,9 +1142,6 @@ ANY11(BGRAToYRow_Any_NEON, BGRAToYRow_NEON, 0, 4, 1, 15)
#ifdef HAS_BGRATOYROW_NEON_DOTPROD
ANY11(BGRAToYRow_Any_NEON_DotProd, BGRAToYRow_NEON_DotProd, 0, 4, 1, 15)
#endif
#ifdef HAS_BGRATOYROW_MSA
ANY11(BGRAToYRow_Any_MSA, BGRAToYRow_MSA, 0, 4, 1, 15)
#endif
#ifdef HAS_BGRATOYROW_LSX
ANY11(BGRAToYRow_Any_LSX, BGRAToYRow_LSX, 0, 4, 1, 15)
#endif
@ -1225,9 +1154,6 @@ ANY11(ABGRToYRow_Any_NEON, ABGRToYRow_NEON, 0, 4, 1, 15)
#ifdef HAS_ABGRTOYROW_NEON_DOTPROD
ANY11(ABGRToYRow_Any_NEON_DotProd, ABGRToYRow_NEON_DotProd, 0, 4, 1, 15)
#endif
#ifdef HAS_ABGRTOYROW_MSA
ANY11(ABGRToYRow_Any_MSA, ABGRToYRow_MSA, 0, 4, 1, 7)
#endif
#ifdef HAS_ABGRTOYROW_LSX
ANY11(ABGRToYRow_Any_LSX, ABGRToYRow_LSX, 0, 4, 1, 15)
#endif
@ -1240,9 +1166,6 @@ ANY11(RGBAToYRow_Any_NEON, RGBAToYRow_NEON, 0, 4, 1, 15)
#ifdef HAS_RGBATOYROW_NEON_DOTPROD
ANY11(RGBAToYRow_Any_NEON_DotProd, RGBAToYRow_NEON_DotProd, 0, 4, 1, 15)
#endif
#ifdef HAS_RGBATOYROW_MSA
ANY11(RGBAToYRow_Any_MSA, RGBAToYRow_MSA, 0, 4, 1, 15)
#endif
#ifdef HAS_RGBATOYROW_LSX
ANY11(RGBAToYRow_Any_LSX, RGBAToYRow_LSX, 0, 4, 1, 15)
#endif
@ -1261,9 +1184,6 @@ ANY11(RGB24ToYJRow_Any_SSSE3, RGB24ToYJRow_SSSE3, 0, 3, 1, 15)
#ifdef HAS_RGB24TOYJROW_NEON
ANY11(RGB24ToYJRow_Any_NEON, RGB24ToYJRow_NEON, 0, 3, 1, 15)
#endif
#ifdef HAS_RGB24TOYROW_MSA
ANY11(RGB24ToYRow_Any_MSA, RGB24ToYRow_MSA, 0, 3, 1, 15)
#endif
#ifdef HAS_RGB24TOYROW_LSX
ANY11(RGB24ToYRow_Any_LSX, RGB24ToYRow_LSX, 0, 3, 1, 15)
#endif
@ -1288,9 +1208,6 @@ ANY11(RAWToYJRow_Any_SSSE3, RAWToYJRow_SSSE3, 0, 3, 1, 15)
#ifdef HAS_RAWTOYJROW_NEON
ANY11(RAWToYJRow_Any_NEON, RAWToYJRow_NEON, 0, 3, 1, 15)
#endif
#ifdef HAS_RAWTOYROW_MSA
ANY11(RAWToYRow_Any_MSA, RAWToYRow_MSA, 0, 3, 1, 15)
#endif
#ifdef HAS_RAWTOYROW_LSX
ANY11(RAWToYRow_Any_LSX, RAWToYRow_LSX, 0, 3, 1, 15)
#endif
@ -1306,9 +1223,6 @@ ANY11(RAWToYJRow_Any_LASX, RAWToYJRow_LASX, 0, 3, 1, 31)
#ifdef HAS_RGB565TOYROW_NEON
ANY11(RGB565ToYRow_Any_NEON, RGB565ToYRow_NEON, 0, 2, 1, 15)
#endif
#ifdef HAS_RGB565TOYROW_MSA
ANY11(RGB565ToYRow_Any_MSA, RGB565ToYRow_MSA, 0, 2, 1, 15)
#endif
#ifdef HAS_RGB565TOYROW_LSX
ANY11(RGB565ToYRow_Any_LSX, RGB565ToYRow_LSX, 0, 2, 1, 15)
#endif
@ -1322,9 +1236,6 @@ ANY11(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, 0, 2, 1, 15)
ANY11(ARGB1555ToYRow_Any_NEON, ARGB1555ToYRow_NEON, 0, 2, 1, 7)
#endif
#endif
#ifdef HAS_ARGB1555TOYROW_MSA
ANY11(ARGB1555ToYRow_Any_MSA, ARGB1555ToYRow_MSA, 0, 2, 1, 15)
#endif
#ifdef HAS_ARGB1555TOYROW_LSX
ANY11(ARGB1555ToYRow_Any_LSX, ARGB1555ToYRow_LSX, 0, 2, 1, 15)
#endif
@ -1344,18 +1255,12 @@ ANY11(YUY2ToYRow_Any_NEON, YUY2ToYRow_NEON, 1, 4, 1, 15)
#ifdef HAS_UYVYTOYROW_NEON
ANY11(UYVYToYRow_Any_NEON, UYVYToYRow_NEON, 1, 4, 1, 15)
#endif
#ifdef HAS_YUY2TOYROW_MSA
ANY11(YUY2ToYRow_Any_MSA, YUY2ToYRow_MSA, 1, 4, 1, 31)
#endif
#ifdef HAS_YUY2TOYROW_LSX
ANY11(YUY2ToYRow_Any_LSX, YUY2ToYRow_LSX, 1, 4, 1, 15)
#endif
#ifdef HAS_YUY2TOYROW_LASX
ANY11(YUY2ToYRow_Any_LASX, YUY2ToYRow_LASX, 1, 4, 1, 31)
#endif
#ifdef HAS_UYVYTOYROW_MSA
ANY11(UYVYToYRow_Any_MSA, UYVYToYRow_MSA, 1, 4, 1, 31)
#endif
#ifdef HAS_UYVYTOYROW_LSX
ANY11(UYVYToYRow_Any_LSX, UYVYToYRow_LSX, 1, 4, 1, 15)
#endif
@ -1377,9 +1282,6 @@ ANY11(SwapUVRow_Any_NEON, SwapUVRow_NEON, 0, 2, 2, 15)
#ifdef HAS_RGB24TOARGBROW_NEON
ANY11(RGB24ToARGBRow_Any_NEON, RGB24ToARGBRow_NEON, 0, 3, 4, 7)
#endif
#ifdef HAS_RGB24TOARGBROW_MSA
ANY11(RGB24ToARGBRow_Any_MSA, RGB24ToARGBRow_MSA, 0, 3, 4, 15)
#endif
#ifdef HAS_RGB24TOARGBROW_LSX
ANY11(RGB24ToARGBRow_Any_LSX, RGB24ToARGBRow_LSX, 0, 3, 4, 15)
#endif
@ -1392,9 +1294,6 @@ ANY11(RAWToARGBRow_Any_NEON, RAWToARGBRow_NEON, 0, 3, 4, 7)
#ifdef HAS_RAWTORGBAROW_NEON
ANY11(RAWToRGBARow_Any_NEON, RAWToRGBARow_NEON, 0, 3, 4, 7)
#endif
#ifdef HAS_RAWTOARGBROW_MSA
ANY11(RAWToARGBRow_Any_MSA, RAWToARGBRow_MSA, 0, 3, 4, 15)
#endif
#ifdef HAS_RAWTOARGBROW_LSX
ANY11(RAWToARGBRow_Any_LSX, RAWToARGBRow_LSX, 0, 3, 4, 15)
#endif
@ -1404,9 +1303,6 @@ ANY11(RAWToARGBRow_Any_LASX, RAWToARGBRow_LASX, 0, 3, 4, 31)
#ifdef HAS_RGB565TOARGBROW_NEON
ANY11(RGB565ToARGBRow_Any_NEON, RGB565ToARGBRow_NEON, 0, 2, 4, 15)
#endif
#ifdef HAS_RGB565TOARGBROW_MSA
ANY11(RGB565ToARGBRow_Any_MSA, RGB565ToARGBRow_MSA, 0, 2, 4, 15)
#endif
#ifdef HAS_RGB565TOARGBROW_LSX
ANY11(RGB565ToARGBRow_Any_LSX, RGB565ToARGBRow_LSX, 0, 2, 4, 15)
#endif
@ -1416,9 +1312,6 @@ ANY11(RGB565ToARGBRow_Any_LASX, RGB565ToARGBRow_LASX, 0, 2, 4, 31)
#ifdef HAS_ARGB1555TOARGBROW_NEON
ANY11(ARGB1555ToARGBRow_Any_NEON, ARGB1555ToARGBRow_NEON, 0, 2, 4, 15)
#endif
#ifdef HAS_ARGB1555TOARGBROW_MSA
ANY11(ARGB1555ToARGBRow_Any_MSA, ARGB1555ToARGBRow_MSA, 0, 2, 4, 15)
#endif
#ifdef HAS_ARGB1555TOARGBROW_LSX
ANY11(ARGB1555ToARGBRow_Any_LSX, ARGB1555ToARGBRow_LSX, 0, 2, 4, 15)
#endif
@ -1428,9 +1321,6 @@ ANY11(ARGB1555ToARGBRow_Any_LASX, ARGB1555ToARGBRow_LASX, 0, 2, 4, 31)
#ifdef HAS_ARGB4444TOARGBROW_NEON
ANY11(ARGB4444ToARGBRow_Any_NEON, ARGB4444ToARGBRow_NEON, 0, 2, 4, 7)
#endif
#ifdef HAS_ARGB4444TOARGBROW_MSA
ANY11(ARGB4444ToARGBRow_Any_MSA, ARGB4444ToARGBRow_MSA, 0, 2, 4, 15)
#endif
#ifdef HAS_ARGB4444TOARGBROW_LSX
ANY11(ARGB4444ToARGBRow_Any_LSX, ARGB4444ToARGBRow_LSX, 0, 2, 4, 15)
#endif
@ -1452,9 +1342,6 @@ ANY11(ARGBUnattenuateRow_Any_AVX2, ARGBUnattenuateRow_AVX2, 0, 4, 4, 7)
#ifdef HAS_ARGBATTENUATEROW_NEON
ANY11(ARGBAttenuateRow_Any_NEON, ARGBAttenuateRow_NEON, 0, 4, 4, 7)
#endif
#ifdef HAS_ARGBATTENUATEROW_MSA
ANY11(ARGBAttenuateRow_Any_MSA, ARGBAttenuateRow_MSA, 0, 4, 4, 7)
#endif
#ifdef HAS_ARGBATTENUATEROW_LSX
ANY11(ARGBAttenuateRow_Any_LSX, ARGBAttenuateRow_LSX, 0, 4, 4, 7)
#endif
@ -1470,9 +1357,6 @@ ANY11(ARGBExtractAlphaRow_Any_AVX2, ARGBExtractAlphaRow_AVX2, 0, 4, 1, 31)
#ifdef HAS_ARGBEXTRACTALPHAROW_NEON
ANY11(ARGBExtractAlphaRow_Any_NEON, ARGBExtractAlphaRow_NEON, 0, 4, 1, 15)
#endif
#ifdef HAS_ARGBEXTRACTALPHAROW_MSA
ANY11(ARGBExtractAlphaRow_Any_MSA, ARGBExtractAlphaRow_MSA, 0, 4, 1, 15)
#endif
#ifdef HAS_ARGBEXTRACTALPHAROW_LSX
ANY11(ARGBExtractAlphaRow_Any_LSX, ARGBExtractAlphaRow_LSX, 0, 4, 1, 15)
#endif
@ -1552,14 +1436,6 @@ ANY11P(I400ToARGBRow_Any_NEON,
4,
7)
#endif
#if defined(HAS_I400TOARGBROW_MSA)
ANY11P(I400ToARGBRow_Any_MSA,
I400ToARGBRow_MSA,
const struct YuvConstants*,
1,
4,
15)
#endif
#if defined(HAS_I400TOARGBROW_LSX)
ANY11P(I400ToARGBRow_Any_LSX,
I400ToARGBRow_LSX,
@ -1593,14 +1469,6 @@ ANY11P(ARGBToRGB565DitherRow_Any_NEON,
2,
7)
#endif
#if defined(HAS_ARGBTORGB565DITHERROW_MSA)
ANY11P(ARGBToRGB565DitherRow_Any_MSA,
ARGBToRGB565DitherRow_MSA,
const uint32_t,
4,
2,
7)
#endif
#if defined(HAS_ARGBTORGB565DITHERROW_LSX)
ANY11P(ARGBToRGB565DitherRow_Any_LSX,
ARGBToRGB565DitherRow_LSX,
@ -1626,9 +1494,6 @@ ANY11P(ARGBShuffleRow_Any_AVX2, ARGBShuffleRow_AVX2, const uint8_t*, 4, 4, 15)
#ifdef HAS_ARGBSHUFFLEROW_NEON
ANY11P(ARGBShuffleRow_Any_NEON, ARGBShuffleRow_NEON, const uint8_t*, 4, 4, 3)
#endif
#ifdef HAS_ARGBSHUFFLEROW_MSA
ANY11P(ARGBShuffleRow_Any_MSA, ARGBShuffleRow_MSA, const uint8_t*, 4, 4, 7)
#endif
#ifdef HAS_ARGBSHUFFLEROW_LSX
ANY11P(ARGBShuffleRow_Any_LSX, ARGBShuffleRow_LSX, const uint8_t*, 4, 4, 7)
#endif
@ -1885,9 +1750,6 @@ ANY11P16(HalfFloat1Row_Any_F16C,
#ifdef HAS_HALFFLOATROW_NEON
ANY11P16(HalfFloatRow_Any_NEON, HalfFloatRow_NEON, uint16_t, uint16_t, 2, 2, 15)
#endif
#ifdef HAS_HALFFLOATROW_MSA
ANY11P16(HalfFloatRow_Any_MSA, HalfFloatRow_MSA, uint16_t, uint16_t, 2, 2, 31)
#endif
#ifdef HAS_BYTETOFLOATROW_NEON
ANY11P16(ByteToFloatRow_Any_NEON, ByteToFloatRow_NEON, uint8_t, float, 1, 3, 7)
#endif
@ -1926,10 +1788,6 @@ ANY11C(UYVYToARGBRow_Any_AVX2, UYVYToARGBRow_AVX2, 1, 4, 4, 31)
ANY11C(YUY2ToARGBRow_Any_NEON, YUY2ToARGBRow_NEON, 1, 4, 4, 7)
ANY11C(UYVYToARGBRow_Any_NEON, UYVYToARGBRow_NEON, 1, 4, 4, 7)
#endif
#if defined(HAS_YUY2TOARGBROW_MSA)
ANY11C(YUY2ToARGBRow_Any_MSA, YUY2ToARGBRow_MSA, 1, 4, 4, 7)
ANY11C(UYVYToARGBRow_Any_MSA, UYVYToARGBRow_MSA, 1, 4, 4, 7)
#endif
#if defined(HAS_YUY2TOARGBROW_LSX)
ANY11C(YUY2ToARGBRow_Any_LSX, YUY2ToARGBRow_LSX, 1, 4, 4, 7)
ANY11C(UYVYToARGBRow_Any_LSX, UYVYToARGBRow_LSX, 1, 4, 4, 7)
@ -1973,9 +1831,6 @@ ANY11I(InterpolateRow_Any_SSSE3,
#ifdef HAS_INTERPOLATEROW_NEON
ANY11I(InterpolateRow_Any_NEON, InterpolateRow_NEON, uint8_t, uint8_t, 1, 1, 15)
#endif
#ifdef HAS_INTERPOLATEROW_MSA
ANY11I(InterpolateRow_Any_MSA, InterpolateRow_MSA, uint8_t, uint8_t, 1, 1, 31)
#endif
#ifdef HAS_INTERPOLATEROW_LSX
ANY11I(InterpolateRow_Any_LSX, InterpolateRow_LSX, uint8_t, uint8_t, 1, 1, 31)
#endif
@ -2060,9 +1915,6 @@ ANY11M(MirrorRow_Any_SSSE3, MirrorRow_SSSE3, 1, 15)
#ifdef HAS_MIRRORROW_NEON
ANY11M(MirrorRow_Any_NEON, MirrorRow_NEON, 1, 31)
#endif
#ifdef HAS_MIRRORROW_MSA
ANY11M(MirrorRow_Any_MSA, MirrorRow_MSA, 1, 63)
#endif
#ifdef HAS_MIRRORROW_LSX
ANY11M(MirrorRow_Any_LSX, MirrorRow_LSX, 1, 31)
#endif
@ -2078,9 +1930,6 @@ ANY11M(MirrorUVRow_Any_SSSE3, MirrorUVRow_SSSE3, 2, 7)
#ifdef HAS_MIRRORUVROW_NEON
ANY11M(MirrorUVRow_Any_NEON, MirrorUVRow_NEON, 2, 31)
#endif
#ifdef HAS_MIRRORUVROW_MSA
ANY11M(MirrorUVRow_Any_MSA, MirrorUVRow_MSA, 2, 7)
#endif
#ifdef HAS_MIRRORUVROW_LSX
ANY11M(MirrorUVRow_Any_LSX, MirrorUVRow_LSX, 2, 7)
#endif
@ -2096,9 +1945,6 @@ ANY11M(ARGBMirrorRow_Any_SSE2, ARGBMirrorRow_SSE2, 4, 3)
#ifdef HAS_ARGBMIRRORROW_NEON
ANY11M(ARGBMirrorRow_Any_NEON, ARGBMirrorRow_NEON, 4, 7)
#endif
#ifdef HAS_ARGBMIRRORROW_MSA
ANY11M(ARGBMirrorRow_Any_MSA, ARGBMirrorRow_MSA, 4, 15)
#endif
#ifdef HAS_ARGBMIRRORROW_LSX
ANY11M(ARGBMirrorRow_Any_LSX, ARGBMirrorRow_LSX, 4, 7)
#endif
@ -2139,9 +1985,6 @@ ANY1(SetRow_Any_LSX, SetRow_LSX, uint8_t, 1, 15)
#ifdef HAS_ARGBSETROW_NEON
ANY1(ARGBSetRow_Any_NEON, ARGBSetRow_NEON, uint32_t, 4, 3)
#endif
#ifdef HAS_ARGBSETROW_MSA
ANY1(ARGBSetRow_Any_MSA, ARGBSetRow_MSA, uint32_t, 4, 3)
#endif
#ifdef HAS_ARGBSETROW_LSX
ANY1(ARGBSetRow_Any_LSX, ARGBSetRow_LSX, uint32_t, 4, 3)
#endif
@ -2175,9 +2018,6 @@ ANY12(SplitUVRow_Any_AVX2, SplitUVRow_AVX2, 0, 2, 0, 31)
#ifdef HAS_SPLITUVROW_NEON
ANY12(SplitUVRow_Any_NEON, SplitUVRow_NEON, 0, 2, 0, 15)
#endif
#ifdef HAS_SPLITUVROW_MSA
ANY12(SplitUVRow_Any_MSA, SplitUVRow_MSA, 0, 2, 0, 31)
#endif
#ifdef HAS_SPLITUVROW_LSX
ANY12(SplitUVRow_Any_LSX, SplitUVRow_LSX, 0, 2, 0, 31)
#endif
@ -2211,11 +2051,6 @@ ANY12(UYVYToUV422Row_Any_NEON, UYVYToUV422Row_NEON, 1, 4, 1, 15)
ANY12(ARGBToUV444Row_Any_NEON_I8MM, ARGBToUV444Row_NEON_I8MM, 0, 4, 0, 7)
ANY12(ARGBToUVJ444Row_Any_NEON_I8MM, ARGBToUVJ444Row_NEON_I8MM, 0, 4, 0, 7)
#endif
#ifdef HAS_YUY2TOUV422ROW_MSA
ANY12(ARGBToUV444Row_Any_MSA, ARGBToUV444Row_MSA, 0, 4, 0, 15)
ANY12(YUY2ToUV422Row_Any_MSA, YUY2ToUV422Row_MSA, 1, 4, 1, 31)
ANY12(UYVYToUV422Row_Any_MSA, UYVYToUV422Row_MSA, 1, 4, 1, 31)
#endif
#ifdef HAS_YUY2TOUV422ROW_LSX
ANY12(ARGBToUV444Row_Any_LSX, ARGBToUV444Row_LSX, 0, 4, 0, 15)
ANY12(YUY2ToUV422Row_Any_LSX, YUY2ToUV422Row_LSX, 1, 4, 1, 15)
@ -2407,9 +2242,6 @@ ANY12S(ARGBToUVRow_Any_SVE2, ARGBToUVRow_SVE2, 0, 4, 1)
#ifdef HAS_ARGBTOUVROW_SME
ANY12S(ARGBToUVRow_Any_SME, ARGBToUVRow_SME, 0, 4, 1)
#endif
#ifdef HAS_ARGBTOUVROW_MSA
ANY12S(ARGBToUVRow_Any_MSA, ARGBToUVRow_MSA, 0, 4, 31)
#endif
#ifdef HAS_ARGBTOUVROW_LSX
ANY12S(ARGBToUVRow_Any_LSX, ARGBToUVRow_LSX, 0, 4, 15)
#endif
@ -2440,9 +2272,6 @@ ANY12S(ABGRToUVJRow_Any_SVE2, ABGRToUVJRow_SVE2, 0, 4, 1)
#ifdef HAS_ABGRTOUVJROW_SME
ANY12S(ABGRToUVJRow_Any_SME, ABGRToUVJRow_SME, 0, 4, 1)
#endif
#ifdef HAS_ARGBTOUVJROW_MSA
ANY12S(ARGBToUVJRow_Any_MSA, ARGBToUVJRow_MSA, 0, 4, 31)
#endif
#ifdef HAS_ARGBTOUVJROW_LSX
ANY12S(ARGBToUVJRow_Any_LSX, ARGBToUVJRow_LSX, 0, 4, 15)
#endif
@ -2461,9 +2290,6 @@ ANY12S(BGRAToUVRow_Any_SVE2, BGRAToUVRow_SVE2, 0, 4, 1)
#ifdef HAS_BGRATOUVROW_SME
ANY12S(BGRAToUVRow_Any_SME, BGRAToUVRow_SME, 0, 4, 1)
#endif
#ifdef HAS_BGRATOUVROW_MSA
ANY12S(BGRAToUVRow_Any_MSA, BGRAToUVRow_MSA, 0, 4, 15)
#endif
#ifdef HAS_BGRATOUVROW_LSX
ANY12S(BGRAToUVRow_Any_LSX, BGRAToUVRow_LSX, 0, 4, 15)
#endif
@ -2479,9 +2305,6 @@ ANY12S(ABGRToUVRow_Any_SVE2, ABGRToUVRow_SVE2, 0, 4, 1)
#ifdef HAS_ABGRTOUVROW_SME
ANY12S(ABGRToUVRow_Any_SME, ABGRToUVRow_SME, 0, 4, 1)
#endif
#ifdef HAS_ABGRTOUVROW_MSA
ANY12S(ABGRToUVRow_Any_MSA, ABGRToUVRow_MSA, 0, 4, 15)
#endif
#ifdef HAS_ABGRTOUVROW_LSX
ANY12S(ABGRToUVRow_Any_LSX, ABGRToUVRow_LSX, 0, 4, 15)
#endif
@ -2497,9 +2320,6 @@ ANY12S(RGBAToUVRow_Any_SVE2, RGBAToUVRow_SVE2, 0, 4, 1)
#ifdef HAS_RGBATOUVROW_SME
ANY12S(RGBAToUVRow_Any_SME, RGBAToUVRow_SME, 0, 4, 1)
#endif
#ifdef HAS_RGBATOUVROW_MSA
ANY12S(RGBAToUVRow_Any_MSA, RGBAToUVRow_MSA, 0, 4, 15)
#endif
#ifdef HAS_RGBATOUVROW_LSX
ANY12S(RGBAToUVRow_Any_LSX, RGBAToUVRow_LSX, 0, 4, 15)
#endif
@ -2509,9 +2329,6 @@ ANY12S(RGB24ToUVRow_Any_NEON, RGB24ToUVRow_NEON, 0, 3, 15)
#ifdef HAS_RGB24TOUVJROW_NEON
ANY12S(RGB24ToUVJRow_Any_NEON, RGB24ToUVJRow_NEON, 0, 3, 15)
#endif
#ifdef HAS_RGB24TOUVROW_MSA
ANY12S(RGB24ToUVRow_Any_MSA, RGB24ToUVRow_MSA, 0, 3, 15)
#endif
#ifdef HAS_RGB24TOUVROW_LSX
ANY12S(RGB24ToUVRow_Any_LSX, RGB24ToUVRow_LSX, 0, 3, 15)
#endif
@ -2524,9 +2341,6 @@ ANY12S(RAWToUVRow_Any_NEON, RAWToUVRow_NEON, 0, 3, 15)
#ifdef HAS_RAWTOUVJROW_NEON
ANY12S(RAWToUVJRow_Any_NEON, RAWToUVJRow_NEON, 0, 3, 15)
#endif
#ifdef HAS_RAWTOUVROW_MSA
ANY12S(RAWToUVRow_Any_MSA, RAWToUVRow_MSA, 0, 3, 15)
#endif
#ifdef HAS_RAWTOUVROW_LSX
ANY12S(RAWToUVRow_Any_LSX, RAWToUVRow_LSX, 0, 3, 15)
#endif
@ -2536,9 +2350,6 @@ ANY12S(RAWToUVRow_Any_LASX, RAWToUVRow_LASX, 0, 3, 31)
#ifdef HAS_RGB565TOUVROW_NEON
ANY12S(RGB565ToUVRow_Any_NEON, RGB565ToUVRow_NEON, 0, 2, 15)
#endif
#ifdef HAS_RGB565TOUVROW_MSA
ANY12S(RGB565ToUVRow_Any_MSA, RGB565ToUVRow_MSA, 0, 2, 15)
#endif
#ifdef HAS_RGB565TOUVROW_LSX
ANY12S(RGB565ToUVRow_Any_LSX, RGB565ToUVRow_LSX, 0, 2, 15)
#endif
@ -2548,9 +2359,6 @@ ANY12S(RGB565ToUVRow_Any_LASX, RGB565ToUVRow_LASX, 0, 2, 31)
#ifdef HAS_ARGB1555TOUVROW_NEON
ANY12S(ARGB1555ToUVRow_Any_NEON, ARGB1555ToUVRow_NEON, 0, 2, 15)
#endif
#ifdef HAS_ARGB1555TOUVROW_MSA
ANY12S(ARGB1555ToUVRow_Any_MSA, ARGB1555ToUVRow_MSA, 0, 2, 15)
#endif
#ifdef HAS_ARGB1555TOUVROW_LSX
ANY12S(ARGB1555ToUVRow_Any_LSX, ARGB1555ToUVRow_LSX, 0, 2, 15)
#endif
@ -2566,18 +2374,12 @@ ANY12S(YUY2ToUVRow_Any_NEON, YUY2ToUVRow_NEON, 1, 4, 15)
#ifdef HAS_UYVYTOUVROW_NEON
ANY12S(UYVYToUVRow_Any_NEON, UYVYToUVRow_NEON, 1, 4, 15)
#endif
#ifdef HAS_YUY2TOUVROW_MSA
ANY12S(YUY2ToUVRow_Any_MSA, YUY2ToUVRow_MSA, 1, 4, 31)
#endif
#ifdef HAS_YUY2TOUVROW_LSX
ANY12S(YUY2ToUVRow_Any_LSX, YUY2ToUVRow_LSX, 1, 4, 15)
#endif
#ifdef HAS_YUY2TOUVROW_LASX
ANY12S(YUY2ToUVRow_Any_LASX, YUY2ToUVRow_LASX, 1, 4, 31)
#endif
#ifdef HAS_UYVYTOUVROW_MSA
ANY12S(UYVYToUVRow_Any_MSA, UYVYToUVRow_MSA, 1, 4, 31)
#endif
#ifdef HAS_UYVYTOUVROW_LSX
ANY12S(UYVYToUVRow_Any_LSX, UYVYToUVRow_LSX, 1, 4, 15)
#endif

3597
source/row_msa.cc
View File

File diff suppressed because it is too large Load Diff

97
source/scale.cc
View File

@ -112,21 +112,6 @@ static void ScalePlaneDown2(int src_width,
}
}
#endif
#if defined(HAS_SCALEROWDOWN2_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ScaleRowDown2 =
filtering == kFilterNone
? ScaleRowDown2_Any_MSA
: (filtering == kFilterLinear ? ScaleRowDown2Linear_Any_MSA
: ScaleRowDown2Box_Any_MSA);
if (IS_ALIGNED(dst_width, 32)) {
ScaleRowDown2 = filtering == kFilterNone ? ScaleRowDown2_MSA
: (filtering == kFilterLinear
? ScaleRowDown2Linear_MSA
: ScaleRowDown2Box_MSA);
}
}
#endif
#if defined(HAS_SCALEROWDOWN2_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ScaleRowDown2 =
@ -317,15 +302,6 @@ static void ScalePlaneDown4(int src_width,
}
}
#endif
#if defined(HAS_SCALEROWDOWN4_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ScaleRowDown4 =
filtering ? ScaleRowDown4Box_Any_MSA : ScaleRowDown4_Any_MSA;
if (IS_ALIGNED(dst_width, 16)) {
ScaleRowDown4 = filtering ? ScaleRowDown4Box_MSA : ScaleRowDown4_MSA;
}
}
#endif
#if defined(HAS_SCALEROWDOWN4_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ScaleRowDown4 =
@ -445,27 +421,6 @@ static void ScalePlaneDown34(int src_width,
}
}
#endif
#if defined(HAS_SCALEROWDOWN34_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
if (dst_width % 48 == 0) {
if (!filtering) {
ScaleRowDown34_0 = ScaleRowDown34_MSA;
ScaleRowDown34_1 = ScaleRowDown34_MSA;
} else {
ScaleRowDown34_0 = ScaleRowDown34_0_Box_MSA;
ScaleRowDown34_1 = ScaleRowDown34_1_Box_MSA;
}
} else {
if (!filtering) {
ScaleRowDown34_0 = ScaleRowDown34_Any_MSA;
ScaleRowDown34_1 = ScaleRowDown34_Any_MSA;
} else {
ScaleRowDown34_0 = ScaleRowDown34_0_Box_Any_MSA;
ScaleRowDown34_1 = ScaleRowDown34_1_Box_Any_MSA;
}
}
}
#endif
#if defined(HAS_SCALEROWDOWN34_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
if (dst_width % 48 == 0) {
@ -694,26 +649,6 @@ static void ScalePlaneDown38(int src_width,
}
}
#endif
#if defined(HAS_SCALEROWDOWN38_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
if (!filtering) {
ScaleRowDown38_3 = ScaleRowDown38_Any_MSA;
ScaleRowDown38_2 = ScaleRowDown38_Any_MSA;
} else {
ScaleRowDown38_3 = ScaleRowDown38_3_Box_Any_MSA;
ScaleRowDown38_2 = ScaleRowDown38_2_Box_Any_MSA;
}
if (dst_width % 12 == 0) {
if (!filtering) {
ScaleRowDown38_3 = ScaleRowDown38_MSA;
ScaleRowDown38_2 = ScaleRowDown38_MSA;
} else {
ScaleRowDown38_3 = ScaleRowDown38_3_Box_MSA;
ScaleRowDown38_2 = ScaleRowDown38_2_Box_MSA;
}
}
}
#endif
#if defined(HAS_SCALEROWDOWN38_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
if (!filtering) {
@ -1015,14 +950,6 @@ static int ScalePlaneBox(int src_width,
}
}
#endif
#if defined(HAS_SCALEADDROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ScaleAddRow = ScaleAddRow_Any_MSA;
if (IS_ALIGNED(src_width, 16)) {
ScaleAddRow = ScaleAddRow_MSA;
}
}
#endif
#if defined(HAS_SCALEADDROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ScaleAddRow = ScaleAddRow_Any_LSX;
@ -1178,14 +1105,6 @@ static int ScalePlaneBilinearDown(int src_width,
InterpolateRow = InterpolateRow_SME;
}
#endif
#if defined(HAS_INTERPOLATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
InterpolateRow = InterpolateRow_Any_MSA;
if (IS_ALIGNED(src_width, 32)) {
InterpolateRow = InterpolateRow_MSA;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
@ -1213,14 +1132,6 @@ static int ScalePlaneBilinearDown(int src_width,
}
}
#endif
#if defined(HAS_SCALEFILTERCOLS_MSA)
if (TestCpuFlag(kCpuHasMSA) && src_width < 32768) {
ScaleFilterCols = ScaleFilterCols_Any_MSA;
if (IS_ALIGNED(dst_width, 16)) {
ScaleFilterCols = ScaleFilterCols_MSA;
}
}
#endif
#if defined(HAS_SCALEFILTERCOLS_LSX)
if (TestCpuFlag(kCpuHasLSX) && src_width < 32768) {
ScaleFilterCols = ScaleFilterCols_Any_LSX;
@ -1430,14 +1341,6 @@ static int ScalePlaneBilinearUp(int src_width,
}
}
#endif
#if defined(HAS_SCALEFILTERCOLS_MSA)
if (filtering && TestCpuFlag(kCpuHasMSA) && src_width < 32768) {
ScaleFilterCols = ScaleFilterCols_Any_MSA;
if (IS_ALIGNED(dst_width, 16)) {
ScaleFilterCols = ScaleFilterCols_MSA;
}
}
#endif
#if defined(HAS_SCALEFILTERCOLS_LSX)
if (filtering && TestCpuFlag(kCpuHasLSX) && src_width < 32768) {
ScaleFilterCols = ScaleFilterCols_Any_LSX;

116
source/scale_any.cc
View File

@ -152,22 +152,6 @@ SDANY(ScaleUVRowDown2Box_Any_NEON,
2,
7)
#endif
#ifdef HAS_SCALEROWDOWN2_MSA
SDANY(ScaleRowDown2_Any_MSA, ScaleRowDown2_MSA, ScaleRowDown2_C, 2, 1, 31)
SDANY(ScaleRowDown2Linear_Any_MSA,
ScaleRowDown2Linear_MSA,
ScaleRowDown2Linear_C,
2,
1,
31)
SDANY(ScaleRowDown2Box_Any_MSA,
ScaleRowDown2Box_MSA,
ScaleRowDown2Box_C,
2,
1,
31)
#endif
#ifdef HAS_SCALEROWDOWN2_LSX
SDANY(ScaleRowDown2_Any_LSX, ScaleRowDown2_LSX, ScaleRowDown2_C, 2, 1, 31)
SDANY(ScaleRowDown2Linear_Any_LSX,
@ -210,15 +194,6 @@ SDANY(ScaleRowDown4Box_Any_NEON,
1,
7)
#endif
#ifdef HAS_SCALEROWDOWN4_MSA
SDANY(ScaleRowDown4_Any_MSA, ScaleRowDown4_MSA, ScaleRowDown4_C, 4, 1, 15)
SDANY(ScaleRowDown4Box_Any_MSA,
ScaleRowDown4Box_MSA,
ScaleRowDown4Box_C,
4,
1,
15)
#endif
#ifdef HAS_SCALEROWDOWN4_LSX
SDANY(ScaleRowDown4_Any_LSX, ScaleRowDown4_LSX, ScaleRowDown4_C, 4, 1, 15)
SDANY(ScaleRowDown4Box_Any_LSX,
@ -289,26 +264,6 @@ SDANY(ScaleRowDown34_1_Box_Any_NEON,
23)
#endif
#endif
#ifdef HAS_SCALEROWDOWN34_MSA
SDANY(ScaleRowDown34_Any_MSA,
ScaleRowDown34_MSA,
ScaleRowDown34_C,
4 / 3,
1,
47)
SDANY(ScaleRowDown34_0_Box_Any_MSA,
ScaleRowDown34_0_Box_MSA,
ScaleRowDown34_0_Box_C,
4 / 3,
1,
47)
SDANY(ScaleRowDown34_1_Box_Any_MSA,
ScaleRowDown34_1_Box_MSA,
ScaleRowDown34_1_Box_C,
4 / 3,
1,
47)
#endif
#ifdef HAS_SCALEROWDOWN34_LSX
SDANY(ScaleRowDown34_Any_LSX,
ScaleRowDown34_LSX,
@ -369,26 +324,6 @@ SDANY(ScaleRowDown38_2_Box_Any_NEON,
1,
11)
#endif
#ifdef HAS_SCALEROWDOWN38_MSA
SDANY(ScaleRowDown38_Any_MSA,
ScaleRowDown38_MSA,
ScaleRowDown38_C,
8 / 3,
1,
11)
SDANY(ScaleRowDown38_3_Box_Any_MSA,
ScaleRowDown38_3_Box_MSA,
ScaleRowDown38_3_Box_C,
8 / 3,
1,
11)
SDANY(ScaleRowDown38_2_Box_Any_MSA,
ScaleRowDown38_2_Box_MSA,
ScaleRowDown38_2_Box_C,
8 / 3,
1,
11)
#endif
#ifdef HAS_SCALEROWDOWN38_LSX
SDANY(ScaleRowDown38_Any_LSX,
ScaleRowDown38_LSX,
@ -450,26 +385,6 @@ SDANY(ScaleARGBRowDown2Box_Any_NEON,
4,
7)
#endif
#ifdef HAS_SCALEARGBROWDOWN2_MSA
SDANY(ScaleARGBRowDown2_Any_MSA,
ScaleARGBRowDown2_MSA,
ScaleARGBRowDown2_C,
2,
4,
3)
SDANY(ScaleARGBRowDown2Linear_Any_MSA,
ScaleARGBRowDown2Linear_MSA,
ScaleARGBRowDown2Linear_C,
2,
4,
3)
SDANY(ScaleARGBRowDown2Box_Any_MSA,
ScaleARGBRowDown2Box_MSA,
ScaleARGBRowDown2Box_C,
2,
4,
3)
#endif
#ifdef HAS_SCALEARGBROWDOWN2_LSX
SDANY(ScaleARGBRowDown2_Any_LSX,
ScaleARGBRowDown2_LSX,
@ -529,18 +444,6 @@ SDAANY(ScaleARGBRowDownEvenBox_Any_NEON,
4,
3)
#endif
#ifdef HAS_SCALEARGBROWDOWNEVEN_MSA
SDAANY(ScaleARGBRowDownEven_Any_MSA,
ScaleARGBRowDownEven_MSA,
ScaleARGBRowDownEven_C,
4,
3)
SDAANY(ScaleARGBRowDownEvenBox_Any_MSA,
ScaleARGBRowDownEvenBox_MSA,
ScaleARGBRowDownEvenBox_C,
4,
3)
#endif
#ifdef HAS_SCALEARGBROWDOWNEVEN_LSX
SDAANY(ScaleARGBRowDownEven_Any_LSX,
ScaleARGBRowDownEven_LSX,
@ -590,9 +493,6 @@ SAROW(ScaleAddRow_Any_AVX2, ScaleAddRow_AVX2, 1, 2, 31)
#ifdef HAS_SCALEADDROW_NEON
SAROW(ScaleAddRow_Any_NEON, ScaleAddRow_NEON, 1, 2, 15)
#endif
#ifdef HAS_SCALEADDROW_MSA
SAROW(ScaleAddRow_Any_MSA, ScaleAddRow_MSA, 1, 2, 15)
#endif
#ifdef HAS_SCALEADDROW_LSX
SAROW(ScaleAddRow_Any_LSX, ScaleAddRow_LSX, 1, 2, 15)
#endif
@ -619,9 +519,6 @@ SAANY(ScaleAddRow_Any_AVX2, ScaleAddRow_AVX2, ScaleAddRow_C, 31)
#ifdef HAS_SCALEADDROW_NEON
SAANY(ScaleAddRow_Any_NEON, ScaleAddRow_NEON, ScaleAddRow_C, 15)
#endif
#ifdef HAS_SCALEADDROW_MSA
SAANY(ScaleAddRow_Any_MSA, ScaleAddRow_MSA, ScaleAddRow_C, 15)
#endif
#ifdef HAS_SCALEADDROW_LSX
SAANY(ScaleAddRow_Any_LSX, ScaleAddRow_LSX, ScaleAddRow_C, 15)
#endif
@ -644,18 +541,12 @@ SAANY(ScaleAddRow_Any_LSX, ScaleAddRow_LSX, ScaleAddRow_C, 15)
#ifdef HAS_SCALEFILTERCOLS_NEON
CANY(ScaleFilterCols_Any_NEON, ScaleFilterCols_NEON, ScaleFilterCols_C, 1, 7)
#endif
#ifdef HAS_SCALEFILTERCOLS_MSA
CANY(ScaleFilterCols_Any_MSA, ScaleFilterCols_MSA, ScaleFilterCols_C, 1, 15)
#endif
#ifdef HAS_SCALEFILTERCOLS_LSX
CANY(ScaleFilterCols_Any_LSX, ScaleFilterCols_LSX, ScaleFilterCols_C, 1, 15)
#endif
#ifdef HAS_SCALEARGBCOLS_NEON
CANY(ScaleARGBCols_Any_NEON, ScaleARGBCols_NEON, ScaleARGBCols_C, 4, 7)
#endif
#ifdef HAS_SCALEARGBCOLS_MSA
CANY(ScaleARGBCols_Any_MSA, ScaleARGBCols_MSA, ScaleARGBCols_C, 4, 3)
#endif
#ifdef HAS_SCALEARGBCOLS_LSX
CANY(ScaleARGBCols_Any_LSX, ScaleARGBCols_LSX, ScaleARGBCols_C, 4, 3)
#endif
@ -666,13 +557,6 @@ CANY(ScaleARGBFilterCols_Any_NEON,
4,
3)
#endif
#ifdef HAS_SCALEARGBFILTERCOLS_MSA
CANY(ScaleARGBFilterCols_Any_MSA,
ScaleARGBFilterCols_MSA,
ScaleARGBFilterCols_C,
4,
7)
#endif
#ifdef HAS_SCALEARGBFILTERCOLS_LSX
CANY(ScaleARGBFilterCols_Any_LSX,
ScaleARGBFilterCols_LSX,

106
source/scale_argb.cc
View File

@ -107,22 +107,6 @@ static void ScaleARGBDown2(int src_width,
: ScaleARGBRowDown2Box_SME;
}
#endif
#if defined(HAS_SCALEARGBROWDOWN2_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ScaleARGBRowDown2 =
filtering == kFilterNone
? ScaleARGBRowDown2_Any_MSA
: (filtering == kFilterLinear ? ScaleARGBRowDown2Linear_Any_MSA
: ScaleARGBRowDown2Box_Any_MSA);
if (IS_ALIGNED(dst_width, 4)) {
ScaleARGBRowDown2 =
filtering == kFilterNone
? ScaleARGBRowDown2_MSA
: (filtering == kFilterLinear ? ScaleARGBRowDown2Linear_MSA
: ScaleARGBRowDown2Box_MSA);
}
}
#endif
#if defined(HAS_SCALEARGBROWDOWN2_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ScaleARGBRowDown2 =
@ -282,16 +266,6 @@ static void ScaleARGBDownEven(int src_width,
}
}
#endif
#if defined(HAS_SCALEARGBROWDOWNEVEN_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ScaleARGBRowDownEven = filtering ? ScaleARGBRowDownEvenBox_Any_MSA
: ScaleARGBRowDownEven_Any_MSA;
if (IS_ALIGNED(dst_width, 4)) {
ScaleARGBRowDownEven =
filtering ? ScaleARGBRowDownEvenBox_MSA : ScaleARGBRowDownEven_MSA;
}
}
#endif
#if defined(HAS_SCALEARGBROWDOWNEVEN_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ScaleARGBRowDownEven = filtering ? ScaleARGBRowDownEvenBox_Any_LSX
@ -386,14 +360,6 @@ static int ScaleARGBBilinearDown(int src_width,
InterpolateRow = InterpolateRow_SME;
}
#endif
#if defined(HAS_INTERPOLATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
InterpolateRow = InterpolateRow_Any_MSA;
if (IS_ALIGNED(clip_src_width, 32)) {
InterpolateRow = InterpolateRow_MSA;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
@ -420,14 +386,6 @@ static int ScaleARGBBilinearDown(int src_width,
}
}
#endif
#if defined(HAS_SCALEARGBFILTERCOLS_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_Any_MSA;
if (IS_ALIGNED(dst_width, 8)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_MSA;
}
}
#endif
#if defined(HAS_SCALEARGBFILTERCOLS_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_Any_LSX;
@ -525,14 +483,6 @@ static int ScaleARGBBilinearUp(int src_width,
InterpolateRow = InterpolateRow_SME;
}
#endif
#if defined(HAS_INTERPOLATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
InterpolateRow = InterpolateRow_Any_MSA;
if (IS_ALIGNED(dst_width, 8)) {
InterpolateRow = InterpolateRow_MSA;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
@ -563,14 +513,6 @@ static int ScaleARGBBilinearUp(int src_width,
}
}
#endif
#if defined(HAS_SCALEARGBFILTERCOLS_MSA)
if (filtering && TestCpuFlag(kCpuHasMSA)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_Any_MSA;
if (IS_ALIGNED(dst_width, 8)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_MSA;
}
}
#endif
#if defined(HAS_SCALEARGBFILTERCOLS_LSX)
if (filtering && TestCpuFlag(kCpuHasLSX)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_Any_LSX;
@ -597,14 +539,6 @@ static int ScaleARGBBilinearUp(int src_width,
}
}
#endif
#if defined(HAS_SCALEARGBCOLS_MSA)
if (!filtering && TestCpuFlag(kCpuHasMSA)) {
ScaleARGBFilterCols = ScaleARGBCols_Any_MSA;
if (IS_ALIGNED(dst_width, 4)) {
ScaleARGBFilterCols = ScaleARGBCols_MSA;
}
}
#endif
#if defined(HAS_SCALEARGBCOLS_LSX)
if (!filtering && TestCpuFlag(kCpuHasLSX)) {
ScaleARGBFilterCols = ScaleARGBCols_Any_LSX;
@ -747,14 +681,6 @@ static int ScaleYUVToARGBBilinearUp(int src_width,
I422ToARGBRow = I422ToARGBRow_SME;
}
#endif
#if defined(HAS_I422TOARGBROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
I422ToARGBRow = I422ToARGBRow_Any_MSA;
if (IS_ALIGNED(src_width, 8)) {
I422ToARGBRow = I422ToARGBRow_MSA;
}
}
#endif
#if defined(HAS_I422TOARGBROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
I422ToARGBRow = I422ToARGBRow_Any_LSX;
@ -809,14 +735,6 @@ static int ScaleYUVToARGBBilinearUp(int src_width,
InterpolateRow = InterpolateRow_SME;
}
#endif
#if defined(HAS_INTERPOLATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
InterpolateRow = InterpolateRow_Any_MSA;
if (IS_ALIGNED(dst_width, 8)) {
InterpolateRow = InterpolateRow_MSA;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
@ -851,14 +769,6 @@ static int ScaleYUVToARGBBilinearUp(int src_width,
}
}
#endif
#if defined(HAS_SCALEARGBFILTERCOLS_MSA)
if (filtering && TestCpuFlag(kCpuHasMSA)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_Any_MSA;
if (IS_ALIGNED(dst_width, 8)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_MSA;
}
}
#endif
#if defined(HAS_SCALEARGBFILTERCOLS_LSX)
if (filtering && TestCpuFlag(kCpuHasLSX)) {
ScaleARGBFilterCols = ScaleARGBFilterCols_Any_LSX;
@ -885,14 +795,6 @@ static int ScaleYUVToARGBBilinearUp(int src_width,
}
}
#endif
#if defined(HAS_SCALEARGBCOLS_MSA)
if (!filtering && TestCpuFlag(kCpuHasMSA)) {
ScaleARGBFilterCols = ScaleARGBCols_Any_MSA;
if (IS_ALIGNED(dst_width, 4)) {
ScaleARGBFilterCols = ScaleARGBCols_MSA;
}
}
#endif
#if defined(HAS_SCALEARGBCOLS_LSX)
if (!filtering && TestCpuFlag(kCpuHasLSX)) {
ScaleARGBFilterCols = ScaleARGBCols_Any_LSX;
@ -1025,14 +927,6 @@ static void ScaleARGBSimple(int src_width,
}
}
#endif
#if defined(HAS_SCALEARGBCOLS_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ScaleARGBCols = ScaleARGBCols_Any_MSA;
if (IS_ALIGNED(dst_width, 4)) {
ScaleARGBCols = ScaleARGBCols_MSA;
}
}
#endif
#if defined(HAS_SCALEARGBCOLS_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
ScaleARGBCols = ScaleARGBCols_Any_LSX;

8
source/scale_common.cc
View File

@ -1667,14 +1667,6 @@ void ScalePlaneVertical(int src_height,
InterpolateRow = InterpolateRow_SME;
}
#endif
#if defined(HAS_INTERPOLATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
InterpolateRow = InterpolateRow_Any_MSA;
if (IS_ALIGNED(dst_width_bytes, 32)) {
InterpolateRow = InterpolateRow_MSA;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;

949
source/scale_msa.cc
View File

@ -1,949 +0,0 @@
/*
* Copyright 2016 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.
*/
#include <assert.h>
#include "libyuv/scale_row.h"
// This module is for GCC MSA
#if !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)
#include "libyuv/macros_msa.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
#define LOAD_INDEXED_DATA(srcp, indx0, out0) \
{ \
out0[0] = srcp[indx0[0]]; \
out0[1] = srcp[indx0[1]]; \
out0[2] = srcp[indx0[2]]; \
out0[3] = srcp[indx0[3]]; \
}
void ScaleARGBRowDown2_MSA(const uint8_t* src_argb,
ptrdiff_t src_stride,
uint8_t* dst_argb,
int dst_width) {
int x;
v16u8 src0, src1, dst0;
(void)src_stride;
for (x = 0; x < dst_width; x += 4) {
src0 = (v16u8)__msa_ld_b((v16i8*)src_argb, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)src_argb, 16);
dst0 = (v16u8)__msa_pckod_w((v4i32)src1, (v4i32)src0);
ST_UB(dst0, dst_argb);
src_argb += 32;
dst_argb += 16;
}
}
void ScaleARGBRowDown2Linear_MSA(const uint8_t* src_argb,
ptrdiff_t src_stride,
uint8_t* dst_argb,
int dst_width) {
int x;
v16u8 src0, src1, vec0, vec1, dst0;
(void)src_stride;
for (x = 0; x < dst_width; x += 4) {
src0 = (v16u8)__msa_ld_b((v16i8*)src_argb, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)src_argb, 16);
vec0 = (v16u8)__msa_pckev_w((v4i32)src1, (v4i32)src0);
vec1 = (v16u8)__msa_pckod_w((v4i32)src1, (v4i32)src0);
dst0 = (v16u8)__msa_aver_u_b((v16u8)vec0, (v16u8)vec1);
ST_UB(dst0, dst_argb);
src_argb += 32;
dst_argb += 16;
}
}
void ScaleARGBRowDown2Box_MSA(const uint8_t* src_argb,
ptrdiff_t src_stride,
uint8_t* dst_argb,
int dst_width) {
int x;
const uint8_t* s = src_argb;
const uint8_t* t = src_argb + src_stride;
v16u8 src0, src1, src2, src3, vec0, vec1, vec2, vec3, dst0;
v8u16 reg0, reg1, reg2, reg3;
v16i8 shuffler = {0, 4, 1, 5, 2, 6, 3, 7, 8, 12, 9, 13, 10, 14, 11, 15};
for (x = 0; x < dst_width; x += 4) {
src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)s, 16);
src2 = (v16u8)__msa_ld_b((v16i8*)t, 0);
src3 = (v16u8)__msa_ld_b((v16i8*)t, 16);
vec0 = (v16u8)__msa_vshf_b(shuffler, (v16i8)src0, (v16i8)src0);
vec1 = (v16u8)__msa_vshf_b(shuffler, (v16i8)src1, (v16i8)src1);
vec2 = (v16u8)__msa_vshf_b(shuffler, (v16i8)src2, (v16i8)src2);
vec3 = (v16u8)__msa_vshf_b(shuffler, (v16i8)src3, (v16i8)src3);
reg0 = __msa_hadd_u_h(vec0, vec0);
reg1 = __msa_hadd_u_h(vec1, vec1);
reg2 = __msa_hadd_u_h(vec2, vec2);
reg3 = __msa_hadd_u_h(vec3, vec3);
reg0 += reg2;
reg1 += reg3;
reg0 = (v8u16)__msa_srari_h((v8i16)reg0, 2);
reg1 = (v8u16)__msa_srari_h((v8i16)reg1, 2);
dst0 = (v16u8)__msa_pckev_b((v16i8)reg1, (v16i8)reg0);
ST_UB(dst0, dst_argb);
s += 32;
t += 32;
dst_argb += 16;
}
}
void ScaleARGBRowDownEven_MSA(const uint8_t* src_argb,
ptrdiff_t src_stride,
int32_t src_stepx,
uint8_t* dst_argb,
int dst_width) {
int x;
int32_t stepx = src_stepx * 4;
int32_t data0, data1, data2, data3;
(void)src_stride;
for (x = 0; x < dst_width; x += 4) {
data0 = LW(src_argb);
data1 = LW(src_argb + stepx);
data2 = LW(src_argb + stepx * 2);
data3 = LW(src_argb + stepx * 3);
SW(data0, dst_argb);
SW(data1, dst_argb + 4);
SW(data2, dst_argb + 8);
SW(data3, dst_argb + 12);
src_argb += stepx * 4;
dst_argb += 16;
}
}
void ScaleARGBRowDownEvenBox_MSA(const uint8_t* src_argb,
ptrdiff_t src_stride,
int src_stepx,
uint8_t* dst_argb,
int dst_width) {
int x;
const uint8_t* nxt_argb = src_argb + src_stride;
int32_t stepx = src_stepx * 4;
int64_t data0, data1, data2, data3;
v16u8 src0 = {0}, src1 = {0}, src2 = {0}, src3 = {0};
v16u8 vec0, vec1, vec2, vec3;
v8u16 reg0, reg1, reg2, reg3, reg4, reg5, reg6, reg7;
v16u8 dst0;
for (x = 0; x < dst_width; x += 4) {
data0 = LD(src_argb);
data1 = LD(src_argb + stepx);
data2 = LD(src_argb + stepx * 2);
data3 = LD(src_argb + stepx * 3);
src0 = (v16u8)__msa_insert_d((v2i64)src0, 0, data0);
src0 = (v16u8)__msa_insert_d((v2i64)src0, 1, data1);
src1 = (v16u8)__msa_insert_d((v2i64)src1, 0, data2);
src1 = (v16u8)__msa_insert_d((v2i64)src1, 1, data3);
data0 = LD(nxt_argb);
data1 = LD(nxt_argb + stepx);
data2 = LD(nxt_argb + stepx * 2);
data3 = LD(nxt_argb + stepx * 3);
src2 = (v16u8)__msa_insert_d((v2i64)src2, 0, data0);
src2 = (v16u8)__msa_insert_d((v2i64)src2, 1, data1);
src3 = (v16u8)__msa_insert_d((v2i64)src3, 0, data2);
src3 = (v16u8)__msa_insert_d((v2i64)src3, 1, data3);
vec0 = (v16u8)__msa_ilvr_b((v16i8)src2, (v16i8)src0);
vec1 = (v16u8)__msa_ilvr_b((v16i8)src3, (v16i8)src1);
vec2 = (v16u8)__msa_ilvl_b((v16i8)src2, (v16i8)src0);
vec3 = (v16u8)__msa_ilvl_b((v16i8)src3, (v16i8)src1);
reg0 = __msa_hadd_u_h(vec0, vec0);
reg1 = __msa_hadd_u_h(vec1, vec1);
reg2 = __msa_hadd_u_h(vec2, vec2);
reg3 = __msa_hadd_u_h(vec3, vec3);
reg4 = (v8u16)__msa_pckev_d((v2i64)reg2, (v2i64)reg0);
reg5 = (v8u16)__msa_pckev_d((v2i64)reg3, (v2i64)reg1);
reg6 = (v8u16)__msa_pckod_d((v2i64)reg2, (v2i64)reg0);
reg7 = (v8u16)__msa_pckod_d((v2i64)reg3, (v2i64)reg1);
reg4 += reg6;
reg5 += reg7;
reg4 = (v8u16)__msa_srari_h((v8i16)reg4, 2);
reg5 = (v8u16)__msa_srari_h((v8i16)reg5, 2);
dst0 = (v16u8)__msa_pckev_b((v16i8)reg5, (v16i8)reg4);
ST_UB(dst0, dst_argb);
src_argb += stepx * 4;
nxt_argb += stepx * 4;
dst_argb += 16;
}
}
void ScaleRowDown2_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,
int dst_width) {
int x;
v16u8 src0, src1, src2, src3, dst0, dst1;
(void)src_stride;
for (x = 0; x < dst_width; x += 32) {
src0 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 16);
src2 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 32);
src3 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 48);
dst0 = (v16u8)__msa_pckod_b((v16i8)src1, (v16i8)src0);
dst1 = (v16u8)__msa_pckod_b((v16i8)src3, (v16i8)src2);
ST_UB2(dst0, dst1, dst, 16);
src_ptr += 64;
dst += 32;
}
}
void ScaleRowDown2Linear_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,
int dst_width) {
int x;
v16u8 src0, src1, src2, src3, vec0, vec1, vec2, vec3, dst0, dst1;
(void)src_stride;
for (x = 0; x < dst_width; x += 32) {
src0 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 16);
src2 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 32);
src3 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 48);
vec0 = (v16u8)__msa_pckev_b((v16i8)src1, (v16i8)src0);
vec2 = (v16u8)__msa_pckev_b((v16i8)src3, (v16i8)src2);
vec1 = (v16u8)__msa_pckod_b((v16i8)src1, (v16i8)src0);
vec3 = (v16u8)__msa_pckod_b((v16i8)src3, (v16i8)src2);
dst0 = __msa_aver_u_b(vec1, vec0);
dst1 = __msa_aver_u_b(vec3, vec2);
ST_UB2(dst0, dst1, dst, 16);
src_ptr += 64;
dst += 32;
}
}
void ScaleRowDown2Box_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,
int dst_width) {
int x;
const uint8_t* s = src_ptr;
const uint8_t* t = src_ptr + src_stride;
v16u8 src0, src1, src2, src3, src4, src5, src6, src7, dst0, dst1;
v8u16 vec0, vec1, vec2, vec3;
for (x = 0; x < dst_width; x += 32) {
src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)s, 16);
src2 = (v16u8)__msa_ld_b((v16i8*)s, 32);
src3 = (v16u8)__msa_ld_b((v16i8*)s, 48);
src4 = (v16u8)__msa_ld_b((v16i8*)t, 0);
src5 = (v16u8)__msa_ld_b((v16i8*)t, 16);
src6 = (v16u8)__msa_ld_b((v16i8*)t, 32);
src7 = (v16u8)__msa_ld_b((v16i8*)t, 48);
vec0 = __msa_hadd_u_h(src0, src0);
vec1 = __msa_hadd_u_h(src1, src1);
vec2 = __msa_hadd_u_h(src2, src2);
vec3 = __msa_hadd_u_h(src3, src3);
vec0 += __msa_hadd_u_h(src4, src4);
vec1 += __msa_hadd_u_h(src5, src5);
vec2 += __msa_hadd_u_h(src6, src6);
vec3 += __msa_hadd_u_h(src7, src7);
vec0 = (v8u16)__msa_srari_h((v8i16)vec0, 2);
vec1 = (v8u16)__msa_srari_h((v8i16)vec1, 2);
vec2 = (v8u16)__msa_srari_h((v8i16)vec2, 2);
vec3 = (v8u16)__msa_srari_h((v8i16)vec3, 2);
dst0 = (v16u8)__msa_pckev_b((v16i8)vec1, (v16i8)vec0);
dst1 = (v16u8)__msa_pckev_b((v16i8)vec3, (v16i8)vec2);
ST_UB2(dst0, dst1, dst, 16);
s += 64;
t += 64;
dst += 32;
}
}
void ScaleRowDown4_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,
int dst_width) {
int x;
v16u8 src0, src1, src2, src3, vec0, vec1, dst0;
(void)src_stride;
for (x = 0; x < dst_width; x += 16) {
src0 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 16);
src2 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 32);
src3 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 48);
vec0 = (v16u8)__msa_pckev_b((v16i8)src1, (v16i8)src0);
vec1 = (v16u8)__msa_pckev_b((v16i8)src3, (v16i8)src2);
dst0 = (v16u8)__msa_pckod_b((v16i8)vec1, (v16i8)vec0);
ST_UB(dst0, dst);
src_ptr += 64;
dst += 16;
}
}
void ScaleRowDown4Box_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,
int dst_width) {
int x;
const uint8_t* s = src_ptr;
const uint8_t* t0 = s + src_stride;
const uint8_t* t1 = s + src_stride * 2;
const uint8_t* t2 = s + src_stride * 3;
v16u8 src0, src1, src2, src3, src4, src5, src6, src7, dst0;
v8u16 vec0, vec1, vec2, vec3;
v4u32 reg0, reg1, reg2, reg3;
for (x = 0; x < dst_width; x += 16) {
src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)s, 16);
src2 = (v16u8)__msa_ld_b((v16i8*)s, 32);
src3 = (v16u8)__msa_ld_b((v16i8*)s, 48);
src4 = (v16u8)__msa_ld_b((v16i8*)t0, 0);
src5 = (v16u8)__msa_ld_b((v16i8*)t0, 16);
src6 = (v16u8)__msa_ld_b((v16i8*)t0, 32);
src7 = (v16u8)__msa_ld_b((v16i8*)t0, 48);
vec0 = __msa_hadd_u_h(src0, src0);
vec1 = __msa_hadd_u_h(src1, src1);
vec2 = __msa_hadd_u_h(src2, src2);
vec3 = __msa_hadd_u_h(src3, src3);
vec0 += __msa_hadd_u_h(src4, src4);
vec1 += __msa_hadd_u_h(src5, src5);
vec2 += __msa_hadd_u_h(src6, src6);
vec3 += __msa_hadd_u_h(src7, src7);
src0 = (v16u8)__msa_ld_b((v16i8*)t1, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)t1, 16);
src2 = (v16u8)__msa_ld_b((v16i8*)t1, 32);
src3 = (v16u8)__msa_ld_b((v16i8*)t1, 48);
src4 = (v16u8)__msa_ld_b((v16i8*)t2, 0);
src5 = (v16u8)__msa_ld_b((v16i8*)t2, 16);
src6 = (v16u8)__msa_ld_b((v16i8*)t2, 32);
src7 = (v16u8)__msa_ld_b((v16i8*)t2, 48);
vec0 += __msa_hadd_u_h(src0, src0);
vec1 += __msa_hadd_u_h(src1, src1);
vec2 += __msa_hadd_u_h(src2, src2);
vec3 += __msa_hadd_u_h(src3, src3);
vec0 += __msa_hadd_u_h(src4, src4);
vec1 += __msa_hadd_u_h(src5, src5);
vec2 += __msa_hadd_u_h(src6, src6);
vec3 += __msa_hadd_u_h(src7, src7);
reg0 = __msa_hadd_u_w(vec0, vec0);
reg1 = __msa_hadd_u_w(vec1, vec1);
reg2 = __msa_hadd_u_w(vec2, vec2);
reg3 = __msa_hadd_u_w(vec3, vec3);
reg0 = (v4u32)__msa_srari_w((v4i32)reg0, 4);
reg1 = (v4u32)__msa_srari_w((v4i32)reg1, 4);
reg2 = (v4u32)__msa_srari_w((v4i32)reg2, 4);
reg3 = (v4u32)__msa_srari_w((v4i32)reg3, 4);
vec0 = (v8u16)__msa_pckev_h((v8i16)reg1, (v8i16)reg0);
vec1 = (v8u16)__msa_pckev_h((v8i16)reg3, (v8i16)reg2);
dst0 = (v16u8)__msa_pckev_b((v16i8)vec1, (v16i8)vec0);
ST_UB(dst0, dst);
s += 64;
t0 += 64;
t1 += 64;
t2 += 64;
dst += 16;
}
}
void ScaleRowDown38_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,
int dst_width) {
int x, width;
uint64_t dst0;
uint32_t dst1;
v16u8 src0, src1, vec0;
v16i8 mask = {0, 3, 6, 8, 11, 14, 16, 19, 22, 24, 27, 30, 0, 0, 0, 0};
(void)src_stride;
assert(dst_width % 3 == 0);
width = dst_width / 3;
for (x = 0; x < width; x += 4) {
src0 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 16);
vec0 = (v16u8)__msa_vshf_b(mask, (v16i8)src1, (v16i8)src0);
dst0 = __msa_copy_u_d((v2i64)vec0, 0);
dst1 = __msa_copy_u_w((v4i32)vec0, 2);
SD(dst0, dst);
SW(dst1, dst + 8);
src_ptr += 32;
dst += 12;
}
}
void ScaleRowDown38_2_Box_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
int x, width;
const uint8_t* s = src_ptr;
const uint8_t* t = src_ptr + src_stride;
uint64_t dst0;
uint32_t dst1;
v16u8 src0, src1, src2, src3, out;
v8u16 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
v4u32 tmp0, tmp1, tmp2, tmp3, tmp4;
v8i16 zero = {0};
v8i16 mask = {0, 1, 2, 8, 3, 4, 5, 9};
v16i8 dst_mask = {0, 2, 16, 4, 6, 18, 8, 10, 20, 12, 14, 22, 0, 0, 0, 0};
v4u32 const_0x2AAA = (v4u32)__msa_fill_w(0x2AAA);
v4u32 const_0x4000 = (v4u32)__msa_fill_w(0x4000);
assert((dst_width % 3 == 0) && (dst_width > 0));
width = dst_width / 3;
for (x = 0; x < width; x += 4) {
src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)s, 16);
src2 = (v16u8)__msa_ld_b((v16i8*)t, 0);
src3 = (v16u8)__msa_ld_b((v16i8*)t, 16);
vec0 = (v8u16)__msa_ilvr_b((v16i8)src2, (v16i8)src0);
vec1 = (v8u16)__msa_ilvl_b((v16i8)src2, (v16i8)src0);
vec2 = (v8u16)__msa_ilvr_b((v16i8)src3, (v16i8)src1);
vec3 = (v8u16)__msa_ilvl_b((v16i8)src3, (v16i8)src1);
vec0 = __msa_hadd_u_h((v16u8)vec0, (v16u8)vec0);
vec1 = __msa_hadd_u_h((v16u8)vec1, (v16u8)vec1);
vec2 = __msa_hadd_u_h((v16u8)vec2, (v16u8)vec2);
vec3 = __msa_hadd_u_h((v16u8)vec3, (v16u8)vec3);
vec4 = (v8u16)__msa_vshf_h(mask, zero, (v8i16)vec0);
vec5 = (v8u16)__msa_vshf_h(mask, zero, (v8i16)vec1);
vec6 = (v8u16)__msa_vshf_h(mask, zero, (v8i16)vec2);
vec7 = (v8u16)__msa_vshf_h(mask, zero, (v8i16)vec3);
vec0 = (v8u16)__msa_pckod_w((v4i32)vec1, (v4i32)vec0);
vec1 = (v8u16)__msa_pckod_w((v4i32)vec3, (v4i32)vec2);
vec0 = (v8u16)__msa_pckod_w((v4i32)vec1, (v4i32)vec0);
tmp0 = __msa_hadd_u_w(vec4, vec4);
tmp1 = __msa_hadd_u_w(vec5, vec5);
tmp2 = __msa_hadd_u_w(vec6, vec6);
tmp3 = __msa_hadd_u_w(vec7, vec7);
tmp4 = __msa_hadd_u_w(vec0, vec0);
vec0 = (v8u16)__msa_pckev_h((v8i16)tmp1, (v8i16)tmp0);
vec1 = (v8u16)__msa_pckev_h((v8i16)tmp3, (v8i16)tmp2);
tmp0 = __msa_hadd_u_w(vec0, vec0);
tmp1 = __msa_hadd_u_w(vec1, vec1);
tmp0 *= const_0x2AAA;
tmp1 *= const_0x2AAA;
tmp4 *= const_0x4000;
tmp0 = (v4u32)__msa_srai_w((v4i32)tmp0, 16);
tmp1 = (v4u32)__msa_srai_w((v4i32)tmp1, 16);
tmp4 = (v4u32)__msa_srai_w((v4i32)tmp4, 16);
vec0 = (v8u16)__msa_pckev_h((v8i16)tmp1, (v8i16)tmp0);
vec1 = (v8u16)__msa_pckev_h((v8i16)tmp4, (v8i16)tmp4);
out = (v16u8)__msa_vshf_b(dst_mask, (v16i8)vec1, (v16i8)vec0);
dst0 = __msa_copy_u_d((v2i64)out, 0);
dst1 = __msa_copy_u_w((v4i32)out, 2);
SD(dst0, dst_ptr);
SW(dst1, dst_ptr + 8);
s += 32;
t += 32;
dst_ptr += 12;
}
}
void ScaleRowDown38_3_Box_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst_ptr,
int dst_width) {
int x, width;
const uint8_t* s = src_ptr;
const uint8_t* t0 = s + src_stride;
const uint8_t* t1 = s + src_stride * 2;
uint64_t dst0;
uint32_t dst1;
v16u8 src0, src1, src2, src3, src4, src5, out;
v8u16 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7;
v4u32 tmp0, tmp1, tmp2, tmp3, tmp4;
v8u16 zero = {0};
v8i16 mask = {0, 1, 2, 8, 3, 4, 5, 9};
v16i8 dst_mask = {0, 2, 16, 4, 6, 18, 8, 10, 20, 12, 14, 22, 0, 0, 0, 0};
v4u32 const_0x1C71 = (v4u32)__msa_fill_w(0x1C71);
v4u32 const_0x2AAA = (v4u32)__msa_fill_w(0x2AAA);
assert((dst_width % 3 == 0) && (dst_width > 0));
width = dst_width / 3;
for (x = 0; x < width; x += 4) {
src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)s, 16);
src2 = (v16u8)__msa_ld_b((v16i8*)t0, 0);
src3 = (v16u8)__msa_ld_b((v16i8*)t0, 16);
src4 = (v16u8)__msa_ld_b((v16i8*)t1, 0);
src5 = (v16u8)__msa_ld_b((v16i8*)t1, 16);
vec0 = (v8u16)__msa_ilvr_b((v16i8)src2, (v16i8)src0);
vec1 = (v8u16)__msa_ilvl_b((v16i8)src2, (v16i8)src0);
vec2 = (v8u16)__msa_ilvr_b((v16i8)src3, (v16i8)src1);
vec3 = (v8u16)__msa_ilvl_b((v16i8)src3, (v16i8)src1);
vec4 = (v8u16)__msa_ilvr_b((v16i8)zero, (v16i8)src4);
vec5 = (v8u16)__msa_ilvl_b((v16i8)zero, (v16i8)src4);
vec6 = (v8u16)__msa_ilvr_b((v16i8)zero, (v16i8)src5);
vec7 = (v8u16)__msa_ilvl_b((v16i8)zero, (v16i8)src5);
vec0 = __msa_hadd_u_h((v16u8)vec0, (v16u8)vec0);
vec1 = __msa_hadd_u_h((v16u8)vec1, (v16u8)vec1);
vec2 = __msa_hadd_u_h((v16u8)vec2, (v16u8)vec2);
vec3 = __msa_hadd_u_h((v16u8)vec3, (v16u8)vec3);
vec0 += __msa_hadd_u_h((v16u8)vec4, (v16u8)vec4);
vec1 += __msa_hadd_u_h((v16u8)vec5, (v16u8)vec5);
vec2 += __msa_hadd_u_h((v16u8)vec6, (v16u8)vec6);
vec3 += __msa_hadd_u_h((v16u8)vec7, (v16u8)vec7);
vec4 = (v8u16)__msa_vshf_h(mask, (v8i16)zero, (v8i16)vec0);
vec5 = (v8u16)__msa_vshf_h(mask, (v8i16)zero, (v8i16)vec1);
vec6 = (v8u16)__msa_vshf_h(mask, (v8i16)zero, (v8i16)vec2);
vec7 = (v8u16)__msa_vshf_h(mask, (v8i16)zero, (v8i16)vec3);
vec0 = (v8u16)__msa_pckod_w((v4i32)vec1, (v4i32)vec0);
vec1 = (v8u16)__msa_pckod_w((v4i32)vec3, (v4i32)vec2);
vec0 = (v8u16)__msa_pckod_w((v4i32)vec1, (v4i32)vec0);
tmp0 = __msa_hadd_u_w(vec4, vec4);
tmp1 = __msa_hadd_u_w(vec5, vec5);
tmp2 = __msa_hadd_u_w(vec6, vec6);
tmp3 = __msa_hadd_u_w(vec7, vec7);
tmp4 = __msa_hadd_u_w(vec0, vec0);
vec0 = (v8u16)__msa_pckev_h((v8i16)tmp1, (v8i16)tmp0);
vec1 = (v8u16)__msa_pckev_h((v8i16)tmp3, (v8i16)tmp2);
tmp0 = __msa_hadd_u_w(vec0, vec0);
tmp1 = __msa_hadd_u_w(vec1, vec1);
tmp0 *= const_0x1C71;
tmp1 *= const_0x1C71;
tmp4 *= const_0x2AAA;
tmp0 = (v4u32)__msa_srai_w((v4i32)tmp0, 16);
tmp1 = (v4u32)__msa_srai_w((v4i32)tmp1, 16);
tmp4 = (v4u32)__msa_srai_w((v4i32)tmp4, 16);
vec0 = (v8u16)__msa_pckev_h((v8i16)tmp1, (v8i16)tmp0);
vec1 = (v8u16)__msa_pckev_h((v8i16)tmp4, (v8i16)tmp4);
out = (v16u8)__msa_vshf_b(dst_mask, (v16i8)vec1, (v16i8)vec0);
dst0 = __msa_copy_u_d((v2i64)out, 0);
dst1 = __msa_copy_u_w((v4i32)out, 2);
SD(dst0, dst_ptr);
SW(dst1, dst_ptr + 8);
s += 32;
t0 += 32;
t1 += 32;
dst_ptr += 12;
}
}
void ScaleAddRow_MSA(const uint8_t* src_ptr, uint16_t* dst_ptr, int src_width) {
int x;
v16u8 src0;
v8u16 dst0, dst1;
v16i8 zero = {0};
assert(src_width > 0);
for (x = 0; x < src_width; x += 16) {
src0 = LD_UB(src_ptr);
dst0 = (v8u16)__msa_ld_h((v8i16*)dst_ptr, 0);
dst1 = (v8u16)__msa_ld_h((v8i16*)dst_ptr, 16);
dst0 += (v8u16)__msa_ilvr_b(zero, (v16i8)src0);
dst1 += (v8u16)__msa_ilvl_b(zero, (v16i8)src0);
ST_UH2(dst0, dst1, dst_ptr, 8);
src_ptr += 16;
dst_ptr += 16;
}
}
void ScaleFilterCols_MSA(uint8_t* dst_ptr,
const uint8_t* src_ptr,
int dst_width,
int x,
int dx) {
int j;
v4i32 vec_x = __msa_fill_w(x);
v4i32 vec_dx = __msa_fill_w(dx);
v4i32 vec_const = {0, 1, 2, 3};
v4i32 vec0, vec1, vec2, vec3, vec4, vec5, vec6, vec7, vec8, vec9;
v4i32 tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
v8u16 reg0, reg1;
v16u8 dst0;
v4i32 const_0xFFFF = __msa_fill_w(0xFFFF);
v4i32 const_0x40 = __msa_fill_w(0x40);
vec0 = vec_dx * vec_const;
vec1 = vec_dx * 4;
vec_x += vec0;
for (j = 0; j < dst_width - 1; j += 16) {
vec2 = vec_x >> 16;
vec6 = vec_x & const_0xFFFF;
vec_x += vec1;
vec3 = vec_x >> 16;
vec7 = vec_x & const_0xFFFF;
vec_x += vec1;
vec4 = vec_x >> 16;
vec8 = vec_x & const_0xFFFF;
vec_x += vec1;
vec5 = vec_x >> 16;
vec9 = vec_x & const_0xFFFF;
vec_x += vec1;
vec6 >>= 9;
vec7 >>= 9;
vec8 >>= 9;
vec9 >>= 9;
LOAD_INDEXED_DATA(src_ptr, vec2, tmp0);
LOAD_INDEXED_DATA(src_ptr, vec3, tmp1);
LOAD_INDEXED_DATA(src_ptr, vec4, tmp2);
LOAD_INDEXED_DATA(src_ptr, vec5, tmp3);
vec2 += 1;
vec3 += 1;
vec4 += 1;
vec5 += 1;
LOAD_INDEXED_DATA(src_ptr, vec2, tmp4);
LOAD_INDEXED_DATA(src_ptr, vec3, tmp5);
LOAD_INDEXED_DATA(src_ptr, vec4, tmp6);
LOAD_INDEXED_DATA(src_ptr, vec5, tmp7);
tmp4 -= tmp0;
tmp5 -= tmp1;
tmp6 -= tmp2;
tmp7 -= tmp3;
tmp4 *= vec6;
tmp5 *= vec7;
tmp6 *= vec8;
tmp7 *= vec9;
tmp4 += const_0x40;
tmp5 += const_0x40;
tmp6 += const_0x40;
tmp7 += const_0x40;
tmp4 >>= 7;
tmp5 >>= 7;
tmp6 >>= 7;
tmp7 >>= 7;
tmp0 += tmp4;
tmp1 += tmp5;
tmp2 += tmp6;
tmp3 += tmp7;
reg0 = (v8u16)__msa_pckev_h((v8i16)tmp1, (v8i16)tmp0);
reg1 = (v8u16)__msa_pckev_h((v8i16)tmp3, (v8i16)tmp2);
dst0 = (v16u8)__msa_pckev_b((v16i8)reg1, (v16i8)reg0);
__msa_st_b(dst0, dst_ptr, 0);
dst_ptr += 16;
}
}
void ScaleARGBCols_MSA(uint8_t* dst_argb,
const uint8_t* src_argb,
int dst_width,
int x,
int dx) {
const uint32_t* src = (const uint32_t*)(src_argb);
uint32_t* dst = (uint32_t*)(dst_argb);
int j;
v4i32 x_vec = __msa_fill_w(x);
v4i32 dx_vec = __msa_fill_w(dx);
v4i32 const_vec = {0, 1, 2, 3};
v4i32 vec0, vec1, vec2;
v4i32 dst0;
vec0 = dx_vec * const_vec;
vec1 = dx_vec * 4;
x_vec += vec0;
for (j = 0; j < dst_width; j += 4) {
vec2 = x_vec >> 16;
x_vec += vec1;
LOAD_INDEXED_DATA(src, vec2, dst0);
__msa_st_w(dst0, dst, 0);
dst += 4;
}
}
void ScaleARGBFilterCols_MSA(uint8_t* dst_argb,
const uint8_t* src_argb,
int dst_width,
int x,
int dx) {
const uint32_t* src = (const uint32_t*)(src_argb);
int j;
v4u32 src0, src1, src2, src3;
v4u32 vec0, vec1, vec2, vec3;
v16u8 reg0, reg1, reg2, reg3, reg4, reg5, reg6, reg7;
v16u8 mult0, mult1, mult2, mult3;
v8u16 tmp0, tmp1, tmp2, tmp3;
v16u8 dst0, dst1;
v4u32 vec_x = (v4u32)__msa_fill_w(x);
v4u32 vec_dx = (v4u32)__msa_fill_w(dx);
v4u32 vec_const = {0, 1, 2, 3};
v16u8 const_0x7f = (v16u8)__msa_fill_b(0x7f);
vec0 = vec_dx * vec_const;
vec1 = vec_dx * 4;
vec_x += vec0;
for (j = 0; j < dst_width - 1; j += 8) {
vec2 = vec_x >> 16;
reg0 = (v16u8)(vec_x >> 9);
vec_x += vec1;
vec3 = vec_x >> 16;
reg1 = (v16u8)(vec_x >> 9);
vec_x += vec1;
reg0 = reg0 & const_0x7f;
reg1 = reg1 & const_0x7f;
reg0 = (v16u8)__msa_shf_b((v16i8)reg0, 0);
reg1 = (v16u8)__msa_shf_b((v16i8)reg1, 0);
reg2 = reg0 ^ const_0x7f;
reg3 = reg1 ^ const_0x7f;
mult0 = (v16u8)__msa_ilvr_b((v16i8)reg0, (v16i8)reg2);
mult1 = (v16u8)__msa_ilvl_b((v16i8)reg0, (v16i8)reg2);
mult2 = (v16u8)__msa_ilvr_b((v16i8)reg1, (v16i8)reg3);
mult3 = (v16u8)__msa_ilvl_b((v16i8)reg1, (v16i8)reg3);
LOAD_INDEXED_DATA(src, vec2, src0);
LOAD_INDEXED_DATA(src, vec3, src1);
vec2 += 1;
vec3 += 1;
LOAD_INDEXED_DATA(src, vec2, src2);
LOAD_INDEXED_DATA(src, vec3, src3);
reg4 = (v16u8)__msa_ilvr_b((v16i8)src2, (v16i8)src0);
reg5 = (v16u8)__msa_ilvl_b((v16i8)src2, (v16i8)src0);
reg6 = (v16u8)__msa_ilvr_b((v16i8)src3, (v16i8)src1);
reg7 = (v16u8)__msa_ilvl_b((v16i8)src3, (v16i8)src1);
tmp0 = __msa_dotp_u_h(reg4, mult0);
tmp1 = __msa_dotp_u_h(reg5, mult1);
tmp2 = __msa_dotp_u_h(reg6, mult2);
tmp3 = __msa_dotp_u_h(reg7, mult3);
tmp0 >>= 7;
tmp1 >>= 7;
tmp2 >>= 7;
tmp3 >>= 7;
dst0 = (v16u8)__msa_pckev_b((v16i8)tmp1, (v16i8)tmp0);
dst1 = (v16u8)__msa_pckev_b((v16i8)tmp3, (v16i8)tmp2);
__msa_st_b(dst0, dst_argb, 0);
__msa_st_b(dst1, dst_argb, 16);
dst_argb += 32;
}
}
void ScaleRowDown34_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* dst,
int dst_width) {
int x;
(void)src_stride;
v16u8 src0, src1, src2, src3;
v16u8 vec0, vec1, vec2;
v16i8 mask0 = {0, 1, 3, 4, 5, 7, 8, 9, 11, 12, 13, 15, 16, 17, 19, 20};
v16i8 mask1 = {5, 7, 8, 9, 11, 12, 13, 15, 16, 17, 19, 20, 21, 23, 24, 25};
v16i8 mask2 = {11, 12, 13, 15, 16, 17, 19, 20,
21, 23, 24, 25, 27, 28, 29, 31};
assert((dst_width % 3 == 0) && (dst_width > 0));
for (x = 0; x < dst_width; x += 48) {
src0 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 16);
src2 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 32);
src3 = (v16u8)__msa_ld_b((v16i8*)src_ptr, 48);
vec0 = (v16u8)__msa_vshf_b(mask0, (v16i8)src1, (v16i8)src0);
vec1 = (v16u8)__msa_vshf_b(mask1, (v16i8)src2, (v16i8)src1);
vec2 = (v16u8)__msa_vshf_b(mask2, (v16i8)src3, (v16i8)src2);
__msa_st_b((v16i8)vec0, dst, 0);
__msa_st_b((v16i8)vec1, dst, 16);
__msa_st_b((v16i8)vec2, dst, 32);
src_ptr += 64;
dst += 48;
}
}
void ScaleRowDown34_0_Box_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* d,
int dst_width) {
const uint8_t* s = src_ptr;
const uint8_t* t = src_ptr + src_stride;
int x;
v16u8 src0, src1, src2, src3, src4, src5, src6, src7, dst0, dst1, dst2;
v16u8 vec0, vec1, vec2, vec3, vec4, vec5;
v16u8 vec6, vec7, vec8, vec9, vec10, vec11;
v8i16 reg0, reg1, reg2, reg3, reg4, reg5;
v8i16 reg6, reg7, reg8, reg9, reg10, reg11;
v16u8 const0 = {3, 1, 1, 1, 1, 3, 3, 1, 1, 1, 1, 3, 3, 1, 1, 1};
v16u8 const1 = {1, 3, 3, 1, 1, 1, 1, 3, 3, 1, 1, 1, 1, 3, 3, 1};
v16u8 const2 = {1, 1, 1, 3, 3, 1, 1, 1, 1, 3, 3, 1, 1, 1, 1, 3};
v16i8 mask0 = {0, 1, 1, 2, 2, 3, 4, 5, 5, 6, 6, 7, 8, 9, 9, 10};
v16i8 mask1 = {10, 11, 12, 13, 13, 14, 14, 15,
16, 17, 17, 18, 18, 19, 20, 21};
v16i8 mask2 = {5, 6, 6, 7, 8, 9, 9, 10, 10, 11, 12, 13, 13, 14, 14, 15};
v8i16 shft0 = {2, 1, 2, 2, 1, 2, 2, 1};
v8i16 shft1 = {2, 2, 1, 2, 2, 1, 2, 2};
v8i16 shft2 = {1, 2, 2, 1, 2, 2, 1, 2};
assert((dst_width % 3 == 0) && (dst_width > 0));
for (x = 0; x < dst_width; x += 48) {
src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)s, 16);
src2 = (v16u8)__msa_ld_b((v16i8*)s, 32);
src3 = (v16u8)__msa_ld_b((v16i8*)s, 48);
src4 = (v16u8)__msa_ld_b((v16i8*)t, 0);
src5 = (v16u8)__msa_ld_b((v16i8*)t, 16);
src6 = (v16u8)__msa_ld_b((v16i8*)t, 32);
src7 = (v16u8)__msa_ld_b((v16i8*)t, 48);
vec0 = (v16u8)__msa_vshf_b(mask0, (v16i8)src0, (v16i8)src0);
vec1 = (v16u8)__msa_vshf_b(mask1, (v16i8)src1, (v16i8)src0);
vec2 = (v16u8)__msa_vshf_b(mask2, (v16i8)src1, (v16i8)src1);
vec3 = (v16u8)__msa_vshf_b(mask0, (v16i8)src2, (v16i8)src2);
vec4 = (v16u8)__msa_vshf_b(mask1, (v16i8)src3, (v16i8)src2);
vec5 = (v16u8)__msa_vshf_b(mask2, (v16i8)src3, (v16i8)src3);
vec6 = (v16u8)__msa_vshf_b(mask0, (v16i8)src4, (v16i8)src4);
vec7 = (v16u8)__msa_vshf_b(mask1, (v16i8)src5, (v16i8)src4);
vec8 = (v16u8)__msa_vshf_b(mask2, (v16i8)src5, (v16i8)src5);
vec9 = (v16u8)__msa_vshf_b(mask0, (v16i8)src6, (v16i8)src6);
vec10 = (v16u8)__msa_vshf_b(mask1, (v16i8)src7, (v16i8)src6);
vec11 = (v16u8)__msa_vshf_b(mask2, (v16i8)src7, (v16i8)src7);
reg0 = (v8i16)__msa_dotp_u_h(vec0, const0);
reg1 = (v8i16)__msa_dotp_u_h(vec1, const1);
reg2 = (v8i16)__msa_dotp_u_h(vec2, const2);
reg3 = (v8i16)__msa_dotp_u_h(vec3, const0);
reg4 = (v8i16)__msa_dotp_u_h(vec4, const1);
reg5 = (v8i16)__msa_dotp_u_h(vec5, const2);
reg6 = (v8i16)__msa_dotp_u_h(vec6, const0);
reg7 = (v8i16)__msa_dotp_u_h(vec7, const1);
reg8 = (v8i16)__msa_dotp_u_h(vec8, const2);
reg9 = (v8i16)__msa_dotp_u_h(vec9, const0);
reg10 = (v8i16)__msa_dotp_u_h(vec10, const1);
reg11 = (v8i16)__msa_dotp_u_h(vec11, const2);
reg0 = __msa_srar_h(reg0, shft0);
reg1 = __msa_srar_h(reg1, shft1);
reg2 = __msa_srar_h(reg2, shft2);
reg3 = __msa_srar_h(reg3, shft0);
reg4 = __msa_srar_h(reg4, shft1);
reg5 = __msa_srar_h(reg5, shft2);
reg6 = __msa_srar_h(reg6, shft0);
reg7 = __msa_srar_h(reg7, shft1);
reg8 = __msa_srar_h(reg8, shft2);
reg9 = __msa_srar_h(reg9, shft0);
reg10 = __msa_srar_h(reg10, shft1);
reg11 = __msa_srar_h(reg11, shft2);
reg0 = reg0 * 3 + reg6;
reg1 = reg1 * 3 + reg7;
reg2 = reg2 * 3 + reg8;
reg3 = reg3 * 3 + reg9;
reg4 = reg4 * 3 + reg10;
reg5 = reg5 * 3 + reg11;
reg0 = __msa_srari_h(reg0, 2);
reg1 = __msa_srari_h(reg1, 2);
reg2 = __msa_srari_h(reg2, 2);
reg3 = __msa_srari_h(reg3, 2);
reg4 = __msa_srari_h(reg4, 2);
reg5 = __msa_srari_h(reg5, 2);
dst0 = (v16u8)__msa_pckev_b((v16i8)reg1, (v16i8)reg0);
dst1 = (v16u8)__msa_pckev_b((v16i8)reg3, (v16i8)reg2);
dst2 = (v16u8)__msa_pckev_b((v16i8)reg5, (v16i8)reg4);
__msa_st_b((v16i8)dst0, d, 0);
__msa_st_b((v16i8)dst1, d, 16);
__msa_st_b((v16i8)dst2, d, 32);
s += 64;
t += 64;
d += 48;
}
}
void ScaleRowDown34_1_Box_MSA(const uint8_t* src_ptr,
ptrdiff_t src_stride,
uint8_t* d,
int dst_width) {
const uint8_t* s = src_ptr;
const uint8_t* t = src_ptr + src_stride;
int x;
v16u8 src0, src1, src2, src3, src4, src5, src6, src7, dst0, dst1, dst2;
v16u8 vec0, vec1, vec2, vec3, vec4, vec5;
v16u8 vec6, vec7, vec8, vec9, vec10, vec11;
v8i16 reg0, reg1, reg2, reg3, reg4, reg5;
v8i16 reg6, reg7, reg8, reg9, reg10, reg11;
v16u8 const0 = {3, 1, 1, 1, 1, 3, 3, 1, 1, 1, 1, 3, 3, 1, 1, 1};
v16u8 const1 = {1, 3, 3, 1, 1, 1, 1, 3, 3, 1, 1, 1, 1, 3, 3, 1};
v16u8 const2 = {1, 1, 1, 3, 3, 1, 1, 1, 1, 3, 3, 1, 1, 1, 1, 3};
v16i8 mask0 = {0, 1, 1, 2, 2, 3, 4, 5, 5, 6, 6, 7, 8, 9, 9, 10};
v16i8 mask1 = {10, 11, 12, 13, 13, 14, 14, 15,
16, 17, 17, 18, 18, 19, 20, 21};
v16i8 mask2 = {5, 6, 6, 7, 8, 9, 9, 10, 10, 11, 12, 13, 13, 14, 14, 15};
v8i16 shft0 = {2, 1, 2, 2, 1, 2, 2, 1};
v8i16 shft1 = {2, 2, 1, 2, 2, 1, 2, 2};
v8i16 shft2 = {1, 2, 2, 1, 2, 2, 1, 2};
assert((dst_width % 3 == 0) && (dst_width > 0));
for (x = 0; x < dst_width; x += 48) {
src0 = (v16u8)__msa_ld_b((v16i8*)s, 0);
src1 = (v16u8)__msa_ld_b((v16i8*)s, 16);
src2 = (v16u8)__msa_ld_b((v16i8*)s, 32);
src3 = (v16u8)__msa_ld_b((v16i8*)s, 48);
src4 = (v16u8)__msa_ld_b((v16i8*)t, 0);
src5 = (v16u8)__msa_ld_b((v16i8*)t, 16);
src6 = (v16u8)__msa_ld_b((v16i8*)t, 32);
src7 = (v16u8)__msa_ld_b((v16i8*)t, 48);
vec0 = (v16u8)__msa_vshf_b(mask0, (v16i8)src0, (v16i8)src0);
vec1 = (v16u8)__msa_vshf_b(mask1, (v16i8)src1, (v16i8)src0);
vec2 = (v16u8)__msa_vshf_b(mask2, (v16i8)src1, (v16i8)src1);
vec3 = (v16u8)__msa_vshf_b(mask0, (v16i8)src2, (v16i8)src2);
vec4 = (v16u8)__msa_vshf_b(mask1, (v16i8)src3, (v16i8)src2);
vec5 = (v16u8)__msa_vshf_b(mask2, (v16i8)src3, (v16i8)src3);
vec6 = (v16u8)__msa_vshf_b(mask0, (v16i8)src4, (v16i8)src4);
vec7 = (v16u8)__msa_vshf_b(mask1, (v16i8)src5, (v16i8)src4);
vec8 = (v16u8)__msa_vshf_b(mask2, (v16i8)src5, (v16i8)src5);
vec9 = (v16u8)__msa_vshf_b(mask0, (v16i8)src6, (v16i8)src6);
vec10 = (v16u8)__msa_vshf_b(mask1, (v16i8)src7, (v16i8)src6);
vec11 = (v16u8)__msa_vshf_b(mask2, (v16i8)src7, (v16i8)src7);
reg0 = (v8i16)__msa_dotp_u_h(vec0, const0);
reg1 = (v8i16)__msa_dotp_u_h(vec1, const1);
reg2 = (v8i16)__msa_dotp_u_h(vec2, const2);
reg3 = (v8i16)__msa_dotp_u_h(vec3, const0);
reg4 = (v8i16)__msa_dotp_u_h(vec4, const1);
reg5 = (v8i16)__msa_dotp_u_h(vec5, const2);
reg6 = (v8i16)__msa_dotp_u_h(vec6, const0);
reg7 = (v8i16)__msa_dotp_u_h(vec7, const1);
reg8 = (v8i16)__msa_dotp_u_h(vec8, const2);
reg9 = (v8i16)__msa_dotp_u_h(vec9, const0);
reg10 = (v8i16)__msa_dotp_u_h(vec10, const1);
reg11 = (v8i16)__msa_dotp_u_h(vec11, const2);
reg0 = __msa_srar_h(reg0, shft0);
reg1 = __msa_srar_h(reg1, shft1);
reg2 = __msa_srar_h(reg2, shft2);
reg3 = __msa_srar_h(reg3, shft0);
reg4 = __msa_srar_h(reg4, shft1);
reg5 = __msa_srar_h(reg5, shft2);
reg6 = __msa_srar_h(reg6, shft0);
reg7 = __msa_srar_h(reg7, shft1);
reg8 = __msa_srar_h(reg8, shft2);
reg9 = __msa_srar_h(reg9, shft0);
reg10 = __msa_srar_h(reg10, shft1);
reg11 = __msa_srar_h(reg11, shft2);
reg0 += reg6;
reg1 += reg7;
reg2 += reg8;
reg3 += reg9;
reg4 += reg10;
reg5 += reg11;
reg0 = __msa_srari_h(reg0, 1);
reg1 = __msa_srari_h(reg1, 1);
reg2 = __msa_srari_h(reg2, 1);
reg3 = __msa_srari_h(reg3, 1);
reg4 = __msa_srari_h(reg4, 1);
reg5 = __msa_srari_h(reg5, 1);
dst0 = (v16u8)__msa_pckev_b((v16i8)reg1, (v16i8)reg0);
dst1 = (v16u8)__msa_pckev_b((v16i8)reg3, (v16i8)reg2);
dst2 = (v16u8)__msa_pckev_b((v16i8)reg5, (v16i8)reg4);
__msa_st_b((v16i8)dst0, d, 0);
__msa_st_b((v16i8)dst1, d, 16);
__msa_st_b((v16i8)dst2, d, 32);
s += 64;
t += 64;
d += 48;
}
}
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
#endif // !defined(LIBYUV_DISABLE_MSA) && defined(__mips_msa)

75
source/scale_uv.cc
View File

@ -155,23 +155,6 @@ static void ScaleUVDown2(int src_width,
}
#endif
#if defined(HAS_SCALEUVROWDOWN2_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ScaleUVRowDown2 =
filtering == kFilterNone
? ScaleUVRowDown2_Any_MSA
: (filtering == kFilterLinear ? ScaleUVRowDown2Linear_Any_MSA
: ScaleUVRowDown2Box_Any_MSA);
if (IS_ALIGNED(dst_width, 2)) {
ScaleUVRowDown2 =
filtering == kFilterNone
? ScaleUVRowDown2_MSA
: (filtering == kFilterLinear ? ScaleUVRowDown2Linear_MSA
: ScaleUVRowDown2Box_MSA);
}
}
#endif
if (filtering == kFilterLinear) {
src_stride = 0;
}
@ -321,16 +304,6 @@ static void ScaleUVDownEven(int src_width,
}
}
#endif
#if defined(HAS_SCALEUVROWDOWNEVEN_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ScaleUVRowDownEven =
filtering ? ScaleUVRowDownEvenBox_Any_MSA : ScaleUVRowDownEven_Any_MSA;
if (IS_ALIGNED(dst_width, 4)) {
ScaleUVRowDownEven =
filtering ? ScaleUVRowDownEvenBox_MSA : ScaleUVRowDownEven_MSA;
}
}
#endif
#if defined(HAS_SCALEUVROWDOWNEVEN_RVV) || defined(HAS_SCALEUVROWDOWN4_RVV)
if (TestCpuFlag(kCpuHasRVV) && !filtering) {
#if defined(HAS_SCALEUVROWDOWNEVEN_RVV)
@ -419,14 +392,6 @@ static int ScaleUVBilinearDown(int src_width,
InterpolateRow = InterpolateRow_SME;
}
#endif
#if defined(HAS_INTERPOLATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
InterpolateRow = InterpolateRow_Any_MSA;
if (IS_ALIGNED(clip_src_width, 32)) {
InterpolateRow = InterpolateRow_MSA;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
@ -452,14 +417,6 @@ static int ScaleUVBilinearDown(int src_width,
ScaleUVFilterCols = ScaleUVFilterCols_NEON;
}
}
#endif
#if defined(HAS_SCALEUVFILTERCOLS_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ScaleUVFilterCols = ScaleUVFilterCols_Any_MSA;
if (IS_ALIGNED(dst_width, 8)) {
ScaleUVFilterCols = ScaleUVFilterCols_MSA;
}
}
#endif
// TODO(fbarchard): Consider not allocating row buffer for kFilterLinear.
// Allocate a row of UV.
@ -545,14 +502,6 @@ static int ScaleUVBilinearUp(int src_width,
InterpolateRow = InterpolateRow_SME;
}
#endif
#if defined(HAS_INTERPOLATEROW_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
InterpolateRow = InterpolateRow_Any_MSA;
if (IS_ALIGNED(dst_width, 16)) {
InterpolateRow = InterpolateRow_MSA;
}
}
#endif
#if defined(HAS_INTERPOLATEROW_LSX)
if (TestCpuFlag(kCpuHasLSX)) {
InterpolateRow = InterpolateRow_Any_LSX;
@ -582,14 +531,6 @@ static int ScaleUVBilinearUp(int src_width,
}
}
#endif
#if defined(HAS_SCALEUVFILTERCOLS_MSA)
if (filtering && TestCpuFlag(kCpuHasMSA)) {
ScaleUVFilterCols = ScaleUVFilterCols_Any_MSA;
if (IS_ALIGNED(dst_width, 16)) {
ScaleUVFilterCols = ScaleUVFilterCols_MSA;
}
}
#endif
#if defined(HAS_SCALEUVCOLS_SSSE3)
if (!filtering && TestCpuFlag(kCpuHasSSSE3) && src_width < 32768) {
ScaleUVFilterCols = ScaleUVCols_SSSE3;
@ -602,14 +543,6 @@ static int ScaleUVBilinearUp(int src_width,
ScaleUVFilterCols = ScaleUVCols_NEON;
}
}
#endif
#if defined(HAS_SCALEUVCOLS_MSA)
if (!filtering && TestCpuFlag(kCpuHasMSA)) {
ScaleUVFilterCols = ScaleUVCols_Any_MSA;
if (IS_ALIGNED(dst_width, 8)) {
ScaleUVFilterCols = ScaleUVCols_MSA;
}
}
#endif
if (!filtering && src_width * 2 == dst_width && x < 0x8000) {
ScaleUVFilterCols = ScaleUVColsUp2_C;
@ -944,14 +877,6 @@ static void ScaleUVSimple(int src_width,
ScaleUVCols = ScaleUVCols_NEON;
}
}
#endif
#if defined(HAS_SCALEUVCOLS_MSA)
if (TestCpuFlag(kCpuHasMSA)) {
ScaleUVCols = ScaleUVCols_Any_MSA;
if (IS_ALIGNED(dst_width, 4)) {
ScaleUVCols = ScaleUVCols_MSA;
}
}
#endif
if (src_width * 2 == dst_width && x < 0x8000) {
ScaleUVCols = ScaleUVColsUp2_C;

34
unit_test/cpu_test.cc
View File

@ -151,15 +151,6 @@ TEST_F(LibYUVBaseTest, TestCpuHas) {
}
#endif // defined(__riscv)
#if defined(__mips__)
int has_mips = TestCpuFlag(kCpuHasMIPS);
if (has_mips) {
int has_msa = TestCpuFlag(kCpuHasMSA);
printf("Has MIPS 0x%x\n", has_mips);
printf("Has MSA 0x%x\n", has_msa);
}
#endif // defined(__mips__)
#if defined(__loongarch__)
int has_loongarch = TestCpuFlag(kCpuHasLOONGARCH);
if (has_loongarch) {
@ -291,18 +282,6 @@ TEST_F(LibYUVBaseTest, TestCompilerMacros) {
#ifdef __llvm__
printf("__llvm__ %d\n", __llvm__);
#endif
#ifdef __mips_msa
printf("__mips_msa %d\n", __mips_msa);
#endif
#ifdef __mips
printf("__mips %d\n", __mips);
#endif
#ifdef __mips_isa_rev
printf("__mips_isa_rev %d\n", __mips_isa_rev);
#endif
#ifdef _MIPS_ARCH_LOONGSON3A
printf("_MIPS_ARCH_LOONGSON3A %d\n", _MIPS_ARCH_LOONGSON3A);
#endif
#ifdef __loongarch__
printf("__loongarch__ %d\n", __loongarch__);
#endif
@ -388,19 +367,6 @@ TEST_F(LibYUVBaseTest, TestLinuxAArch64) {
}
#endif
TEST_F(LibYUVBaseTest, DISABLED_TestLinuxMipsMsa) {
if (FileExists("../../unit_test/testdata/mips.txt")) {
printf("Note: testing to load \"../../unit_test/testdata/mips.txt\"\n");
EXPECT_EQ(0, MipsCpuCaps("../../unit_test/testdata/mips.txt"));
EXPECT_EQ(kCpuHasMSA, MipsCpuCaps("../../unit_test/testdata/mips_msa.txt"));
EXPECT_EQ(kCpuHasMSA,
MipsCpuCaps("../../unit_test/testdata/mips_loongson2k.txt"));
} else {
printf("WARNING: unable to load \"../../unit_test/testdata/mips.txt\"\n");
}
}
TEST_F(LibYUVBaseTest, DISABLED_TestLinuxRVV) {
if (FileExists("../../unit_test/testdata/riscv64.txt")) {
printf("Note: testing to load \"../../unit_test/testdata/riscv64.txt\"\n");

7
unit_test/testdata/mips_msa.txt vendored
View File

@ -1,7 +0,0 @@
system type : generic-loongson-machine
machine : loongson,generic
processor : 0
isa : mips1 mips2 mips3 mips4 mips5 mips32r1 mips32r2 mips64r1 mips64r2
ASEs implemented : vz msa
shadow register sets : 1

7
unit_test/unit_test.cc
View File

@ -90,17 +90,10 @@ static int TestCpuEnv(int cpu_info) {
cpu_info &= ~libyuv::kCpuHasSME;
}
#endif
#if defined(__mips__) && defined(__linux__)
if (TestEnv("LIBYUV_DISABLE_MSA")) {
cpu_info &= ~libyuv::kCpuHasMSA;
}
#endif
#if defined(__longarch__) && defined(__linux__)
if (TestEnv("LIBYUV_DISABLE_LSX")) {
cpu_info &= ~libyuv::kCpuHasLSX;
}
#endif
#if defined(__longarch__) && defined(__linux__)
if (TestEnv("LIBYUV_DISABLE_LASX")) {
cpu_info &= ~libyuv::kCpuHasLASX;
}

9
util/cpuid.c
View File

@ -113,15 +113,6 @@ int main(int argc, const char* argv[]) {
}
#endif // defined(__riscv)
#if defined(__mips__)
int has_mips = TestCpuFlag(kCpuHasMIPS);
if (has_mips) {
int has_msa = TestCpuFlag(kCpuHasMSA);
printf("Has MIPS 0x%x\n", has_mips);
printf("Has MSA 0x%x\n", has_msa);
}
#endif // defined(__mips__)
#if defined(__loongarch__)
int has_loongarch = TestCpuFlag(kCpuHasLOONGARCH);
if (has_loongarch) {