coffee/libyuv
1
0
Fork 0
mirror of https://chromium.googlesource.com/libyuv/libyuv synced 2026-02-12 21:29:51 +08:00
Code Issues Packages Projects Releases Wiki Activity

Style fixes for mips version of SplitUV for nv12/21

Browse Source 
BUG=126
TEST=lint passes and rotate_test and cpu_test on try bot pass.
Review URL: https://webrtc-codereview.appspot.com/884004

git-svn-id: http://libyuv.googlecode.com/svn/trunk@418 16f28f9a-4ce2-e073-06de-1de4eb20be90
This commit is contained in:
fbarchard@google.com 2012-10-14 06:41:17 +00:00
parent c4163acb7b
commit bb6bddc9fb
7 changed files with 158 additions and 209 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
README.chromium
View File

@ -1,6 +1,6 @@
Name: libyuv Name: libyuv
URL: http://code.google.com/p/libyuv/ URL: http://code.google.com/p/libyuv/
Version: 417 Version: 418
License: BSD License: BSD
License File: LICENSE License File: LICENSE

3
include/libyuv/row.h
View File

@ -271,7 +271,8 @@ void ARGBMirrorRow_C(const uint8* src, uint8* dst, int width);
void SplitUV_SSE2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int pix); void SplitUV_SSE2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int pix);
void SplitUV_NEON(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int pix); void SplitUV_NEON(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int pix);
void SplitUV_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int pix); void SplitUV_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
int pix);
void SplitUV_C(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int pix); void SplitUV_C(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int pix);
void CopyRow_SSE2(const uint8* src, uint8* dst, int count); void CopyRow_SSE2(const uint8* src, uint8* dst, int count);

2
include/libyuv/version.h
View File

@ -11,6 +11,6 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_ // NOLINT #ifndef INCLUDE_LIBYUV_VERSION_H_ // NOLINT
#define INCLUDE_LIBYUV_VERSION_H_ #define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 417 #define LIBYUV_VERSION 418
#endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT #endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT

2
source/convert.cc
View File

@ -380,7 +380,7 @@ static int X420ToI420(const uint8* src_y,
SplitUV = SplitUV_SSE2; SplitUV = SplitUV_SSE2;
} }
#elif defined(HAS_SPLITUV_MIPS_DSPR2) #elif defined(HAS_SPLITUV_MIPS_DSPR2)
if (TestCpuFlag(kCpuHasMIPS) && TestCpuFlag(kCpuHasMIPS_DSPR2)){ if (TestCpuFlag(kCpuHasMIPS_DSPR2)) {
SplitUV = SplitUV_MIPS_DSPR2; SplitUV = SplitUV_MIPS_DSPR2;
} }
#endif #endif

42
source/cpu_id.cc
View File

@ -101,41 +101,38 @@ static const int kXCR_XFEATURE_ENABLED_MASK = 0;
// For Arm, but public to allow testing on any CPU // For Arm, but public to allow testing on any CPU
LIBYUV_API LIBYUV_API
int ArmCpuCaps(const char* cpuinfo_name) { int ArmCpuCaps(const char* cpuinfo_name) {
int flags = 0; FILE* f = fopen(cpuinfo_name, "r");
FILE* fin = fopen(cpuinfo_name, "r"); if (f) {
if (fin) {
char buf[512]; char buf[512];
while (fgets(buf, 511, fin)) { while (fgets(buf, 511, f)) {
if (memcmp(buf, "Features", 8) == 0) { if (memcmp(buf, "Features", 8) == 0) {
char* p = strstr(buf, " neon"); char* p = strstr(buf, " neon");
if (p && (p[5] == ' ' || p[5] == '\n')) { if (p && (p[5] == ' ' || p[5] == '\n')) {
flags |= kCpuHasNEON; fclose(f);
break; return kCpuHasNEON;
} }
} }
} }
fclose(fin); fclose(f);
} }
return flags; return 0;
} }
static int MipsCpuCaps(const char *search_string) { static int MipsCpuCaps(const char* search_string) {
int flags = 0; const char* file_name = "/proc/cpuinfo";
const char *file_name = "/proc/cpuinfo";
char cpuinfo_line[256]; char cpuinfo_line[256];
FILE *f = NULL; FILE* f = NULL;
if ((f = fopen(file_name, "r")) != NULL) {
if ((f = fopen (file_name, "r")) != NULL) { while (fgets(cpuinfo_line, sizeof(cpuinfo_line), f) != NULL) {
while (fgets (cpuinfo_line, sizeof (cpuinfo_line), f) != NULL) { if (strstr(cpuinfo_line, search_string) != NULL) {
if (strstr (cpuinfo_line, search_string) != NULL) { fclose(f);
flags |= kCpuHasMIPS_DSP; return kCpuHasMIPS_DSP;
fclose (f);
return flags;
} }
} }
fclose(f);
} }
/* Did not find string in the proc file, or not Linux ELF. */ /* Did not find string in the proc file, or not Linux ELF. */
return flags; return 0;
} }
// CPU detect function for SIMD instruction sets. // CPU detect function for SIMD instruction sets.
@ -197,9 +194,9 @@ int InitCpuFlags(void) {
if (TestEnv("LIBYUV_DISABLE_AVX2")) { if (TestEnv("LIBYUV_DISABLE_AVX2")) {
cpu_info_ &= ~kCpuHasAVX2; cpu_info_ &= ~kCpuHasAVX2;
} }
#elif defined (__mips__) && defined(__linux__) #elif defined(__mips__) && defined(__linux__)
// linux mips parse text file for dsp detect. // linux mips parse text file for dsp detect.
cpu_info_ = MipsCpuCaps("dsp"); // set kCpuHasMIPS_DSP cpu_info_ = MipsCpuCaps("dsp"); // set kCpuHasMIPS_DSP.
#if defined(__mips_dspr2) #if defined(__mips_dspr2)
cpu_info_ |= kCpuHasMIPS_DSPR2; cpu_info_ |= kCpuHasMIPS_DSPR2;
#endif #endif
@ -214,7 +211,6 @@ int InitCpuFlags(void) {
if (getenv("LIBYUV_DISABLE_MIPS_DSPR2")) { if (getenv("LIBYUV_DISABLE_MIPS_DSPR2")) {
cpu_info_ &= ~kCpuHasMIPS_DSPR2; cpu_info_ &= ~kCpuHasMIPS_DSPR2;
} }
#elif defined(__arm__) #elif defined(__arm__)
#if defined(__linux__) && (defined(__ARM_NEON__) || defined(LIBYUV_NEON)) #if defined(__linux__) && (defined(__ARM_NEON__) || defined(LIBYUV_NEON))
// linux arm parse text file for neon detect. // linux arm parse text file for neon detect.

313
source/row_mips.cc
View File

@ -1,5 +1,5 @@
/* /*
* Copyright (c) 2011 The LibYuv project authors. All Rights Reserved. * Copyright (c) 2012 The LibYuv project authors. All Rights Reserved.
* *
* Use of this source code is governed by a BSD-style license * 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 * that can be found in the LICENSE file in the root of the source
@ -9,6 +9,7 @@
*/ */
#include "libyuv/row.h" #include "libyuv/row.h"
#ifdef __cplusplus #ifdef __cplusplus
namespace libyuv { namespace libyuv {
extern "C" { extern "C" {
@ -16,191 +17,141 @@ extern "C" {
#if !defined(YUV_DISABLE_ASM) && defined(__mips__) #if !defined(YUV_DISABLE_ASM) && defined(__mips__)
#ifdef HAS_SPLITUV_MIPS_DSPR2 #ifdef HAS_SPLITUV_MIPS_DSPR2
void SplitUV_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int width) {
__asm__ __volatile__(
".set push \n\t"
".set noreorder \n\t"
"srl $t4, %[width], 4 \n\t" // how many multiplies of 16 8bits
"blez $t4, 2f \n\t"
" andi %[width], %[width], 0xf \n\t" // residual
"andi $t0, %[src_uv], 0x3 \n\t"
"andi $t1, %[dst_u], 0x3 \n\t"
"andi $t2, %[dst_v], 0x3 \n\t"
"or $t0, $t0, $t1 \n\t"
"or $t0, $t0, $t2 \n\t"
"beqz $t0, 12f \n\t" // if src and dsts are aligned
" nop \n\t"
// src and dst are unaligned
"1: \n\t"
"addiu $t4, $t4, -1 \n\t"
"lwr $t0, 0(%[src_uv]) \n\t"
"lwl $t0, 3(%[src_uv]) \n\t" // t0 = V1 | U1 | V0 | U0
"lwr $t1, 4(%[src_uv]) \n\t"
"lwl $t1, 7(%[src_uv]) \n\t" // t1 = V3 | U3 | V2 | U2
"lwr $t2, 8(%[src_uv]) \n\t"
"lwl $t2, 11(%[src_uv]) \n\t" // t2 = V5 | U5 | V4 | U4
"lwr $t3, 12(%[src_uv]) \n\t"
"lwl $t3, 15(%[src_uv]) \n\t" // t3 = V7 | U7 | V6 | U6
"lwr $t5, 16(%[src_uv]) \n\t"
"lwl $t5, 19(%[src_uv]) \n\t" // t5 = V9 | U9 | V8 | U8
"lwr $t6, 20(%[src_uv]) \n\t"
"lwl $t6, 23(%[src_uv]) \n\t" // t6 = V11 | U11 | V10 | U10
"lwr $t7, 24(%[src_uv]) \n\t"
"lwl $t7, 27(%[src_uv]) \n\t" // t7 = V13 | U13 | V12 | U12
"lwr $t8, 28(%[src_uv]) \n\t"
"lwl $t8, 31(%[src_uv]) \n\t" // t8 = V15 | U15 | V14 | U14
"precrq.qb.ph $t9, $t1, $t0 \n\t" // t9 = V3 | V2 | V1 | V0
"precr.qb.ph $t0, $t1, $t0 \n\t" // t0 = U3 | U2 | U1 | U0
"precrq.qb.ph $t1, $t3, $t2 \n\t" // t1 = V7 | V6 | V5 | V4
"precr.qb.ph $t2, $t3, $t2 \n\t" // t2 = U7 | U6 | U5 | U4
"precrq.qb.ph $t3, $t6, $t5 \n\t" // t3 = V11 | V10 | V9 | V8
"precr.qb.ph $t5, $t6, $t5 \n\t" // t5 = U11 | U10 | U9 | U8
"precrq.qb.ph $t6, $t8, $t7 \n\t" // t6 = V15 | V14 | V13 | V12
"precr.qb.ph $t7, $t8, $t7 \n\t" // t7 = U15 | U14 | U13 | U12
"addiu %[src_uv], %[src_uv], 32 \n\t"
"swr $t9, 0(%[dst_v]) \n\t"
"swl $t9, 3(%[dst_v]) \n\t"
"swr $t0, 0(%[dst_u]) \n\t"
"swl $t0, 3(%[dst_u]) \n\t"
"swr $t1, 4(%[dst_v]) \n\t"
"swl $t1, 7(%[dst_v]) \n\t"
"swr $t2, 4(%[dst_u]) \n\t"
"swl $t2, 7(%[dst_u]) \n\t"
"swr $t3, 8(%[dst_v]) \n\t"
"swl $t3, 11(%[dst_v]) \n\t"
"swr $t5, 8(%[dst_u]) \n\t"
"swl $t5, 11(%[dst_u]) \n\t"
"swr $t6, 12(%[dst_v]) \n\t"
"swl $t6, 15(%[dst_v]) \n\t"
"swr $t7, 12(%[dst_u]) \n\t"
"swl $t7, 15(%[dst_u]) \n\t"
"addiu %[dst_u], %[dst_u], 16 \n\t"
"bgtz $t4, 1b \n\t"
" addiu %[dst_v], %[dst_v], 16 \n\t"
"beqz %[width], 3f \n\t"
" nop \n\t"
"b 2f \n\t"
" nop \n\t"
// src and dst are aligned
"12: \n\t"
"addiu $t4, $t4, -1 \n\t"
"lw $t0, 0(%[src_uv]) \n\t" // t0 = V1 | U1 | V0 | U0
"lw $t1, 4(%[src_uv]) \n\t" // t1 = V3 | U3 | V2 | U2
"lw $t2, 8(%[src_uv]) \n\t" // t2 = V5 | U5 | V4 | U4
"lw $t3, 12(%[src_uv]) \n\t" // t3 = V7 | U7 | V6 | U6
"lw $t5, 16(%[src_uv]) \n\t" // t5 = V9 | U9 | V8 | U8
"lw $t6, 20(%[src_uv]) \n\t" // t6 = V11 | U11 | V10 | U10
"lw $t7, 24(%[src_uv]) \n\t" // t7 = V13 | U13 | V12 | U12
"lw $t8, 28(%[src_uv]) \n\t" // t8 = V15 | U15 | V14 | U14
"addiu %[src_uv], %[src_uv], 32 \n\t"
"precrq.qb.ph $t9, $t1, $t0 \n\t" // t9 = V3 | V2 | V1 | V0
"precr.qb.ph $t0, $t1, $t0 \n\t" // t0 = U3 | U2 | U1 | U0
"precrq.qb.ph $t1, $t3, $t2 \n\t" // t1 = V7 | V6 | V5 | V4
"precr.qb.ph $t2, $t3, $t2 \n\t" // t2 = U7 | U6 | U5 | U4
"precrq.qb.ph $t3, $t6, $t5 \n\t" // t3 = V11 | V10 | V9 | V8
"precr.qb.ph $t5, $t6, $t5 \n\t" // t5 = U11 | U10 | U9 | U8
"precrq.qb.ph $t6, $t8, $t7 \n\t" // t6 = V15 | V14 | V13 | V12
"precr.qb.ph $t7, $t8, $t7 \n\t" // t7 = U15 | U14 | U13 | U12
"sw $t9, 0(%[dst_v]) \n\t"
"sw $t0, 0(%[dst_u]) \n\t"
"sw $t1, 4(%[dst_v]) \n\t"
"sw $t2, 4(%[dst_u]) \n\t"
"sw $t3, 8(%[dst_v]) \n\t"
"sw $t5, 8(%[dst_u]) \n\t"
"sw $t6, 12(%[dst_v]) \n\t"
"sw $t7, 12(%[dst_u]) \n\t"
"addiu %[dst_v], %[dst_v], 16 \n\t"
"bgtz $t4, 12b \n\t"
" addiu %[dst_u], %[dst_u], 16 \n\t"
"beqz %[width], 3f \n\t"
" nop \n\t"
"2: \n\t"
"lbu $t0, 0(%[src_uv]) \n\t"
"lbu $t1, 1(%[src_uv]) \n\t"
"addiu %[src_uv], %[src_uv], 2 \n\t"
"addiu %[width], %[width], -1 \n\t"
"sb $t0, 0(%[dst_u]) \n\t"
"sb $t1, 0(%[dst_v]) \n\t"
"addiu %[dst_u], %[dst_u], 1 \n\t"
"bgtz %[width], 2b \n\t"
" addiu %[dst_v], %[dst_v], 1 \n\t"
"3: \n\t"
".set pop \n\t"
: [src_uv] "+r" (src_uv), [width] "+r" (width),
[dst_u] "+r" (dst_u), [dst_v] "+r" (dst_v)
:
: "t0", "t1","t2", "t3", "t4", "t5", "t6", "t7", "t8", "t9"
);
}
#endif // HAS_SPLITUV_MIPS_DSPR2
#ifdef HAS_SPLITUV_MIPS_DSPR2
// Reads 16 pairs of UV and write even values to dst_u and odd to dst_v
// Alignment requirement: 16 bytes for pointers, and multiple of 16 pixels.
void SplitUV_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, int width) {
asm volatile (
".p2align 2 \n"
"1: \n"
"vld2.u8 {q0, q1}, [%0]! \n" // load 16 pairs of UV
"subs %3, %3, #16 \n" // 16 processed per loop
"vst1.u8 {q0}, [%1]! \n" // store U
"vst1.u8 {q1}, [%2]! \n" // Store V
"bgt 1b \n"
: "+r"(src_uv), // %0
"+r"(dst_u), // %1
"+r"(dst_v), // %2
"+r"(width) // %3 // Output registers
: // Input registers
: "memory", "cc", "q0", "q1" // Clobber List
);
}
#endif // HAS_SPLITUV_MIPS_DSPR2
#ifdef HAS_SPLITUV_MIPS_DSPR2
// Reads 4 pairs of UV and write even values to dst_u and odd to dst_v
// Alignment requirement: 4 bytes for pointers, and multiple of 4 pixels.
void SplitUV_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v, void SplitUV_MIPS_DSPR2(const uint8* src_uv, uint8* dst_u, uint8* dst_v,
int width) { int width) {
asm volatile (
".set push \n" __asm__ __volatile__ (
".set noreorder \n" ".set push \n"
".p2align 2 \n" ".set noreorder \n"
"1: \n"
"lw $t0, 0(%[src_uv]) \n" // V1 | U1 | V0 | U0 "srl $t4, %[width], 4 \n" // multiplies of 16
"lw $t1, 4(%[src_uv]) \n" // V3 | U3 | V2 | U2 "blez $t4, 2f \n"
"addiu %[width], %[width], -4 \n" " andi %[width], %[width], 0xf \n" // residual
"addiu %[src_uv], %[src_uv], 8 \n" "andi $t0, %[src_uv], 0x3 \n"
"precr.qb.ph $t2, $t1, $t0 \n" // U3 | U2 | U1 | U0 "andi $t1, %[dst_u], 0x3 \n"
"precrq.qb.ph $t3, $t1, $t0 \n" // V3 | V2 | V1 | V0 "andi $t2, %[dst_v], 0x3 \n"
"sw $t2, 0(%[dst_u]) \n" "or $t0, $t0, $t1 \n"
"sw $t3, 0(%[dst_v]) \n" "or $t0, $t0, $t2 \n"
"addiu %[dst_u], %[dst_u], 4 \n"
"bgtz %[width], 1b \n" "beqz $t0, 12f \n" // test if aligned
" addiu %[dst_v], %[dst_v], 4 \n" " nop \n"
".set pop \n"
: [src_uv] "+r" (src_uv), // src and dst are unaligned
[width] "+r" (width), "1: \n"
[dst_u] "+r" (dst_u), "addiu $t4, $t4, -1 \n"
[dst_v] "+r" (dst_v) "lwr $t0, 0(%[src_uv]) \n"
: "lwl $t0, 3(%[src_uv]) \n" // V1 | U1 | V0 | U0
: "t0", "t1","t2", "t3", "lwr $t1, 4(%[src_uv]) \n"
"lwl $t1, 7(%[src_uv]) \n" // V3 | U3 | V2 | U2
"lwr $t2, 8(%[src_uv]) \n"
"lwl $t2, 11(%[src_uv]) \n" // V5 | U5 | V4 | U4
"lwr $t3, 12(%[src_uv]) \n"
"lwl $t3, 15(%[src_uv]) \n" // V7 | U7 | V6 | U6
"lwr $t5, 16(%[src_uv]) \n"
"lwl $t5, 19(%[src_uv]) \n" // V9 | U9 | V8 | U8
"lwr $t6, 20(%[src_uv]) \n"
"lwl $t6, 23(%[src_uv]) \n" // V11 | U11 | V10 | U10
"lwr $t7, 24(%[src_uv]) \n"
"lwl $t7, 27(%[src_uv]) \n" // V13 | U13 | V12 | U12
"lwr $t8, 28(%[src_uv]) \n"
"lwl $t8, 31(%[src_uv]) \n" // V15 | U15 | V14 | U14
"precrq.qb.ph $t9, $t1, $t0 \n" // V3 | V2 | V1 | V0
"precr.qb.ph $t0, $t1, $t0 \n" // U3 | U2 | U1 | U0
"precrq.qb.ph $t1, $t3, $t2 \n" // V7 | V6 | V5 | V4
"precr.qb.ph $t2, $t3, $t2 \n" // U7 | U6 | U5 | U4
"precrq.qb.ph $t3, $t6, $t5 \n" // V11 | V10 | V9 | V8
"precr.qb.ph $t5, $t6, $t5 \n" // U11 | U10 | U9 | U8
"precrq.qb.ph $t6, $t8, $t7 \n" // V15 | V14 | V13 | V12
"precr.qb.ph $t7, $t8, $t7 \n" // U15 | U14 | U13 | U12
"addiu %[src_uv], %[src_uv], 32 \n"
"swr $t9, 0(%[dst_v]) \n"
"swl $t9, 3(%[dst_v]) \n"
"swr $t0, 0(%[dst_u]) \n"
"swl $t0, 3(%[dst_u]) \n"
"swr $t1, 4(%[dst_v]) \n"
"swl $t1, 7(%[dst_v]) \n"
"swr $t2, 4(%[dst_u]) \n"
"swl $t2, 7(%[dst_u]) \n"
"swr $t3, 8(%[dst_v]) \n"
"swl $t3, 11(%[dst_v]) \n"
"swr $t5, 8(%[dst_u]) \n"
"swl $t5, 11(%[dst_u]) \n"
"swr $t6, 12(%[dst_v]) \n"
"swl $t6, 15(%[dst_v]) \n"
"swr $t7, 12(%[dst_u]) \n"
"swl $t7, 15(%[dst_u]) \n"
"addiu %[dst_u], %[dst_u], 16 \n"
"bgtz $t4, 1b \n"
" addiu %[dst_v], %[dst_v], 16 \n"
"beqz %[width], 3f \n"
" nop \n"
"b 2f \n"
" nop \n"
// src and dst are aligned
"12: \n"
"addiu $t4, $t4, -1 \n"
"lw $t0, 0(%[src_uv]) \n" // V1 | U1 | V0 | U0
"lw $t1, 4(%[src_uv]) \n" // V3 | U3 | V2 | U2
"lw $t2, 8(%[src_uv]) \n" // V5 | U5 | V4 | U4
"lw $t3, 12(%[src_uv]) \n" // V7 | U7 | V6 | U6
"lw $t5, 16(%[src_uv]) \n" // V9 | U9 | V8 | U8
"lw $t6, 20(%[src_uv]) \n" // V11 | U11 | V10 | U10
"lw $t7, 24(%[src_uv]) \n" // V13 | U13 | V12 | U12
"lw $t8, 28(%[src_uv]) \n" // V15 | U15 | V14 | U14
"addiu %[src_uv], %[src_uv], 32 \n"
"precrq.qb.ph $t9, $t1, $t0 \n" // V3 | V2 | V1 | V0
"precr.qb.ph $t0, $t1, $t0 \n" // U3 | U2 | U1 | U0
"precrq.qb.ph $t1, $t3, $t2 \n" // V7 | V6 | V5 | V4
"precr.qb.ph $t2, $t3, $t2 \n" // U7 | U6 | U5 | U4
"precrq.qb.ph $t3, $t6, $t5 \n" // V11 | V10 | V9 | V8
"precr.qb.ph $t5, $t6, $t5 \n" // U11 | U10 | U9 | U8
"precrq.qb.ph $t6, $t8, $t7 \n" // V15 | V14 | V13 | V12
"precr.qb.ph $t7, $t8, $t7 \n" // U15 | U14 | U13 | U12
"sw $t9, 0(%[dst_v]) \n"
"sw $t0, 0(%[dst_u]) \n"
"sw $t1, 4(%[dst_v]) \n"
"sw $t2, 4(%[dst_u]) \n"
"sw $t3, 8(%[dst_v]) \n"
"sw $t5, 8(%[dst_u]) \n"
"sw $t6, 12(%[dst_v]) \n"
"sw $t7, 12(%[dst_u]) \n"
"addiu %[dst_v], %[dst_v], 16 \n"
"bgtz $t4, 12b \n"
" addiu %[dst_u], %[dst_u], 16 \n"
"beqz %[width], 3f \n"
" nop \n"
"2: \n"
"lbu $t0, 0(%[src_uv]) \n"
"lbu $t1, 1(%[src_uv]) \n"
"addiu %[src_uv], %[src_uv], 2 \n"
"addiu %[width], %[width], -1 \n"
"sb $t0, 0(%[dst_u]) \n"
"sb $t1, 0(%[dst_v]) \n"
"addiu %[dst_u], %[dst_u], 1 \n"
"bgtz %[width], 2b \n"
" addiu %[dst_v], %[dst_v], 1 \n"
"3: \n"
".set pop \n"
: [src_uv] "+r" (src_uv),
[width] "+r" (width),
[dst_u] "+r" (dst_u),
[dst_v] "+r" (dst_v)
:
: "t0", "t1", "t2", "t3",
"t4", "t5", "t6", "t7", "t8", "t9"
); );
} }
#endif // HAS_SPLITUV_MIPS_DSPR2 #endif // HAS_SPLITUV_MIPS_DSPR2
#endif // __mips__ #endif // __mips__
#ifdef __cplusplus #ifdef __cplusplus

3
source/row_neon.cc
View File

@ -909,7 +909,8 @@ void I422ToUYVYRow_NEON(const uint8* src_y,
} }
#ifdef HAS_ARGBTOARGB4444ROW_NEON #ifdef HAS_ARGBTOARGB4444ROW_NEON
void ARGBToARGB4444Row_NEON(const uint8* src_argb, uint8* dst_argb4444, int pix) { void ARGBToARGB4444Row_NEON(const uint8* src_argb, uint8* dst_argb4444,
int pix) {
asm volatile ( asm volatile (
"vmov.u8 d4, #0x0f \n" // bits to clear with vbic. "vmov.u8 d4, #0x0f \n" // bits to clear with vbic.
".p2align 2 \n" ".p2align 2 \n"