Skip to content

L
libyuv
Commit dbe48143 authored by fbarchard@google.com's avatar fbarchard@google.com
Browse files
Options

Move scale row functions to scale_win etc 

BUG=none
TEST=untested
R=tpsiaki@google.com

Review URL: https://webrtc-codereview.appspot.com/4509005

git-svn-id: http://libyuv.googlecode.com/svn/trunk@879 16f28f9a-4ce2-e073-06de-1de4eb20be90
parent 9cd5bc0c
Expand all Hide whitespace changes
Inline Side-by-side
Showing with 133 additions and 149 deletions
Android.mk
View file @ dbe48143
......@@ -29,6 +29,7 @@ LOCAL_SRC_FILES := \
source/scale_argb.cc \
source/scale_common.cc \
source/scale_mips.cc \
source/scale_posix.cc \
source/video_common.cc
# TODO(fbarchard): Enable mjpeg encoder.
......
README.chromium
View file @ dbe48143
Name: libyuv
URL: http://code.google.com/p/libyuv/
Version: 877
Version: 879
License: BSD
License File: LICENSE
......
include/libyuv/scale_row.h
View file @ dbe48143
This diff is collapsed.
include/libyuv/version.h
View file @ dbe48143
......@@ -11,6 +11,6 @@
#ifndef INCLUDE_LIBYUV_VERSION_H_ // NOLINT
#define INCLUDE_LIBYUV_VERSION_H_
#define LIBYUV_VERSION 877
#define LIBYUV_VERSION 879
#endif // INCLUDE_LIBYUV_VERSION_H_ NOLINT
libyuv.gyp
View file @ dbe48143
......@@ -74,6 +74,7 @@
'include/libyuv/row.h',
'include/libyuv/scale.h',
'include/libyuv/scale_argb.h',
'include/libyuv/scale_row.h',
'include/libyuv/version.h',
'include/libyuv/video_common.h',
......@@ -107,10 +108,12 @@
'source/row_win.cc',
'source/scale.cc',
'source/scale_argb.cc',
'source/scale_argb_neon.cc',
'source/scale_argb_neon.cc', # Deprecated.
'source/scale_common.cc',
'source/scale_mips.cc',
'source/scale_neon.cc',
'source/scale_posix.cc',
'source/scale_win.cc',
'source/video_common.cc',
],
},
......
linux.mk
View file @ dbe48143
......@@ -28,6 +28,7 @@ LOCAL_OBJ_FILES := \
source/scale_argb.o \
source/scale_common.o \
source/scale_mips.o \
source/scale_posix.o \
source/video_common.o
.cc.o:
......
source/row_mips.cc
View file @ dbe48143
......@@ -15,6 +15,9 @@ namespace libyuv {
extern "C" {
#endif
// The following are available on Mips platforms:
#if !defined(LIBYUV_DISABLE_MIPS) && defined(__mips__)
#ifdef HAS_COPYROW_MIPS
void CopyRow_MIPS(const uint8* src, uint8* dst, int count) {
__asm__ __volatile__ (
......@@ -968,6 +971,8 @@ void InterpolateRows_MIPS_DSPR2(uint8* dst_ptr, const uint8* src_ptr,
}
#endif // __mips_dsp_rev >= 2
#endif // defined(__mips__)
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
......
source/row_posix.cc
View file @ dbe48143
......@@ -10,14 +10,12 @@
#include "libyuv/row.h"
#include "libyuv/basic_types.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// This module is for GCC x86 and x64
// This module is for GCC x86 and x64.
#if !defined(LIBYUV_DISABLE_X86) && (defined(__x86_64__) || defined(__i386__))
// TODO(nfullagar): For Native Client: When new toolchain becomes available,
......
source/scale.cc
View file @ dbe48143
This diff is collapsed.
source/scale_argb.cc
View file @ dbe48143
This diff is collapsed.
source/scale_argb_neon.cc
View file @ dbe48143
/*
* Copyright 2012 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.
*/
// This file is deprecated.
#include "libyuv/basic_types.h"
#include "libyuv/row.h"
#ifdef __cplusplus
namespace libyuv {
extern "C" {
#endif
// This module is for GCC Neon
#if !defined(LIBYUV_DISABLE_NEON) && defined(__ARM_NEON__)
void ScaleARGBRowDown2_NEON(const uint8* src_ptr, ptrdiff_t /* src_stride */,
uint8* dst, int dst_width) {
asm volatile (
"1: \n"
// load even pixels into q0, odd into q1
"vld2.32 {q0, q1}, [%0]! \n"
"vld2.32 {q2, q3}, [%0]! \n"
"subs %2, %2, #8 \n" // 8 processed per loop
"vst1.8 {q1}, [%1]! \n" // store odd pixels
"vst1.8 {q3}, [%1]! \n"
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(dst), // %1
"+r"(dst_width) // %2
:
: "memory", "cc", "q0", "q1", "q2", "q3" // Clobber List
);
}
void ScaleARGBRowDown2Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width) {
asm volatile (
// change the stride to row 2 pointer
"add %1, %1, %0 \n"
"1: \n"
"vld4.8 {d0, d2, d4, d6}, [%0]! \n" // load 8 ARGB pixels.
"vld4.8 {d1, d3, d5, d7}, [%0]! \n" // load next 8 ARGB pixels.
"subs %3, %3, #8 \n" // 8 processed per loop.
"vpaddl.u8 q0, q0 \n" // B 16 bytes -> 8 shorts.
"vpaddl.u8 q1, q1 \n" // G 16 bytes -> 8 shorts.
"vpaddl.u8 q2, q2 \n" // R 16 bytes -> 8 shorts.
"vpaddl.u8 q3, q3 \n" // A 16 bytes -> 8 shorts.
"vld4.8 {d16, d18, d20, d22}, [%1]! \n" // load 8 more ARGB pixels.
"vld4.8 {d17, d19, d21, d23}, [%1]! \n" // load last 8 ARGB pixels.
"vpadal.u8 q0, q8 \n" // B 16 bytes -> 8 shorts.
"vpadal.u8 q1, q9 \n" // G 16 bytes -> 8 shorts.
"vpadal.u8 q2, q10 \n" // R 16 bytes -> 8 shorts.
"vpadal.u8 q3, q11 \n" // A 16 bytes -> 8 shorts.
"vrshrn.u16 d0, q0, #2 \n" // downshift, round and pack
"vrshrn.u16 d1, q1, #2 \n"
"vrshrn.u16 d2, q2, #2 \n"
"vrshrn.u16 d3, q3, #2 \n"
"vst4.8 {d0, d1, d2, d3}, [%2]! \n"
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(src_stride), // %1
"+r"(dst), // %2
"+r"(dst_width) // %3
:
: "memory", "cc", "q0", "q1", "q2", "q3", "q8", "q9", "q10", "q11"
);
}
// Reads 4 pixels at a time.
// Alignment requirement: src_argb 4 byte aligned.
void ScaleARGBRowDownEven_NEON(const uint8* src_argb, ptrdiff_t, int src_stepx,
uint8* dst_argb, int dst_width) {
asm volatile (
"mov r12, %3, lsl #2 \n"
".p2align 2 \n"
"1: \n"
"vld1.32 {d0[0]}, [%0], r12 \n"
"vld1.32 {d0[1]}, [%0], r12 \n"
"vld1.32 {d1[0]}, [%0], r12 \n"
"vld1.32 {d1[1]}, [%0], r12 \n"
"subs %2, %2, #4 \n" // 4 pixels per loop.
"vst1.8 {q0}, [%1]! \n"
"bgt 1b \n"
: "+r"(src_argb), // %0
"+r"(dst_argb), // %1
"+r"(dst_width) // %2
: "r"(src_stepx) // %3
: "memory", "cc", "r12", "q0"
);
}
// Reads 4 pixels at a time.
// Alignment requirement: src_argb 4 byte aligned.
void ScaleARGBRowDownEvenBox_NEON(const uint8* src_argb, ptrdiff_t src_stride,
int src_stepx,
uint8* dst_argb, int dst_width) {
asm volatile (
"mov r12, %4, lsl #2 \n"
"add %1, %1, %0 \n"
".p2align 2 \n"
"1: \n"
"vld1.8 {d0}, [%0], r12 \n" // Read 4 2x2 blocks -> 2x1
"vld1.8 {d1}, [%1], r12 \n"
"vld1.8 {d2}, [%0], r12 \n"
"vld1.8 {d3}, [%1], r12 \n"
"vld1.8 {d4}, [%0], r12 \n"
"vld1.8 {d5}, [%1], r12 \n"
"vld1.8 {d6}, [%0], r12 \n"
"vld1.8 {d7}, [%1], r12 \n"
"vaddl.u8 q0, d0, d1 \n"
"vaddl.u8 q1, d2, d3 \n"
"vaddl.u8 q2, d4, d5 \n"
"vaddl.u8 q3, d6, d7 \n"
"vswp.8 d1, d2 \n" // ab_cd -> ac_bd
"vswp.8 d5, d6 \n" // ef_gh -> eg_fh
"vadd.u16 q0, q0, q1 \n" // (a+b)_(c+d)
"vadd.u16 q2, q2, q3 \n" // (e+f)_(g+h)
"vrshrn.u16 d0, q0, #2 \n" // first 2 pixels.
"vrshrn.u16 d1, q2, #2 \n" // next 2 pixels.
"subs %3, %3, #4 \n" // 4 pixels per loop.
"vst1.8 {q0}, [%2]! \n"
"bgt 1b \n"
: "+r"(src_argb), // %0
"+r"(src_stride), // %1
"+r"(dst_argb), // %2
"+r"(dst_width) // %3
: "r"(src_stepx) // %4
: "memory", "cc", "r12", "q0", "q1", "q2", "q3"
);
}
#endif // __ARM_NEON__
#ifdef __cplusplus
} // extern "C"
} // namespace libyuv
#endif
source/scale_common.cc
View file @ dbe48143
This diff is collapsed.
source/scale_neon.cc
View file @ dbe48143
......@@ -8,7 +8,6 @@
* be found in the AUTHORS file in the root of the source tree.
*/
#include "libyuv/basic_types.h"
#include "libyuv/row.h"
#ifdef __cplusplus
......@@ -16,7 +15,7 @@ namespace libyuv {
extern "C" {
#endif
// This module is for GCC Neon
// This module is for GCC Neon.
#if !defined(LIBYUV_DISABLE_NEON) && defined(__ARM_NEON__)
// NEON downscalers with interpolation.
......@@ -546,6 +545,123 @@ void ScaleFilterRows_NEON(uint8* dst_ptr,
: "q0", "q1", "d4", "d5", "q13", "q14", "memory", "cc"
);
}
void ScaleARGBRowDown2_NEON(const uint8* src_ptr, ptrdiff_t /* src_stride */,
uint8* dst, int dst_width) {
asm volatile (
"1: \n"
// load even pixels into q0, odd into q1
"vld2.32 {q0, q1}, [%0]! \n"
"vld2.32 {q2, q3}, [%0]! \n"
"subs %2, %2, #8 \n" // 8 processed per loop
"vst1.8 {q1}, [%1]! \n" // store odd pixels
"vst1.8 {q3}, [%1]! \n"
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(dst), // %1
"+r"(dst_width) // %2
:
: "memory", "cc", "q0", "q1", "q2", "q3" // Clobber List
);
}
void ScaleARGBRowDown2Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
uint8* dst, int dst_width) {
asm volatile (
// change the stride to row 2 pointer
"add %1, %1, %0 \n"
"1: \n"
"vld4.8 {d0, d2, d4, d6}, [%0]! \n" // load 8 ARGB pixels.
"vld4.8 {d1, d3, d5, d7}, [%0]! \n" // load next 8 ARGB pixels.
"subs %3, %3, #8 \n" // 8 processed per loop.
"vpaddl.u8 q0, q0 \n" // B 16 bytes -> 8 shorts.
"vpaddl.u8 q1, q1 \n" // G 16 bytes -> 8 shorts.
"vpaddl.u8 q2, q2 \n" // R 16 bytes -> 8 shorts.
"vpaddl.u8 q3, q3 \n" // A 16 bytes -> 8 shorts.
"vld4.8 {d16, d18, d20, d22}, [%1]! \n" // load 8 more ARGB pixels.
"vld4.8 {d17, d19, d21, d23}, [%1]! \n" // load last 8 ARGB pixels.
"vpadal.u8 q0, q8 \n" // B 16 bytes -> 8 shorts.
"vpadal.u8 q1, q9 \n" // G 16 bytes -> 8 shorts.
"vpadal.u8 q2, q10 \n" // R 16 bytes -> 8 shorts.
"vpadal.u8 q3, q11 \n" // A 16 bytes -> 8 shorts.
"vrshrn.u16 d0, q0, #2 \n" // downshift, round and pack
"vrshrn.u16 d1, q1, #2 \n"
"vrshrn.u16 d2, q2, #2 \n"
"vrshrn.u16 d3, q3, #2 \n"
"vst4.8 {d0, d1, d2, d3}, [%2]! \n"
"bgt 1b \n"
: "+r"(src_ptr), // %0
"+r"(src_stride), // %1
"+r"(dst), // %2
"+r"(dst_width) // %3
:
: "memory", "cc", "q0", "q1", "q2", "q3", "q8", "q9", "q10", "q11"
);
}
// Reads 4 pixels at a time.
// Alignment requirement: src_argb 4 byte aligned.
void ScaleARGBRowDownEven_NEON(const uint8* src_argb, ptrdiff_t, int src_stepx,
uint8* dst_argb, int dst_width) {
asm volatile (
"mov r12, %3, lsl #2 \n"
".p2align 2 \n"
"1: \n"
"vld1.32 {d0[0]}, [%0], r12 \n"
"vld1.32 {d0[1]}, [%0], r12 \n"
"vld1.32 {d1[0]}, [%0], r12 \n"
"vld1.32 {d1[1]}, [%0], r12 \n"
"subs %2, %2, #4 \n" // 4 pixels per loop.
"vst1.8 {q0}, [%1]! \n"
"bgt 1b \n"
: "+r"(src_argb), // %0
"+r"(dst_argb), // %1
"+r"(dst_width) // %2
: "r"(src_stepx) // %3
: "memory", "cc", "r12", "q0"
);
}
// Reads 4 pixels at a time.
// Alignment requirement: src_argb 4 byte aligned.
void ScaleARGBRowDownEvenBox_NEON(const uint8* src_argb, ptrdiff_t src_stride,
int src_stepx,
uint8* dst_argb, int dst_width) {
asm volatile (
"mov r12, %4, lsl #2 \n"
"add %1, %1, %0 \n"
".p2align 2 \n"
"1: \n"
"vld1.8 {d0}, [%0], r12 \n" // Read 4 2x2 blocks -> 2x1
"vld1.8 {d1}, [%1], r12 \n"
"vld1.8 {d2}, [%0], r12 \n"
"vld1.8 {d3}, [%1], r12 \n"
"vld1.8 {d4}, [%0], r12 \n"
"vld1.8 {d5}, [%1], r12 \n"
"vld1.8 {d6}, [%0], r12 \n"
"vld1.8 {d7}, [%1], r12 \n"
"vaddl.u8 q0, d0, d1 \n"
"vaddl.u8 q1, d2, d3 \n"
"vaddl.u8 q2, d4, d5 \n"
"vaddl.u8 q3, d6, d7 \n"
"vswp.8 d1, d2 \n" // ab_cd -> ac_bd
"vswp.8 d5, d6 \n" // ef_gh -> eg_fh
"vadd.u16 q0, q0, q1 \n" // (a+b)_(c+d)
"vadd.u16 q2, q2, q3 \n" // (e+f)_(g+h)
"vrshrn.u16 d0, q0, #2 \n" // first 2 pixels.
"vrshrn.u16 d1, q2, #2 \n" // next 2 pixels.
"subs %3, %3, #4 \n" // 4 pixels per loop.
"vst1.8 {q0}, [%2]! \n"
"bgt 1b \n"
: "+r"(src_argb), // %0
"+r"(src_stride), // %1
"+r"(dst_argb), // %2
"+r"(dst_width) // %3
: "r"(src_stepx) // %4
: "memory", "cc", "r12", "q0", "q1", "q2", "q3"
);
}
#endif // __ARM_NEON__
#ifdef __cplusplus
......
source/scale_posix.cc 0 → 100644
View file @ dbe48143
This diff is collapsed.
source/scale_win.cc 0 → 100644
View file @ dbe48143
This diff is collapsed.
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment