2967bf355e
This commit bring all up-to-date changes from master that are applicable to nextgenv2. Due to the remove VP10 code in master, we had to cherry pick the following commits to get those changes: Add default flags for arm64/armv8 builds Allows building simple targets with sane default flags. For example, using the Android arm64 toolchain from the NDK: https://developer.android.com/ndk/guides/standalone_toolchain.html ./build/tools/make-standalone-toolchain.sh --arch=arm64 \ --platform=android-24 --install-dir=/tmp/arm64 CROSS=/tmp/arm64/bin/aarch64-linux-android- \ ~/libvpx/configure --target=arm64-linux-gcc --disable-multithread BUG=webm:1143 vpx_lpf_horizontal_4_sse2: Remove dead load. Change-Id: I51026c52baa1f0881fcd5b68e1fdf08a2dc0916e Fail early when android target does not include --sdk-path Change-Id: I07e7e63476a2e32e3aae123abdee8b7bbbdc6a8c configure: clean up var style and set_all usage Use quotes whenever possible and {} always for variables. Replace multiple set_all calls with *able_feature(). Conflicts: build/make/configure.sh vp9-svc: Remove some unneeded code/comment. datarate_test,DatarateTestLarge: normalize bits type quiets a msvc warning: conversion from 'const int64_t' to 'size_t', possible loss of data mips added p6600 cpu support Removed -funroll-loops psnr.c: use int64_t for sum of differences Since the values can be negative. *.asm: normalize label format add a trailing ':', though it's optional with the tools we support, it's more common to use it to mark a label. this also quiets the orphan-labels warning with nasm/yasm. BUG=b/29583530 Prevent negative variance Due to rounding, hbd variance may become negative. This commit put in check and clamp of negative values to 0. configure: remove old visual studio support (<2010) BUG=b/29583530 Conflicts: configure configure: restore vs_version variable inadvertently lost in the final patchset of: 078dff7 configure: remove old visual studio support (<2010) this prevents an empty CONFIG_VS_VERSION and avoids make failure Require x86inc.asm Force enable x86inc.asm when building for x86. Previously there were compatibility issues so a flag was added to simplify disabling this code. The known issues have been resolved and x86inc.asm is the preferred abstraction layer (over x86_abi_support.asm). BUG=b:29583530 convolve_test: fix byte offsets in hbd build CONVERT_TO_BYTEPTR(x) was corrected in: 003a9d2 Port metric computation changes from nextgenv2 to use the more common (x) within the expansion. offsets should occur after converting the pointer to the desired type. + factorized some common expressions Conflicts: test/convolve_test.cc vpx_dsp: remove x86inc.asm distinction BUG=b:29583530 Conflicts: vpx_dsp/vpx_dsp.mk vpx_dsp/vpx_dsp_rtcd_defs.pl vpx_dsp/x86/highbd_variance_sse2.c vpx_dsp/x86/variance_sse2.c test: remove x86inc.asm distinction BUG=b:29583530 Conflicts: test/vp9_subtract_test.cc configure: remove x86inc.asm distinction BUG=b:29583530 Change-Id: I59a1192142e89a6a36b906f65a491a734e603617 Update vpx subpixel 1d filter ssse3 asm Speed test shows the new vertical filters have degradation on Celeron Chromebook. Added "X86_SUBPIX_VFILTER_PREFER_SLOW_CELERON" to control the vertical filters activated code. Now just simply active the code without degradation on Celeron. Later there should be 2 set of vertical filters ssse3 functions, and let jump table to choose based on CPU type. improve vpx_filter_block1d* based on replace paddsw+psrlw to pmulhrsw Make set_reference control API work in VP9 Moved the API patch from NextGenv2. An example was included. To try it, for example, run the following command: $ examples/vpx_cx_set_ref vp9 352 288 in.yuv out.ivf 4 30 Conflicts: examples.mk examples/vpx_cx_set_ref.c test/cx_set_ref.sh vp9/decoder/vp9_decoder.c deblock filter : moved from vp8 code branch The deblocking filters used in vp8 have been moved to vpx_dsp for use by both vp8 and vp9. vpx_thread.[hc]: update webp source reference + drop the blob hash, the updated reference will be updated in the commit message BUG=b/29583578 vpx_thread: use native windows cond var if available BUG=b/29583578 original webp change: commit 110ad5835ecd66995d0e7f66dca1b90dea595f5a Author: James Zern <jzern@google.com> Date: Mon Nov 23 19:49:58 2015 -0800 thread: use native windows cond var if available Vista / Server 2008 and up. no speed difference observed. 100644 blob 4fc372b7bc6980a9ed3618c8cce5b67ed7b0f412 src/utils/thread.c 100644 blob 840831185502d42a3246e4b7ff870121c8064791 src/utils/thread.h vpx_thread: use InitializeCriticalSectionEx if available BUG=b/29583578 original webp change: commit 63fadc9ffacc77d4617526a50c696d21d558a70b Author: James Zern <jzern@google.com> Date: Mon Nov 23 20:38:46 2015 -0800 thread: use InitializeCriticalSectionEx if available Windows Vista / Server 2008 and up 100644 blob f84207d89b3a6bb98bfe8f3fa55cad72dfd061ff src/utils/thread.c 100644 blob 840831185502d42a3246e4b7ff870121c8064791 src/utils/thread.h vpx_thread: use WaitForSingleObjectEx if available BUG=b/29583578 original webp change: commit 0fd0e12bfe83f16ce4f1c038b251ccbc13c62ac2 Author: James Zern <jzern@google.com> Date: Mon Nov 23 20:40:26 2015 -0800 thread: use WaitForSingleObjectEx if available Windows XP and up 100644 blob d58f74e5523dbc985fc531cf5f0833f1e9157cf0 src/utils/thread.c 100644 blob 840831185502d42a3246e4b7ff870121c8064791 src/utils/thread.h vpx_thread: use CreateThread for windows phone BUG=b/29583578 original webp change: commit d2afe974f9d751de144ef09d31255aea13b442c0 Author: James Zern <jzern@google.com> Date: Mon Nov 23 20:41:26 2015 -0800 thread: use CreateThread for windows phone _beginthreadex is unavailable for winrt/uwp Change-Id: Ie7412a568278ac67f0047f1764e2521193d74d4d 100644 blob 93f7622797f05f6acc1126e8296c481d276e4047 src/utils/thread.c 100644 blob 840831185502d42a3246e4b7ff870121c8064791 src/utils/thread.h vp9_postproc.c missing extern. BUG=webm:1256 deblock: missing const on extern const. postproc - move filling of noise buffer to vpx_dsp. Fix encoder crashes for odd size input clean-up vp9_intrapred_test remove tuple and overkill VP9IntraPredBase class. postproc: noise style fixes. gtest-all.cc: quiet an unused variable warning under windows / mingw builds vp9_intrapred_test: follow-up cleanup address few comments from ce050afaf3e288895c3bee4160336e2d2133b6ea Change-Id: I3eece7efa9335f4210303993ef6c1857ad5c29c8
732 lines
29 KiB
C
732 lines
29 KiB
C
/*
|
|
* Copyright (c) 2014 The WebM 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 <emmintrin.h> // SSE2
|
|
|
|
#include "./vpx_config.h"
|
|
#include "./vpx_dsp_rtcd.h"
|
|
|
|
#include "vpx_ports/mem.h"
|
|
|
|
typedef uint32_t (*high_variance_fn_t) (const uint16_t *src, int src_stride,
|
|
const uint16_t *ref, int ref_stride,
|
|
uint32_t *sse, int *sum);
|
|
|
|
uint32_t vpx_highbd_calc8x8var_sse2(const uint16_t *src, int src_stride,
|
|
const uint16_t *ref, int ref_stride,
|
|
uint32_t *sse, int *sum);
|
|
|
|
uint32_t vpx_highbd_calc16x16var_sse2(const uint16_t *src, int src_stride,
|
|
const uint16_t *ref, int ref_stride,
|
|
uint32_t *sse, int *sum);
|
|
|
|
static void highbd_8_variance_sse2(const uint16_t *src, int src_stride,
|
|
const uint16_t *ref, int ref_stride,
|
|
int w, int h, uint32_t *sse, int *sum,
|
|
high_variance_fn_t var_fn, int block_size) {
|
|
int i, j;
|
|
|
|
*sse = 0;
|
|
*sum = 0;
|
|
|
|
for (i = 0; i < h; i += block_size) {
|
|
for (j = 0; j < w; j += block_size) {
|
|
unsigned int sse0;
|
|
int sum0;
|
|
var_fn(src + src_stride * i + j, src_stride,
|
|
ref + ref_stride * i + j, ref_stride, &sse0, &sum0);
|
|
*sse += sse0;
|
|
*sum += sum0;
|
|
}
|
|
}
|
|
}
|
|
|
|
static void highbd_10_variance_sse2(const uint16_t *src, int src_stride,
|
|
const uint16_t *ref, int ref_stride,
|
|
int w, int h, uint32_t *sse, int *sum,
|
|
high_variance_fn_t var_fn, int block_size) {
|
|
int i, j;
|
|
uint64_t sse_long = 0;
|
|
int32_t sum_long = 0;
|
|
|
|
for (i = 0; i < h; i += block_size) {
|
|
for (j = 0; j < w; j += block_size) {
|
|
unsigned int sse0;
|
|
int sum0;
|
|
var_fn(src + src_stride * i + j, src_stride,
|
|
ref + ref_stride * i + j, ref_stride, &sse0, &sum0);
|
|
sse_long += sse0;
|
|
sum_long += sum0;
|
|
}
|
|
}
|
|
*sum = ROUND_POWER_OF_TWO(sum_long, 2);
|
|
*sse = (uint32_t)ROUND_POWER_OF_TWO(sse_long, 4);
|
|
}
|
|
|
|
static void highbd_12_variance_sse2(const uint16_t *src, int src_stride,
|
|
const uint16_t *ref, int ref_stride,
|
|
int w, int h, uint32_t *sse, int *sum,
|
|
high_variance_fn_t var_fn, int block_size) {
|
|
int i, j;
|
|
uint64_t sse_long = 0;
|
|
int32_t sum_long = 0;
|
|
|
|
for (i = 0; i < h; i += block_size) {
|
|
for (j = 0; j < w; j += block_size) {
|
|
unsigned int sse0;
|
|
int sum0;
|
|
var_fn(src + src_stride * i + j, src_stride,
|
|
ref + ref_stride * i + j, ref_stride, &sse0, &sum0);
|
|
sse_long += sse0;
|
|
sum_long += sum0;
|
|
}
|
|
}
|
|
*sum = ROUND_POWER_OF_TWO(sum_long, 4);
|
|
*sse = (uint32_t)ROUND_POWER_OF_TWO(sse_long, 8);
|
|
}
|
|
|
|
|
|
#define HIGH_GET_VAR(S) \
|
|
void vpx_highbd_get##S##x##S##var_sse2(const uint8_t *src8, int src_stride, \
|
|
const uint8_t *ref8, int ref_stride, \
|
|
uint32_t *sse, int *sum) { \
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
|
|
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8); \
|
|
vpx_highbd_calc##S##x##S##var_sse2(src, src_stride, ref, ref_stride, \
|
|
sse, sum); \
|
|
} \
|
|
\
|
|
void vpx_highbd_10_get##S##x##S##var_sse2(const uint8_t *src8, int src_stride, \
|
|
const uint8_t *ref8, int ref_stride, \
|
|
uint32_t *sse, int *sum) { \
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
|
|
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8); \
|
|
vpx_highbd_calc##S##x##S##var_sse2(src, src_stride, ref, ref_stride, \
|
|
sse, sum); \
|
|
*sum = ROUND_POWER_OF_TWO(*sum, 2); \
|
|
*sse = ROUND_POWER_OF_TWO(*sse, 4); \
|
|
} \
|
|
\
|
|
void vpx_highbd_12_get##S##x##S##var_sse2(const uint8_t *src8, int src_stride, \
|
|
const uint8_t *ref8, int ref_stride, \
|
|
uint32_t *sse, int *sum) { \
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
|
|
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8); \
|
|
vpx_highbd_calc##S##x##S##var_sse2(src, src_stride, ref, ref_stride, \
|
|
sse, sum); \
|
|
*sum = ROUND_POWER_OF_TWO(*sum, 4); \
|
|
*sse = ROUND_POWER_OF_TWO(*sse, 8); \
|
|
}
|
|
|
|
HIGH_GET_VAR(16);
|
|
HIGH_GET_VAR(8);
|
|
|
|
#undef HIGH_GET_VAR
|
|
|
|
#define VAR_FN(w, h, block_size, shift) \
|
|
uint32_t vpx_highbd_8_variance##w##x##h##_sse2( \
|
|
const uint8_t *src8, int src_stride, \
|
|
const uint8_t *ref8, int ref_stride, uint32_t *sse) { \
|
|
int sum; \
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
|
|
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8); \
|
|
highbd_8_variance_sse2(src, src_stride, ref, ref_stride, w, h, sse, &sum, \
|
|
vpx_highbd_calc##block_size##x##block_size##var_sse2, \
|
|
block_size); \
|
|
return *sse - (((int64_t)sum * sum) >> shift); \
|
|
} \
|
|
\
|
|
uint32_t vpx_highbd_10_variance##w##x##h##_sse2( \
|
|
const uint8_t *src8, int src_stride, \
|
|
const uint8_t *ref8, int ref_stride, uint32_t *sse) { \
|
|
int sum; \
|
|
int64_t var; \
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
|
|
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8); \
|
|
highbd_10_variance_sse2( \
|
|
src, src_stride, ref, ref_stride, w, h, sse, &sum, \
|
|
vpx_highbd_calc##block_size##x##block_size##var_sse2, block_size); \
|
|
var = (int64_t)(*sse) - (((int64_t)sum * sum) >> shift); \
|
|
return (var >= 0) ? (uint32_t)var : 0; \
|
|
} \
|
|
\
|
|
uint32_t vpx_highbd_12_variance##w##x##h##_sse2( \
|
|
const uint8_t *src8, int src_stride, \
|
|
const uint8_t *ref8, int ref_stride, uint32_t *sse) { \
|
|
int sum; \
|
|
int64_t var; \
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
|
|
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8); \
|
|
highbd_12_variance_sse2( \
|
|
src, src_stride, ref, ref_stride, w, h, sse, &sum, \
|
|
vpx_highbd_calc##block_size##x##block_size##var_sse2, block_size); \
|
|
var = (int64_t)(*sse) - (((int64_t)sum * sum) >> shift); \
|
|
return (var >= 0) ? (uint32_t)var : 0; \
|
|
}
|
|
|
|
VAR_FN(64, 64, 16, 12);
|
|
VAR_FN(64, 32, 16, 11);
|
|
VAR_FN(32, 64, 16, 11);
|
|
VAR_FN(32, 32, 16, 10);
|
|
VAR_FN(32, 16, 16, 9);
|
|
VAR_FN(16, 32, 16, 9);
|
|
VAR_FN(16, 16, 16, 8);
|
|
VAR_FN(16, 8, 8, 7);
|
|
VAR_FN(8, 16, 8, 7);
|
|
VAR_FN(8, 8, 8, 6);
|
|
|
|
#undef VAR_FN
|
|
|
|
unsigned int vpx_highbd_8_mse16x16_sse2(const uint8_t *src8, int src_stride,
|
|
const uint8_t *ref8, int ref_stride,
|
|
unsigned int *sse) {
|
|
int sum;
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8);
|
|
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
|
|
highbd_8_variance_sse2(src, src_stride, ref, ref_stride, 16, 16,
|
|
sse, &sum, vpx_highbd_calc16x16var_sse2, 16);
|
|
return *sse;
|
|
}
|
|
|
|
unsigned int vpx_highbd_10_mse16x16_sse2(const uint8_t *src8, int src_stride,
|
|
const uint8_t *ref8, int ref_stride,
|
|
unsigned int *sse) {
|
|
int sum;
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8);
|
|
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
|
|
highbd_10_variance_sse2(src, src_stride, ref, ref_stride, 16, 16,
|
|
sse, &sum, vpx_highbd_calc16x16var_sse2, 16);
|
|
return *sse;
|
|
}
|
|
|
|
unsigned int vpx_highbd_12_mse16x16_sse2(const uint8_t *src8, int src_stride,
|
|
const uint8_t *ref8, int ref_stride,
|
|
unsigned int *sse) {
|
|
int sum;
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8);
|
|
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
|
|
highbd_12_variance_sse2(src, src_stride, ref, ref_stride, 16, 16,
|
|
sse, &sum, vpx_highbd_calc16x16var_sse2, 16);
|
|
return *sse;
|
|
}
|
|
|
|
unsigned int vpx_highbd_8_mse8x8_sse2(const uint8_t *src8, int src_stride,
|
|
const uint8_t *ref8, int ref_stride,
|
|
unsigned int *sse) {
|
|
int sum;
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8);
|
|
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
|
|
highbd_8_variance_sse2(src, src_stride, ref, ref_stride, 8, 8,
|
|
sse, &sum, vpx_highbd_calc8x8var_sse2, 8);
|
|
return *sse;
|
|
}
|
|
|
|
unsigned int vpx_highbd_10_mse8x8_sse2(const uint8_t *src8, int src_stride,
|
|
const uint8_t *ref8, int ref_stride,
|
|
unsigned int *sse) {
|
|
int sum;
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8);
|
|
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
|
|
highbd_10_variance_sse2(src, src_stride, ref, ref_stride, 8, 8,
|
|
sse, &sum, vpx_highbd_calc8x8var_sse2, 8);
|
|
return *sse;
|
|
}
|
|
|
|
unsigned int vpx_highbd_12_mse8x8_sse2(const uint8_t *src8, int src_stride,
|
|
const uint8_t *ref8, int ref_stride,
|
|
unsigned int *sse) {
|
|
int sum;
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8);
|
|
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
|
|
highbd_12_variance_sse2(src, src_stride, ref, ref_stride, 8, 8,
|
|
sse, &sum, vpx_highbd_calc8x8var_sse2, 8);
|
|
return *sse;
|
|
}
|
|
|
|
// The 2 unused parameters are place holders for PIC enabled build.
|
|
// These definitions are for functions defined in
|
|
// highbd_subpel_variance_impl_sse2.asm
|
|
#define DECL(w, opt) \
|
|
int vpx_highbd_sub_pixel_variance##w##xh_##opt(const uint16_t *src, \
|
|
ptrdiff_t src_stride, \
|
|
int x_offset, int y_offset, \
|
|
const uint16_t *dst, \
|
|
ptrdiff_t dst_stride, \
|
|
int height, \
|
|
unsigned int *sse, \
|
|
void *unused0, void *unused);
|
|
#define DECLS(opt) \
|
|
DECL(8, opt); \
|
|
DECL(16, opt)
|
|
|
|
DECLS(sse2);
|
|
|
|
#undef DECLS
|
|
#undef DECL
|
|
|
|
#define FN(w, h, wf, wlog2, hlog2, opt, cast) \
|
|
uint32_t vpx_highbd_8_sub_pixel_variance##w##x##h##_##opt(const uint8_t *src8, \
|
|
int src_stride, \
|
|
int x_offset, \
|
|
int y_offset, \
|
|
const uint8_t *dst8, \
|
|
int dst_stride, \
|
|
uint32_t *sse_ptr) { \
|
|
uint32_t sse; \
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
|
|
uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); \
|
|
int se = vpx_highbd_sub_pixel_variance##wf##xh_##opt(src, src_stride, \
|
|
x_offset, y_offset, \
|
|
dst, dst_stride, h, \
|
|
&sse, NULL, NULL); \
|
|
if (w > wf) { \
|
|
unsigned int sse2; \
|
|
int se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt(src + 16, \
|
|
src_stride, \
|
|
x_offset, y_offset, \
|
|
dst + 16, \
|
|
dst_stride, \
|
|
h, &sse2, \
|
|
NULL, NULL); \
|
|
se += se2; \
|
|
sse += sse2; \
|
|
if (w > wf * 2) { \
|
|
se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt(src + 32, src_stride, \
|
|
x_offset, y_offset, \
|
|
dst + 32, dst_stride, \
|
|
h, &sse2, NULL, NULL); \
|
|
se += se2; \
|
|
sse += sse2; \
|
|
se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt( \
|
|
src + 48, src_stride, x_offset, y_offset, \
|
|
dst + 48, dst_stride, h, &sse2, NULL, NULL); \
|
|
se += se2; \
|
|
sse += sse2; \
|
|
} \
|
|
} \
|
|
*sse_ptr = sse; \
|
|
return sse - ((cast se * se) >> (wlog2 + hlog2)); \
|
|
} \
|
|
\
|
|
uint32_t vpx_highbd_10_sub_pixel_variance##w##x##h##_##opt( \
|
|
const uint8_t *src8, int src_stride, int x_offset, int y_offset, \
|
|
const uint8_t *dst8, int dst_stride, uint32_t *sse_ptr) { \
|
|
uint32_t sse; \
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
|
|
uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); \
|
|
int se = vpx_highbd_sub_pixel_variance##wf##xh_##opt(src, src_stride, \
|
|
x_offset, y_offset, \
|
|
dst, dst_stride, \
|
|
h, &sse, NULL, NULL); \
|
|
if (w > wf) { \
|
|
uint32_t sse2; \
|
|
int se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt(src + 16, \
|
|
src_stride, \
|
|
x_offset, y_offset, \
|
|
dst + 16, \
|
|
dst_stride, \
|
|
h, &sse2, \
|
|
NULL, NULL); \
|
|
se += se2; \
|
|
sse += sse2; \
|
|
if (w > wf * 2) { \
|
|
se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt(src + 32, src_stride, \
|
|
x_offset, y_offset, \
|
|
dst + 32, dst_stride, \
|
|
h, &sse2, NULL, NULL); \
|
|
se += se2; \
|
|
sse += sse2; \
|
|
se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt(src + 48, src_stride, \
|
|
x_offset, y_offset, \
|
|
dst + 48, dst_stride, \
|
|
h, &sse2, NULL, NULL); \
|
|
se += se2; \
|
|
sse += sse2; \
|
|
} \
|
|
} \
|
|
se = ROUND_POWER_OF_TWO(se, 2); \
|
|
sse = ROUND_POWER_OF_TWO(sse, 4); \
|
|
*sse_ptr = sse; \
|
|
return sse - ((cast se * se) >> (wlog2 + hlog2)); \
|
|
} \
|
|
\
|
|
uint32_t vpx_highbd_12_sub_pixel_variance##w##x##h##_##opt( \
|
|
const uint8_t *src8, int src_stride, int x_offset, int y_offset, \
|
|
const uint8_t *dst8, int dst_stride, uint32_t *sse_ptr) { \
|
|
int start_row; \
|
|
uint32_t sse; \
|
|
int se = 0; \
|
|
uint64_t long_sse = 0; \
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
|
|
uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); \
|
|
for (start_row = 0; start_row < h; start_row +=16) { \
|
|
uint32_t sse2; \
|
|
int height = h - start_row < 16 ? h - start_row : 16; \
|
|
int se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt( \
|
|
src + (start_row * src_stride), src_stride, \
|
|
x_offset, y_offset, dst + (start_row * dst_stride), \
|
|
dst_stride, height, &sse2, NULL, NULL); \
|
|
se += se2; \
|
|
long_sse += sse2; \
|
|
if (w > wf) { \
|
|
se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt( \
|
|
src + 16 + (start_row * src_stride), src_stride, \
|
|
x_offset, y_offset, dst + 16 + (start_row * dst_stride), \
|
|
dst_stride, height, &sse2, NULL, NULL); \
|
|
se += se2; \
|
|
long_sse += sse2; \
|
|
if (w > wf * 2) { \
|
|
se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt( \
|
|
src + 32 + (start_row * src_stride), src_stride, \
|
|
x_offset, y_offset, dst + 32 + (start_row * dst_stride), \
|
|
dst_stride, height, &sse2, NULL, NULL); \
|
|
se += se2; \
|
|
long_sse += sse2; \
|
|
se2 = vpx_highbd_sub_pixel_variance##wf##xh_##opt( \
|
|
src + 48 + (start_row * src_stride), src_stride, \
|
|
x_offset, y_offset, dst + 48 + (start_row * dst_stride), \
|
|
dst_stride, height, &sse2, NULL, NULL); \
|
|
se += se2; \
|
|
long_sse += sse2; \
|
|
}\
|
|
} \
|
|
} \
|
|
se = ROUND_POWER_OF_TWO(se, 4); \
|
|
sse = (uint32_t)ROUND_POWER_OF_TWO(long_sse, 8); \
|
|
*sse_ptr = sse; \
|
|
return sse - ((cast se * se) >> (wlog2 + hlog2)); \
|
|
}
|
|
|
|
#define FNS(opt) \
|
|
FN(64, 64, 16, 6, 6, opt, (int64_t)); \
|
|
FN(64, 32, 16, 6, 5, opt, (int64_t)); \
|
|
FN(32, 64, 16, 5, 6, opt, (int64_t)); \
|
|
FN(32, 32, 16, 5, 5, opt, (int64_t)); \
|
|
FN(32, 16, 16, 5, 4, opt, (int64_t)); \
|
|
FN(16, 32, 16, 4, 5, opt, (int64_t)); \
|
|
FN(16, 16, 16, 4, 4, opt, (int64_t)); \
|
|
FN(16, 8, 16, 4, 3, opt, (int64_t)); \
|
|
FN(8, 16, 8, 3, 4, opt, (int64_t)); \
|
|
FN(8, 8, 8, 3, 3, opt, (int64_t)); \
|
|
FN(8, 4, 8, 3, 2, opt, (int64_t));
|
|
|
|
|
|
FNS(sse2);
|
|
|
|
#undef FNS
|
|
#undef FN
|
|
|
|
// The 2 unused parameters are place holders for PIC enabled build.
|
|
#define DECL(w, opt) \
|
|
int vpx_highbd_sub_pixel_avg_variance##w##xh_##opt(const uint16_t *src, \
|
|
ptrdiff_t src_stride, \
|
|
int x_offset, int y_offset, \
|
|
const uint16_t *dst, \
|
|
ptrdiff_t dst_stride, \
|
|
const uint16_t *sec, \
|
|
ptrdiff_t sec_stride, \
|
|
int height, \
|
|
unsigned int *sse, \
|
|
void *unused0, void *unused);
|
|
#define DECLS(opt1) \
|
|
DECL(16, opt1) \
|
|
DECL(8, opt1)
|
|
|
|
DECLS(sse2);
|
|
#undef DECL
|
|
#undef DECLS
|
|
|
|
#define FN(w, h, wf, wlog2, hlog2, opt, cast) \
|
|
uint32_t vpx_highbd_8_sub_pixel_avg_variance##w##x##h##_##opt( \
|
|
const uint8_t *src8, int src_stride, int x_offset, int y_offset, \
|
|
const uint8_t *dst8, int dst_stride, uint32_t *sse_ptr, \
|
|
const uint8_t *sec8) { \
|
|
uint32_t sse; \
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
|
|
uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); \
|
|
uint16_t *sec = CONVERT_TO_SHORTPTR(sec8); \
|
|
int se = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
|
|
src, src_stride, x_offset, \
|
|
y_offset, dst, dst_stride, sec, w, h, &sse, NULL, NULL); \
|
|
if (w > wf) { \
|
|
uint32_t sse2; \
|
|
int se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
|
|
src + 16, src_stride, x_offset, y_offset, \
|
|
dst + 16, dst_stride, sec + 16, w, h, &sse2, NULL, NULL); \
|
|
se += se2; \
|
|
sse += sse2; \
|
|
if (w > wf * 2) { \
|
|
se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
|
|
src + 32, src_stride, x_offset, y_offset, \
|
|
dst + 32, dst_stride, sec + 32, w, h, &sse2, NULL, NULL); \
|
|
se += se2; \
|
|
sse += sse2; \
|
|
se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
|
|
src + 48, src_stride, x_offset, y_offset, \
|
|
dst + 48, dst_stride, sec + 48, w, h, &sse2, NULL, NULL); \
|
|
se += se2; \
|
|
sse += sse2; \
|
|
} \
|
|
} \
|
|
*sse_ptr = sse; \
|
|
return sse - ((cast se * se) >> (wlog2 + hlog2)); \
|
|
} \
|
|
\
|
|
uint32_t vpx_highbd_10_sub_pixel_avg_variance##w##x##h##_##opt( \
|
|
const uint8_t *src8, int src_stride, int x_offset, int y_offset, \
|
|
const uint8_t *dst8, int dst_stride, uint32_t *sse_ptr, \
|
|
const uint8_t *sec8) { \
|
|
uint32_t sse; \
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
|
|
uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); \
|
|
uint16_t *sec = CONVERT_TO_SHORTPTR(sec8); \
|
|
int se = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
|
|
src, src_stride, x_offset, \
|
|
y_offset, dst, dst_stride, \
|
|
sec, w, h, &sse, NULL, NULL); \
|
|
if (w > wf) { \
|
|
uint32_t sse2; \
|
|
int se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
|
|
src + 16, src_stride, \
|
|
x_offset, y_offset, \
|
|
dst + 16, dst_stride, \
|
|
sec + 16, w, h, &sse2, \
|
|
NULL, NULL); \
|
|
se += se2; \
|
|
sse += sse2; \
|
|
if (w > wf * 2) { \
|
|
se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
|
|
src + 32, src_stride, \
|
|
x_offset, y_offset, \
|
|
dst + 32, dst_stride, \
|
|
sec + 32, w, h, &sse2, \
|
|
NULL, NULL); \
|
|
se += se2; \
|
|
sse += sse2; \
|
|
se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
|
|
src + 48, src_stride, \
|
|
x_offset, y_offset, \
|
|
dst + 48, dst_stride, \
|
|
sec + 48, w, h, &sse2, \
|
|
NULL, NULL); \
|
|
se += se2; \
|
|
sse += sse2; \
|
|
} \
|
|
} \
|
|
se = ROUND_POWER_OF_TWO(se, 2); \
|
|
sse = ROUND_POWER_OF_TWO(sse, 4); \
|
|
*sse_ptr = sse; \
|
|
return sse - ((cast se * se) >> (wlog2 + hlog2)); \
|
|
} \
|
|
\
|
|
uint32_t vpx_highbd_12_sub_pixel_avg_variance##w##x##h##_##opt( \
|
|
const uint8_t *src8, int src_stride, int x_offset, int y_offset, \
|
|
const uint8_t *dst8, int dst_stride, uint32_t *sse_ptr, \
|
|
const uint8_t *sec8) { \
|
|
int start_row; \
|
|
uint32_t sse; \
|
|
int se = 0; \
|
|
uint64_t long_sse = 0; \
|
|
uint16_t *src = CONVERT_TO_SHORTPTR(src8); \
|
|
uint16_t *dst = CONVERT_TO_SHORTPTR(dst8); \
|
|
uint16_t *sec = CONVERT_TO_SHORTPTR(sec8); \
|
|
for (start_row = 0; start_row < h; start_row +=16) { \
|
|
uint32_t sse2; \
|
|
int height = h - start_row < 16 ? h - start_row : 16; \
|
|
int se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
|
|
src + (start_row * src_stride), src_stride, x_offset, \
|
|
y_offset, dst + (start_row * dst_stride), dst_stride, \
|
|
sec + (start_row * w), w, height, &sse2, NULL, NULL); \
|
|
se += se2; \
|
|
long_sse += sse2; \
|
|
if (w > wf) { \
|
|
se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
|
|
src + 16 + (start_row * src_stride), src_stride, \
|
|
x_offset, y_offset, \
|
|
dst + 16 + (start_row * dst_stride), dst_stride, \
|
|
sec + 16 + (start_row * w), w, height, &sse2, NULL, NULL); \
|
|
se += se2; \
|
|
long_sse += sse2; \
|
|
if (w > wf * 2) { \
|
|
se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
|
|
src + 32 + (start_row * src_stride), src_stride, \
|
|
x_offset, y_offset, \
|
|
dst + 32 + (start_row * dst_stride), dst_stride, \
|
|
sec + 32 + (start_row * w), w, height, &sse2, NULL, NULL); \
|
|
se += se2; \
|
|
long_sse += sse2; \
|
|
se2 = vpx_highbd_sub_pixel_avg_variance##wf##xh_##opt( \
|
|
src + 48 + (start_row * src_stride), src_stride, \
|
|
x_offset, y_offset, \
|
|
dst + 48 + (start_row * dst_stride), dst_stride, \
|
|
sec + 48 + (start_row * w), w, height, &sse2, NULL, NULL); \
|
|
se += se2; \
|
|
long_sse += sse2; \
|
|
} \
|
|
} \
|
|
} \
|
|
se = ROUND_POWER_OF_TWO(se, 4); \
|
|
sse = (uint32_t)ROUND_POWER_OF_TWO(long_sse, 8); \
|
|
*sse_ptr = sse; \
|
|
return sse - ((cast se * se) >> (wlog2 + hlog2)); \
|
|
}
|
|
|
|
|
|
#define FNS(opt1) \
|
|
FN(64, 64, 16, 6, 6, opt1, (int64_t)); \
|
|
FN(64, 32, 16, 6, 5, opt1, (int64_t)); \
|
|
FN(32, 64, 16, 5, 6, opt1, (int64_t)); \
|
|
FN(32, 32, 16, 5, 5, opt1, (int64_t)); \
|
|
FN(32, 16, 16, 5, 4, opt1, (int64_t)); \
|
|
FN(16, 32, 16, 4, 5, opt1, (int64_t)); \
|
|
FN(16, 16, 16, 4, 4, opt1, (int64_t)); \
|
|
FN(16, 8, 16, 4, 3, opt1, (int64_t)); \
|
|
FN(8, 16, 8, 4, 3, opt1, (int64_t)); \
|
|
FN(8, 8, 8, 3, 3, opt1, (int64_t)); \
|
|
FN(8, 4, 8, 3, 2, opt1, (int64_t));
|
|
|
|
FNS(sse2);
|
|
|
|
#undef FNS
|
|
#undef FN
|
|
|
|
void vpx_highbd_upsampled_pred_sse2(uint16_t *comp_pred,
|
|
int width, int height,
|
|
const uint8_t *ref8,
|
|
int ref_stride) {
|
|
int i, j;
|
|
int stride = ref_stride << 3;
|
|
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
|
|
|
|
if (width >= 8) {
|
|
// read 8 points at one time
|
|
for (i = 0; i < height; i++) {
|
|
for (j = 0; j < width; j+= 8) {
|
|
__m128i s0 = _mm_cvtsi32_si128(*(const uint32_t *)ref);
|
|
__m128i s1 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 8));
|
|
__m128i s2 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 16));
|
|
__m128i s3 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 24));
|
|
__m128i s4 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 32));
|
|
__m128i s5 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 40));
|
|
__m128i s6 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 48));
|
|
__m128i s7 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 56));
|
|
__m128i t0, t1, t2, t3;
|
|
|
|
t0 = _mm_unpacklo_epi16(s0, s1);
|
|
t1 = _mm_unpacklo_epi16(s2, s3);
|
|
t2 = _mm_unpacklo_epi16(s4, s5);
|
|
t3 = _mm_unpacklo_epi16(s6, s7);
|
|
t0 = _mm_unpacklo_epi32(t0, t1);
|
|
t2 = _mm_unpacklo_epi32(t2, t3);
|
|
t0 = _mm_unpacklo_epi64(t0, t2);
|
|
|
|
_mm_storeu_si128((__m128i *)(comp_pred), t0);
|
|
comp_pred += 8;
|
|
ref += 64; // 8 * 8;
|
|
}
|
|
ref += stride - (width << 3);
|
|
}
|
|
} else {
|
|
// read 4 points at one time
|
|
for (i = 0; i < height; i++) {
|
|
for (j = 0; j < width; j+= 4) {
|
|
__m128i s0 = _mm_cvtsi32_si128(*(const uint32_t *)ref);
|
|
__m128i s1 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 8));
|
|
__m128i s2 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 16));
|
|
__m128i s3 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 24));
|
|
__m128i t0, t1;
|
|
|
|
t0 = _mm_unpacklo_epi16(s0, s1);
|
|
t1 = _mm_unpacklo_epi16(s2, s3);
|
|
t0 = _mm_unpacklo_epi32(t0, t1);
|
|
|
|
_mm_storel_epi64((__m128i *)(comp_pred), t0);
|
|
comp_pred += 4;
|
|
ref += 4 * 8;
|
|
}
|
|
ref += stride - (width << 3);
|
|
}
|
|
}
|
|
}
|
|
|
|
void vpx_highbd_comp_avg_upsampled_pred_sse2(uint16_t *comp_pred,
|
|
const uint8_t *pred8,
|
|
int width, int height,
|
|
const uint8_t *ref8,
|
|
int ref_stride) {
|
|
const __m128i one = _mm_set1_epi16(1);
|
|
int i, j;
|
|
int stride = ref_stride << 3;
|
|
uint16_t *pred = CONVERT_TO_SHORTPTR(pred8);
|
|
uint16_t *ref = CONVERT_TO_SHORTPTR(ref8);
|
|
|
|
if (width >= 8) {
|
|
// read 8 points at one time
|
|
for (i = 0; i < height; i++) {
|
|
for (j = 0; j < width; j+= 8) {
|
|
__m128i s0 = _mm_cvtsi32_si128(*(const uint32_t *)ref);
|
|
__m128i s1 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 8));
|
|
__m128i s2 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 16));
|
|
__m128i s3 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 24));
|
|
__m128i s4 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 32));
|
|
__m128i s5 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 40));
|
|
__m128i s6 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 48));
|
|
__m128i s7 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 56));
|
|
__m128i p0 = _mm_loadu_si128((const __m128i *)pred);
|
|
__m128i t0, t1, t2, t3;
|
|
|
|
t0 = _mm_unpacklo_epi16(s0, s1);
|
|
t1 = _mm_unpacklo_epi16(s2, s3);
|
|
t2 = _mm_unpacklo_epi16(s4, s5);
|
|
t3 = _mm_unpacklo_epi16(s6, s7);
|
|
t0 = _mm_unpacklo_epi32(t0, t1);
|
|
t2 = _mm_unpacklo_epi32(t2, t3);
|
|
t0 = _mm_unpacklo_epi64(t0, t2);
|
|
|
|
p0 = _mm_adds_epu16(t0, p0);
|
|
p0 = _mm_adds_epu16(p0, one);
|
|
p0 = _mm_srli_epi16(p0, 1);
|
|
|
|
_mm_storeu_si128((__m128i *)(comp_pred), p0);
|
|
comp_pred += 8;
|
|
pred += 8;
|
|
ref += 8 * 8;
|
|
}
|
|
ref += stride - (width << 3);
|
|
}
|
|
} else {
|
|
// read 4 points at one time
|
|
for (i = 0; i < height; i++) {
|
|
for (j = 0; j < width; j+= 4) {
|
|
__m128i s0 = _mm_cvtsi32_si128(*(const uint32_t *)ref);
|
|
__m128i s1 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 8));
|
|
__m128i s2 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 16));
|
|
__m128i s3 = _mm_cvtsi32_si128(*(const uint32_t *)(ref + 24));
|
|
__m128i p0 = _mm_loadl_epi64((const __m128i *)pred);
|
|
__m128i t0, t1;
|
|
|
|
t0 = _mm_unpacklo_epi16(s0, s1);
|
|
t1 = _mm_unpacklo_epi16(s2, s3);
|
|
t0 = _mm_unpacklo_epi32(t0, t1);
|
|
|
|
p0 = _mm_adds_epu16(t0, p0);
|
|
p0 = _mm_adds_epu16(p0, one);
|
|
p0 = _mm_srli_epi16(p0, 1);
|
|
|
|
_mm_storel_epi64((__m128i *)(comp_pred), p0);
|
|
comp_pred += 4;
|
|
pred += 4;
|
|
ref += 4 * 8;
|
|
}
|
|
ref += stride - (width << 3);
|
|
}
|
|
}
|
|
}
|