generic-library/vpx
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
vpx/vp9/common/vp9_thread_common.h

64 lines
2.1 KiB
C
Raw Normal View History

vp9 decoder: row-based multi-threaded loopfilter Implemented parallel loopfiltering, which uses existing tile- decoding threads. Each thread works on one row, and when that row is loopfiltered, it moves to next unattended row. To ensure the correct filtering order, threads are synchronized and one superblock is filtered only if the superblocks it depends on are filtered already. To reduce synchronization overhead and speed up the decoder, we use nsync > 1 for high resolution. Performance tests: 1. on desktop: 8-tile 4k video using 8 threads, speedup: 70% - 80% 4-tile HD video using 4 threads, speedup: ~35% 2. on mobile device(Nexus 7): 4-tile 1080p video using 4 threads, speedup: 18% - 25% 4-tile 1080p video using 2 threads, speedup: 10% - 15% Change-Id: If54b4a11960dd706c22d5ad145ad94156031f36a
2013-12-27 15:25:54 -08:00
/*
* 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.
*/
Fix a macro definition to be consistent with the head file name. Change-Id: I9634332a2b3fac7e7f3b7ef58821ea7c81c5c813
2015-09-29 09:34:42 -07:00
#ifndef VP9_COMMON_VP9_THREAD_COMMON_H_
#define VP9_COMMON_VP9_THREAD_COMMON_H_
vp9 decoder: row-based multi-threaded loopfilter Implemented parallel loopfiltering, which uses existing tile- decoding threads. Each thread works on one row, and when that row is loopfiltered, it moves to next unattended row. To ensure the correct filtering order, threads are synchronized and one superblock is filtered only if the superblocks it depends on are filtered already. To reduce synchronization overhead and speed up the decoder, we use nsync > 1 for high resolution. Performance tests: 1. on desktop: 8-tile 4k video using 8 threads, speedup: 70% - 80% 4-tile HD video using 4 threads, speedup: ~35% 2. on mobile device(Nexus 7): 4-tile 1080p video using 4 threads, speedup: 18% - 25% 4-tile 1080p video using 2 threads, speedup: 10% - 15% Change-Id: If54b4a11960dd706c22d5ad145ad94156031f36a
2013-12-27 15:25:54 -08:00
#include "./vpx_config.h"
vp9_ethread: add parallel loopfilter 1. Added row-based loopfilter in encoder; 2. Moved common multi-threaded loopfilter functions from decoder to common; 3. Merged multi-threaded loopfilter code, and made encoder/ decoder call same function to reduce code duplication. Encoder tests showed that 1% - 2% speedup was seen for good-quality 2-pass mode(at speed 3); 1% - 3% speedup using 2 threads and 4% - 6% speedup using 4 threads were seen for real-time mode(at speed 7). Change-Id: I8a4ac51c2ad9bab9fa7b864e90743931c53ec1c4
2015-01-06 14:14:26 -08:00
#include "vp9/common/vp9_loopfilter.h"
Rename vpx_thread to vpx_util Change the dir name to include more util tools. Change-Id: Id5b16062803ce5eed872fe2edb36d7e56b32eed8
2015-07-02 10:01:09 -07:00
#include "vpx_util/vpx_thread.h"
vp9 decoder: row-based multi-threaded loopfilter Implemented parallel loopfiltering, which uses existing tile- decoding threads. Each thread works on one row, and when that row is loopfiltered, it moves to next unattended row. To ensure the correct filtering order, threads are synchronized and one superblock is filtered only if the superblocks it depends on are filtered already. To reduce synchronization overhead and speed up the decoder, we use nsync > 1 for high resolution. Performance tests: 1. on desktop: 8-tile 4k video using 8 threads, speedup: 70% - 80% 4-tile HD video using 4 threads, speedup: ~35% 2. on mobile device(Nexus 7): 4-tile 1080p video using 4 threads, speedup: 18% - 25% 4-tile 1080p video using 2 threads, speedup: 10% - 15% Change-Id: If54b4a11960dd706c22d5ad145ad94156031f36a
2013-12-27 15:25:54 -08:00
vp9: add extern "C" to headers Change-Id: I1b6927ad820f99340985b094d415aaab14defaf4
2015-09-09 21:04:27 -07:00
#ifdef __cplusplus
extern "C" {
#endif
vp9 decoder: row-based multi-threaded loopfilter Implemented parallel loopfiltering, which uses existing tile- decoding threads. Each thread works on one row, and when that row is loopfiltered, it moves to next unattended row. To ensure the correct filtering order, threads are synchronized and one superblock is filtered only if the superblocks it depends on are filtered already. To reduce synchronization overhead and speed up the decoder, we use nsync > 1 for high resolution. Performance tests: 1. on desktop: 8-tile 4k video using 8 threads, speedup: 70% - 80% 4-tile HD video using 4 threads, speedup: ~35% 2. on mobile device(Nexus 7): 4-tile 1080p video using 4 threads, speedup: 18% - 25% 4-tile 1080p video using 2 threads, speedup: 10% - 15% Change-Id: If54b4a11960dd706c22d5ad145ad94156031f36a
2013-12-27 15:25:54 -08:00
struct VP9Common;
Make encoder and decoder share common thread function Moved vp9_accumulate_frame_counts to vp9_thread_common.c to eliminate the duplicate code. Change-Id: I9cf506d729603c8bf1494b4c86a3b7d47af1917a
2015-02-06 11:37:45 -08:00
struct FRAME_COUNTS;
vp9 decoder: row-based multi-threaded loopfilter Implemented parallel loopfiltering, which uses existing tile- decoding threads. Each thread works on one row, and when that row is loopfiltered, it moves to next unattended row. To ensure the correct filtering order, threads are synchronized and one superblock is filtered only if the superblocks it depends on are filtered already. To reduce synchronization overhead and speed up the decoder, we use nsync > 1 for high resolution. Performance tests: 1. on desktop: 8-tile 4k video using 8 threads, speedup: 70% - 80% 4-tile HD video using 4 threads, speedup: ~35% 2. on mobile device(Nexus 7): 4-tile 1080p video using 4 threads, speedup: 18% - 25% 4-tile 1080p video using 2 threads, speedup: 10% - 15% Change-Id: If54b4a11960dd706c22d5ad145ad94156031f36a
2013-12-27 15:25:54 -08:00
// Loopfilter row synchronization
typedef struct VP9LfSyncData {
#if CONFIG_MULTITHREAD
pthread_mutex_t *mutex_;
pthread_cond_t *cond_;
#endif
// Allocate memory to store the loop-filtered superblock index in each row.
int *cur_sb_col;
// The optimal sync_range for different resolution and platform should be
// determined by testing. Currently, it is chosen to be a power-of-2 number.
int sync_range;
vp9: fix m/t loop filter invalid free store the number of allocated rows in VP9LfSync, the calculated values can not be relied on when dealing with corrupt material. Change-Id: I13b8bcec9738c299a71df726772ab7ac05511e5b
2014-08-28 17:11:31 -07:00
int rows;
move LFWorkerData allocation to VP9LfSync this removes an assumption that worker->data1 would be pointing to a TileWorkerData allocation. additionally, within the multi-threaded loopfilter pass VP9LfSync as a parameter to the worker hook, removing the need for a shadow pointer in LFWorkerData. Change-Id: Ic7b2faa34e3eb59dbcb8a7c67f333448fa047c88
2014-08-31 13:16:37 -07:00
// Row-based parallel loopfilter data
LFWorkerData *lfdata;
int num_workers;
vp9 decoder: row-based multi-threaded loopfilter Implemented parallel loopfiltering, which uses existing tile- decoding threads. Each thread works on one row, and when that row is loopfiltered, it moves to next unattended row. To ensure the correct filtering order, threads are synchronized and one superblock is filtered only if the superblocks it depends on are filtered already. To reduce synchronization overhead and speed up the decoder, we use nsync > 1 for high resolution. Performance tests: 1. on desktop: 8-tile 4k video using 8 threads, speedup: 70% - 80% 4-tile HD video using 4 threads, speedup: ~35% 2. on mobile device(Nexus 7): 4-tile 1080p video using 4 threads, speedup: 18% - 25% 4-tile 1080p video using 2 threads, speedup: 10% - 15% Change-Id: If54b4a11960dd706c22d5ad145ad94156031f36a
2013-12-27 15:25:54 -08:00
} VP9LfSync;
// Allocate memory for loopfilter row synchronization.
vp9_ethread: add parallel loopfilter 1. Added row-based loopfilter in encoder; 2. Moved common multi-threaded loopfilter functions from decoder to common; 3. Merged multi-threaded loopfilter code, and made encoder/ decoder call same function to reduce code duplication. Encoder tests showed that 1% - 2% speedup was seen for good-quality 2-pass mode(at speed 3); 1% - 3% speedup using 2 threads and 4% - 6% speedup using 4 threads were seen for real-time mode(at speed 7). Change-Id: I8a4ac51c2ad9bab9fa7b864e90743931c53ec1c4
2015-01-06 14:14:26 -08:00
void vp9_loop_filter_alloc(VP9LfSync *lf_sync, struct VP9Common *cm, int rows,
move LFWorkerData allocation to VP9LfSync this removes an assumption that worker->data1 would be pointing to a TileWorkerData allocation. additionally, within the multi-threaded loopfilter pass VP9LfSync as a parameter to the worker hook, removing the need for a shadow pointer in LFWorkerData. Change-Id: Ic7b2faa34e3eb59dbcb8a7c67f333448fa047c88
2014-08-31 13:16:37 -07:00
int width, int num_workers);
vp9 decoder: row-based multi-threaded loopfilter Implemented parallel loopfiltering, which uses existing tile- decoding threads. Each thread works on one row, and when that row is loopfiltered, it moves to next unattended row. To ensure the correct filtering order, threads are synchronized and one superblock is filtered only if the superblocks it depends on are filtered already. To reduce synchronization overhead and speed up the decoder, we use nsync > 1 for high resolution. Performance tests: 1. on desktop: 8-tile 4k video using 8 threads, speedup: 70% - 80% 4-tile HD video using 4 threads, speedup: ~35% 2. on mobile device(Nexus 7): 4-tile 1080p video using 4 threads, speedup: 18% - 25% 4-tile 1080p video using 2 threads, speedup: 10% - 15% Change-Id: If54b4a11960dd706c22d5ad145ad94156031f36a
2013-12-27 15:25:54 -08:00
// Deallocate loopfilter synchronization related mutex and data.
vp9: fix m/t loop filter invalid free store the number of allocated rows in VP9LfSync, the calculated values can not be relied on when dealing with corrupt material. Change-Id: I13b8bcec9738c299a71df726772ab7ac05511e5b
2014-08-28 17:11:31 -07:00
void vp9_loop_filter_dealloc(VP9LfSync *lf_sync);
vp9 decoder: row-based multi-threaded loopfilter Implemented parallel loopfiltering, which uses existing tile- decoding threads. Each thread works on one row, and when that row is loopfiltered, it moves to next unattended row. To ensure the correct filtering order, threads are synchronized and one superblock is filtered only if the superblocks it depends on are filtered already. To reduce synchronization overhead and speed up the decoder, we use nsync > 1 for high resolution. Performance tests: 1. on desktop: 8-tile 4k video using 8 threads, speedup: 70% - 80% 4-tile HD video using 4 threads, speedup: ~35% 2. on mobile device(Nexus 7): 4-tile 1080p video using 4 threads, speedup: 18% - 25% 4-tile 1080p video using 2 threads, speedup: 10% - 15% Change-Id: If54b4a11960dd706c22d5ad145ad94156031f36a
2013-12-27 15:25:54 -08:00
// Multi-threaded loopfilter that uses the tile threads.
vp9/common: apply clang-format Change-Id: Ie0f150fdcfcbf7c4db52d3a08bc8238ed1c72e3b
2016-07-26 16:52:55 -07:00
void vp9_loop_filter_frame_mt(YV12_BUFFER_CONFIG *frame, struct VP9Common *cm,
vp9_ethread: add parallel loopfilter 1. Added row-based loopfilter in encoder; 2. Moved common multi-threaded loopfilter functions from decoder to common; 3. Merged multi-threaded loopfilter code, and made encoder/ decoder call same function to reduce code duplication. Encoder tests showed that 1% - 2% speedup was seen for good-quality 2-pass mode(at speed 3); 1% - 3% speedup using 2 threads and 4% - 6% speedup using 4 threads were seen for real-time mode(at speed 7). Change-Id: I8a4ac51c2ad9bab9fa7b864e90743931c53ec1c4
2015-01-06 14:14:26 -08:00
struct macroblockd_plane planes[MAX_MB_PLANE],
vp9/common: apply clang-format Change-Id: Ie0f150fdcfcbf7c4db52d3a08bc8238ed1c72e3b
2016-07-26 16:52:55 -07:00
int frame_filter_level, int y_only,
int partial_frame, VPxWorker *workers,
int num_workers, VP9LfSync *lf_sync);
vp9 decoder: row-based multi-threaded loopfilter Implemented parallel loopfiltering, which uses existing tile- decoding threads. Each thread works on one row, and when that row is loopfiltered, it moves to next unattended row. To ensure the correct filtering order, threads are synchronized and one superblock is filtered only if the superblocks it depends on are filtered already. To reduce synchronization overhead and speed up the decoder, we use nsync > 1 for high resolution. Performance tests: 1. on desktop: 8-tile 4k video using 8 threads, speedup: 70% - 80% 4-tile HD video using 4 threads, speedup: ~35% 2. on mobile device(Nexus 7): 4-tile 1080p video using 4 threads, speedup: 18% - 25% 4-tile 1080p video using 2 threads, speedup: 10% - 15% Change-Id: If54b4a11960dd706c22d5ad145ad94156031f36a
2013-12-27 15:25:54 -08:00
vp9_accumulate_frame_counts: pass counts directly Change-Id: Ic3c6cfba5b1867c335f2834da936e20caec8597a
2015-08-19 19:25:38 -07:00
void vp9_accumulate_frame_counts(struct FRAME_COUNTS *accum,
const struct FRAME_COUNTS *counts, int is_dec);
Make encoder and decoder share common thread function Moved vp9_accumulate_frame_counts to vp9_thread_common.c to eliminate the duplicate code. Change-Id: I9cf506d729603c8bf1494b4c86a3b7d47af1917a
2015-02-06 11:37:45 -08:00
vp9: add extern "C" to headers Change-Id: I1b6927ad820f99340985b094d415aaab14defaf4
2015-09-09 21:04:27 -07:00
#ifdef __cplusplus
} // extern "C"
#endif
Fix a macro definition to be consistent with the head file name. Change-Id: I9634332a2b3fac7e7f3b7ef58821ea7c81c5c813
2015-09-29 09:34:42 -07:00
#endif // VP9_COMMON_VP9_THREAD_COMMON_H_
Reference in New Issue Copy Permalink