virtualenv/FFmpeg
1
0
Fork 0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2025-08-10 06:10:52 +02:00
Code Issues Releases Activity

vulkan_decode: halve execution pool size

Browse Source 
Determined experimentally, on various videos and hardware.
On Intel, using less resources in-flight is around 15% faster,
with similar results on Nvidia hardware.
This commit is contained in:
Lynne
2023-06-07 02:59:55 +02:00
parent 5f1be341c2
commit 24c4307b80
1 changed files with 3 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
libavcodec/vulkan_decode.c
View File

@@ -1105,8 +1105,9 @@ int ff_vk_decode_init(AVCodecContext *avctx)
session_create.pVideoProfile = &prof->profile_list.pProfiles[0]; session_create.pVideoProfile = &prof->profile_list.pProfiles[0];
/* Create decode exec context. /* Create decode exec context.
* 4 async contexts per thread seems like a good number. */ * 2 async contexts per thread was experimentally determined to be optimal
err = ff_vk_exec_pool_init(s, &qf_dec, &ctx->exec_pool, 4*avctx->thread_count, * for a majority of streams. */
err = ff_vk_exec_pool_init(s, &qf_dec, &ctx->exec_pool, 2*avctx->thread_count,
nb_q, VK_QUERY_TYPE_RESULT_STATUS_ONLY_KHR, 0, nb_q, VK_QUERY_TYPE_RESULT_STATUS_ONLY_KHR, 0,
session_create.pVideoProfile); session_create.pVideoProfile);
if (err < 0) if (err < 0)