[FFmpeg-devel] [PATCH v2 01/12] ffv1enc_vulkan: disable autodetection of async_depth
Lynne
dev at lynne.ee
Mon Feb 24 10:04:14 EET 2025
The issue is that this could consume gigabytes of VRAM at higher
resolutions for not that much of a speedup.
Automatic detection was not a good idea as we can't know how much
VRAM is actually free.
Just remove it.
---
libavcodec/ffv1enc_vulkan.c | 7 +------
1 file changed, 1 insertion(+), 6 deletions(-)
diff --git a/libavcodec/ffv1enc_vulkan.c b/libavcodec/ffv1enc_vulkan.c
index 3d7ee073aa..6a12ee2055 100644
--- a/libavcodec/ffv1enc_vulkan.c
+++ b/libavcodec/ffv1enc_vulkan.c
@@ -1631,11 +1631,6 @@ static av_cold int vulkan_encode_ffv1_init(AVCodecContext *avctx)
max_heap_size = max_heap_size - (max_heap_size >> 3);
}
- if (!fv->async_depth) {
- fv->async_depth = FFMIN(fv->qf->num, FFMAX(max_heap_size / maxsize, 1));
- fv->async_depth = FFMAX(fv->async_depth, 1);
- }
-
av_log(avctx, AV_LOG_INFO, "Async buffers: %zuMiB per context, %zuMiB total, depth: %i\n",
maxsize / (1024*1024),
(fv->async_depth * maxsize) / (1024*1024),
@@ -1820,7 +1815,7 @@ static const AVOption vulkan_encode_ffv1_options[] = {
{ .i64 = 0 }, 0, 1, VE },
{ "async_depth", "Internal parallelization depth", OFFSET(async_depth), AV_OPT_TYPE_INT,
- { .i64 = 0 }, 0, INT_MAX, VE },
+ { .i64 = 1 }, 1, INT_MAX, VE },
{ NULL }
};
--
2.47.2
More information about the ffmpeg-devel
mailing list