frontents/va: Use PIPE_USAGE_STAGING for coded buffer

Coded buffer will only be read on CPU, setting
PIPE_USAGE_STAGING instead of PIPE_USAGE_STREAM
makes the CPU reads much faster.

On 6700XT this reduces the CPU copy by around
3ms to 0.3 ms on average while under high GPU
load - real-time game streaming.

Signed-off-by: David Rosca <nowrep@gmail.com>
Reviewed-by: Ruijing Dong <ruijing.dong@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/20989>

diff --git a/src/gallium/frontends/va/picture_h264_enc.c b/src/gallium/frontends/va/picture_h264_enc.c
index 5159cdd..79e23d4 100644 (file)
--- a/src/gallium/frontends/va/picture_h264_enc.c
+++ b/src/gallium/frontends/va/picture_h264_enc.c
@@ -53,7 +53,7 @@ vlVaHandleVAEncPictureParameterBufferTypeH264(vlVaDriver *drv, vlVaContext *cont
    coded_buf = handle_table_get(drv->htab, h264->coded_buf);
    if (!coded_buf->derived_surface.resource)
       coded_buf->derived_surface.resource = pipe_buffer_create(drv->pipe->screen, PIPE_BIND_VERTEX_BUFFER,
-                                            PIPE_USAGE_STREAM, coded_buf->size);
+                                            PIPE_USAGE_STAGING, coded_buf->size);
    context->coded_buf = coded_buf;
 
    if (context->desc.h264enc.is_ltr)

diff --git a/src/gallium/frontends/va/picture_hevc_enc.c b/src/gallium/frontends/va/picture_hevc_enc.c
index 44429c2..5826d75 100644 (file)
--- a/src/gallium/frontends/va/picture_hevc_enc.c
+++ b/src/gallium/frontends/va/picture_hevc_enc.c
@@ -56,7 +56,7 @@ vlVaHandleVAEncPictureParameterBufferTypeHEVC(vlVaDriver *drv, vlVaContext *cont
 
    if (!coded_buf->derived_surface.resource)
       coded_buf->derived_surface.resource = pipe_buffer_create(drv->pipe->screen, PIPE_BIND_VERTEX_BUFFER,
-                                            PIPE_USAGE_STREAM, coded_buf->size);
+                                            PIPE_USAGE_STAGING, coded_buf->size);
 
    context->coded_buf = coded_buf;
    context->desc.h265enc.pic.log2_parallel_merge_level_minus2 = h265->log2_parallel_merge_level_minus2;