nir: add image_block_size_per_thread_agx info

This is like shared_size but for spatial data instead, for compute shaders that
use the tilebuffer.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30633>
This commit is contained in:
Alyssa Rosenzweig 2024-08-07 13:06:06 -04:00
parent e754e54f88
commit b9eff53daa

View file

@ -458,6 +458,15 @@ typedef struct shader_info {
*/
bool has_cooperative_matrix:1;
/**
* Number of bytes of shared imageblock memory per thread. Currently,
* this requires that the workgroup size is 32x32x1 and that
* shared_size = 0. These requirements could be lifted in the future.
* However, there is no current OpenGL/Vulkan API support for
* imageblocks. This is only used internally to accelerate blit/copy.
*/
uint8_t image_block_size_per_thread_agx;
/**
* pointer size is:
* AddressingModelLogical: 0 (default)