radv: Add on-demand compilation of built-in shaders.
In environments where we cannot cache, e.g. Android (no homedir),
ChromeOS (readonly rootfs) or sandboxes (cannot open cache), the
startup cost of creating a device in radv is rather high, due
to compiling all possible built-in pipelines up front. This meant
depending on the CPU a 1-4 sec cost of creating a Device.
For CTS this cost is unacceptable, and likely for starting random
apps too.
So if there is no cache, with this patch radv will compile shaders
on demand. Once there is a cache from the first run, even if
incomplete, the driver knows that it can likely write the cache
and precompiles everything.
Note that I did not switch the buffer and itob/btoi compute pipelines
to on-demand, since you cannot really do anything in Vulkan without
them and there are only a few.
This reduces the CTS runtime for the no caches scenario on my
threadripper from 32 minutes to 8 minutes.
Reviewed-by: Dave Airlie <airlied@redhat.com>
diff --git a/src/amd/vulkan/radv_meta.h b/src/amd/vulkan/radv_meta.h
index 35067f6..f8d48f4 100644
--- a/src/amd/vulkan/radv_meta.h
+++ b/src/amd/vulkan/radv_meta.h
@@ -58,34 +58,34 @@
VkRect2D render_area;
};
-VkResult radv_device_init_meta_clear_state(struct radv_device *device);
+VkResult radv_device_init_meta_clear_state(struct radv_device *device, bool on_demand);
void radv_device_finish_meta_clear_state(struct radv_device *device);
-VkResult radv_device_init_meta_resolve_state(struct radv_device *device);
+VkResult radv_device_init_meta_resolve_state(struct radv_device *device, bool on_demand);
void radv_device_finish_meta_resolve_state(struct radv_device *device);
-VkResult radv_device_init_meta_depth_decomp_state(struct radv_device *device);
+VkResult radv_device_init_meta_depth_decomp_state(struct radv_device *device, bool on_demand);
void radv_device_finish_meta_depth_decomp_state(struct radv_device *device);
-VkResult radv_device_init_meta_fast_clear_flush_state(struct radv_device *device);
+VkResult radv_device_init_meta_fast_clear_flush_state(struct radv_device *device, bool on_demand);
void radv_device_finish_meta_fast_clear_flush_state(struct radv_device *device);
-VkResult radv_device_init_meta_blit_state(struct radv_device *device);
+VkResult radv_device_init_meta_blit_state(struct radv_device *device, bool on_demand);
void radv_device_finish_meta_blit_state(struct radv_device *device);
-VkResult radv_device_init_meta_blit2d_state(struct radv_device *device);
+VkResult radv_device_init_meta_blit2d_state(struct radv_device *device, bool on_demand);
void radv_device_finish_meta_blit2d_state(struct radv_device *device);
VkResult radv_device_init_meta_buffer_state(struct radv_device *device);
void radv_device_finish_meta_buffer_state(struct radv_device *device);
-VkResult radv_device_init_meta_query_state(struct radv_device *device);
+VkResult radv_device_init_meta_query_state(struct radv_device *device, bool on_demand);
void radv_device_finish_meta_query_state(struct radv_device *device);
-VkResult radv_device_init_meta_resolve_compute_state(struct radv_device *device);
+VkResult radv_device_init_meta_resolve_compute_state(struct radv_device *device, bool on_demand);
void radv_device_finish_meta_resolve_compute_state(struct radv_device *device);
-VkResult radv_device_init_meta_resolve_fragment_state(struct radv_device *device);
+VkResult radv_device_init_meta_resolve_fragment_state(struct radv_device *device, bool on_demand);
void radv_device_finish_meta_resolve_fragment_state(struct radv_device *device);
void radv_meta_save(struct radv_meta_saved_state *saved_state,