OpenHarmony/third_party_mesa3d
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

rusticl/kernel: skip adding global id offsets if not used

Browse Source 
This allows us to shrink the kernel input buffer quite significantly as
the offset is a vec aligned size_t3 value.

Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Emma Anholt <emma@anholt.net>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25303>
This commit is contained in:
Karol Herbst
2023-09-19 18:07:27 +02:00
committed by Marge Bot
parent a9f408bd52
commit 33ca3a999c
1 changed files with 21 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

41
src/gallium/frontends/rusticl/core/kernel.rs
View File

@@ -518,21 +518,28 @@ fn lower_and_optimize_nir(
); );
nir.extract_constant_initializers(); nir.extract_constant_initializers();
// TODO 32 bit devices // run before gather info
// add vars for global offsets nir_pass!(nir, nir_lower_system_values);
internal_args.push(InternalKernelArg { let mut compute_options = nir_lower_compute_system_values_options::default();
kind: InternalKernelArgType::GlobalWorkOffsets, compute_options.set_has_base_global_invocation_id(true);
offset: 0, nir_pass!(nir, nir_lower_compute_system_values, &compute_options);
size: (3 * dev.address_bits() / 8) as usize, nir.gather_info();
});
if nir.reads_sysval(gl_system_value::SYSTEM_VALUE_BASE_GLOBAL_INVOCATION_ID) {
internal_args.push(InternalKernelArg {
kind: InternalKernelArgType::GlobalWorkOffsets,
offset: 0,
size: (3 * dev.address_bits() / 8) as usize,
});
lower_state.base_global_invoc_id_loc = args.len() + internal_args.len() - 1;
nir.add_var(
nir_variable_mode::nir_var_uniform,
unsafe { glsl_vector_type(address_bits_base_type, 3) },
lower_state.base_global_invoc_id_loc,
"base_global_invocation_id",
);
}
lower_state.base_global_invoc_id_loc = args.len() + internal_args.len() - 1;
nir.add_var(
nir_variable_mode::nir_var_uniform,
unsafe { glsl_vector_type(address_bits_base_type, 3) },
lower_state.base_global_invoc_id_loc,
"base_global_invocation_id",
);
if nir.has_constant() { if nir.has_constant() {
internal_args.push(InternalKernelArg { internal_args.push(InternalKernelArg {
kind: InternalKernelArgType::ConstantBuffer, kind: InternalKernelArgType::ConstantBuffer,
@@ -562,12 +569,6 @@ fn lower_and_optimize_nir(
); );
} }
// run before gather info
nir_pass!(nir, nir_lower_system_values);
let mut compute_options = nir_lower_compute_system_values_options::default();
compute_options.set_has_base_global_invocation_id(true);
nir_pass!(nir, nir_lower_compute_system_values, &compute_options);
nir.gather_info();
if nir.num_images() > 0 || nir.num_textures() > 0 { if nir.num_images() > 0 || nir.num_textures() > 0 {
let count = nir.num_images() + nir.num_textures(); let count = nir.num_images() + nir.num_textures();
internal_args.push(InternalKernelArg { internal_args.push(InternalKernelArg {