AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

aco: Add Primitive Ordered Pixel Shading scheduling rules

Browse Source 
Implementing the acquire/release semantics of fragment shader interlock
ordered section in Vulkan, and preventing reordering of memory accesses
requiring primitive ordering out of the ordered section.

Also, the ordered section should be as short as possible, so not reordering
the instructions awaiting overlapped waves upwards, and the exit from the
ordered section downwards.

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Signed-off-by: Vitaliy Triang3l Kuzmin <triang3l@yandex.ru>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22250>
This commit is contained in:
Vitaliy Triang3l Kuzmin
2023-04-03 21:27:47 +03:00
committed by Marge Bot
parent 6082e126eb
commit 2194e8bd82
2 changed files with 33 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/amd/compiler/aco_ir.cpp
+11
View File
@@ -223,6 +223,17 @@ init_program(Program* program, Stage stage, const struct aco_shader_info* info,
memory_sync_info
get_sync_info(const Instruction* instr)
{
/* Primitive Ordered Pixel Shading barriers necessary for accesses to memory shared between
* overlapping waves in the queue family.
*/
if (instr->opcode == aco_opcode::p_pops_gfx9_overlapped_wave_wait_done ||
(instr->opcode == aco_opcode::s_wait_event &&
!(instr->sopp().imm & wait_event_imm_dont_wait_export_ready))) {
return memory_sync_info(storage_buffer | storage_image, semantic_acquire, scope_queuefamily);
} else if (instr->opcode == aco_opcode::p_pops_gfx9_ordered_section_done) {
return memory_sync_info(storage_buffer | storage_image, semantic_release, scope_queuefamily);
}
switch (instr->format) {
case Format::SMEM: return instr->smem().sync;
case Format::MUBUF: return instr->mubuf().sync;