diff options
| author | 16-Bit-Dog <67922228+16-Bit-Dog@users.noreply.github.com> | 2025-10-10 13:09:24 -0400 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2025-10-10 17:09:24 +0000 |
| commit | 1e0908bd7107dfbdac912b693c3ab9bd6e1dc8b3 (patch) | |
| tree | cc39d2e18abc954fb76f9a54b11a8d492685c6e2 /tests/language-feature/pointer/coherent-load-store-groupshared.slang | |
| parent | b4023f715885ada9a2777ea3b0d6d9739860b39b (diff) | |
Addition of `Load`/`Store` coherent operations (#8395)
Fixes: https://github.com/shader-slang/slang/issues/7634
Duplicate of PR https://github.com/shader-slang/slang/pull/8052
Primary Changes:
* Added `storeCoherent` and `loadCoherent` for coherent load/store via
pointers. This is backed by `IRMemoryScopeAttr` which is an `IRAttr`
attached to `IRLoad` and `IRStore`
* Logic in `source\slang\slang-emit-spirv.cpp` for load/store emitting
has been reworked to be less messy and more maintainable
* Add to `hlsl.meta.slang` coop vector and coop matrix coherent
load/store operations
Secondary Changes:
* Added a missing load/store test for coop matrix:
`tests\cooperative-matrix\load-store-pointer.slang`
---------
Co-authored-by: ArielG-NV <aglasroth@nvidia.com>
Co-authored-by: ArielG-NV <159081215+ArielG-NV@users.noreply.github.com>
Co-authored-by: slangbot <186143334+slangbot@users.noreply.github.com>
Co-authored-by: Nathan V. Morrical <natemorrical@gmail.com>
Diffstat (limited to 'tests/language-feature/pointer/coherent-load-store-groupshared.slang')
| -rw-r--r-- | tests/language-feature/pointer/coherent-load-store-groupshared.slang | 26 |
1 files changed, 26 insertions, 0 deletions
diff --git a/tests/language-feature/pointer/coherent-load-store-groupshared.slang b/tests/language-feature/pointer/coherent-load-store-groupshared.slang new file mode 100644 index 000000000..2e537ef01 --- /dev/null +++ b/tests/language-feature/pointer/coherent-load-store-groupshared.slang @@ -0,0 +1,26 @@ +//TEST:SIMPLE(filecheck=SPIRV):-stage compute -entry computeMain -target spirv -capability vk_mem_model +//TEST(compute):COMPARE_COMPUTE(filecheck-buffer=CHECK):-vk -emit-spirv-directly -capability vk_mem_model + +// Tests if we pass-through and handle groupshared address space pointers correctly. +// Ensure SPIRV emits coherent operations here +// SPIRV: MakePointerAvailable|NonPrivatePointer +// SPIRV: MakePointerVisible|NonPrivatePointer + +// CHECK: 2 +// CHECK-NEXT: 1 +// CHECK-NEXT: 0 + +//TEST_INPUT:ubuffer(data=[0 0 0], stride=4):out,name=outputBuffer +RWStructuredBuffer<int> outputBuffer; + +groupshared int[32] shared; + +#define THREAD_GROUP_SIZE 3 +[numthreads(THREAD_GROUP_SIZE, 1, 1)] +void computeMain(uint3 group_thread_id: SV_GroupThreadID) +{ + Ptr<int, Access::ReadWrite, AddressSpace::GroupShared> ptr = __getAddress(shared[0]); + storeCoherent<4, MemoryScope::Workgroup>(ptr + group_thread_id.x, (int)group_thread_id.x); + AllMemoryBarrierWithGroupSync(); + outputBuffer[group_thread_id.x] = loadCoherent<4, MemoryScope::Workgroup>(ptr + THREAD_GROUP_SIZE - group_thread_id.x - 1); +}
\ No newline at end of file |