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>

1 files changed, 38 insertions, 0 deletions

diff --git a/tests/cooperative-vector/coherent-load-store-pointer.slang b/tests/cooperative-vector/coherent-load-store-pointer.slang
new file mode 100644
index 000000000..40efeee1a
--- /dev/null
+++ b/tests/cooperative-vector/coherent-load-store-pointer.slang
@@ -0,0 +1,38 @@
+//TEST:SIMPLE(filecheck=SPIRV):-stage compute -entry computeMain -target spirv
+
+// coherent CoopVec operations crash the Nvidia driver.
+//TEST(compute):COMPARE_COMPUTE(filecheck-buffer=CHECK):-vk -render-feature cooperative-vector -emit-spirv-directly
+
+// Ensure SPIRV emits coherent operations here
+// SPIRV: MakePointerVisible
+// SPIRV: MakePointerAvailable
+
+//TEST_INPUT: set inputBuffer = ubuffer(data=[1 2 3 4 5 6 7 8 9 10 11 12], stride=4);
+uniform int32_t* inputBuffer;
+
+//TEST_INPUT: set outputBuffer = out ubuffer(data=[0 0 0 0 0 0 0 0], stride=4);
+uniform int32_t* outputBuffer;
+
+// CHECK: 9
+// CHECK-NEXT: A
+// CHECK-NEXT: B
+// CHECK-NEXT: C
+// CHECK-NEXT: 1
+// CHECK-NEXT: 2
+// CHECK-NEXT: 3
+// CHECK-NEXT: 4
+
+[shader("compute")]
+[numthreads(1, 1, 1)]
+void computeMain()
+{
+    //// First half of input.
+    let a = coopVecLoadCoherent<4, int32_t>(inputBuffer, 0, MemoryScope::Device);
+    //// Second half of input.
+    let b = coopVecLoadCoherent<4, int32_t>(inputBuffer + 4, 4 * 4, MemoryScope::Device);
+    //// Store second half of input to first half of output buffer.
+    b.storeCoherent(outputBuffer, 0, MemoryScope::Device);
+    //// Store first half of input to second half of output buffer.
+    a.storeCoherent(outputBuffer, 4 * 4, MemoryScope::Device);
+}
+