Cleanup atomic intrinsics. (#5324)

* Cleanup atomic intrinsics.

* Fix.

* Fix glsl.

* Remove hacky intrinsic expansion logic for glsl image atomics.

* Fix all tests.

* Fix.

* Add `InterlockedAddF16Emulated`.

* Fix glsl intrinsic.

* Fix.

1 files changed, 57 insertions, 0 deletions

diff --git a/tests/metal/atomic-byteaddressbuffer.slang b/tests/metal/atomic-byteaddressbuffer.slang
new file mode 100644
index 000000000..677f80dbf
--- /dev/null
+++ b/tests/metal/atomic-byteaddressbuffer.slang
@@ -0,0 +1,57 @@
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=CHK):-slang -compute -dx12 -profile cs_6_0 -use-dxil -shaderobj -output-using-type
+//TEST(compute, vulkan):COMPARE_COMPUTE_EX(filecheck-buffer=CHK):-vk -compute -shaderobj -output-using-type
+//TEST(compute):COMPARE_COMPUTE_EX(filecheck-buffer=CHK):-cuda -compute -shaderobj -output-using-type
+//TEST:SIMPLE(filecheck=LIB):-target metallib -entry computeMain -stage compute -DMETAL
+
+//TEST_INPUT:ubuffer(data=[0 0 0 0 0]):name=uintBuffer
+RWByteAddressBuffer uintBuffer;
+
+//TEST_INPUT: ubuffer(data=[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 ], stride=4):out,name outputBuffer
+RWStructuredBuffer<float> outputBuffer;
+
+[numthreads(1,1,1)]
+void computeMain()
+{
+    uintBuffer.InterlockedAdd(0, 1);
+    int oldValue;
+    //LIB: call {{.*}}.atomic.global.add.u.i32
+    uintBuffer.InterlockedAdd(0, 1, oldValue);
+    // CHK: 1
+    outputBuffer[0] = oldValue;
+
+    uintBuffer.InterlockedAdd(0, 1, oldValue);
+    // CHK: 2
+    outputBuffer[1] = (int)oldValue;
+
+    uintBuffer.InterlockedCompareExchange(0, 3, 4, oldValue);
+    // CHK: 3
+    outputBuffer[2] = (int)oldValue;
+
+    uintBuffer.InterlockedOr(0, 3, oldValue);
+    // CHK: 4
+    outputBuffer[3] = oldValue; // 4
+
+    uintBuffer.InterlockedExchange(0, 4, oldValue);
+    // CHK: 7
+    outputBuffer[4] = oldValue; // 7
+
+    uintBuffer.InterlockedMin(0, 3, oldValue);
+    // CHK: 4
+    outputBuffer[5] = oldValue; // 4
+
+    uintBuffer.InterlockedMax(0, 4, oldValue);
+    // CHK: 3
+    outputBuffer[6] = oldValue; // 3
+
+    uintBuffer.InterlockedAnd(0, 7, oldValue);
+    // CHK: 4
+    outputBuffer[7] = oldValue; // 4
+
+    uintBuffer.InterlockedXor(0, 7, oldValue);
+    // CHK: 4
+    outputBuffer[8] = oldValue; // 4
+
+    // CHK: 3
+    outputBuffer[9] = uintBuffer.Load(0);
+
+}
\ No newline at end of file