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.

3 files changed, 63 insertions, 7 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
diff --git a/tests/metal/atomic-intrinsics.slang b/tests/metal/atomic-intrinsics.slang
index 5d47db913..afa0e5365 100644
--- a/tests/metal/atomic-intrinsics.slang
+++ b/tests/metal/atomic-intrinsics.slang
@@ -1,8 +1,7 @@
 //TEST:SIMPLE(filecheck=MTL):-target metal -entry computeMain -stage compute -DMETAL
 //TEST:SIMPLE(filecheck=LIB):-target metallib -entry computeMain -stage compute -DMETAL
 //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 -emit-spirv-directly -compute -shaderobj -output-using-type
-//TEST(compute, vulkan):COMPARE_COMPUTE_EX(filecheck-buffer=CHK):-vk -emit-spirv-via-glsl -compute -shaderobj -output-using-type
+//TEST(compute, vulkan):COMPARE_COMPUTE_EX(filecheck-buffer=CHK):-vk -compute -shaderobj -output-using-type
 
 //DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-cpu -compute -shaderobj -output-using-type
 //DISABLE_TEST(compute):COMPARE_COMPUTE_EX:-cuda -compute -shaderobj -output-using-type
@@ -36,22 +35,22 @@ void computeMain(uint groupIndex : SV_GroupIndex)
     float val = 0.0f;
 
     // InterlockedAdd
-    //MTL: atomic_uint threadgroup* {{.*}}shareMemUI
+    //MTL: atomic_uint threadgroup*{{.*}}shareMemUI
     //LIB: call {{.*}}.atomic.local.add.u.i32
     InterlockedAdd(shareMemUI[idx], uint(1));
     val += shareMemUI[idx];
 
-    //MTL: atomic_int threadgroup* {{.*}}shareMemI
+    //MTL: atomic_int threadgroup*{{.*}}shareMemI
     //LIB: call {{.*}}.atomic.local.add.s.i32
     InterlockedAdd(shareMemI[idx],  2);
     val += shareMemI[idx];
 
-    //MTL: atomic_uint device* {{.*}}uintBuffer
+    //MTL: atomic_uint device*{{.*}}uintBuffer
     //LIB: call {{.*}}.atomic.global.add.u.i32
     InterlockedAdd(uintBuffer[idx], 1);
     val += uintBuffer[idx];
 
-    //MTL: atomic_int device* {{.*}}intBuffer
+    //MTL: atomic_int device*{{.*}}intBuffer
     //LIB: call {{.*}}.atomic.global.add.s.i32
     InterlockedAdd(intBuffer[idx], 2);
     val += intBuffer[idx];
diff --git a/tests/metal/atomic-texture-buffer.slang b/tests/metal/atomic-texture-buffer.slang
index 3e4eda94b..1db156364 100644
--- a/tests/metal/atomic-texture-buffer.slang
+++ b/tests/metal/atomic-texture-buffer.slang
@@ -2,7 +2,7 @@
 //TEST:SIMPLE(filecheck=METAL_FLOAT): -target metal -stage compute -entry computeMain -DFLOAT
 //TEST:SIMPLE(filecheck=METALLIB): -target metallib -stage compute -entry computeMain
 
-// METAL_FLOAT: 'float' atomic texture operations are disallowed with Metal target's
+// METAL_FLOAT: floating point atomic operation
 
 //METALLIB: @computeMain