[llvm] [SPIRV] Do not use OpTypeRuntimeArray in Kernel env. (PR #149522)

[Victor Lomuller via llvm-commits]

================
@@ -828,9 +828,11 @@ SPIRVType *SPIRVGlobalRegistry::getOpTypeArray(uint32_t NumElems,
          "Invalid array element type");
   SPIRVType *SpvTypeInt32 = getOrCreateSPIRVIntegerType(32, MIRBuilder);
   SPIRVType *ArrayType = nullptr;
-  if (NumElems != 0) {
-    Register NumElementsVReg =
-        buildConstantInt(NumElems, MIRBuilder, SpvTypeInt32, EmitIR);
+  const SPIRVSubtarget &ST =
+      cast<SPIRVSubtarget>(MIRBuilder.getMF().getSubtarget());
+  if (NumElems != 0 || !ST.isShader()) {
+    Register NumElementsVReg = buildConstantInt(
+        NumElems ? NumElems : 1, MIRBuilder, SpvTypeInt32, EmitIR);
----------------
Naghasan wrote:

> But examples like https://godbolt.org/z/zaTT7nTvr are already UB

This isn't UB at all, it is outside the core specs (C++ says 0 is not valid) but it is a valid compiler extension (add `-pedantic` and you get a `warning: zero size arrays are an extension [-Wzero-length-array]`). And `0` as first offset makes it perfectly fine as well (you don't move the pointer). And in other languages, there is valid reasons to have this construct.

> BTW, @Naghasan do you happen to know, why OpTypeRuntimeArray is not allowed for Kernel? It's not looking like VLA to me.

This isn't linked to VLA, but a Vulkan runtime construct that isn't mappable in OpenCL

https://github.com/llvm/llvm-project/pull/149522