[Mlir-commits] [mlir] [mlir][gpu] Deprecate gpu::Serialization* passes. (PR #65857)

[Guray Ozen]

================
@@ -78,11 +78,13 @@ void mlir::sparse_tensor::buildSparseCompiler(
 
   // Finalize GPU code generation.
   if (gpuCodegen) {
-#if MLIR_GPU_TO_CUBIN_PASS_ENABLE
-    pm.addNestedPass<gpu::GPUModuleOp>(createGpuSerializeToCubinPass(
-        options.gpuTriple, options.gpuChip, options.gpuFeatures));
-#endif
+    GpuNVVMAttachTargetOptions nvvmTargetOptions;
+    nvvmTargetOptions.triple = options.gpuTriple;
+    nvvmTargetOptions.chip = options.gpuChip;
+    nvvmTargetOptions.features = options.gpuFeatures;
+    pm.addPass(createGpuNVVMAttachTarget(nvvmTargetOptions));
     pm.addPass(createGpuToLLVMConversionPass());
+    pm.addPass(createGpuModuleToBinaryPass());
----------------
grypp wrote:

> So, is the issue with cuMemAlloc(&ptr, sizeBytes) for zero size known? 

Yes this is a known failure. 

> And should we just do something about that in the library wrapper?

Let's put an if statement that checks zero size in the runtime. 

https://github.com/llvm/llvm-project/pull/65857