[llvm] [AMDGPU][NewPM] Port "AMDGPUResourceUsageAnalysis" to NPM (PR #130959)

[via llvm-commits]

================
@@ -93,6 +93,13 @@ FUNCTION_PASS_WITH_PARAMS(
     parseAMDGPUAtomicOptimizerStrategy, "strategy=dpp|iterative|none")
 #undef FUNCTION_PASS_WITH_PARAMS
 
+#ifndef MACHINE_FUNCTION_ANALYSIS
+#define MACHINE_FUNCTION_ANALYSIS(NAME, CREATE_PASS)
+#endif
+
+MACHINE_FUNCTION_ANALYSIS("amdgpu-resource-usage", AMDGPUResourceUsageAnalysis(static_cast<const GCNTargetMachine *>(this)))
+#undef MACHINE_FUNCTION_ANALYSIS
----------------
paperchalice wrote:

Constructor uses `const TargetMachine*`.

https://github.com/llvm/llvm-project/pull/130959