[Codegen][CUDA] Remove handling for block ID ranges (#20009)

We have this old LLVM plugin running around for setting the maximum
value of block ID intrinsics when going to CUDA. This functionality has
been upstream since LLVM PR #107659 , and we've been setting the
relevant attributes in our compile pipeline.

compiler/plugins/target/CUDA/BUILD.bazel

@@ -21,10 +21,6 @@ iree_compiler_cc_library(
     name = "CUDA",
     srcs = [
         "CUDATarget.cpp",
-        "SetBlockIdsRangePass.cpp",
-    ],
-    hdrs = [
-        "SetBlockIdsRangePass.h",
     ],
     deps = [
         "//compiler/src/iree/compiler/Codegen",

compiler/plugins/target/CUDA/CMakeLists.txt

@@ -20,11 +20,8 @@ iree_compiler_register_plugin(
 iree_cc_library(
   NAME
     CUDA
-  HDRS
-    "SetBlockIdsRangePass.h"
   SRCS
     "CUDATarget.cpp"
-    "SetBlockIdsRangePass.cpp"
   DEPS
     LLVMAnalysis
     LLVMBitReader

compiler/plugins/target/CUDA/CUDATarget.cpp

@@ -4,7 +4,6 @@
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 
-#include "./SetBlockIdsRangePass.h"
 #include "iree/compiler/Codegen/Common/Passes.h"
 #include "iree/compiler/Codegen/Dialect/Codegen/IR/IREECodegenDialect.h"
 #include "iree/compiler/Codegen/Dialect/GPU/TargetUtils/KnownTargets.h"
@@ -369,7 +368,6 @@ static void optimizeModule(llvm::Module &module,
 
   mpm.addPass(llvm::VerifierPass());
   llvm::FunctionPassManager fpm;
-  fpm.addPass(llvm::SetBlockIdsRangePass(maxWorkgroupSize));
   mpm.addPass(createModuleToFunctionPassAdaptor(std::move(fpm)));
   mpm.addPass(pb.buildPerModuleDefaultPipeline(ol));
   mpm.addPass(llvm::VerifierPass());