Update error messages.

In D98447#2622788, @whchung wrote:

Also, in our downstream work, we did a rebase recently and we found we have to disable multi-threading when we run MLIR passes.
To emit HSACO for AMD GPU platform, lld is used, and it seems lld itself doesn't seem to be thread-safe.

Another fix.

In D98447#2622643, @whchung wrote:

@csigg May I understand how you configure the build?

With this cmake command:

cmake -G Ninja ../llvm \
    -DLLVM_ENABLE_PROJECTS="mlir;lld" \
    -DLLVM_BUILD_EXAMPLES=ON \
    -DLLVM_TARGETS_TO_BUILD="X86;AMDGPU" \
    -DCMAKE_BUILD_TYPE=Release \
    -DLLVM_ENABLE_ASSERTIONS=ON \
    -DBUILD_SHARED_LIBS=ON \
    -DLLVM_BUILD_LLVM_DYLIB=ON \
    -DMLIR_ROCM_RUNNER_ENABLED=1

Fix dependencies for libMLIR.so, fix tests.

Simplify dependencies.

In D98447#2622038, @whchung wrote:

@csigg May I understand the goal for this changeset is to merge everything under mlir-cpu-runner?

Another fix.

Various fixes.

Rebase.

Do not depend on D98396.

Polish.

Rebase.

Rebase.

Fix nit.

I'm wondering though, wouldn't it be cleaner to implement the copy c'tor for Pass::Option so that passes (like SerializeToBlobPass here) can use the default copy c'tor?

This change should be ready for review again.

Rebase, cmake fixes.

Fix.

Move LLVM dependencies.

Reduce public #includes.

Use base class added in D98279.

Change isa string to reference.

Fix file comment.

Remove accidental mlir-cuda-runner change.

Add options for triple, chip, features.

Rebase

In D98168#2614317, @bondhugula wrote:

In D98168#2613946, @csigg wrote:

Remove triple, target and feature options.

Could you include this in the commit summary please?

Address reviewer comments.

Fix.

Remove triple, target and feature options.

Move option to GpuToLLVMConversionPass (from Base) to remove the duplicate forward declaration of getDefaultGpuBinaryAnnotation().

Switch to context push/pop.

Simplify.

Needs C++14, not C++11.

Fix clang-tidy, hopefully.

Rebase.