How does ProGraML capture Loop optimization (loop unroll) information during graph generation? about programl

[ChrisCummins]

• Where the loop optimization information be included in LLVM IR, is it only in Metadata? I guess Yes after referring to LLVM-loop-unrolling.

That looks right to me.

• Does ProGraML use the Metadata information in LLVM IR to construct control flow, data flow, and call flow (as structural information)?

It doesn't automatically add metadata info to instructions. You would need to extend ProgramGraphBuilder::AddLlvmInstruction() to pull out any extra metadata you want and embed it in the graph as node features. Here is an example of how we do this for instruction profiling data:

https://github.com/ChrisCummins/ProGraML/blob/development/programl/ir/llvm/internal/program_graph_builder.cc#L307-L319

I'd happily accept a patch that also records loop unroll information, as I think that would be useful.

ProGraML/programl/ir/llvm/internal/program_graph_builder.cc

Lines 307 to 319 in cd7d293

	#if PROGRAML_LLVM_VERSION_MAJOR > 3
	// Add profiling information features, if available.
	uint64_t profTotalWeight;
	if (instruction->extractProfTotalWeight(profTotalWeight)) {
	graph::AddScalarFeature(node, "llvm_profile_total_weight", profTotalWeight);
	}
	uint64_t profTrueWeight;
	uint64_t profFalseWeight;
	if (instruction->extractProfMetadata(profTrueWeight, profFalseWeight)) {
	graph::AddScalarFeature(node, "llvm_profile_true_weight", profTrueWeight);
	graph::AddScalarFeature(node, "llvm_profile_false_weight", profFalseWeight);
	}
	#endif

Cheers,
Chris