We already have information about static alloca stack locations in our
side table. Emitting instructions for them is inefficient, and it only
happens when the address of the alloca has been materialized within the
current block, which isn't often.
Details
Diff Detail
- Build Status
Buildable 8839 Build 8839: arc lint + arc unit
Event Timeline
Code LGTM, I have some comments for the testcases inline.
llvm/test/DebugInfo/WebAssembly/dbg-declare.ll | ||
---|---|---|
4 | When removing the check for the DEBUG_VALUE, could you add a check for the actual variable location being emitted via the MMI side table? llc -filetype=obj | llvm-dwarfdump - and a check for a DW_AT_name / DW_AT_location pair should do the trick. | |
llvm/test/DebugInfo/X86/dbg-declare-alloca.ll | ||
8 | I would prefer something that checks the dwarf or assembler output for an actual debug location to avoid future bitrot. | |
32 | We usually strip out all non-essential attributes. (Usually at least everything in quotes :-) | |
llvm/test/DebugInfo/X86/dbg-declare.ll | ||
7 | ditto here, checking that we still emit the location would be even better |
llvm/test/DebugInfo/WebAssembly/dbg-declare.ll | ||
---|---|---|
4 | llvm-dwarfdump doesn't appear to know how to parse wasm object files, so I can't add that check. I can add the DW_AT_location checks to the other tests, though. |
llvm/test/DebugInfo/WebAssembly/dbg-declare.ll | ||
---|---|---|
4 | Sorry, I overlooked that this was WebAssembly. Makes sense, thanks! |
When removing the check for the DEBUG_VALUE, could you add a check for the actual variable location being emitted via the MMI side table? llc -filetype=obj | llvm-dwarfdump - and a check for a DW_AT_name / DW_AT_location pair should do the trick.