code_llvm: print source locations #22819
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
tkelman
reviewed
Jul 15, 2017
src/disasm.cpp
Outdated
| #include "debuginfo.h" | ||
|
|
||
| class LineNumberAnnotatedWriter : public AssemblyAnnotationWriter { | ||
| DILocation *InstrLoc = nullptr; |
tkelman
reviewed
Jul 15, 2017
src/disasm.cpp
Outdated
| @@ -494,4 +721,15 @@ void jl_dump_asm_internal(uintptr_t Fptr, size_t Fsize, int64_t slide, | |||
| DisInfo.createSymbols(); | |||
| } | |||
| } | |||
|
|
|||
| extern "C" JL_DLLEXPORT | |||
| LLVMDisasmContextRef jl_LLVMCreateDisasm(const char *TripleName, void *DisInfo, int TagType, LLVMOpInfoCallback GetOpInfo, LLVMSymbolLookupCallback SymbolLookUp) | |||
tkelman
reviewed
Jul 16, 2017
src/disasm.cpp
Outdated
|
|
||
| void addSubprogram(const Function *F, DISubprogram *SP) | ||
| { | ||
| Subprogram[F] = SP; |
Supplements the printing of LLVM IR with source location information
|
I may have gone slightly overboard in fixing Tony's latest review comment, and now print out all available line info. The purpose of this is so that we can see whether it looks right when making changes to the debugging info, and to help with matching up source code with emitted instructions: |
|
Can we have a test for that new feature? |
|
We don't have a good way to unit test these, and it depends on LLVM optimizations and upstream representations, so we've been reluctant to commit too much to the exact representation this function returns ( Line 3 in cdaa5a7 |
maleadt
approved these changes
Jul 24, 2017
| LinePrinter.emit_finish(Out); | ||
| } | ||
|
|
||
| // print an llvm IR acquired from jl_get_llvmf |
There was a problem hiding this comment.
You shouldn't approve the PR if you request changes... 😉
There was a problem hiding this comment.
I think print (as in -to-string) is sufficiently clear as a verb, in context here.
| // print an llvm IR acquired from jl_get_llvmf | ||
| // warning: this takes ownership of, and destroys, f->getParent() | ||
| extern "C" JL_DLLEXPORT | ||
| const jl_value_t *jl_dump_function_ir(void *f, bool strip_ir_metadata, bool dump_module) |
There was a problem hiding this comment.
maybe confine the NFC code movements to a separate commit next time, this is kinda annoying to review with GitHub's diff view.
| return 0; | ||
| } | ||
|
|
||
| // print a native disassembly for the function starting at fptr |
|
Huh, how did I miss I don't like the current DILineInfoPrinter format though, pretty verbose and the braces are hard to follow. Maybe some indentation and/or structure characters and/or frame numbers, eg. define i64 @julia_sum_61779(%jl_value_t addrspace(10)* dereferenceable(40)) #0 {
top:
; frame 0 (sum @ reduce.jl:363)
; └frame 1 (countnz @ reduce.jl:725)
; └frame 2 (count @ reduce.jl:701)
; └frame 3 (eachindex @ abstractarray.jl:763)
; └frame 4 (indices1 @ abstractarray.jl:73)
; └frame 5 (indices @ abstractarray.jl:66)
%1 = addrspacecast %jl_value_t addrspace(10)* %0 to %jl_value_t addrspace(11)*
%2 = bitcast %jl_value_t addrspace(11)* %1 to %jl_value_t addrspace(10)* addrspace(11)*
%3 = getelementptr %jl_value_t addrspace(10)*, %jl_value_t addrspace(10)* addrspace(11)* %2, i64 3
%4 = bitcast %jl_value_t addrspace(10)* addrspace(11)* %3 to i64 addrspace(11)*
%5 = load i64, i64 addrspace(11)* %4, align 8
; └end frame 5 (indices)
; └end frame 4 (indices1)
; └end frame 3 (eachindex)
%6 = icmp slt i64 %5, 1
br i1 %6, label %L26, label %if.lr.ph
if.lr.ph: ; preds = %top
; reduce.jl:702 (frame 2)
%7 = bitcast %jl_value_t addrspace(11)* %1 to i8* addrspace(11)*
%8 = load i8*, i8* addrspace(11)* %7, align 8 |
|
That looks nice, although we don't have the nice pretty syntax coloring 😛 OK to merge, and encourage further experimentation in future PRs? |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Supplements the printing of LLVM IR with source location information.
With inspiration from #19342