[llvm-dev] Dumping compiled module as LLVM IR bitcode after certain optimization passes

[Alex Susu via llvm-dev]

Hello.
Is there a possibility to instruct at command line the clang or the opt tools to emit
LLVM bitcode file after a certain optimization pass such as LoopVectorize (or all
optimization passes)? (This question was asked also at
http://stackoverflow.com/questions/17904440/dump-ir-after-each-llvm-optimization-each-pass-both-llvm-ir-passes-and-backen
.)
As far as I can see giving:
clang -save-temps ...
generates a .bc file for the compiled module just before passing the module to the
optimization passes (opt).

An alternative I have is to programmatically generate the corresponding bitcode file
at the right time after finishing the pass in discussion - see, for example,
http://stackoverflow.com/questions/13903686/writing-module-to-bc-bitcode-file, methods
WriteBitcodeToFile() or WriteModule() .

Thank you,
Alex
_______________________________________________
LLVM Developers mailing list
llvm...@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev

[Mehdi Amini via llvm-dev]

Hi,

Usually we disable the llvm optimizations (clang -mllvm disable-llvm-optzns ) and then use opt to run the passes till the point of interest. I don’t know of a shortcut in clang (the new PassManager will make it possible to pass an arbitrary pipeline with a string, but not there yet).

—
Mehdi

[Mehdi Amini via llvm-dev]

Also, if I have a small test-case, using `clang -mllvm -print-after-all` is convenient to get the IR at any point.
Adding "-filter-print-funcs” can help reducing the amount in the output to a single function (or a list of functions), for example:
`clang -mllvm -filter-print-funcs=_Z3foov -mllvm -print-after-all`.

(There is a print-after option but it does not work in clang).

—
Mehdi