I have a working patch proposal to bug 2606. However because: 1) I'm relatively new at this for LLVM work. 2) The patch involves an area that affects all JIT scenarios. 3) The patch may conflict with design approaches and affect performance. 4) There is a lot of intense work going on right now to meet deadlines by the LLVM team, and not much time available for guidance. What is the best way to approach this. Should I submit the patch as a comment to 2606, or should I submit the patch here. Basically the patch involves forcing a stub emission for functions defined across module boundaries (without adding to pending), even though lazy compilation could be turned off. Among the changes, the lazy compilation check in JITCompilerFn(...) is disabled. I'm still adding my own unit tests, but I'd like to know how to approach such a discussion for changes to code areas that I'm not responsible for. Thanks in advance Garrison
Anton Korobeynikov
2010-Feb-14 20:01 UTC
[LLVMdev] Patch proposal submission process query
Hello, Garrison> What is the best way to approach this. Should I submit the patch as a comment > to 2606, or should I submit the patch here.If you feel that your patch will (possibly) touch many current use-cases, then the best way of doing the things would be the following: 1. Submit a design proposal here. 2. Discuss & collect the opinions until the process will converge 3. Implement & submit the patch for review (either here or to llvm-commits) -- With best regards, Anton Korobeynikov Faculty of Mathematics and Mechanics, Saint Petersburg State University