A GCC Parallelization Bottleneck Might Get Addressed This Summer
A student is proposing parallelization improvements to the GCC code compiler this year as part of Google's Summer of Code initiative.
Currently GCC suffers from a bottleneck when trying to deal with huge source files that can approach hundreds of thousands of lines of code. While not too common, there are large source files in the GCC code-base itself that run into this bottleneck even for 128 core build servers.
Student developer Giuliano Belinassi is hoping to work on addressing this parallelization bottleneck in the GCC build process by reworking some GIMPLE passes and being able to parallelize the expand_all_functions behavior. The technical plans for those interested can be found via the proposal, which has also been brought up on their mailing list.
Given today's rising core counts, we are certainly exciting over the prospects of any parallelization improvements to common software such as GCC. This project though will still need to await review and approval by Google before it can be potentially green-lit for GSoC 2019 this summer.
Currently GCC suffers from a bottleneck when trying to deal with huge source files that can approach hundreds of thousands of lines of code. While not too common, there are large source files in the GCC code-base itself that run into this bottleneck even for 128 core build servers.
Student developer Giuliano Belinassi is hoping to work on addressing this parallelization bottleneck in the GCC build process by reworking some GIMPLE passes and being able to parallelize the expand_all_functions behavior. The technical plans for those interested can be found via the proposal, which has also been brought up on their mailing list.
Given today's rising core counts, we are certainly exciting over the prospects of any parallelization improvements to common software such as GCC. This project though will still need to await review and approval by Google before it can be potentially green-lit for GSoC 2019 this summer.
1 Comment