On Jun 9, 2020, Thomas Schwinge <tho...@codesourcery.com> wrote: > Are you able to easily create/suggest patches for these? (You're > probably not set up for offloading compilation...)
I can try, but I can certainly use help, if not in coding, at least with testing. > Can you suggest > how/where to adjust: producer-side (GCC driver, 'mkoffload's?), or > consumer-side (testsuite: offload tree scanning machinery etc., or have > to put some '-dumpbase' or similar into all such test case files?)? Given that files we used to output have now moved to /tmp, I suspect we're failing to pass -dumpbase at some point, so the offloading compiler ends up naming dump files after the temporary outputs rather than as per the requested/expected names inferred from the main compilation. I guess mkoffload might need some tweaking; I see pieces of lto-wrapper to call it, and the offload compiler, that I suspected might require tweaking for the dumpbase revamp as well. Let's see how far I can get by just looking at the code ;-) If I were to get something like a -v compile and link session, from someone all set for offloading compilation, with the command line passed to lto-wrapper and the full commands it runs, I might be able to figure out the dynamics more readily (hint, hint ;-) Thanks in advance, -- Alexandre Oliva, freedom fighter he/him https://FSFLA.org/blogs/lxo/ Free Software Evangelist Stallman was right, but he's left :( GNU Toolchain Engineer Live long and free, and prosper ethically
