> > Fine business! I had someone reach out off list to do some testing on >> mips64 so maybe we should wait a bit to hear back. The diff is still OK >> with me. Thanks again. >> > > Took me a while to even get here, and "here" is "not finished yet". First > it bombed on 4G default datasize ulimit, so I bumped that, then I didn't > notice that crash left a 4G clang core file so next build failed due to > filled up pobj/, now I am where zig2 will not start, it immediate goes > ":out of memory" without any flags or so: > > On my build systems I usually have the data size at like 8GB, 12GB, 16GB > depending on > how much memory I have. From briefly watching the build at times 4GB will > definitely > not be a enough. I'm also not sure if the build once it gets past the > initial bootstrap > respects such a value of MAKE_JOBS. > I think ninja handled MAKE_JOBS nicely on my 16core box, its just so few sources before it goes into 4+G linking (and lto?) work that it's barely noticeable. I think that part spends some 30-40 minutes hogging a single core on the 1.6GHz cpu, whereas the 6-7 sources before zig1 take at most a few minutes to pass by, serially or in parallel.
-- May the most significant bit of your life be positive.
