[bug #44555] "make" fails to use parallelism

Follow-up Comment #16, bug #44555 (project make): Hi! I am sad to say, I missed that this bug had been fixed at the time. Thank you so much for incorporating the fix! I can confirm that using make 4.2.1 fixes this problem for us, and we no longer have to carry this patch locally. Additionally,

[bug #44555] "make" fails to use parallelism

Update of bug #44555 (project make): Status:None => Fixed Assigned to:None => psmith Open/Closed:Open => Closed Fixed Release:

[bug #44555] "make" fails to use parallelism

Follow-up Comment #14, bug #44555 (project make): This has become off-topic for the current bug report and anyway it was discussed exhaustively back then; details are in mail archives. But in any case, whether you agree or disagree after reading, it's water well past the bridge by now. __

[bug #44555] "make" fails to use parallelism

Follow-up Comment #13, bug #44555 (project make): "The new feature allowing for pattern-rules...added a lot of complexity and had some performance impacts" Why then was it pushed, given that it changed a convention of file-order vs. best-fit order? ___

[bug #44555] "make" fails to use parallelism

Follow-up Comment #12, bug #44555 (project make): Bisecting between 3.81 and current Git HEAD will not be very productive. 3.81 was released almost 10 years ago. There have been many performance regressions, and many performance improvements over that time. It's likely that some of the regressi

[bug #44555] "make" fails to use parallelism

Follow-up Comment #11, bug #44555 (project make): Following up on Jeff's comment, there are multiple performance regressions after make 3.81. The fork/vfork is the most significant one, but not the only one. These perf regressions are serious problems and upgrading from 3.81 would still cost a sma

[bug #44555] "make" fails to use parallelism

Follow-up Comment #10, bug #44555 (project make): OK, I'll put it back for the next release. ___ Reply to this item at: ___ Message sent via/by Savannah

[bug #44555] "make" fails to use parallelism

Follow-up Comment #9, bug #44555 (project make): While it doesn't account for the full 2.5+ minutes difference, "strace -c"'s summary of syscalls shows that vfork() in a good version of make is less than .02s, while fork() (show as clone()) in a bad version is about 75s (547us/call), a difference

[bug #44555] "make" fails to use parallelism

Follow-up Comment #8, bug #44555 (project make): I see absolutely the same results as noted. And there is NO DOUBT that the fork/vfork is to blame! I tested today on any single commit since 3.75(over 1000 versions!), and every single one of those versions up to 94735f0ad7f67c56afa1513381c73e8f62

[bug #44555] "make" fails to use parallelism

Follow-up Comment #7, bug #44555 (project make): (I am the original bug submitter and poster of the early anonymous comments) Is there additional information I can furnish to help in the resolution of this bug? Were you able to reproduce the problem with the script back in comment #2? _

[bug #44555] "make" fails to use parallelism

Follow-up Comment #6, bug #44555 (project make): I can confirm the bug report and that reverting 94735f0 solves the problem. For us this is a very critical problem, since build times are increased so drastically. ___ Reply to this item at:

[bug #44555] "make" fails to use parallelism

Follow-up Comment #5, bug #44555 (project make): Well, I don't want to move back to vfork() "just because it works", without understanding the situation better. The reasons we switched away from vfork() (portability and correctness) are still issues. I will need to investigate this more deeply t

[bug #44555] "make" fails to use parallelism

Follow-up Comment #4, bug #44555 (project make): We've now deployed our patched GNU make with the patch from comment #2 and the results are positive--the performance regression seen in GNU make 4.x is fixed. Is there any additional information I can provide to allow you to incorporate this fix in

[bug #44555] "make" fails to use parallelism

Follow-up Comment #3, bug #44555 (project make): To reply to a few points I didn't specifically address in my last comment: * make isn't actually dropping into single-job mode, as can be seen from "make --debug=j", but for some reason it's having trouble starting as many jobs as requested. * no,

[bug #44555] "make" fails to use parallelism

Follow-up Comment #2, bug #44555 (project make): I share your surprise that fork vs vfork could possibly have an important performance impact, but *it's what git bisect told me *testing immediately before the commit, I did not encounter the problem *testing at the tip of master with the commit rev

[bug #44555] "make" fails to use parallelism

Follow-up Comment #1, bug #44555 (project make): How do you know that it falls back to one job at a time? Does make actually print out the message it usually does when it gives up on parallelism? Or, do you just observe the build taking a long time and watch the commands being invoked, and see t

[bug #44555] "make" fails to use parallelism

URL: Summary: "make" fails to use parallelism Project: make Submitted by: None Submitted on: Mon 16 Mar 2015 07:33:02 PM UTC Severity: 3 - Normal Item Group: Bug