teresajohnson wrote:
I just tracked down slowness running clang tests locally to this change. It's
very noticeable if I run a single test, or a single directory of tests.
When I ctrl-c the stack trace includes:
```
clang_repl_cmd = subprocess.run(
[clang_repl_exe, "-orc-runtime", "-oop-executor"],
```
I ran a single clang test 5 times each with current HEAD and with a "return
False" right at the start of have_host_out_of_process_jit_feature_support.
Here are the user time comparisons:
```
time.out.off: User time (seconds): 0.38
time.out.off: User time (seconds): 0.36
time.out.off: User time (seconds): 0.34
time.out.off: User time (seconds): 0.40
time.out.off: User time (seconds): 0.36
time.out.on: User time (seconds): 17.80
time.out.on: User time (seconds): 17.83
time.out.on: User time (seconds): 17.80
time.out.on: User time (seconds): 17.64
time.out.on: User time (seconds): 17.71
```
Can something be done to prevent this overhead?
https://github.com/llvm/llvm-project/pull/144064
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
