-
Notifications
You must be signed in to change notification settings - Fork 18k
cmd/compile: performance regression on SSA #15934
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Comments
This looks like a problem with placing spills conservatively. Quoting from a comment I put in regalloc.go:
In this case, the I'm surprised the old compiler does this optimization. |
See also #15845 , the fixes for these two bugs may be related. |
It seems part of the problem then is related to #14758 as well. However, if you remove the |
Weird. I see no substantive differences in the assembly with go1.6 and tip after I take the append out. |
...and I get only ~4% slower with tip. |
I'll close this down since parts of the problem are captured in other issues. I'll reopen if I can produce a more succinct reproduction of a slow down. |
Using
go1.7beta1
Consider the following benchmark:
When running without SSA:
When running with SSA:
This is a 25% decrease in performance and might be worth investigating why SSA is doing worse.
/cc @randall77
The text was updated successfully, but these errors were encountered: