Opened 2 years ago

Closed 2 years ago

Last modified 2 years ago

#14240 closed bug (invalid)

CSE’ing w/w’ed code regresses program runtime

Reported by: nomeata Owned by: nomeata
Priority: normal Milestone:
Component: Compiler Version: 8.2.1
Keywords: Cc:
Operating System: Unknown/Multiple Architecture: Unknown/Multiple
Type of failure: Runtime performance bug Test Case:
Blocked By: Blocking:
Related Tickets: Differential Rev(s):
Wiki Page:

Description

Commit changeset:fe04f378/ghc, which fixed #14186, causes two runtime regressions:

nofib/time/integer 1.529 	+ 4.19% 1.593 	seconds
nofib/time/lambda  1.041 	+ 7.3%  1.117 	seconds

Sigh. I guess it up to me to track that one down.

Change History (3)

comment:1 Changed 2 years ago by nomeata

Owner: set to nomeata

comment:2 Changed 2 years ago by simonpj

Wow that is surprising. I wonder if there is any allocation change, or only runtime...

comment:3 Changed 2 years ago by nomeata

Resolution: invalid
Status: newclosed

changeset:28a115e/ghc “fixed” the lambda regression. I conclude that this is a layout issue…

I really wish we could get a better handle on these things. Whenever the RTS is touched, even by completely minuscule changes, a few benchmarks go up or down by 3 to 5 runtime.

Also, there are no Core changes in the affected benchmarks (did not test the libraries, though).

Version 1, edited 2 years ago by nomeata (previous) (next) (diff)
Note: See TracTickets for help on using tickets.