#9661: Branchless ==# is compiled to branchy code -------------------------------------+------------------------------------- Reporter: dfeuer | Owner: Type: feature | Status: new request | Milestone: Priority: normal | Version: 7.9 Component: Compiler | Keywords: Resolution: | Architecture: Unknown/Multiple Operating System: | Difficulty: Unknown Unknown/Multiple | Blocked By: Type of failure: Runtime | Related Tickets: performance bug | Test Case: | Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- Comment (by rwbarton): Replying to [comment:8 dfeuer]:
With the exception of some bottom-of-the-stack `base` code with module dependency nightmares, people who use such primops are usually trying to do something specific for performance reasons, and we should be careful not to interfere excessively if we can avoid doing so.
Your preferred `==#` can be implemented in terms of mine, I believe,
Everyone who uses `==` on `Int` is using `==#`, that was the point of my Haskell `f` example above. Let's not lose track of the big picture here. like this:
{{{#!hs bartonEq x y = case feuerEq x y of 1# -> 1# _ -> 0# }}}
Going the other way around, however, is impossible. If you prefer, we
can give `feuerEq` a different name so current code will continue to work the same. I don't understand what `bartonEq` or `feuerEq` is or what the difference between them is supposed to be. They look equal to me.
In any case, I would like to see some benchmarks showing that there are performance gains to be had in real programs from control over branchlessness before expending any effort on implementation.
That ship has long since sailed—the massive breaking primBool change was undertaken specifically to support this sort of thing.
No, it was undertaken to allow GHC to find better code to generate under some circumstances. It was not undertaken to give the user control over whether GHC produces branchless code or branching code. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/9661#comment:9 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler