#8033: add AVX register support to llvm calling convention -------------------------------------+------------------------------------ Reporter: carter | Owner: carter Type: task | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.7 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Unknown/Multiple Type of failure: None/Unknown | Difficulty: Unknown Test Case: | Blocked By: Blocking: | Related Tickets: -------------------------------------+------------------------------------ Comment (by gmainland): I think the optimal solution would be to have Float and Double arguments passed in registers on X86_32. Per my comment on [http://www.haskell.org/pipermail/ghc-devs/2013-July/001648.html ghc- devs], we want code compiled with the LLVM back-end to be able to inter- operate with code compiled by the native back-end. Since the native codegen passes all Float and Double parameters on the stack, we would need to change the behavior of the native codegen too---it's not enough to change the LLVM calling convention. I don't think this would be all that much work. If we really care about 32-bit performance, it's the right thing to do. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/8033#comment:16 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler