#10394: LLVM mangler doesn't mangle AVX instructions -------------------------------------+------------------------------------- Reporter: dobenour | Owner: Type: bug | Status: new Priority: normal | Milestone: 7.12.1 Component: Compiler (LLVM) | Version: 7.11 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Type of failure: Runtime crash | Unknown/Multiple Blocked By: | Test Case: Related Tickets: | Blocking: | Differential Revisions: -------------------------------------+------------------------------------- Changes (by rwbarton): * cc: bgamari (added) Comment: In fact the CPP around the AVX rewriting stuff was broken anyways (should have been `#ifdef REWRITE_AVX` not `#if REWRITE_AVX`) so even if `x86_64_TARGET_ARCH` was set, the result would have been a CPP error. It looks like this never actually worked. Makes me wonder if we even need the AVX rewriting nowadays; or maybe we've told LLVM not to use these instructions, and could now allow it to; or maybe we can tell it not to assume the stack is 32-byte aligned. Your rewrite looks generally sensible, though I haven't examined it closely. cc'ing Ben since I think he did the conversion to use prefix data for tables-next-to-code. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/10394#comment:1 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler