#13960: Ticks exhausted with 8.0.2 -------------------------------------+------------------------------------- Reporter: tom-bop | Owner: (none) Type: bug | Status: infoneeded Priority: normal | Milestone: Component: Compiler | Version: 8.0.2 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by bgamari): Thanks tom-bop! What is happening here is that GHC is inlining the guts of `toStrict (toLazyByteString x)` into each call site of `fromString` (since this is what `bytestring` requests). In the case that we are calling `fromString` on a dynamically computed `String` this might make sense since we could possibly fuse the encoding logic and buffer write into the producer. However, in this particular case the `String` is a literal (that is, produced by `unpackCString#`). I really don't think there is any good reason to inline here. Moreover, in principle `Builder` should handle string literals with a simple `memcpy`. Unfortunately, it can't currently do this (easily) since string literals don't have an exposed size and we treat `'\0'` a bit funnily (encoding it as `'\xC0\x80'`, which the `memcpy` would need to undo). Ultimately I think the right solution would be to simply treat literals properly. -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/13960#comment:12 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler