#10844: CallStack should not be inlined -------------------------------------+------------------------------------- Reporter: nomeata | Owner: gridaphobe Type: task | Status: patch Priority: normal | Milestone: Component: Compiler | Version: 7.10.2 Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: | Unknown/Multiple Type of failure: None/Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Phab:D1259 Wiki Page: | -------------------------------------+------------------------------------- Comment (by gridaphobe): Thanks, that makes sense. After adding the `cheapBuild` patch (only for string literals) I'm finally seeing a nice win on binary sizes. Unfortunately it comes at the cost of runtime performance, which seems like the wrong tradeoff; I'd rather have larger binaries if they run faster. Here's a selection of the more interesting summary results from nofib. I haven't had a chance to investigate where the performance loss is coming from. {{{ NoFib Results -------------------------------------------------------------------------------- Program Size Allocs Runtime Elapsed TotalMem -------------------------------------------------------------------------------- atom -1.0% 0.0% +0.9% +1.6% 0.0% binary-trees -0.9% -0.0% +9.4% +9.2% 0.0% bspt -1.0% -0.3% 0.004 0.013 -33.3% cacheprof -1.9% +2.0% -1.3% -1.8% +0.9% circsim -1.0% +0.0% +4.9% +5.1% 0.0% constraints -1.0% 0.0% +0.9% +1.1% 0.0% cryptarithm1 -1.0% 0.0% +6.5% +7.7% 0.0% cse -1.0% +1.4% 0.000 0.008 0.0% fannkuch-redux -1.0% 0.0% +3.0% +3.1% 0.0% fibheaps +12.2% -0.0% 0.017 0.025 0.0% fish -1.0% +3.1% 0.006 0.014 0.0% hidden -1.0% +0.0% +1.5% +1.7% 0.0% integer -1.0% 0.0% +1.7% +2.0% 0.0% k-nucleotide -0.9% -0.0% +2.7% +2.9% 0.0% lcss -1.0% 0.0% -0.9% -2.1% 0.0% power -0.9% +0.0% -4.1% -3.7% 0.0% scs -0.7% +0.1% +1.3% +0.8% 0.0% -------------------------------------------------------------------------------- Min -1.9% -2.0% -4.1% -3.7% -33.3% Max +12.2% +3.5% +9.4% +9.2% +0.9% Geometric Mean -0.9% +0.1% +1.4% +1.4% -0.4% }}} -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/10844#comment:17 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler