Help with optimization/profiling

To look at core I like to put this at the top of the file:

{-# OPTIONS_GHC -ddump-simpl -dsuppress-all -dno-suppress-type-signatures -dno-typeable-binds -ddump-to-file #-}

Afterwards I do a find . -name "*.dump-simpl" which shows that on my machine it generated the dump at this location:

% find . -name "*.dump-simpl"
./dist-newstyle/build/aarch64-osx/ghc-9.10.3/hs-plugui-1.0.1/x/main/build/main/main-tmp/app/Main.dump-simpl

That file contains the Core dump which is the intermediate language that GHC has applied most of its optimizations to.

I do notice that it doesn’t give much extra information in this case. There is quite some aggressive inlining which you can work around by adding {-# NOINLINE #-} to imageBytes, generateImage and encodePalettedPng, which doesn’t really change the running time.

In this case I think there’s not much to win from looking atimageBytes because it is basically just calling two library functions.

So optimizing this further would basically mean optimizing the underlying libraries.