... | ... | @@ -500,6 +500,16 @@ |
|
|
stg to really verify these thunks.
|
|
|
|
|
|
*** The Fix
|
|
|
- Note taken on [2021-07-25 Sun 17:09] \\
|
|
|
Adding some strictness produced the following ticky profile:
|
|
|
#+begin_src shell
|
|
|
17371793 1949818584 0 4 iEMM $saboveNest{v r4Uk} (GHC.Utils.Ppr) (fun)
|
|
|
13224056 1116014480 0 3 MEM beside{v r14Y} (GHC.Utils.Ppr) (fun)
|
|
|
11442227 691716208 0 3 i.M Data.IntMap.Internal.$winsert{v rg6b} (fun)
|
|
|
1163713 280712792 0 5 SMMSM GHC.Core.Opt.Simplify.simplExpr2{v roJf} (fun)
|
|
|
#+end_src
|
|
|
Still need to check time with nofib
|
|
|
|
|
|
- Note taken on [2021-07-25 Sun 16:43] \\
|
|
|
Update. I built a stage 2 profiled compiler with the previously mentioned patch
|
|
|
but the build went into an infinite loop. I've narrowed down the cause to adding
|
... | ... | |