Sebastian Graf

Error gutter shows up to the right of line numbers instead of to the left

3

I've been using commit https://github.com/kak-lsp/kak-lsp/commit/354b46e3cf56f0da35b444941a701ca4c1135aa8 (somewhere between v8 and v9) for some time and finally switched to v9 stable this week. But now I noticed that kak-lsp inserts its error...

unsafeCoerce# will be gone in GHC 8.12

3

Just a heads up: I realised through a broken build with a recent GHC HEAD that `basement` uses `unsafeCoerce#` (rather than `unsafeCoerce`). That primop will be gone as of 8.12,...

Some thoughts from the perspective of a Haskeller while reading https://github.com/Kindelia/HVM/blob/master/HOW.md

7

Some notes I took while reading your introduction to your exciting project from my perspective as someone familiar with performant GHC-Haskell code. Of interest to me is *how exactly everything...

"List Fold (Sequential)" Haskell benchmark isn't entirely faithful

6

I appreciate the work and thought you put into this project! It got me quite excited, to be honest; now I really want to know if/how interaction nets can actually...

ext-STG interpreter: Wrong output on NoFib's `bernoulli`

2

Here's a complete reproducer: ```hs import Data.Ratio import System.Environment import Control.Monad import Data.Char import Data.List hash :: String -> Int hash = foldl' (\acc c -> ord c + acc*31)...

StgState allocations dominate

5

Here's a profile of a simplified benchmark case of NoFib's `bernoulli` after #8 has been fixed: ``` COST CENTRE MODULE SRC %time %alloc lookupEnvSO Stg.Interpreter.Base lib/Stg/Interpreter/Base.hs:(631,1)-(649,21) 6.1 3.4 evalStackContinuation.\ Stg.Interpreter...

evalPrimOp is a bottle-neck due to linear search

3

While the implementation of, e.g., `...ByteArray.evalPrimOp` is rather direct and elegant at the moment ```hs newtype Name = BS8.ByteString evalPrimOp :: PrimOpEval -> Name -> [Atom] -> Type -> Maybe...

`temci report` should include (part of) error output, if the output format permits

3

[Velcom's output format](https://github.com/IPDSnelting/velcom/blob/main/docs/bench_repo_specification.md#output-format) allows the specification of `error` measurements. Temci's report functionality should make use of that. Currently, I have to read both the `exec` as well as the `report`...

[Here](https://temci.readthedocs.io/en/latest/temci_exec.html#number-of-runs) I see the following: > # Number of runs > >The number of runs per block is either fixed by the `runs` settings that apply or is between the...

I have a run block (description `build`) that has `run_config.runs=1`, but I still get output like the following: ``` Benchmark 5 to 10 times Benchmark 1 out of 5 to...