Fix benchmark build #196

sjakobi · 2019-12-20T06:25:13Z

Based on work by @hsyl20. (Thanks!)

Fixes #52.

TODO:

~~Fix deprecation warnings.~~

… which currently fails with BenchAll.hs:173:55: No instance for (Random Word8) arising from a use of `random' Possible fix: add an instance declaration for (Random Word8) In the second argument of `(.)', namely `random' In the second argument of `S.unfoldrN', namely `(Just . random)' In the second argument of `(.)', namely `S.unfoldrN 10000 (Just . random)'

blaze-builder-0.3.* is incompatible with GHC >= 8.4 With --allow-newer it fails to build with: /tmp/stack26342/blaze-builder-0.3.3.4/Blaze/ByteString/Builder/Internal/Types.hs:71:10: error: • No instance for (Semigroup Builder) arising from the superclasses of an instance declaration • In the instance declaration for ‘Monoid Builder’ | 71 | instance Monoid Builder where | ^^^^^^^^^^^^^^

sjakobi · 2019-12-20T06:29:17Z

At this point benchmarks build with GHC 7.2 up and including to 8.2.

For GHC >= 8.4 there is no build plan due to incompatibility with blaze-builder-0.3.*.

bytestring/bench/bench-bytestring.cabal

Lines 43 to 46 in 95fe6bd

    
             -- There is no point using blaze-builder-0.4+ as that 
        
             -- version is just a wrapper around bytestring itself, 
        
             -- abd so a comparison benchmark would have no purpose. 
        
           , blaze-builder == 0.3.*

I'm not yet sure how to best resolve that.

bench/bench-bytestring.cabal

sjakobi · 2019-12-20T08:02:19Z

bench-bytestring-builder produces a bunch of deprecation warnings for Data.ByteString.findSubstrings:

bytestring/bench/BenchAll.hs

Lines 282 to 354 in 95fe6bd

    
           , bgroup "substrings" 
        
             [ bgroup "easy" 
        
               [ bench "easy1"    . nf (uncurry S.findSubstrings) 
        
                                 $ easySubstrings 1 1000000 
        
               , bench "easy4"    . nf (uncurry S.findSubstrings) 
        
                                 $ easySubstrings 4 1000000 
        
               , bench "easy16"   . nf (uncurry S.findSubstrings) 
        
                                 $ easySubstrings 16 1000000 
        
               , bench "easy64"   . nf (uncurry S.findSubstrings) 
        
                                 $ easySubstrings 64 1000000 
        
               , bench "easy128"  . nf (uncurry S.findSubstrings) 
        
                                 $ easySubstrings 128 1000000 
        
               , bench "easy1024" . nf (uncurry S.findSubstrings) 
        
                                 $ easySubstrings 1024 1000000 
        
               ] 
        
             , bgroup "random" 
        
               [ bench "random1"    . nf (uncurry S.findSubstrings) 
        
                                 $ randomSubstrings 1 1000000 
        
               , bench "random4"    . nf (uncurry S.findSubstrings) 
        
                                 $ randomSubstrings 4 1000000 
        
               , bench "random16"   . nf (uncurry S.findSubstrings) 
        
                                 $ randomSubstrings 16 1000000 
        
               , bench "random64"   . nf (uncurry S.findSubstrings) 
        
                                 $ randomSubstrings 64 1000000 
        
               , bench "random128"  . nf (uncurry S.findSubstrings) 
        
                                 $ randomSubstrings 128 1000000 
        
               , bench "random1024" . nf (uncurry S.findSubstrings) 
        
                                 $ randomSubstrings 1024 1000000 
        
               ] 
        
             , bgroup "hard" 
        
               [ bench "hard1"    . nf (uncurry S.findSubstrings) 
        
                                 $ hardSubstrings 1 1000000 
        
               , bench "hard4"    . nf (uncurry S.findSubstrings) 
        
                                 $ hardSubstrings 4 1000000 
        
               , bench "hard16"   . nf (uncurry S.findSubstrings) 
        
                                 $ hardSubstrings 16 1000000 
        
               , bench "hard64"   . nf (uncurry S.findSubstrings) 
        
                                 $ hardSubstrings 64 1000000 
        
               , bench "hard128"  . nf (uncurry S.findSubstrings) 
        
                                 $ hardSubstrings 128 1000000 
        
               , bench "hard1024" . nf (uncurry S.findSubstrings) 
        
                                 $ hardSubstrings 1024 1000000 
        
               ] 
        
             , bgroup "pathological" 
        
               [ bench "pathological1"    . nf (uncurry S.findSubstrings) 
        
                                 $ pathologicalSubstrings 1 1000000 
        
               , bench "pathological4"    . nf (uncurry S.findSubstrings) 
        
                                 $ pathologicalSubstrings 4 1000000 
        
               , bench "pathological16"   . nf (uncurry S.findSubstrings) 
        
                                 $ pathologicalSubstrings 16 1000000 
        
               , bench "pathological64"   . nf (uncurry S.findSubstrings) 
        
                                 $ pathologicalSubstrings 64 1000000 
        
               , bench "pathological128"  . nf (uncurry S.findSubstrings) 
        
                                 $ pathologicalSubstrings 128 1000000 
        
               , bench "pathological1024" . nf (uncurry S.findSubstrings) 
        
                                 $ pathologicalSubstrings 1024 1000000 
        
               ] 
        
             , bgroup "html" 
        
               [ bench "html1"    . nfIO . fmap (uncurry S.findSubstrings) 
        
                                 $ htmlSubstrings wikiPage 1 1000000 
        
               , bench "html4"    . nfIO . fmap (uncurry S.findSubstrings) 
        
                                 $ htmlSubstrings wikiPage 4 1000000 
        
               , bench "html16"   . nfIO . fmap (uncurry S.findSubstrings) 
        
                                 $ htmlSubstrings wikiPage 16 1000000 
        
               , bench "html64"   . nfIO . fmap (uncurry S.findSubstrings) 
        
                                 $ htmlSubstrings wikiPage 64 1000000 
        
               , bench "html128"  . nfIO . fmap (uncurry S.findSubstrings) 
        
                                 $ htmlSubstrings wikiPage 128 1000000 
        
               , bench "html1024" . nfIO . fmap (uncurry S.findSubstrings) 
        
                                 $ htmlSubstrings wikiPage 1024 1000000 
        
               ] 
        
             ]

#181 removes findSubstrings and these benchmarks, so I'd suggest not fixing the warnings for now.

sjakobi · 2019-12-20T08:05:24Z

bench-bytestring builds and runs with GHC 7.2 up to and including 8.8 now – please review! :)

hsyl20 · 2019-12-20T08:13:32Z

LGTM. Thanks!

sjakobi · 2019-12-22T16:14:38Z

bench/bench-bytestring.cabal

+  if impl(ghc >= 8.4)
+     build-depends: blaze-builder >= 0.3 && < 0.5
+  else
+     build-depends: blaze-builder == 0.3.*


On second thought, it seemed suboptimal to disable so many benchmarks when building with GHC-8.4+. Instead, I have relaxed the version constraint on blaze-builder when building with those GHC versions.

The comparison with blaze-builder isn't going to be very useful when using GHC-8.4+, but it shouldn't hurt either.

cartazio · 2020-05-06T19:34:43Z

@sjakobi whats needed to make this good to merge? lets catsup on the deb :)

cartazio · 2020-05-06T19:34:51Z

*debt

sjakobi · 2020-05-06T19:38:18Z

@cartazio I believe this PR is ready to merge.

cartazio · 2020-05-06T19:40:43Z

Cool.

sjakobi · 2020-05-06T19:42:09Z

Many thanks, @cartazio! :)

hsyl20 and others added 3 commits December 20, 2019 07:22

Fix benchmarks (rebased)

b56dc7a

sjakobi commented Dec 20, 2019

View reviewed changes

bench/bench-bytestring.cabal Outdated Show resolved Hide resolved

Turn off building bench-bytestring-builder for GHC >= 8.4

bc8cc8b

sjakobi force-pushed the hsyl20-fix-benchmarks branch from bed3386 to bc8cc8b Compare December 20, 2019 07:49

sjakobi changed the title ~~WIP: Fix benchmark build~~ Fix benchmark build Dec 20, 2019

sjakobi added 2 commits December 22, 2019 16:38

Use CPP to make bench-bytestring-builder buildable with GHC-8.4+

cd9121a

Simply relax the bounds on blaze-builder for GHC >= 8.4

8cf327a

sjakobi commented Dec 22, 2019

View reviewed changes

sjakobi mentioned this pull request Feb 3, 2020

Implement count with SSE 4.2 and AVX2 #202

Merged

sjakobi mentioned this pull request May 4, 2020

∞-loop, space==bound in primMapByteStringBounded #204

Merged

cartazio merged commit 38ad056 into haskell:master May 6, 2020

sjakobi deleted the hsyl20-fix-benchmarks branch May 6, 2020 19:41

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix benchmark build #196

Fix benchmark build #196

sjakobi commented Dec 20, 2019 •

edited

Loading

sjakobi commented Dec 20, 2019

sjakobi commented Dec 20, 2019

sjakobi commented Dec 20, 2019

hsyl20 commented Dec 20, 2019

sjakobi Dec 22, 2019

cartazio commented May 6, 2020

cartazio commented May 6, 2020

sjakobi commented May 6, 2020

cartazio commented May 6, 2020

sjakobi commented May 6, 2020

Fix benchmark build #196

Fix benchmark build #196

Conversation

sjakobi commented Dec 20, 2019 • edited Loading

sjakobi commented Dec 20, 2019

sjakobi commented Dec 20, 2019

sjakobi commented Dec 20, 2019

hsyl20 commented Dec 20, 2019

sjakobi Dec 22, 2019

Choose a reason for hiding this comment

cartazio commented May 6, 2020

cartazio commented May 6, 2020

sjakobi commented May 6, 2020

cartazio commented May 6, 2020

sjakobi commented May 6, 2020

sjakobi commented Dec 20, 2019 •

edited

Loading