Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Address review comments, add more tests, fixup siphash
this is just the old implementation of siphash. with some tweaks to make it more streamable (still not great but good enough for benchmarks). -- squash log Use capi if available for offset Add more capi calls Flip expectation and actual args it showed up wrongly in the test suite Add more capi, add TODO Revert Hashable/Class to upstream master trying to contain changes to just that module Fix regression (once more) Don't expose the old functions Fix capi issues. Apprantly half of these functions weren't defined in the header. Tell people to use sse instead of asking Apply clang format Add siphash to benchmarks Add initializeState function, this will capture initizlie don't know how to do finalize yet Make it compile again (apparntly k1 and salt were used interchanbly) that's not a good idea Clean out the c code a bit. It was really strange. Make it build.. again? re-add header file I guess this will just keep on bein inconsistent remove cbits *.h from benchamark cabal file OMG that worked, lazy text now bytearray man, I doubted myself every step along the way, maybe I can do this? Draft for doing bytestring Make some headway on the regession tests Make testsuite pass once more Put in rounds as an argument Add compression round as args Add some common sense tests on previous state fails of course, not sure what's happening here Import <$> I guess different chars can also produce same hashes. Cleanup C-api for a bit it's still utterly broken however. I guess I need to try it out in C and see why the state isn't changing. Chunck -> Chunk Get rid of more c code idk maybe this do something with the sse stuff (after reading ghci comment) Figured out why no compression (lol) Update 64bit text once more Better explain 0 Split lines Lazy text can be big as well don't funroll by hand, add comments explaing what's going on Use properties instead of hardcoded test cases for statefullness Add prefix based tests as well hmm not so ez for text this also makes me doubt the bytestring implementaiton. I should check if with larger arbitrary instances that doesn't fall over (text tends to have larger bytestring representations then bytestring) Guess we really did solve the bytestring implementation This should pass CI
- Loading branch information