This project is mirrored from https://github.com/haskell/hoopl. Pull mirroring updated .
  1. 08 Oct, 2018 2 commits
  2. 03 Jun, 2018 1 commit
  3. 13 Mar, 2018 1 commit
  4. 11 Feb, 2017 4 commits
  5. 07 Feb, 2017 1 commit
  6. 04 Feb, 2017 1 commit
  7. 28 Jan, 2017 1 commit
  8. 26 Dec, 2016 1 commit
  9. 25 Dec, 2016 1 commit
  10. 15 Nov, 2016 1 commit
  11. 06 Aug, 2016 1 commit
  12. 28 Jul, 2016 1 commit
  13. 04 Apr, 2016 1 commit
  14. 30 Mar, 2016 1 commit
  15. 01 Feb, 2016 2 commits
  16. 18 Jan, 2016 5 commits
  17. 21 Dec, 2015 1 commit
  18. 06 Dec, 2015 1 commit
  19. 03 Dec, 2015 2 commits
  20. 01 Nov, 2015 1 commit
  21. 31 Oct, 2015 2 commits
  22. 17 Oct, 2015 1 commit
  23. 09 Aug, 2015 1 commit
  24. 08 Aug, 2015 3 commits
  25. 12 Jul, 2015 1 commit
    • Michal Terepeta's avatar
      test/Simplify: don't crash on divide by zero · 4dccb474
      Michal Terepeta authored
      Currently test/Simplify would not check for 0 before trying to do
      constant folding, which is quite bad since it simply crashes the
      program.
      
      The fix is to simply avoid evaluating the expression if there is a
      division by zero.
      4dccb474
  26. 05 Jul, 2015 1 commit
    • Michal Terepeta's avatar
      Fix dead code elimination used in tests · 782ffec3
      Michal Terepeta authored
      The helper functions to fold over instructions and expressions in
      test/OptSupport have a bug where they would not recurse for complex
      expressions. And since testing/Live module depends on it, complex
      expressions would not be fully analyzed, leading to removal of code
      that is *not* dead, e.g.,
        var1 = m[0]
        var1 = m[var1] + m[var1]
      would get rewritten to just
        var1 = m[var1] + m[var1]
      which is clearly wrong.
      
      Note that this bug affects only Hoopl's tests.
      782ffec3
  27. 30 Jun, 2015 1 commit
    • Michal Terepeta's avatar
      Fix parsing of memory accesses in tests · 19526cc2
      Michal Terepeta authored
      Apparently the current parser would fail on any memory accesses, both
      loads and stores. In case of the former it's enough to try parsing the
      memory accesses before variables. For stores, we can use backtracking
      so that we don't consume the initial "m" of memory assignemnt thinking
      that it's a variable identifier.
      19526cc2