Skip to content
Snippets Groups Projects

Repository graph

You can move around the graph by using the arrow keys.
Select Git revision
  • 120-implementation-issue-with-alignment
  • adjust-ltemp-fix-scalars
  • admissible-inames-finder-avoid-pulling-in-extraneous-constraints
  • arguments-as-basestorage
  • bad_dessign
  • barrier-insertion-use-self-dependencies
  • barrier-insertion-with-finer-dependency-tracking
  • better-offset-treatment
  • c_pragma
  • c_vec_tag
  • cache-improvements
  • cache-invoker-and-type-inf
  • cl-function-descriptions
  • count-function-fix
  • count-granularity2
  • cpu-mutilevel-scan
  • deeply_bureaucratic_transform_addressing
  • do-not-uniquify-in-executor
  • dont-approximate-hw-loop-lbound
  • elif-else
  • v2017.2
  • v2017.1
  • v2016.2
  • v2016.1.1
  • v2016.1
25 results
Created with Raphaël 2.2.01Apr31Mar302925222116151413121093129Feb28272625242318121187630Jan292826252423222119181514716Dec872130Nov2928272521426Oct22181716141398130Sep232221201714138128Aug272423212019171312111097228Jul262524222018151412976532130Jun292827262524232221201917161211105428May2524231413121173230Apr242319181613129625Mar2221141086428Feb242220191817161513121098653226Jan252019Accept MatchExpressionBase in parse_stack_match()Fix schedule issue reportingloopy.{context_matching->match}, prettier programmatic matchesFix, test 'from loopy import *'Add InstructionBase.assignee_name[CL backend] Fix enabling of fp64 extensionAnnotate vector dtypes with targetAllow running CLI as python -m loopyDeal with Fortran 'real' typeSwitch 'breadth' tests to use POCLDoc reference fixesAdd atomic ops to tutorialAllow atomics on local variablesIntroduce a notion of 'scope' for temporary variables, in preparation for global temporariesSkip int64 test on old PyOpenCLType system fixesSupport generating atomic operations (at least for OpenCL 1.x)Fix tracing after type system apocalypseFix pow() code generation for new type systemFix another bare-numpy-type escapeImprove distinguishability between numpy dtypes and loopy typesSome missing NumpyDtype wrappingMore type system fixery regarding vectorsMark atomics unvectorizableBump data model to void broken cachesFix image data fetch in brave new type systemSome missing 'NumpyType' wrappers in nbodyFix types in stats gatheringMore numpy type system falloutFix mis-implemented __ne__Enable persistent hashing for loopy typesEnable/define language representation of atomic opsRemove debug printMerge remote-tracking branch 'upstream/master'replaced operation tuples with TypedOp classadded StridedGmemAccess class, now used as key in gmem access dictsnow calculating strides greater than 1For 1D arrays, allow undetermined shape in shape determinationAdd Python sparse matvec examplesubscript counter only looking for lid0 now, if not found, setting stride to maxsize