opt-in-boostable-into to main · Andreas Klöckner / loopy · GitLab

Commits on Source 4556

0494c442 replaced consec, uniform, and nonconsec subscript types with DataAccess class
Mar 12, 2016
09420000 subscript counter only looking for lid0 now, if not found, setting stride to maxsize
Mar 13, 2016
2dbeb387 now calculating strides greater than 1
Mar 15, 2016
0573d545 added StridedGmemAccess class, now used as key in gmem access dicts
Mar 15, 2016
c0dcf557 replaced operation tuples with TypedOp class
Mar 15, 2016
0f2a2519 Merge remote-tracking branch 'upstream/master'
Mar 16, 2016
b3d9498c temporary fix for stride counting when min tag axis >0, and added variable...
Apr 01, 2016
138c0b1e added variable name to StridedGmemAccess (still work to do)
Apr 01, 2016
e16469d7 adding local mem access counter
Apr 01, 2016
84163f0c Implement a generator for duplication options in case of non-schedulable kernels
Jul 12, 2016
45113adc Simplify removal of common inames
Jul 13, 2016
1cd034c4 [bugfix] Reintroduce common inames in the result
Jul 13, 2016
5afe7de9 Add additional test for duplication options
Jul 13, 2016
0a426c35 [bugfix] Carry all common inames through recursion correctly
Jul 25, 2016
9c83b608 Work in feedback from code review
Jul 27, 2016
bc1e1d10 Merge branch 'master' of github.com:dokempf/loopy
Jul 27, 2016
51ab3064 PEP8 fixes
Jul 27, 2016
5645e1b9 Minor doc tweak
Jul 27, 2016
28aa4305 needs_iname_duplication -> has_schedulable_iname_nesting
Jul 28, 2016
c0542007 Discard any parallel inames from the duplication options algorithm
Jul 28, 2016
a66fce04 Add a preschedule check for schedulable iname nesting!
Aug 01, 2016
ffe24c1b Add new functions to __all__ in __init__.py
Aug 01, 2016
0260cc7b Make generation of hints statement robust
Aug 01, 2016
ad35d2dd Correctly go into recursion when having common inames
Aug 01, 2016
c4794071 Exclude instructions without iname dependencies in dependency extraction
Aug 01, 2016
35e64eee pep8
Aug 01, 2016
d26918f4 Add 'boostable_into' field to insn deps
Aug 08, 2016
13a49768 Use boostable_into information only after iname duplication was deemed necessary without.
Aug 19, 2016
45bd98f8 Remove empty set from insn_deps (again)
Aug 19, 2016
35241a0a First discard parallel inames, then try use_boostable_into
Aug 19, 2016
56001eae Discard duplication options that do not affect an instruction
Aug 19, 2016
f1e4852c Emit a warning if falling into boostable_into branch
Aug 19, 2016
07fb77a1 Only build CentOS binaries for master [ci skip]
Aug 19, 2016
81e86ded Sort domain parameters on kernel creation for better determinism
Aug 27, 2016
0d069a65 commenting out debug print
Sep 01, 2016
8011a92f Fix loopy type handling in type inference
Sep 08, 2016
fc19d7c9 Make kernel execution a function of the target
Sep 14, 2016
94361fd7 Add loopy.compiled compatibility shim
Sep 14, 2016
6a0a8039 Stop using CompiledKernel.code in auto_test
Sep 14, 2016
42de8af3 Fix target replacement logic for kernel arg workarounds
Sep 14, 2016
87b722a3 Fix target execution prototype
Sep 14, 2016
b104125d Merge branch 'master' into detect-fix-unschedulable
Sep 14, 2016
fc14cb71 Restore library_for_test (oops) (!2 )
Sep 16, 2016
0616712d Reset kernel executor cache after unpickling
Sep 18, 2016
2edb8713 Fix SEM Laplacian test with detect-fix-unschedulable
Sep 21, 2016
3653cffa Merge branch 'master' into detect-fix-unschedulable
Sep 21, 2016
68b6e6d2 Merge branch 'detect-fix-unschedulable' into 'master'
Sep 21, 2016
4b6e9645 Fix kernel unpickling wrt _kernel_executor_cache
Sep 21, 2016
279b7949 Fix missing cases in make_options
Sep 28, 2016
5f4ff941 Make get_required_predicates succeed even for empty kernels
Sep 29, 2016
393a11cc Fix interaction between CSE expansion and iname duplication
Oct 04, 2016
dfecf6a7 Introduce kernel option to ignore boostable_into field in iname duplication
Oct 05, 2016
a3005e1f Some forced_iname_deps -> within_inames rename tracking
Oct 05, 2016
e374b76f Improve printing of instructions with id=None
Oct 05, 2016
aab1e81e Improve integration between tagged CSEs and (iname duplication and reductions)
Oct 05, 2016
66a0d06e Fix recursion in processing of 'if()' in FunctionToPrimitiveMapper
Oct 05, 2016
9c529e3c Add a regression test guarding against a cache collision
Oct 05, 2016
c912e553 fixing merge conflicts after summer16 changes
Oct 06, 2016
3cd46527 Merge remote-tracking branch 'upstream/master'
Oct 06, 2016
c0467e83 Add option for kernel creation with sequential dependencies
Oct 10, 2016
fea129c0 Clarify fusion docs
Oct 11, 2016
990422d4 Add bastardized LBM example from Loic
Oct 11, 2016
904cbe97 Improve documentation of add_prefetch
Oct 11, 2016
515f1ed8 ignore_boostable_into_field -> ignore_boostable_into
Oct 12, 2016
c86a05b2 Also use ignore_boostable_into in the scheduler
Oct 12, 2016
85b01330 Fix encoding in setup.py
Oct 12, 2016
1ca76c5a Trying to re-preprocess already-preprocessed kernel: don't complain, just return it
Oct 12, 2016
2ad06808 Improve an instruction creation error message
Oct 12, 2016
95db3151 Allow id_prefix,dep in with blocks
Oct 12, 2016
a452d2be Update LBM test
Oct 12, 2016
e5824356 Add FAQ about group dependencies
Oct 12, 2016
88d15686 Fix updating for InstructionBase instances passed into make_kernel
Oct 12, 2016
1a3e4259 TypedOp -> Op
Oct 15, 2016
fadc3518 merged StridedGmemAccess and LmemAccess into MemAccess, also reformatted long lines of code
Oct 15, 2016
eeb027a6 allowing for data types entered as np.float64 and converting to np.dtype(np.float64)
Oct 16, 2016
7ef4a9ac Merge remote-tracking branch 'upstream/master'
Oct 16, 2016
b58de0e7 Move instruction classes to their own file
Oct 17, 2016
9b1b1829 Add barrier and nop instructions
Oct 17, 2016
2db20d67 combined get_lmem_access_poly and get_gmem_access_poly into get_mem_access_poly
Oct 17, 2016
1a9b133b Merge remote-tracking branch 'upstream/master'
Oct 17, 2016
4d95281d Merge branch 'allow-ignoring-boostable-into' into 'master'
Oct 18, 2016
3a68e4de Improve documentation/interface of alias_temporaries
Oct 18, 2016
2e562728 Merge gitlab.tiker.net:inducer/loopy
Oct 18, 2016
f0c17f3a ilp in loopy.transform, not loopy.
Oct 19, 2016
9b4a8ffb Update compyte
Oct 19, 2016
e913af86 Fix with_transformed_expressions for special instructions
Oct 19, 2016
9c010eb6 Toward common subexpression elimination in subscripts: generate C (and related...
Oct 19, 2016
79793b19 Extract var shape guessing into a separate function
Oct 19, 2016
d000aa4c Merge porter:src/loopy
Oct 19, 2016
bf0409c0 Add missing future import
Oct 19, 2016
aa184b9a Fix codegen differences from switch to C ASTs
Oct 19, 2016
8c5365c9 Turn predicates into full-fledged expressions (Github issue #37)
Oct 19, 2016
ccfb7df2 Predicates are no longer strings: do not intern
Oct 19, 2016
8426f030 Fix class in super() for CInstruction.read_dependency_names
Oct 19, 2016
1f67ed05 read_dependency_names: do not attempt to modify frozenset
Oct 19, 2016
c8b2928f Tweak instruction imports
Oct 19, 2016
0de75a94 Take rhss into account, too, when figuring out temporary shapes
Oct 19, 2016
4d8b6013 Generalize guess_var_shape
Oct 19, 2016
a90e6ddd Add support for nops
Oct 19, 2016
48958263 Make temporary elimination aware of base_storage (Issue #36)
Oct 20, 2016

4,456 additional commits have been omitted to prevent performance issues.