Commits on Source
747
6e60b129
get_sub_group_info: 0x1020 -> 0x2010
Oct 31, 2020
e88b62a4
Merge pull request #394 from zachjweiner/fix-get_sub_group_info-version-check
Oct 31, 2020
4302d21e
Skip enqueue_fill_buffer for large arrays on NVIDIA
Nov 01, 2020
af14ee81
Directly link issue 395
Nov 01, 2020
32907a84
add large array test
Nov 01, 2020
84b9b7be
fix quotes
Nov 01, 2020
0aebd7bc
Don't exceed max_mem_alloc_size in test
Nov 01, 2020
128b33d5
Merge pull request #396 from zachjweiner/circumvent-nvidia-enqueue_fill_buffer-bug
Nov 01, 2020
f95b67ea
Fix doc reference to enqueue_copy in subst.rst
Nov 08, 2020
7342d7e1
Merge pull request #397 from inducer/fix-enqueue_copy-ref
Nov 08, 2020
b96ca61c
Array.__repr__: Make CL arrays visually distinguishable from numpy arrays
Nov 11, 2020
46005444
Merge pull request #398 from inducer/array-distinctive-repr
Nov 11, 2020
1c75f604
Add minor clarification to the instructions for getting the vendors dir under conda
Nov 17, 2020
4117be1b
Merge pull request #400 from zachjweiner/conda-pip-notes-in-install-doc
Nov 17, 2020
3ad1f427
Array: Handle no-queue in __str__, __repr__ more gracefully
Nov 18, 2020
859af50c
Improve no-queue printed representation: add shape, dtype
Nov 18, 2020
9e3512cb
Merge pull request #401 from inducer/array-str-repr-handle-no-queue
Nov 18, 2020
fd32e6ce
Bump version
Nov 18, 2020
a5c92a93
Fix MANIFEST.in
Nov 18, 2020
65353e5b
Handle contiguity and reshape for empty arrays like numpy
Nov 18, 2020
108b5071
Merge pull request #402 from inducer/empty-array-reshape-contig
Nov 18, 2020
31a96823
Skip rectangular copy tests on Apple
Nov 18, 2020
d12e39dd
Merge pull request #403 from inducer/skip-rectangular-copy-tests-on-apple
Nov 19, 2020
ff2f0ca7
Bump version
Nov 19, 2020
1772383c
Add arm64 support for Apple Silicon
Nov 21, 2020
33914e08
Merge branch 'geohot-master' into 'master'
Nov 22, 2020
18157c09
Merge pull request #405 from geohot/master
Nov 21, 2020
310a0da9
Merge branch 'master' of
ssh://github.com/inducer/pyopencl
Nov 21, 2020
478c2f88
Fix a build failure (kernel cloning introduced in 2.1, not 2.0)
Nov 27, 2020
3d7a1010
Merge pull request #408 from pocl/master
Nov 27, 2020
545ab4b2
Flake8 doc/conf
Dec 01, 2020
9ab1b582
Switch to furo doc theme
Dec 01, 2020
74a83022
Merge branch 'furo-doc-theme' into 'master'
Dec 02, 2020
7064465c
Stop using pytools.{any,all}
Dec 07, 2020
18229e3e
Merge branch 'stop-using-pytools-any-all' into 'master'
Dec 08, 2020
7f45ab37
Drop usage of six
Dec 19, 2020
f2f6f6fe
Merge pull request #412 from inducer/drop-six
Dec 19, 2020
54a91c25
Put a deadline on Ranlux deprecation
Dec 31, 2020
ecb964b2
Fix RNG event management
Dec 31, 2020
12f7c2ae
Fix .event -> .events typo
Dec 31, 2020
dd73975e
Remove another now-unnecessary add_event in clrandom.rand
Dec 31, 2020
986d2efd
Merge pull request #414 from inducer/fix-rng-evt-mgmt
Dec 31, 2020
31e1c91f
Set autoclass_content = "class", autodoc_typehints = "description", remove...
Jan 04, 2021
673912dc
Merge branch 'inducer-master-patch-80983' into 'master'
Jan 04, 2021
bed21942
Add ARRAY_KERNEL_EXEC_HOOK
Jan 07, 2021
b0d7fcac
Ignore not-callable pylint warning on ARRAY_KERNEL_EXEC_HOOK
Jan 08, 2021
8b00bfac
Add a note that ARRAY_KERNEL_EXEC_HOOK is undocumented
Jan 08, 2021
de5802eb
Merge pull request #415 from inducer/array-kernel-exec-hook
Jan 15, 2021
be9fd38f
Create special-case set_arg wrappers for common integer/float types
Jan 15, 2021
afaad630
Bump version
Jan 15, 2021
c9f439e0
Avoid malloc in enqueue_nd_range_kernel
Jan 15, 2021
c6aad43e
Privatize cl.array.{splay,Array.get_sizes}, cache _get_sizes results
Jan 15, 2021
36780052
Fix a few remaining invocations of splay
Jan 15, 2021
39d4fadb
Avoid a few calls to len() in enqueue_nd_range_kernel
Jan 15, 2021
cace1cc1
Actually pass a device to _splay on invocations outside pyopencl.array
Jan 15, 2021
ee989c8a
Revert "Create special-case set_arg wrappers for common integer/float types"
Jan 17, 2021
8c63f87f
Remove a pypy compatibility hack for checking for object-array-ness
Jan 17, 2021
05050cd2
Inline _make_strides in pyopencl.array
Jan 17, 2021
a8c0184d
Cache result of _get_common_dtype
Jan 17, 2021
eb47983a
Avoid generator expression in checking for negative shape entries
Jan 17, 2021
50705a96
Do not recompute flags in arrays made by _new_like_me
Jan 17, 2021
b7c052f9
Update compyte for equal_strides fast path
Jan 17, 2021
f1db98f9
Invoker gen: check for _KERNEL_ARG_CLASSES before None
Jan 17, 2021
66d0d25c
Make a kernel-specific class to override __call__ and avoid an indirect call on kernel enqueue
Jan 17, 2021
0e0683f6
Do not cache result in _get_common_dtype if obj2 is of type int
Jan 17, 2021
6f759230
Fix overriding of custom set_args in KernelWithOverriddenCall
Jan 17, 2021
75e460b7
Streamline invoker code (and its generation)
Jan 17, 2021
03e6e62f
Rename KernelWithOverriddenCall->KernelWithCustomEnqueue
Jan 17, 2021
7e9df904
Turn off generation of debug helper code in invoker with sys.flag.optimize
Jan 17, 2021
95b290f3
kernel_get_work_group_info: Use Device.int_ptr in cache key
Jan 17, 2021
17c47958
Rewrite first_arg_dependent_memoize without decorator package, drop dep
Jan 17, 2021
5eb3f3b9
Revert "Rewrite first_arg_dependent_memoize without decorator package, drop dep"
Jan 18, 2021
429ba485
Use positional args in _cl.enqueue_nd_range_kernel in invoker to avoid pybind slowness with kwargs
Jan 18, 2021
b8542b5d
Remove unused generate_buffer_arg_setter
Jan 18, 2021
e2ba9014
Invoker generator: globalize some py_codegen imports
Jan 18, 2021
852e4139
Wrapper: switch some C++ args from py::object to py::handle
Jan 18, 2021
676e30f5
Make creation of KernelWithCustomEnqueue easier to read (thanks
@kaushikcfd
)
Jan 18, 2021
6f6a54a5
Introduce _set_arg{,_buf}_multi for batched argument setting
Jan 18, 2021
f80c664f
Pack indices and args into the same tuple for set_arg*multi
Jan 19, 2021
a80586c8
Introduce Kernel._set_arg_buf_pack_multi for inlined buffer packing+arg setting
Jan 19, 2021
9ea089aa
Tweak examples to emphasize importance of retaining kernels
Jan 19, 2021
d6e0e1fb
Fix lenght assert on args_and_indices in invoker generation
Jan 19, 2021
60d5f39a
Drop absurd non-error-checking for typechar
Jan 19, 2021
3070eb99
Fix scope leakage in add_buf_arg
Jan 19, 2021
89c38431
Make VectorArg, ScalarArg comparable and add persistent-dict key generation for them
Jan 20, 2021
5dee9035
Incorporate argument-processing bits of kernel_runner into generated invoker
Jan 20, 2021
f0308dce
Add POCL Nvidia jobs to Gitlab CI
Jan 20, 2021
73faf922
Xfail a few tests in test_wrapper on POCL+Nvidia
Jan 20, 2021
0abbbaff
Make {Scalar,Vector,Other}Arg hashable
Jan 20, 2021
4117c4f0
Merge branch 'master' into speed-up-enqeue
Jan 20, 2021
385a8fef
Xfail bitonic argsort on pocl nvidia
Jan 20, 2021
50ee7ae8
Xfail test_array test_get_async on Pocl GPU
Jan 20, 2021
72793885
Xfail ranlux on on POCL + Nvidia
Jan 20, 2021
67984a7a
Xfail rect copies on Pocl GPU
Jan 20, 2021
2f39f702
Elementwise capture_as implementation: don't return after capture_call
Jan 20, 2021
d280d498
Xfail a few more ranlux tests on pocl+nvidia
Jan 20, 2021
69423255
Merge pull request #416 from inducer/speed-up-enqeue
Jan 20, 2021
86a793ec
Fix VectorArg.__hash__
Jan 20, 2021
ee3e68b9
Merge pull request #419 from inducer/fix-vectorarg-hash
Jan 21, 2021
bc770816
Undo accidental compyte reversion
Jan 21, 2021
647 additional commits have been omitted to prevent performance issues.
Loading
Loading