2f34370891eff5328f0b7787d5830857fa5566d3 to main · Andreas Klöckner / loopy · GitLab

Commits on Source 5040

886baacc Minor tweaks to dependencies-non-printing
Nov 30, 2015
8cb367db More kernel printing/tutorial tweaks
Nov 30, 2015
3d911f26 Restructure transforms code
Nov 30, 2015
bbfc3fa7 More transform shuffling
Nov 30, 2015
1c732a8f Second half of previous transform jostling commit
Nov 30, 2015
0b5d34f0 Add missing import
Nov 30, 2015
4412cab0 Initial version of distributive law transform
Nov 30, 2015
a4bf021a Use Subscript.index_tuple in more places
Dec 01, 2015
fe92b233 Leave a note about a possible error check
Dec 01, 2015
dfd763c2 Improve error checking in ArrayChanger
Dec 01, 2015
73b3ab13 Working version of the distributive law transform
Dec 01, 2015
3864505d Add test for distributive law transform
Dec 01, 2015
ad7baaf2 Add initial version of gNUMA test
Dec 01, 2015
e9943a32 Add opt levels, make gNUMA test executable
Dec 02, 2015
f83de483 Minor loopy pyinstaller tweak
Dec 07, 2015
dea64f61 First steps towards an ISPC target
Dec 07, 2015
5710fca2 Finish ISPC backend
Dec 07, 2015
cd4df8a7 Use cgen from git
Dec 07, 2015
c4c1b6d3 Flatten target submodules, tweak CUDA backend
Dec 07, 2015
8dc0ea33 Support choosing target from the command line
Dec 07, 2015
b7651913 Fix set_default_target initial setup
Dec 08, 2015
8ebcc066 Fix CL image arguments
Dec 08, 2015
6211cd14 Fix test failure: revert OpenCL barrier spacing style
Dec 08, 2015
ad7288f5 ExpressionInstruction -> Assignment
Dec 16, 2015
61a833a8 Import cleanup
Jan 07, 2016
405c789e Pass kernel to get_{global,local}_axis_expr, use signed indices on CUDA
Jan 07, 2016
fa3d3d66 CUDA target: generate launch bounds, extern C
Jan 07, 2016
19c24c8c Bump cache version
Jan 07, 2016
8ba03bf4 Use cgen from git in old PyOpenCL build
Jan 07, 2016
0ebd02f1 Introduce placeholders for hw axes, rather than using target-specific expressions
Jan 07, 2016
6b391673 Add OCCA codegen mode to ISPC target
Jan 07, 2016
c6965254 ISPC backend: Cast program indices to signed
Jan 14, 2016
bcece167 Update tutorial for better isl printing as of 0.16
Jan 14, 2016
910a2990 Fix up MANIFEST.in for release
Jan 15, 2016
ccb5fc74 Bump version
Jan 15, 2016
5dc6ba52 Fix up declared depdendencies
Jan 15, 2016
e1293b7f Fix up extras dependency declarations
Jan 15, 2016
851d5e27 Fix codegen from (immutable) tuples in Min/Max
Jan 18, 2016
166ddd3a Only generate static workgroup size info if workgroup sizes are actually static
Jan 18, 2016
14523cf7 Bump pymbolic depdency to fix dealing with Min/Max
Jan 18, 2016
2410fe2c Implement chunk_iname
Jan 18, 2016
c51c8aa0 get_op_poly doc fix
Jan 19, 2016
37eb5a8b Add sum_ops_to_dtypes
Jan 19, 2016
be4aba12 Add find_unused_axis_tag
Jan 19, 2016
f7de86e7 Improve tag_inames interface: strings, like.INAME, unused.[gl]
Jan 19, 2016
6e649f7b Folds in statistics module
Jan 21, 2016
92bf9408 Implement access footprint gathering
Jan 22, 2016
520ccbb7 Document fuse_kernels
Jan 22, 2016
be8c7735 Rename instruction attributes for more compatibility with pymbolic.imperative
Jan 23, 2016
838ba11d Track rename of insn_deps in rest of loopy. This caused a fluke-y test...
Jan 23, 2016
b16209ce Deal with changed interface on fuse_instruction_streams_with_unique_ids
Jan 24, 2016
37ef167d Make flake8 happy
Jan 25, 2016
67b9e4b4 Fix dependency printing logic
Jan 25, 2016
02e1ec2a Buffering: Fix usage of 1D arrays
Jan 25, 2016
8b29c09a Deal with absence of indices in array_buffer_map's in-footprint checking
Jan 25, 2016
64f43157 Automatic axis assignment: Detect, avoid unbounded axes
Jan 25, 2016
f65ebfd7 assignment_to_subst: complain if nothing matched
Jan 25, 2016
3c391614 precompute: complain if subst rule not found
Jan 25, 2016
1980520b Prevent infinite recursion in automatic axis assignment
Jan 25, 2016
7882e8ef Add store_instruction=False to buffer_array
Jan 25, 2016
59e2b821 Bump data model version
Jan 25, 2016
8ba1f9ef Fix kernel.domains comparison
Jan 26, 2016
b0b38c2f Fix buffer_array caching
Jan 26, 2016
eeac7ac9 Bump cache version
Jan 26, 2016
c8a3ad34 Flake 8 happiness
Jan 26, 2016
621770bd Accept and convexify domains on kernel creation
Jan 26, 2016
21bc1cc4 Understand Fortran 'double precision' type
Jan 28, 2016
b3eae237 Binary build fixes
Jan 29, 2016
34f304e6 Fix automated binary build
Jan 29, 2016
07d38823 Pin PyInstaller version for Py2.6
Jan 29, 2016
f1cf40a1 More autobuild tweaks
Jan 29, 2016
7068fe59 Tweak pyinstaller version selection
Jan 29, 2016
27478133 Docker binary build tweak
Jan 29, 2016
455027d5 Tweak pyinstaller spec file
Jan 29, 2016
0fbb9795 Use Py2.7 on Centos
Jan 29, 2016
69c94f14 Centos Docker build fix
Jan 29, 2016
02ee2222 More build fixes
Jan 29, 2016
78e69b89 Use SCL feature in Docker build
Jan 29, 2016
a7339a99 Fix SCL usage
Jan 29, 2016
6d198e2d Re-Fix SCL usage
Jan 29, 2016
dde57232 Introduce C target w/Fortran ABI
Jan 29, 2016
2b43d0aa Recognize more file extensions for Fortran
Jan 29, 2016
ccc0582d More build tweakery
Jan 30, 2016
07b49f8d ISPC/CUDA targets: call superclass constructor
Jan 30, 2016
4830e952 Py2.6/2.7 bug compatibility: no list comprehensions in function with exec
Jan 30, 2016
2fe494a1 Use up-to-date pyinstaller
Jan 30, 2016
8d806ee6 Fix CI binary build
Jan 30, 2016
842f346f Adjust hello-world example to make sense on ISPC
Jan 30, 2016
908a5481 Release doc adjustments
Jan 30, 2016
0c2216ca Recognize .py as loopy lang
Jan 30, 2016
0e9ad9b7 Point to binary from docs
Jan 30, 2016
b3375f95 Bump version
Jan 30, 2016
7f3442f5 uniform loads and stores now only counted once per thread group
Jan 30, 2016
3df264bf Merge remote-tracking branch 'upstream/master'
Jan 30, 2016
f16c6bef created memoized insn count function
Feb 06, 2016
f6688709 working on memoizing, get_dram_access_poly currently broken
Feb 07, 2016
c3085dea Merge branch 'master' of github.com:jdsteve2/loopy into uniform
Feb 08, 2016
27b94c94 Make loopy kernel Python-hashable
Feb 08, 2016
8a747238 Finish up uniform load counting
Feb 08, 2016
32ae06ce Add rename_argument
Feb 11, 2016

4,940 additional commits have been omitted to prevent performance issues.