https://github.com/halide/Halide

sort by:
Revision Author Date Message Commit Date
11ef7b7 Fix inlining condition and add debug printing for long inlined producer chains 10 August 2020, 15:37:46 UTC
d3149f3 Use estimate_all for depthwise_separable_conv and conv_layer 10 August 2020, 15:06:37 UTC
34f90eb Print name of next node to be scheduled in CYOS mode 10 August 2020, 15:06:07 UTC
6d8ff1b Fix serial tile sizing 10 August 2020, 14:53:09 UTC
27734a7 Don't include output files in benchmark args 09 August 2020, 15:26:11 UTC
d3158c6 Rename iir_blur_generator -> iir_blur 08 August 2020, 22:57:25 UTC
af954fb Use each app's actual input images/parameters when benchmarking autoscheduler samples 08 August 2020, 22:56:50 UTC
085a31a Let the cost model decide if a state is doing excessive recompute 08 August 2020, 22:48:28 UTC
45bee63 Update iir_blur 08 August 2020, 22:47:28 UTC
b7750b2 Merge branch 'standalone_autoscheduler_gpu' of https://github.com/halide/Halide into standalone_autoscheduler_gpu 08 August 2020, 17:49:43 UTC
ef7cc0c Add host_dirty call to synthetic input buffers in rungen 08 August 2020, 17:47:33 UTC
644270b Consider options with less parallelism if there are no alternatives 08 August 2020, 15:20:32 UTC
0862bad Merge branch 'standalone_autoscheduler_gpu' of https://github.com/halide/Halide into standalone_autoscheduler_gpu 08 August 2020, 02:40:31 UTC
38b8122 Add compute_capability to default target 07 August 2020, 15:40:15 UTC
2db1fda Fix interpolate naming in collect_stats.sh 07 August 2020, 13:42:37 UTC
475659f Fix IRPrinter test 06 August 2020, 17:49:54 UTC
ecbe859 Replace normalize with output in interpolate_generator.cpp 06 August 2020, 16:39:18 UTC
778cf09 Move -march to a variable in Makefile 06 August 2020, 16:34:01 UTC
02f8387 Add test_state to Makefile 06 August 2020, 16:31:30 UTC
1c5d058 Compute num_realizations for unscheduled producers 06 August 2020, 16:30:45 UTC
7fc8d57 Only consider valid compute locations for unscheduled producers 04 August 2020, 20:09:09 UTC
52312ef Handle chains of inlined stages when staging producers 04 August 2020, 16:55:00 UTC
337f317 Fix tiling of stages with 1 dimension 03 August 2020, 23:39:13 UTC
b3e7805 Rename features: *_per_vector -> *_per_point 03 August 2020, 18:37:48 UTC
01797f9 Remove local features 03 August 2020, 18:33:17 UTC
a6e2856 Fix LoopNestParser's inline checking 03 August 2020, 14:27:11 UTC
fec9c34 Don't wait for benchmark queue to finish when in training mode 03 August 2020, 13:41:24 UTC
84d7fee Reset weights 03 August 2020, 13:20:58 UTC
189c291 Assume unscheduled producers are promoted to registers 03 August 2020, 13:20:15 UTC
f15198f Fix LoopNestParser inline condition 03 August 2020, 12:54:31 UTC
e9cfa25 Tidy up handling of func inlining decisions 03 August 2020, 12:53:45 UTC
010d077 Add Util.h 02 August 2020, 20:28:39 UTC
121041a Update working set index in cost_model_generator.cpp 02 August 2020, 20:26:51 UTC
f7fdf20 Remove per_vector features 02 August 2020, 20:00:30 UTC
72f909b Skip stages with 0 requests in extract_features.py 02 August 2020, 19:57:53 UTC
ec11fa5 Remove trace from metrics comparison 02 August 2020, 19:55:44 UTC
19ab5c1 Add loop nest parser test 02 August 2020, 19:54:57 UTC
426c1a5 Fix working_set index in cost_model_generator.cpp 02 August 2020, 19:51:05 UTC
9d4e5ed Add CYOS_FROM_FILE mode 31 July 2020, 22:44:51 UTC
654fb16 Update Makefile to include tests 31 July 2020, 14:12:24 UTC
4a4d45e Don't exit early from generate_gpu_tilings when a single option exceeds the thread limit 31 July 2020, 13:42:41 UTC
494382d Fix vectorized loads 31 July 2020, 13:37:18 UTC
501b188 Handle vectorized shared memory loads 28 July 2020, 21:26:36 UTC
f4cc439 Fix staging of inlined producers 28 July 2020, 16:44:23 UTC
3ebd4ea Prune splits of small, odd number extents 28 July 2020, 15:35:36 UTC
2459572 Sort metric comparisons by factor 28 July 2020, 01:56:46 UTC
b4888ea Account for serial loops outside a func's realization in points_accessed_per_thread 28 July 2020, 01:53:29 UTC
5ff1648 Update GlobalMemInfo.h -> GPUMemInfo.h in autoscheduler.inc 27 July 2020, 21:58:01 UTC
f0baefb Rename output file in compare_with_metrics.sh 27 July 2020, 21:57:00 UTC
9c30c24 Add depthwise_separable_conv to collect_stats.sh 27 July 2020, 21:56:24 UTC
5c1a889 Fix lookup key in compare_with_metrics.py 27 July 2020, 21:56:01 UTC
2e2d8b3 Fix thread loop index for nested stages 27 July 2020, 21:41:51 UTC
de84c10 Retain all Jacobians when analyzing memory access edges 27 July 2020, 19:21:18 UTC
06940fb Prune some states with dynamic local memory allocations 27 July 2020, 02:59:45 UTC
c51919d Rename GlobalMemInfo.h -> GPUMemInfo.h 27 July 2020, 01:19:00 UTC
ac85d64 Fix local mem feature names in cost_model_generator.cpp 26 July 2020, 16:32:34 UTC
0c9d180 Move compute_location debug printing 26 July 2020, 16:29:05 UTC
48238b3 Update local mem features to use MemInfo and an AccessAccumulator 26 July 2020, 16:28:25 UTC
b7e9a57 Add local metrics to compare_with_metrics.py 25 July 2020, 15:23:44 UTC
cc72f59 Handle local memory allocations that can be stored in registers 24 July 2020, 21:37:59 UTC
ecafcc3 Store benchmark queue in a directory instead of searching for samples 24 July 2020, 17:09:52 UTC
255890c Use get_host_target to get the target once 24 July 2020, 14:46:27 UTC
aa9ed0b Add metric comparison option to generate_autotune_results.sh 24 July 2020, 14:16:04 UTC
4bf074a Add VectorReduce to ExprBranching 24 July 2020, 13:59:09 UTC
a1bacf4 Merge branch 'standalone_autoscheduler_gpu' of https://github.com/halide/Halide into standalone_autoscheduler_gpu 24 July 2020, 03:29:14 UTC
256d40f Correct a few more Block statements 22 July 2020, 22:35:55 UTC
0954a84 Merge remote-tracking branch 'origin/abadams/unordered_blocks' into standalone_autoscheduler_gpu 22 July 2020, 22:32:18 UTC
18a1385 Add test that gpu barriers aren't inserted in unordered blocks 22 July 2020, 22:13:52 UTC
37e16c8 Revert to old block printing for C backend 22 July 2020, 22:13:41 UTC
25fbec8 Add an "ordering" flag to the Block node, and use it to emit aliasing metadata 22 July 2020, 21:55:25 UTC
3a84528 Merge pull request #5131 from halide/srj-hvx-check Improve HVX codegen error reporting 22 July 2020, 17:23:44 UTC
8918446 Merge pull request #5133 from halide/alexreinking-patch-1 Get rid of stale Travis CI build info from README. 21 July 2020, 23:16:16 UTC
daf7aa7 Get rid of stale Travis CI build info from README. 21 July 2020, 22:38:25 UTC
54f854e Update CodeGen_Hexagon.cpp 21 July 2020, 18:44:09 UTC
0fb2489 Update CodeGen_Hexagon.cpp 21 July 2020, 18:41:56 UTC
26bfbb4 Update CodeGen_Hexagon.cpp 21 July 2020, 18:06:00 UTC
3237682 Update CodeGen_Hexagon.cpp 21 July 2020, 17:34:16 UTC
10b8ba4 Fix stride indices 21 July 2020, 15:35:41 UTC
4fcae20 Improve HVX codegen error reporting If you try to compile HVX standalone code with HL_TARGET=hexagon-32-noos, you will die because necessary glue functions are defined in hvx_64 or hvx_128 but not 'baseline' hvx. Add an assertion check with a helpful error meesage to avoid just segfaulting deep inside LLVM. 20 July 2020, 23:47:49 UTC
41a756e Merge pull request #5129 from halide/srj-mkdir Add a couple of missing 'mkdir' usages in Makefile 20 July 2020, 21:55:21 UTC
96ca7f3 Merge branch 'master' into srj-mkdir 20 July 2020, 21:55:08 UTC
a30f220 Merge pull request #5128 from halide/srj-llvm Fix for trunk LLVM 20 July 2020, 21:54:54 UTC
937f797 Add a couple of missing 'mkdir' usages in Makefile 20 July 2020, 21:09:28 UTC
06f535f Add 'factor' to compare_with_metrics.py 20 July 2020, 20:44:18 UTC
ee2ca29 Handle tail warps in extract_features.py 20 July 2020, 20:43:45 UTC
df33922 Fix for trunk LLVM PrintMachineCode has been removed in LLVM 12/trunk 20 July 2020, 20:25:13 UTC
554e1dd Merge pull request #5125 from halide/abadams/rungenmain_error Add an error message if you forget to compile RunGenMain with a registration file 20 July 2020, 18:10:59 UTC
767d3e9 Fix total number of decisions 20 July 2020, 18:01:57 UTC
250cb44 Fix fractional memory access strides 20 July 2020, 18:01:20 UTC
c723ebf Fix conv_layer Var naming 20 July 2020, 02:48:34 UTC
10e7fb0 Move common testing functionality to test/test.h 18 July 2020, 19:26:54 UTC
2c03bfa Handle inlined stages in extract_features.py 18 July 2020, 19:20:39 UTC
a879b50 Add debug printing to cost model 18 July 2020, 19:20:09 UTC
54d84e3 Add timeout to benchmark loop and only search for samples in current batch 18 July 2020, 19:16:13 UTC
eb99441 Merge pull request #5126 from halide/shoaibkamil/llvm_clone_tag Update README to suggest cloning a release of LLVM, not a branch 17 July 2020, 21:58:04 UTC
a900b96 Add bounds along a single edge chain for when computing global/shared load features 17 July 2020, 21:16:05 UTC
50c947b Update README to suggest cloning a release of LLVM, not a branch 17 July 2020, 20:48:39 UTC
43f94b3 Add an error message if you forget to compile RunGenMain with a registration file 17 July 2020, 20:38:21 UTC
c7393ad Merge pull request #5122 from halide/srj-clangfmt Upgrade clang-format to v10 16 July 2020, 17:45:20 UTC
33ecc3f Upgrade clang-format to v10 Upgrade the clang-format checks to clang-format-10, and reformat code accordingly. Also add a way tp specify the clang-format version for `make format`; it defaults to the version of Clang for the current LLVM, but since clang-format doesn't provide stable formatting across versions, this might be wrong. 16 July 2020, 17:44:49 UTC
back to top