72c91b1 | Andrew Adams | 14 January 2020, 01:09:30 UTC | Switch to 16 entries per line | 14 January 2020, 01:09:30 UTC |
e643e18 | Andrew Adams | 14 January 2020, 00:59:51 UTC | Use string literals instead of byte arrays for embedded blobs | 14 January 2020, 00:59:51 UTC |
c6af8a5 | Andrew Adams | 14 January 2020, 00:18:02 UTC | Merge pull request #4522 from halide/reschedule_nl_means Reschedule and benchmark nl_means app | 14 January 2020, 00:18:02 UTC |
c7de85e | Steven Johnson | 13 January 2020, 21:14:08 UTC | Merge pull request #4523 from halide/fix_build_after_divide_by_zero_merge Remove reference to deleted intrinsics | 13 January 2020, 21:14:08 UTC |
98d1d15 | Andrew Adams | 13 January 2020, 21:13:56 UTC | Remove reference to deleted intrinsics | 13 January 2020, 21:13:56 UTC |
60a5792 | Steven Johnson | 13 January 2020, 20:55:55 UTC | Merge pull request #4439 from halide/define_div_by_zero Define division by zero to return zero | 13 January 2020, 20:55:55 UTC |
90f382d | Andrew Adams | 13 January 2020, 19:18:56 UTC | Delete dangling thought | 13 January 2020, 19:18:56 UTC |
f0116c5 | Andrew Adams | 13 January 2020, 19:07:55 UTC | Extend comments on div/mod | 13 January 2020, 19:07:55 UTC |
684fba6 | Andrew Adams | 13 January 2020, 18:50:49 UTC | Add clarifying comment | 13 January 2020, 18:50:49 UTC |
500206e | Steven Johnson | 11 January 2020, 01:48:20 UTC | Merge pull request #4520 from halide/srj-tablecall Make the Simplifier handle PureExtern functions via LUT | 11 January 2020, 01:48:20 UTC |
b346332 | Andrew Adams | 10 January 2020, 23:47:12 UTC | Merge pull request #4461 from halide/fix-fft Fix small FFTs. | 10 January 2020, 23:47:12 UTC |
5cbd5a0 | Andrew Adams | 10 January 2020, 23:01:01 UTC | Reschedule and benchmark nl_means app | 10 January 2020, 23:01:01 UTC |
b7f5cc5 | Andrew Adams | 10 January 2020, 21:25:49 UTC | Merge pull request #4517 from halide/dynamic_allocation_in_gpu_kernel Permit more memory types for GPU intermediate data | 10 January 2020, 21:25:49 UTC |
cf459c6 | Steven Johnson | 10 January 2020, 18:43:29 UTC | Remove unnecessary lambda wrappers | 10 January 2020, 18:43:29 UTC |
f7f72cf | Andrew Adams | 10 January 2020, 18:43:22 UTC | Explain some magic numbers | 10 January 2020, 18:43:22 UTC |
53e8e42 | Steven Johnson | 10 January 2020, 02:02:25 UTC | Update Simplify_Call.cpp | 10 January 2020, 02:02:25 UTC |
6fc998e | Steven Johnson | 10 January 2020, 01:58:22 UTC | Update Simplify_Call.cpp | 10 January 2020, 01:58:22 UTC |
e05874a | Steven Johnson | 10 January 2020, 01:51:11 UTC | Update Simplify_Call.cpp | 10 January 2020, 01:51:11 UTC |
c034bbe | Steven Johnson | 10 January 2020, 01:42:55 UTC | Update Simplify_Call.cpp | 10 January 2020, 01:42:55 UTC |
f63df58 | Steven Johnson | 10 January 2020, 01:34:24 UTC | Make the Simplifier handle PureExtern functions via LUT There were a number of PureExtern functions we didn't handle (notably sin and cos, but many others too); rather than add even more if-else clauses, I refactored the PureExtern case into a set of LUTs to reduce redundant code. Note that the actual motivation for this change was to ensure that sin(CONSTANT) and cos(CONSTANT) got simplified to a real constant (rather than putting us at the mercy of LLVM). (With more cleverness we could make it a single LUT, but my first approach involved template-fu that was overly complex to read; I suspect that this approach is a better balance of compile time vs code complexity, but would welcome someone more clever to simplify it further.) | 10 January 2020, 01:34:24 UTC |
bfb761c | Andrew Adams | 09 January 2020, 23:56:40 UTC | Merge pull request #4519 from halide/update_gpu_app_schedules Update benchmark comments in apps | 09 January 2020, 23:56:40 UTC |
026599f | Steven Johnson | 09 January 2020, 23:16:25 UTC | Merge pull request #4488 from halide/aligned-read Use pointer access for aligned load/store in OpenCL codegen | 09 January 2020, 23:16:25 UTC |
b9a5939 | Andrew Adams | 09 January 2020, 22:58:59 UTC | Merge remote-tracking branch 'origin/master' into define_div_by_zero | 09 January 2020, 22:58:59 UTC |
dc21ceb | Andrew Adams | 09 January 2020, 22:24:10 UTC | Use smaller thread blocks in GPU test | 09 January 2020, 22:24:10 UTC |
e0e4bdb | Andrew Adams | 09 January 2020, 20:58:03 UTC | Update benchmark comments in apps Benchmarked a few more apps on the same machine as the recently-rescheduled ones, to get a consistent set of benchmarks for people to refer too. Also rescheduled stencil chain, including writing a GPU schedule. It's a very simple GPU schedule but I was unable to get a better one. | 09 January 2020, 20:58:03 UTC |
34ba826 | Andrew Adams | 09 January 2020, 17:47:30 UTC | It's not safe to preserve all pure intrinsics in bounds expressions | 09 January 2020, 17:47:30 UTC |
2860f3d | Steven Johnson | 08 January 2020, 23:45:34 UTC | Merge branch 'master' into aligned-read | 08 January 2020, 23:45:34 UTC |
206c726 | Andrew Adams | 08 January 2020, 23:11:07 UTC | quiet_div can be treated as pure for bounds inference as can other pure intrinsics | 08 January 2020, 23:11:07 UTC |
a36e80d | Andrew Adams | 08 January 2020, 21:50:24 UTC | Merge pull request #4518 from halide/fix_solve_canonicalization Fix solve canonicalization | 08 January 2020, 21:50:24 UTC |
c939937 | Andrew Adams | 08 January 2020, 21:25:07 UTC | Simplify interpolate schedule This dodges a pathological correlated subexpression blowing out shared. Longer term we probably need to do exact tracking of shared memory needs instead. | 08 January 2020, 21:25:07 UTC |
0a9c1dc | Andrew Adams | 08 January 2020, 17:58:48 UTC | Address some review comments | 08 January 2020, 17:58:48 UTC |
6158323 | Steven Johnson | 08 January 2020, 17:26:42 UTC | Merge branch 'master' into fix_solve_canonicalization | 08 January 2020, 17:26:42 UTC |
81327fb | Steven Johnson | 08 January 2020, 17:26:22 UTC | Merge pull request #4513 from halide/srj-hvx-nocopy inject_host_dev_buffer_copies() should be a no-op for t.arch = Hexagon | 08 January 2020, 17:26:22 UTC |
9c3bed3 | Andrew Adams | 08 January 2020, 00:06:38 UTC | Fix alignment issue The heap slice size per block needs to be a multiple of the largest type bit-width, or the indexing gets sloppy. | 08 January 2020, 00:06:38 UTC |
890adff | Andrew Adams | 07 January 2020, 23:57:18 UTC | Only run GPU test when GPU target enabled | 07 January 2020, 23:57:18 UTC |
4a68d55 | Andrew Adams | 07 January 2020, 23:54:43 UTC | Add test | 07 January 2020, 23:54:43 UTC |
b63cab8 | Andrew Adams | 07 January 2020, 23:50:54 UTC | Fix order of canonicalization and unpacking LHS This could cause incorrect results from the solver | 07 January 2020, 23:50:54 UTC |
e4f9e74 | Steven Johnson | 07 January 2020, 23:26:41 UTC | Merge branch 'master' into aligned-read | 07 January 2020, 23:26:41 UTC |
c34331f | Steven Johnson | 07 January 2020, 23:26:30 UTC | Merge branch 'master' into fix-fft | 07 January 2020, 23:26:30 UTC |
131498d | Steven Johnson | 07 January 2020, 23:04:49 UTC | Merge branch 'master' into srj-hvx-nocopy | 07 January 2020, 23:04:49 UTC |
969b86c | Steven Johnson | 07 January 2020, 23:04:39 UTC | Fix in inject_host_dev_buffer_copies() instead | 07 January 2020, 23:04:39 UTC |
7d02f29 | Andrew Adams | 07 January 2020, 22:39:06 UTC | revert cuda malloc/free injection | 07 January 2020, 22:39:06 UTC |
09f5628 | Andrew Adams | 07 January 2020, 22:13:12 UTC | Merge remote-tracking branch 'origin/master' into dynamic_allocation_in_gpu_kernel | 07 January 2020, 22:13:12 UTC |
067724b | Andrew Adams | 07 January 2020, 22:12:14 UTC | Merge pull request #4515 from halide/cblas_fix Reinsert check for cblas.h | 07 January 2020, 22:12:14 UTC |
8d18445 | Andrew Adams | 07 January 2020, 22:11:26 UTC | Update existing gpu memory tests to exercise new functionality | 07 January 2020, 22:11:26 UTC |
02ccdda | Andrew Adams | 07 January 2020, 22:11:07 UTC | Reschedule IIR blur to use global for the intermediate | 07 January 2020, 22:11:07 UTC |
8f87d00 | Andrew Adams | 07 January 2020, 22:10:54 UTC | Allocate a single global heap buffer and stripe it | 07 January 2020, 22:10:54 UTC |
0ff833b | Andrew Adams | 07 January 2020, 18:09:04 UTC | Reinsert check for cblas.h But add more search locations | 07 January 2020, 18:09:04 UTC |
8ddf7ef | Dillon Sharlet | 07 January 2020, 05:30:35 UTC | Merge pull request #4465 from halide/deprecate-hvx-v60 Remove support for Hexagon v60 architecture. | 07 January 2020, 05:30:35 UTC |
f560272 | Steven Johnson | 06 January 2020, 22:53:49 UTC | make format | 06 January 2020, 22:53:49 UTC |
2a4aeb6 | Steven Johnson | 06 January 2020, 22:53:11 UTC | make format | 06 January 2020, 22:53:11 UTC |
abe59f9 | Steven Johnson | 06 January 2020, 22:46:19 UTC | Merge branch 'master' into fix-fft | 06 January 2020, 22:46:19 UTC |
a18946c | Steven Johnson | 06 January 2020, 22:45:43 UTC | Merge branch 'master' into aligned-read | 06 January 2020, 22:45:43 UTC |
3c6f839 | Steven Johnson | 06 January 2020, 22:07:37 UTC | Never call inject_host_dev_buffer_copies() if t.arch = Hexagon | 06 January 2020, 22:07:37 UTC |
bfc4dff | Steven Johnson | 06 January 2020, 19:55:49 UTC | Merge branch 'master' into deprecate-hvx-v60 | 06 January 2020, 19:55:49 UTC |
65c05b4 | Steven Johnson | 06 January 2020, 19:55:20 UTC | Merge pull request #4511 from halide/srj-clang-fmt-fixes Followup fixes for clang-format checker | 06 January 2020, 19:55:20 UTC |
089010f | Steven Johnson | 06 January 2020, 18:58:17 UTC | Followup fixes for clang-format checker Exclude dirs via .clang-format-ignore. Also special-case PyStub.cpp since it has a confusing structure that varies results between clang-format 9 and 10. | 06 January 2020, 18:58:17 UTC |
ddaec3e | Steven Johnson | 06 January 2020, 18:41:35 UTC | Update .clang-format-ignore | 06 January 2020, 18:41:35 UTC |
f32bf3b | Steven Johnson | 06 January 2020, 18:39:59 UTC | More clang-format fixes | 06 January 2020, 18:39:59 UTC |
55f15c0 | Steven Johnson | 06 January 2020, 17:44:54 UTC | Update .clang-format-ignore | 06 January 2020, 17:44:54 UTC |
2480603 | Steven Johnson | 06 January 2020, 17:19:18 UTC | Add .clang-format-ignore | 06 January 2020, 17:19:18 UTC |
71897cb | Steven Johnson | 06 January 2020, 17:10:08 UTC | Update presubmit.yml | 06 January 2020, 17:10:08 UTC |
7f4cda7 | Steven Johnson | 06 January 2020, 17:08:20 UTC | Update presubmit.yml | 06 January 2020, 17:08:20 UTC |
be1852b | Steven Johnson | 06 January 2020, 17:06:27 UTC | Fix clang-format stuff | 06 January 2020, 17:06:27 UTC |
96f6465 | Steven Johnson | 06 January 2020, 17:00:51 UTC | Merge branch 'master' into deprecate-hvx-v60 | 06 January 2020, 17:00:51 UTC |
1d58da0 | Andrew Adams | 06 January 2020, 16:51:22 UTC | Merge remote-tracking branch 'origin/master' into define_div_by_zero | 06 January 2020, 16:51:22 UTC |
ec03c4c | Steven Johnson | 04 January 2020, 19:35:37 UTC | Merge pull request #4503 from halide/srj-presubmit Run clang-format as a Presubmit check | 04 January 2020, 19:35:37 UTC |
d05d29a | Steven Johnson | 04 January 2020, 01:48:52 UTC | Merge pull request #4510 from halide/srj-license Update LICENSE.txt for cmdline.h | 04 January 2020, 01:48:52 UTC |
d48d5fd | Steven Johnson | 04 January 2020, 01:00:24 UTC | Update LICENSE.txt for cmdline.h apps/support/cmdline.h is BSD 3-Clause (which is different from the bulk of Halide), so call this out specifically. | 04 January 2020, 01:00:24 UTC |
e0ec130 | dsharletg | 04 January 2020, 00:20:07 UTC | Restore .clang-format | 04 January 2020, 00:20:07 UTC |
f1884aa | Steven Johnson | 03 January 2020, 23:54:24 UTC | Merge pull request #4507 from halide/srj-msan-boundsinference BoundsInference needs to mark buffer shapes too | 03 January 2020, 23:54:24 UTC |
cee51dd | Steven Johnson | 03 January 2020, 23:53:14 UTC | Merge branch 'master' into srj-presubmit | 03 January 2020, 23:53:14 UTC |
60b039d | Steven Johnson | 03 January 2020, 23:52:08 UTC | Merge pull request #4508 from halide/srj-format More clang-format fixes | 03 January 2020, 23:52:08 UTC |
e5d26ed | Andrew Adams | 03 January 2020, 23:38:57 UTC | Roll back most runtime changes, and allocate a buffer instead | 03 January 2020, 23:38:57 UTC |
35009a4 | dsharletg | 03 January 2020, 23:34:22 UTC | Remove Elfsigner log. | 03 January 2020, 23:34:22 UTC |
23c74bb | dsharletg | 03 January 2020, 23:29:09 UTC | Deprecate v60. | 03 January 2020, 23:29:09 UTC |
443048d | dsharletg | 03 January 2020, 23:28:48 UTC | Switch architecture from v60 to v62. | 03 January 2020, 23:28:48 UTC |
462e994 | dsharletg | 03 January 2020, 23:19:23 UTC | Merge branch 'master' of https://github.com/halide/Halide into deprecate-hvx-v60 | 03 January 2020, 23:19:23 UTC |
7577c35 | Steven Johnson | 03 January 2020, 22:39:53 UTC | Merge pull request #4509 from halide/srj-makefix Remove --no-builtin-variables from MAKEFLAGS | 03 January 2020, 22:39:53 UTC |
7643f25 | Steven Johnson | 03 January 2020, 21:16:11 UTC | Update msan_aottest.cpp | 03 January 2020, 21:16:11 UTC |
4883188 | Steven Johnson | 03 January 2020, 20:50:36 UTC | update | 03 January 2020, 20:50:36 UTC |
571fd5b | Steven Johnson | 03 January 2020, 20:32:23 UTC | Back out recent Makefile change | 03 January 2020, 20:32:23 UTC |
b112d71 | Steven Johnson | 03 January 2020, 20:17:07 UTC | Trigger change | 03 January 2020, 20:17:07 UTC |
2f2b527 | Steven Johnson | 03 January 2020, 20:11:37 UTC | Trigger change | 03 January 2020, 20:11:37 UTC |
e543f30 | Steven Johnson | 03 January 2020, 19:49:01 UTC | More clang-format fixes - make our .clang-format style compatible with clang-format-9 - re-run 'make format' to bring everything back into conformance with the slightly-tweaked style | 03 January 2020, 19:49:01 UTC |
9002aa7 | Steven Johnson | 03 January 2020, 19:10:47 UTC | BoundsInference needs to mark buffer shapes too BoundsInference.cpp was missing the logic to mark newly-created buffer shapes as initialized for MSAN, so pipelines with `.define_extern()` usage could get erroneous MSAN failures. | 03 January 2020, 19:10:47 UTC |
7c012e5 | Steven Johnson | 03 January 2020, 18:51:00 UTC | 'make format' | 03 January 2020, 18:51:00 UTC |
5a79822 | Steven Johnson | 03 January 2020, 18:31:15 UTC | Merge branch 'master' into srj-presubmit | 03 January 2020, 18:31:15 UTC |
f8d2bf2 | Steven Johnson | 03 January 2020, 18:14:48 UTC | Merge branch 'master' into define_div_by_zero | 03 January 2020, 18:14:48 UTC |
cf3234a | Steven Johnson | 03 January 2020, 18:13:06 UTC | Merge branch 'master' into fix-fft | 03 January 2020, 18:13:06 UTC |
ca5370d | Steven Johnson | 03 January 2020, 18:11:51 UTC | Merge branch 'master' into aligned-read | 03 January 2020, 18:11:51 UTC |
5490043 | Steven Johnson | 03 January 2020, 18:10:45 UTC | Merge pull request #4501 from halide/clang_format_makefile_target Clang format makefile target | 03 January 2020, 18:10:45 UTC |
4f41061 | Steven Johnson | 03 January 2020, 18:09:43 UTC | Merge pull request #4506 from halide/srj-no_bounds Delete auto_schedule_no_bounds.cpp | 03 January 2020, 18:09:43 UTC |
e8a6fcd | Steven Johnson | 03 January 2020, 18:09:06 UTC | Merge pull request #4505 from halide/srj-cmake-xml2 Fix CMake breakage for recent LLVM | 03 January 2020, 18:09:06 UTC |
58074e1 | Andrew Adams | 03 January 2020, 17:37:15 UTC | Merge branch 'clang_format_makefile_target' of https://github.com/halide/Halide into clang_format_makefile_target | 03 January 2020, 17:37:15 UTC |
5b4b3b3 | Andrew Adams | 03 January 2020, 17:37:07 UTC | Merge remote-tracking branch 'origin/master' into clang_format_makefile_target | 03 January 2020, 17:37:07 UTC |
fde9d2a | Andrew Adams | 03 January 2020, 17:36:37 UTC | Merge remote-tracking branch 'origin/malloc_in_cuda_kernels' into dynamic_allocation_in_gpu_kernel | 03 January 2020, 17:36:37 UTC |
7103c28 | Steven Johnson | 03 January 2020, 02:01:58 UTC | Delete auto_schedule_no_bounds.cpp It no longer fails, and thus its success is a failure | 03 January 2020, 02:01:58 UTC |
211fd60 | Andrew Adams | 03 January 2020, 01:46:05 UTC | Call malloc/free in cuda kernels | 03 January 2020, 01:46:05 UTC |
d1c4ff2 | Steven Johnson | 03 January 2020, 01:44:21 UTC | Merge pull request #4504 from halide/srj-makefile Remove useless definition in Makefile | 03 January 2020, 01:44:21 UTC |