| d7cf9bc | Steven Johnson | 09 October 2020, 16:25:38 UTC | Merge branch 'master' into abadams/nested_vectorization_tweaks | 09 October 2020, 16:25:38 UTC |
| fa54197 | Steven Johnson | 08 October 2020, 21:23:52 UTC | Remove the ADD_[U]INT64_T_SUFFIX macros Modify codegen for C-like backends to just emit integer constants with the correct suffix for the backend, rather than wrapping in a giant macro; the macro approach worked but lordy was it painful to read. | 09 October 2020, 16:24:11 UTC |
| 4070346 | Steven Johnson | 08 October 2020, 23:51:03 UTC | Merge branch 'master' into abadams/nested_vectorization_tweaks | 08 October 2020, 23:51:03 UTC |
| a1c9d89 | Steven Johnson | 08 October 2020, 17:39:19 UTC | Fix for trunk LLVM | 08 October 2020, 22:50:01 UTC |
| 72f8c54 | Steven Johnson | 08 October 2020, 20:24:06 UTC | Tickle Buildbots | 08 October 2020, 20:24:06 UTC |
| db2411a | Steven Johnson | 08 October 2020, 18:41:24 UTC | Tickle Buildbots | 08 October 2020, 18:41:24 UTC |
| bd92176 | Andrew Adams | 07 October 2020, 19:32:03 UTC | Merge remote-tracking branch 'origin/master' into abadams/nested_vectorization_tweaks | 07 October 2020, 19:32:03 UTC |
| a189fd4 | Andrew Adams | 07 October 2020, 18:36:52 UTC | Merge pull request #5335 from jlaxson/gpu-test-workgroup-size Reduce GPU Tile Size in Tests | 07 October 2020, 18:36:52 UTC |
| a85407a | Steven Johnson | 06 October 2020, 20:30:10 UTC | Merge branch 'master' into abadams/nested_vectorization_tweaks | 06 October 2020, 20:30:10 UTC |
| b1fd538 | Steven Johnson | 05 October 2020, 22:41:32 UTC | Make C++ backend requirement for C++11 explicit Some C++11 features had crept into our C++ backend codegen; make this explicit and check for the correct version at the top of the generated file. (Then remove the stray checks for C++11 version elsewhere.) | 06 October 2020, 16:56:23 UTC |
| 719d457 | Andrew Adams | 06 October 2020, 16:42:40 UTC | Merge remote-tracking branch 'origin/master' into abadams/nested_vectorization_tweaks | 06 October 2020, 16:42:40 UTC |
| ab35a73 | Andrew Adams | 06 October 2020, 16:42:28 UTC | Try to work around MSVC weirdness | 06 October 2020, 16:42:28 UTC |
| d012df7 | Andrew Adams | 06 October 2020, 16:35:49 UTC | Merge pull request #5331 from rootjalex/master fix bounds inference bug for bounded interval / unbounded interval | 06 October 2020, 16:35:49 UTC |
| 2d15069 | Andrew Adams | 06 October 2020, 16:34:55 UTC | Merge pull request #5334 from halide/abadams/braces_around_statements Require braces around if/while bodies. | 06 October 2020, 16:34:55 UTC |
| 8c12c67 | John Laxson | 06 October 2020, 02:00:54 UTC | Reduce GPU Tile Size | 06 October 2020, 02:00:54 UTC |
| ecad269 | Andrew Adams | 05 October 2020, 23:56:18 UTC | Use switch statement instead of if sequence | 05 October 2020, 23:56:18 UTC |
| 325490b | Andrew Adams | 05 October 2020, 23:53:09 UTC | Require braces around if/while bodies. | 05 October 2020, 23:53:09 UTC |
| a783daf | Andrew Adams | 05 October 2020, 23:07:06 UTC | Merge branch 'abadams/nested_vectorization_tweaks' of https://github.com/halide/Halide into abadams/nested_vectorization_tweaks | 05 October 2020, 23:07:06 UTC |
| 7244d1e | Andrew Adams | 05 October 2020, 23:07:00 UTC | Make constructor explicit | 05 October 2020, 23:07:00 UTC |
| eb279fc | Alexander Root | 05 October 2020, 22:18:10 UTC | update comment on bounds bug | 05 October 2020, 22:18:10 UTC |
| 408a0e2 | Alexander Root | 05 October 2020, 21:45:48 UTC | fix bounds inference bug for bounded interval / unbounded interval | 05 October 2020, 21:45:48 UTC |
| 387acbf | Steven Johnson | 05 October 2020, 18:47:32 UTC | Merge branch 'master' into abadams/nested_vectorization_tweaks | 05 October 2020, 18:47:32 UTC |
| 68f66fe | Shoaib Kamil | 05 October 2020, 17:55:33 UTC | Fix 32-bit Windows vcvars command (#5330) `x64_x86` = build using 64-bit compiler, targeting 32-bit x86 (https://docs.microsoft.com/en-us/cpp/build/building-on-the-command-line?view=vs-2019). We had it backwards before. | 05 October 2020, 17:55:33 UTC |
| 276372e | Andrew Adams | 05 October 2020, 17:47:53 UTC | Add missing is-multiple check before bitcasting to wider type in shuffle codegen | 05 October 2020, 17:47:53 UTC |
| 31d0cd0 | Andrew Adams | 05 October 2020, 17:34:14 UTC | Drive-by addition of braces to if statement | 05 October 2020, 17:34:14 UTC |
| d97ee65 | Andrew Adams | 05 October 2020, 16:35:59 UTC | Address review comments | 05 October 2020, 16:35:59 UTC |
| 86fa657 | Steven Johnson | 02 October 2020, 22:02:38 UTC | registerPassBuilderCallbacks is only available in LLVM 12+ | 04 October 2020, 18:16:12 UTC |
| 2d3ebec | Ankit Aggarwal | 01 October 2020, 00:38:14 UTC | Formatting. | 04 October 2020, 18:16:12 UTC |
| ea583f3 | Ankit Aggarwal | 26 August 2020, 18:20:45 UTC | Add TargetMachine::registerPassBuilderCallbacks Add call to TargetMachine::registerPassBuilderCallbacks to allow targets to add passes to the pass pipeline using the New Pass Manager | 04 October 2020, 18:16:12 UTC |
| fa9da79 | Steven Johnson | 03 October 2020, 17:16:31 UTC | Merge pull request #5324 from halide/srj/inject_buffer InjectBufferCopiesForInputsAndOutputs should check for unexpected Call nodes. | 03 October 2020, 17:16:31 UTC |
| 3d19071 | Steven Johnson | 03 October 2020, 04:18:10 UTC | Merge pull request #5315 from halide/srj/abort Kill halide_abort() | 03 October 2020, 04:18:10 UTC |
| fa4d11b | Steven Johnson | 03 October 2020, 00:33:29 UTC | InjectBufferCopiesForInputsAndOutputs should have an assertion for Call nodes | 03 October 2020, 00:33:29 UTC |
| 41426c4 | Andrew Adams | 03 October 2020, 00:26:35 UTC | Avoid nested ternary op to pacify older clang-format | 03 October 2020, 00:26:35 UTC |
| dcdf48c | Andrew Adams | 03 October 2020, 00:14:09 UTC | Update schedule | 03 October 2020, 00:14:09 UTC |
| 097d6ce | Andrew Adams | 03 October 2020, 00:03:57 UTC | Better comments | 03 October 2020, 00:03:57 UTC |
| edeb32f | Andrew Adams | 02 October 2020, 23:58:11 UTC | Revert accidental change to some json | 02 October 2020, 23:58:11 UTC |
| fe4801a | Andrew Adams | 02 October 2020, 22:18:02 UTC | Clean up nested vectorization performance test | 02 October 2020, 22:18:02 UTC |
| dccb856 | Andrew Adams | 02 October 2020, 22:17:46 UTC | Add pmaddwd for avx512 | 02 October 2020, 22:17:46 UTC |
| 908e626 | Steven Johnson | 02 October 2020, 22:14:00 UTC | Kill halide_abort() This was added long ago as an attempt to work around issues with the Windows Debug runtime (in which calling `abort()` would produce an "Abort, Retry, Ignore" dialog). We no long do debug builds of any sort on our buildbots, so let's lose all this mess to simplify our world a bit. | 02 October 2020, 22:14:00 UTC |
| be7303f | Andrew Adams | 02 October 2020, 21:18:11 UTC | Better pmaddwd recognition for VectorReduce in x86 backend | 02 October 2020, 21:18:11 UTC |
| dba05a7 | Andrew Adams | 02 October 2020, 21:17:50 UTC | more ramp-of-ramp simplification | 02 October 2020, 21:17:50 UTC |
| 96137b6 | Andrew Adams | 02 October 2020, 21:17:28 UTC | Add detection for transpose-shuffles | 02 October 2020, 21:17:28 UTC |
| a1206b5 | Steven Johnson | 02 October 2020, 21:11:49 UTC | Merge pull request #5322 from halide/srj/runtime-warn Make sure the runtime compiler settings in CMake match those in Make | 02 October 2020, 21:11:49 UTC |
| bacd284 | Steven Johnson | 02 October 2020, 18:47:48 UTC | Make sure the runtime compiler settings in CMake match those in Make Mainly, we weren't setting any of the warning flags, so CMake builds were more forgiving than Make. | 02 October 2020, 21:11:28 UTC |
| 14b567e | Steven Johnson | 02 October 2020, 18:49:18 UTC | Merge pull request #5312 from halide/vksnk/align_loads Don't try to align loads if alignment is not divisible by the size of the load | 02 October 2020, 18:49:18 UTC |
| ac56ec9 | Steven Johnson | 02 October 2020, 16:56:02 UTC | Merge pull request #5308 from halide/build/shared-llvm-fix Fix linking to shared LLVM. | 02 October 2020, 16:56:02 UTC |
| ff5c2ad | Steven Johnson | 02 October 2020, 01:07:11 UTC | Merge branch 'master' into vksnk/align_loads | 02 October 2020, 01:07:11 UTC |
| 08ab4a5 | Steven Johnson | 01 October 2020, 22:35:18 UTC | Merge pull request #5321 from halide/docs/readme-homebrew Add package manager info to README.md | 01 October 2020, 22:35:18 UTC |
| 2768ee3 | Alex Reinking | 01 October 2020, 21:52:23 UTC | Add package manager info to README.md | 01 October 2020, 21:52:23 UTC |
| 70e98d2 | Steven Johnson | 01 October 2020, 21:21:11 UTC | Merge branch 'master' into build/shared-llvm-fix | 01 October 2020, 21:21:11 UTC |
| 17e1ec6 | Steven Johnson | 01 October 2020, 21:20:57 UTC | Merge branch 'master' into vksnk/align_loads | 01 October 2020, 21:20:57 UTC |
| 3b47c0e | Marcos Slomp | 01 October 2020, 21:19:11 UTC | [d3d12] allocation cache + bugfixes (#5298) * refactoring to remove wait/sync points from kernel dispatch * debugging and bugfixes * refactoring wait/sync procedures * refactoring buffer signal checkpoints * improved time tracing * improved device selection and additional trace scoping features * addressing clang format issues * more clang format complaints... ¯\_(ツ)_/¯ * more clang format... * clang format... * nullptr -> NULL (0) * addressing code review comments * scoping the kernel argument setup code * addressing code review comments * clang format... * buffer allocation cache * rearranging wait/sync points in the allocation cache strategy * cleanup and refactoring * bugfix: must reset descriptor binder state when recycling it for a new command frame * releasing cached resources on device shutdown * reworking device crop release since allocation cache has been implemented * improved trace info, comments and asserts * refactorig device creation * tracking checkpoints in device<->device transfers * better debug dump report scope * removing old code * clang format * unused variable * Fix python_correctness_boundary_conditions * atomic clarity * refactoring of allocation cache (with local toggle) * Tickle Buildbots * Modify load/store codegen to support load/store from/to shared mem * adding trace-level support * Tickle buildbot * addressing code review * refactoring trace errors/warnings and context halting condition * clang format * build fix * resetting d3dd12_frame struct fields after release * reverting accidetal changes (corrupted git index during stage) * adding remarks with regards to device creation quirks in d3d12 * fixing build (windows cross-compilation on linux) * removing double semi-colon (clang-format) * Tickle the buildbots * Fix signed/unsigned mismatch in d3d12compute.cpp * Remove space in cast Co-authored-by: Marcos Slomp <slomp@adobe.com> Co-authored-by: Shoaib Kamil <kamil@adobe.com> Co-authored-by: Steven Johnson <srj@google.com> Co-authored-by: Shoaib Kamil <shoaibkamil@gmail.com> | 01 October 2020, 21:19:11 UTC |
| 0f8e9cc | Steven Johnson | 01 October 2020, 21:00:55 UTC | Tickle the buildbots | 01 October 2020, 21:00:55 UTC |
| cb9589d | Steven Johnson | 01 October 2020, 21:00:25 UTC | Tickle the buildbots | 01 October 2020, 21:00:25 UTC |
| 3431d52 | Steven Johnson | 01 October 2020, 16:53:33 UTC | Merge branch 'master' into build/shared-llvm-fix | 01 October 2020, 16:53:33 UTC |
| 87466ad | Steven Johnson | 01 October 2020, 16:53:03 UTC | Merge branch 'master' into vksnk/align_loads | 01 October 2020, 16:53:03 UTC |
| a957053 | Steven Johnson | 01 October 2020, 16:43:58 UTC | Merge pull request #5317 from halide/srj/runtime-errors Make halide_assert() use do...while(0) idiom | 01 October 2020, 16:43:58 UTC |
| 7da643b | Steven Johnson | 01 October 2020, 00:23:50 UTC | Make halide_assert() use do.while(0) idiom This is the canonical form for statement-like macros in C. Added some missing semicolons that were detected by this (and fixed formatting). | 01 October 2020, 00:23:50 UTC |
| f501810 | Steven Johnson | 30 September 2020, 23:07:36 UTC | Merge branch 'master' into build/shared-llvm-fix | 30 September 2020, 23:07:36 UTC |
| ab3a541 | Steven Johnson | 30 September 2020, 23:06:56 UTC | Merge branch 'master' into vksnk/align_loads | 30 September 2020, 23:06:56 UTC |
| 6fc5bc8 | Steven Johnson | 30 September 2020, 23:05:10 UTC | Merge pull request #5307 from halide/shoaibkamil/host_supports_target_device Fix host_supports_target_device() | 30 September 2020, 23:05:10 UTC |
| f7b5382 | Shoaib Kamil | 30 September 2020, 17:58:06 UTC | Update DeviceInterface.cpp | 30 September 2020, 17:58:06 UTC |
| 2c6e683 | Volodymyr Kysenko | 30 September 2020, 17:47:31 UTC | Don't try to align loads if alignment is not divisible by the size of the load | 30 September 2020, 17:49:48 UTC |
| 69be4d4 | Alex Reinking | 30 September 2020, 16:39:20 UTC | Fix linking to shared LLVM. Fixes #5304. | 30 September 2020, 17:17:47 UTC |
| c7935de | Shoaib Kamil | 30 September 2020, 13:35:53 UTC | Fix host_supports_target_device() | 30 September 2020, 13:35:53 UTC |
| dc89424 | Steven Johnson | 30 September 2020, 01:22:20 UTC | Merge pull request #5303 from halide/srj/win-32 Fix runtime build rules in Makefile | 30 September 2020, 01:22:20 UTC |
| 9cd9048 | Andrew Adams | 30 September 2020, 00:47:19 UTC | Fixes for ARM | 30 September 2020, 00:47:19 UTC |
| d1ea4c3 | Steven Johnson | 29 September 2020, 23:43:04 UTC | Merge pull request #5301 from halide/srj/init-index Avoid possibly-uninitialized use of RVar::_index | 29 September 2020, 23:43:04 UTC |
| df83329 | Andrew Adams | 29 September 2020, 22:08:54 UTC | Add WIP test for nested vectorization perf | 29 September 2020, 22:08:54 UTC |
| ffc036f | Andrew Adams | 29 September 2020, 22:08:37 UTC | Improvements to nested vectorization simplification and codegen | 29 September 2020, 22:08:37 UTC |
| 69aeb9e | Steven Johnson | 29 September 2020, 19:12:53 UTC | Fix runtime build rules in Makefile: - one of the Windows-specific runtime files had a 32-vs-64 glitch - CMake now uses `fno-threadsafe-statics` (instead of `-std=gnu++98`) to disable thread-safe static initialization; as a result, this allowed C++11 code requirements to creep in (via d3d12compute.cpp), but we didn't notice because the Makefile wasn't properly building that file due to the 32-vs-64 glitch. Fixed by updating the Makefile to use this flag instead (which was an overdue fix anyway). | 29 September 2020, 19:44:46 UTC |
| 97cefb9 | Steven Johnson | 29 September 2020, 18:11:26 UTC | Avoid possibly-uninitialized use of RVar::_index One of the armbots warned that this field could be used uninitialized; I can't replicate anywhere else, but indeed, the string-only ctor of RVar left this uninitialized. Defaulted it to -1 and added an explicit check in _var(). (Yes, the call to `at()` will fail when out of range, but explicit checking is better IMHO.) | 29 September 2020, 18:37:35 UTC |
| f6c607b | Alex Reinking | 25 September 2020, 23:53:16 UTC | Replace large code model build option with target feature. (#5216) | 25 September 2020, 23:53:16 UTC |
| a239951 | Steven Johnson | 24 September 2020, 18:55:20 UTC | Merge pull request #5294 from halide/srj/dupnames Check for duplicated Parameter/Buffer names in InferArguments (Issue #5292) | 24 September 2020, 18:55:20 UTC |
| c110bec | Steven Johnson | 23 September 2020, 00:54:30 UTC | Check for duplicated Parameter/Buffer names in InferArguments | 24 September 2020, 18:54:59 UTC |
| a4e4052 | Marcos Slomp | 24 September 2020, 17:12:45 UTC | [d3d12] recycling "frame" resources and removing superfluous sync points (#5293) * refactoring to remove wait/sync points from kernel dispatch * debugging and bugfixes * refactoring wait/sync procedures * refactoring buffer signal checkpoints * improved time tracing * improved device selection and additional trace scoping features * addressing clang format issues * more clang format complaints... ¯\_(ツ)_/¯ * more clang format... * clang format... * nullptr -> NULL (0) * addressing code review comments * scoping the kernel argument setup code * addressing code review comments * clang format... Co-authored-by: Marcos Slomp <slomp@adobe.com> | 24 September 2020, 17:12:45 UTC |
| ee2cb21 | Steven Johnson | 23 September 2020, 21:47:31 UTC | Merge pull request #5283 from halide/wabt-bundle Bundle wabt objects into libHalide | 23 September 2020, 21:47:31 UTC |
| 44817ce | Andrew Adams | 23 September 2020, 20:15:35 UTC | Merge pull request #5295 from halide/abadams/fix_generate_output_snippets Rename LINES to INTERESTING_LINES | 23 September 2020, 20:15:35 UTC |
| d638d81 | Andrew Adams | 23 September 2020, 19:34:55 UTC | Rename LINES to INTERESTING_LINES Some terminals treat LINES as a special var, breaking this script | 23 September 2020, 19:34:55 UTC |
| 21e3f96 | Steven Johnson | 22 September 2020, 16:40:28 UTC | Merge pull request #5290 from NewProggie/patch-1 Fix typo in lesson 21 | 22 September 2020, 16:40:28 UTC |
| 671530a | Kai Wolf | 22 September 2020, 07:32:06 UTC | Fix typo in lesson 21 | 22 September 2020, 07:32:06 UTC |
| f5a764f | Volodymyr Kysenko | 21 September 2020, 16:52:03 UTC | Merge pull request #4873 from halide/vksnk/vector-ramp Support for multi-dim vectorization | 21 September 2020, 16:52:03 UTC |
| 7178b83 | Alex Reinking | 21 September 2020, 02:22:03 UTC | Re-enable CUDAVectorize tests. Fixes #4554. (#5286) | 21 September 2020, 02:22:03 UTC |
| f256e8f | Volodymyr Kysenko | 19 September 2020, 20:24:58 UTC | Address review comments | 19 September 2020, 20:24:58 UTC |
| ab54b55 | Alex Reinking | 18 September 2020, 23:36:49 UTC | bundle wabt objects into libHalide | 18 September 2020, 23:40:03 UTC |
| 26ebef3 | Steven Johnson | 17 September 2020, 22:23:39 UTC | Merge pull request #5279 from halide/srj-tidy Appease clang-tidy | 17 September 2020, 22:23:39 UTC |
| b6da613 | Steven Johnson | 17 September 2020, 22:17:44 UTC | Merge pull request #5275 from halide/srj/simplify-if-then-else Simplify Call::if_then_else | 17 September 2020, 22:17:44 UTC |
| bd0a6a6 | Steven Johnson | 17 September 2020, 17:41:20 UTC | Merge pull request #5277 from mirsadm/mm/new_line_fix Missing new line causes OpenCL programs to fail to compile when targetting cl_half/cl_doubles | 17 September 2020, 17:41:20 UTC |
| afd6e93 | Volodymyr Kysenko | 17 September 2020, 05:28:55 UTC | Merge branch 'master' into vksnk/vector-ramp | 17 September 2020, 05:28:55 UTC |
| e0f66bc | Steven Johnson | 16 September 2020, 21:42:22 UTC | Appease clang-tidy Remove two unused 'using' declarations. | 16 September 2020, 21:42:22 UTC |
| e53d04e | Mirsad Makalic | 16 September 2020, 07:22:20 UTC | -Missing new line | 16 September 2020, 19:11:12 UTC |
| f2aa702 | Steven Johnson | 16 September 2020, 16:45:44 UTC | Update Simplify_Call.cpp | 16 September 2020, 16:45:44 UTC |
| bd40529 | Steven Johnson | 16 September 2020, 01:15:25 UTC | Update Simplify_Call.cpp | 16 September 2020, 01:15:25 UTC |
| c63e425 | Steven Johnson | 16 September 2020, 00:22:38 UTC | Simplify Call::if_then_else According to @abadams, Call::if_then_else promises to evaluate exactly one (but not both) of the true and false exprs, so we should simplify cases in which the condition is provably true or false. | 16 September 2020, 00:22:38 UTC |
| db901f7 | Alex Reinking | 15 September 2020, 21:31:54 UTC | Packaging/fix windows (#5274) * fix cmake dir location * fix windows packaging again | 15 September 2020, 21:31:54 UTC |
| 58d78de | Alex Reinking | 15 September 2020, 21:25:03 UTC | fix cmake dir location (#5273) | 15 September 2020, 21:25:03 UTC |
| aef3f1b | Steven Johnson | 15 September 2020, 18:03:50 UTC | Merge pull request #5228 from halide/refactor/autoschedulers Refactor the autoschedulers to their own directory. | 15 September 2020, 18:03:50 UTC |
| 2ed3e14 | Andrew Adams | 15 September 2020, 00:35:51 UTC | More expedient hackery to handle rpath on OS X issues | 15 September 2020, 00:35:51 UTC |
| 40c9dc6 | Andrew Adams | 14 September 2020, 21:09:19 UTC | Merge | 14 September 2020, 21:09:19 UTC |
| 5ceeffb | Andrew Adams | 14 September 2020, 21:07:32 UTC | Makefile rpath fixes for os x | 14 September 2020, 21:07:32 UTC |
