71a86c9 | Steven Johnson | 20 August 2020, 22:39:50 UTC | Merge branch 'master' into vksnk/compute_with_async | 20 August 2020, 22:39:50 UTC |
25f8231 | Steven Johnson | 20 August 2020, 22:37:30 UTC | Allow compile_to_multitarget() to emit object files (#5183) * Allow compile_to_multitarget() to emit object files (Issue #5169) * Update Module.cpp * Update Module.cpp * Smarten compile_to_multitarget * c_source should be single, not multi * Fix apps/linear_algebra * Revert "Fix apps/linear_algebra" This reverts commit 01c15b40c86893ae30820dec33e7057f85b15bc6. * Update Module.cpp * Fixes * Don't substitute _ for - Co-authored-by: Alex Reinking <alex.reinking@gmail.com> | 20 August 2020, 22:37:30 UTC |
4abb189 | Steven Johnson | 19 August 2020, 18:31:01 UTC | Merge branch 'master' into vksnk/compute_with_async | 19 August 2020, 18:31:01 UTC |
d1f34da | Steven Johnson | 19 August 2020, 18:29:04 UTC | fix for trunk llvm, try #2 (#5198) Previous fix broke LLVM 11 (I was too eager to land, sorry) | 19 August 2020, 18:29:04 UTC |
640214d | Steven Johnson | 19 August 2020, 18:13:10 UTC | Fix for trunk LLVM (#5197) | 19 August 2020, 18:13:10 UTC |
eba0790 | Volodymyr Kysenko | 19 August 2020, 06:04:57 UTC | format | 19 August 2020, 06:04:57 UTC |
2a3cce1 | Volodymyr Kysenko | 19 August 2020, 06:01:06 UTC | Huh, do I have different clang-format | 19 August 2020, 06:01:06 UTC |
f8feb6f | Volodymyr Kysenko | 19 August 2020, 04:59:57 UTC | Update CMake | 19 August 2020, 04:59:57 UTC |
5734295 | Volodymyr Kysenko | 19 August 2020, 04:47:53 UTC | Update comments | 19 August 2020, 04:47:53 UTC |
fada648 | Volodymyr Kysenko | 19 August 2020, 04:30:50 UTC | Add a check that all functions scheduled with compute_with have matching async | 19 August 2020, 04:30:50 UTC |
365b2e3 | Volodymyr Kysenko | 19 August 2020, 03:32:11 UTC | Extend ForkAsyncProducers to support functions scheduled with compute_with | 19 August 2020, 03:32:11 UTC |
ef37487 | Volodymyr Kysenko | 17 August 2020, 18:02:27 UTC | Merge pull request #5185 from halide/vksnk/compute_with_store_at Fix #5178: Fix the case when functions from the fused group have different store_levels | 17 August 2020, 18:02:27 UTC |
a0d3237 | Andrew Adams | 17 August 2020, 17:14:40 UTC | Merge pull request #5187 from halide/abadams/reschedule_bgu Reschedule BGU to fix performance regression | 17 August 2020, 17:14:40 UTC |
9669817 | Andrew Adams | 16 August 2020, 20:54:08 UTC | Reschedule BGU to fix performance regression BGU on CUDA had regressed from its stated performance due to the atomic floating point adds being compiled to CAS loops due to complex indexing expressions diverging on the LHS and RHS of the +=. Inlining less stuff into the += operations makes it succeed again, and the schedule was improved with a few other tweaks. Longer-term we need a first-class way to represent += so that we're not sensitive to this sort of divergence. | 16 August 2020, 20:54:08 UTC |
e280037 | Volodymyr Kysenko | 15 August 2020, 02:35:42 UTC | Handle the case when the same function is build multiple times | 15 August 2020, 02:35:42 UTC |
051d674 | Volodymyr Kysenko | 15 August 2020, 00:42:44 UTC | make format | 15 August 2020, 00:42:44 UTC |
277b5db | Volodymyr Kysenko | 15 August 2020, 00:41:50 UTC | Fix the case when functions from the fused group have different store_levels | 15 August 2020, 00:41:50 UTC |
9f55e10 | Andrew Adams | 14 August 2020, 21:38:39 UTC | Merge pull request #5182 from halide/abadams/reschedule_stencil_chain Add memory staging to stencil chain | 14 August 2020, 21:38:39 UTC |
3177019 | Steven Johnson | 14 August 2020, 21:05:40 UTC | Don't allow Target strings without complete arch-bits-os (#5181) * Don't allow Target strings without complete arch-bits-os We previously accepted 'incomplete' Target strings (filling in host attributes for arch-bits-os if unspecified); we thought this would be a convenience, but in practice, this is usually indicative of an error or typo. This changes to make the Target(string) ctor assert-fail if the resulting target has an unspecified arch-bits-os. * Update target.py * Update Target.cpp * Update Target.cpp | 14 August 2020, 21:05:40 UTC |
b7cf1a1 | Andrew Adams | 14 August 2020, 17:43:07 UTC | Merge branch 'abadams/reschedule_stencil_chain' of https://github.com/halide/Halide into abadams/reschedule_stencil_chain | 14 August 2020, 17:43:07 UTC |
39c1a9a | Andrew Adams | 14 August 2020, 17:42:56 UTC | Explanatory comments for .in() usage | 14 August 2020, 17:42:56 UTC |
2a46538 | Andrew Adams | 14 August 2020, 16:06:12 UTC | Merge pull request #5184 from halide/abadams/fix_potential_gpu_deadlock Fix a source of GPU barrier deadlocks | 14 August 2020, 16:06:12 UTC |
cd9a0ae | Steven Johnson | 13 August 2020, 22:12:20 UTC | Merge branch 'master' into abadams/reschedule_stencil_chain | 13 August 2020, 22:12:20 UTC |
1d49c70 | Alex Reinking | 13 August 2020, 20:47:47 UTC | Merge pull request #5135 from halide/cpack Flesh out CPack packaging for releases. | 13 August 2020, 20:47:47 UTC |
e3606cc | Andrew Adams | 13 August 2020, 17:37:30 UTC | Fix GPU barrier deadlocks Partition loops shouldn't mess with serial loops containing thread barriers, potentially causing warp divergence and deadlock (seen in some obscure lens blur schedules). Also we were generating too many thread barriers in a branch where the base mutator class was accidentally always mutating something, so there's a change to FuseGPUThreadLoops to make it more bug-resistant. Without these additional barriers I have been unable to come up with a case where a barrier ends up somewhere that would deadlock, so no test. | 13 August 2020, 18:41:13 UTC |
98a116a | Steven Johnson | 13 August 2020, 16:31:11 UTC | Clean up is-jit-compiled checks in Pipeline (#5172) * Clean up is-jit-compiled checks in Pipeline Because WebAssembly is a special beast, the way it is 'jitted' is special, and the checks to avoid redundant jitting needed extra logic in compile_jit(). Unfortunately there was another place in Pipeline that also needed this special casing. This PR adds a `get_compiled_jit_target()` bottleneck to consolidate this. * defined() -> has_unknowns() | 13 August 2020, 16:31:11 UTC |
fa1abba | Steven Johnson | 13 August 2020, 16:30:46 UTC | Add infer_input_bounds(vector<int>) (#5174) * Add infer_input_bounds(vector<int>) Add a variant of infer_input_bounds() that takes an explicit vector of int, rather than the up-to-4-int version that is a holdover from the buffer_t days; deprecate the old version; convert all existing code to use the new one. Note that I'm using a new overload (with an initializer-list) as a way to subvert the mis-binding of `{}` and `{1}` to the deprecated function; this adds a trivial amount of overhead but (I think) allows us to ensure that converted code probably avoids the deprecated method. * Update Func.cpp | 13 August 2020, 16:30:46 UTC |
f6dcdde | Andrew Adams | 13 August 2020, 02:08:43 UTC | Merge pull request #5177 from halide/abadams/fix_stencil_chain_gpu_schedule Schedule last stage of stencil chain on GPU too | 13 August 2020, 02:08:43 UTC |
bc066f9 | Andrew Adams | 12 August 2020, 22:45:47 UTC | Add memory staging to stencil chain | 12 August 2020, 22:45:47 UTC |
f7528c2 | Steven Johnson | 12 August 2020, 21:06:03 UTC | Merge pull request #5176 from halide/srj-hvx-codegen Remove unnecessary call to halide.hexagon.pack.vh in CG_HVX | 12 August 2020, 21:06:03 UTC |
d1592d1 | Steven Johnson | 12 August 2020, 17:08:02 UTC | Merge pull request #5155 from halide/abadams/add_missing_boundary_condition_overload Add missing overload for boundary conditions on a buffer | 12 August 2020, 17:08:02 UTC |
2e2649f | Andrew Adams | 11 August 2020, 19:54:08 UTC | Merge remote-tracking branch 'origin/master' into abadams/add_missing_boundary_condition_overload | 11 August 2020, 19:54:08 UTC |
b8ad19f | Andrew Adams | 11 August 2020, 19:12:57 UTC | Schedule last stage of stencil chain on GPU too | 11 August 2020, 19:12:57 UTC |
528b46b | Steven Johnson | 11 August 2020, 01:15:17 UTC | Remove unnecessary call to halide.hexagon.pack.vh in CG_HVX In the degenerate case of shuffle_vector() calling vlut() to shuffle a vector that is wider than 256 elements, the code was incorrectly using halide.hexagon.pack.vh on a vector-of-bool; this used to be necessary, but hasn't been for a while, so clearly this code path wasn't being exercised. Remove the halide.hexagon.pack.vh and added a test case to exercise that path. Also, drive-by removal of #include "EliminateBoolVectors.h" from CGHVX since it is no longer used there. | 11 August 2020, 16:59:31 UTC |
26b5be4 | Andrew Adams | 09 August 2020, 16:05:40 UTC | Update fft app to use new boundary condition syntax | 09 August 2020, 16:05:40 UTC |
52da814 | Andrew Adams | 09 August 2020, 00:46:32 UTC | Merge pull request #5165 from halide/abadams/rungen_set_host_dirty Make sure to set_host_dirty in rungen | 09 August 2020, 00:46:32 UTC |
aa92f5c | Andrew Adams | 08 August 2020, 18:36:47 UTC | Make sure to set_host_dirty in rungen Otherwise synthetic inputs like 'random' end up being 'zero' | 08 August 2020, 18:36:47 UTC |
41e10e0 | Andrew Adams | 07 August 2020, 19:07:06 UTC | Merge branch 'abadams/add_missing_boundary_condition_overload' of https://github.com/halide/Halide into abadams/add_missing_boundary_condition_overload | 07 August 2020, 19:07:06 UTC |
46ae5ff | Andrew Adams | 07 August 2020, 19:06:57 UTC | Fix boundary condition in blur app | 07 August 2020, 19:06:57 UTC |
49d0476 | Steven Johnson | 07 August 2020, 04:31:25 UTC | Merge pull request #5162 from halide/srj-wasm-shell-version Update WASM_SHELL_VERSION | 07 August 2020, 04:31:25 UTC |
640c324 | Steven Johnson | 07 August 2020, 04:31:10 UTC | Merge pull request #5161 from halide/srj-infer-input-bounds Add a Target to the args of infer_input_bounds() | 07 August 2020, 04:31:10 UTC |
cfc125d | Alex Reinking | 07 August 2020, 01:10:26 UTC | Merge branch 'master' into cpack | 07 August 2020, 01:10:26 UTC |
4e2c25f | Alex Reinking | 07 August 2020, 01:08:47 UTC | Merge pull request #5163 from halide/srj-blur-fix Fix apps/blur on Hexagon | 07 August 2020, 01:08:47 UTC |
62814b1 | Alex Reinking | 07 August 2020, 00:54:55 UTC | Add Windows support for bundling LLVM. | 07 August 2020, 00:54:55 UTC |
8127ba9 | Steven Johnson | 06 August 2020, 23:22:07 UTC | Merge branch 'master' into abadams/add_missing_boundary_condition_overload | 06 August 2020, 23:22:07 UTC |
32c6fdb | Steven Johnson | 06 August 2020, 23:21:30 UTC | Merge pull request #5158 from halide/abadams/fix_nl_means_estimates Fix incorrect estimates for nl_means autoscheduler | 06 August 2020, 23:21:30 UTC |
fabd9b7 | Steven Johnson | 06 August 2020, 22:21:54 UTC | Fix apps/blur on Hexagon Don't rely on HALIDE_RUNTIME_HEXAGON; examine the metadata to see how we were compiled | 06 August 2020, 22:21:54 UTC |
ad9feb5 | Alex Reinking | 06 August 2020, 17:57:56 UTC | Fixing WASM by explicitly importing LLD. Consolidating LLVM deps finding. | 06 August 2020, 17:57:56 UTC |
e6d35ee | Alex Reinking | 06 August 2020, 08:05:03 UTC | Something with WABT is broken... trying to fix. | 06 August 2020, 08:05:03 UTC |
73b8560 | Alex Reinking | 06 August 2020, 03:41:21 UTC | Improve readability. | 06 August 2020, 03:41:21 UTC |
47f608b | Alex Reinking | 05 August 2020, 23:46:07 UTC | Major packaging improvements. 1. Support selecting library type in `find_package`. 2. Support linking to shared LLVM or bundling static LLVM in libHalide.a 3. Add packaging scripts. | 05 August 2020, 23:46:07 UTC |
a95e858 | Alex Reinking | 05 August 2020, 23:40:53 UTC | Fix export names for plugins and Python bindings | 05 August 2020, 23:40:53 UTC |
ad1f4fa | Alex Reinking | 05 August 2020, 23:38:41 UTC | Set library version to 10.0.0 | 05 August 2020, 23:38:41 UTC |
7ffc2e5 | Alex Reinking | 05 August 2020, 23:37:33 UTC | Rename WITH_EXCEPTIONS to HALIDE_WITH_EXCEPTIONS, like HALIDE_ENABLE_RTTI | 05 August 2020, 23:37:33 UTC |
cc4f8d1 | Alex Reinking | 05 August 2020, 23:33:22 UTC | Add VS output path to .gitignore, fix IDEA. | 05 August 2020, 23:33:22 UTC |
b89ec19 | Alex Reinking | 05 August 2020, 23:33:02 UTC | Run CMake formatter | 05 August 2020, 23:33:02 UTC |
ada32bc | Alex Reinking | 05 August 2020, 23:28:16 UTC | Merge pull request #5157 from halide/rename-to-nvptx Rename WITH_PTX to WITH_NVPTX so all WITH_* macros match LLVM component names | 05 August 2020, 23:28:16 UTC |
8b8a9ea | Steven Johnson | 05 August 2020, 23:04:18 UTC | Update WASM_SHELL_VERSION 8.6.325 is the minimum version that fixes some wasm-simd bugs that affect us. | 05 August 2020, 23:04:18 UTC |
3355a33 | Steven Johnson | 05 August 2020, 22:30:50 UTC | Add a Target to the args of infer_input_bounds() Currently, if the code isn't already jitted (or is jitted with NoBoundsQuery enabled), infer_input_bounds() falls back to whatever is in HL_JIT_TARGET. This is suboptimal, because it hides the fact that infer_input_bounds() relies on the value of that env var. It can also cause unintentional incorrect re-jits; e.g. if simd_op_check is run for HL_TARGET=wasm-32-wasmrt (and HL_JIT_TARGET not set at all), the infer_input_bounds() call can end up re-jitting for native code instead of WebAssembly. | 05 August 2020, 22:30:50 UTC |
09518b4 | Alex Reinking | 05 August 2020, 17:22:16 UTC | Merge branch 'master' into rename-to-nvptx | 05 August 2020, 17:22:16 UTC |
e3a573b | Alex Reinking | 05 August 2020, 17:21:43 UTC | Update TARGET_PTX to TARGET_NVPTX internally. | 05 August 2020, 17:21:43 UTC |
2e7b5ac | Volodymyr Kysenko | 04 August 2020, 22:53:21 UTC | Merge pull request #5159 from halide/vksnk/ucon-no-asserts Fix 'unused var' warning for user_context generated by CodeGen_C with no_asserts flag | 04 August 2020, 22:53:21 UTC |
d65846b | Volodymyr Kysenko | 04 August 2020, 21:07:19 UTC | Fix 'unused var' warning for user_context generated by CodeGen_C with no_asserts flag | 04 August 2020, 21:08:20 UTC |
fe2e78e | Andrew Adams | 04 August 2020, 20:57:01 UTC | Fix incorrect estimates for nl_means autoscheduler We run it on the parrot image, which is 5 megapixels. | 04 August 2020, 20:57:01 UTC |
74829ea | Alex Reinking | 04 August 2020, 20:18:38 UTC | Make Makefile use WITH_NVPTX in interface. | 04 August 2020, 20:18:38 UTC |
89f9a92 | Alex Reinking | 04 August 2020, 20:10:53 UTC | Rename WITH_PTX to WITH_NVPTX so all WITH_* macros match LLVM component names | 04 August 2020, 20:10:53 UTC |
6da8008 | Andrew Adams | 04 August 2020, 17:22:05 UTC | clang-format | 04 August 2020, 17:22:05 UTC |
3433d49 | Andrew Adams | 04 August 2020, 17:07:13 UTC | Add missing overload for boundary conditions on a buffer The following syntax did not work due to a lack of appropriate overload: repeat_edge(buffer, {{0, W}, {0, H}}) The version that worked was: repeat_edge(buffer, 0, W, 0, H) but that syntax has fallen out of favor This PR adds an overload to make the former syntax work, and explicitly deprecates the latter. Also removed inline qualifiers on some templated functions, because they're already inline, and changed an arg to be by const ref instead of pointlessly by value. | 04 August 2020, 17:07:13 UTC |
682d73f | Steven Johnson | 03 August 2020, 21:28:07 UTC | Merge pull request #5153 from halide/srj-tidy2 Fix clang-tidy presubmit | 03 August 2020, 21:28:07 UTC |
34f105d | Steven Johnson | 03 August 2020, 21:21:33 UTC | Fix clang-tidy presubmit Need to run `sudo apt-get update` before `sudo apt-get install` to avoid possibly stale install info. | 03 August 2020, 21:21:33 UTC |
0d6d7e0 | Steven Johnson | 30 July 2020, 20:11:35 UTC | Merge pull request #5150 from halide/srj-roundingmode Remove RoundingMode.h | 30 July 2020, 20:11:35 UTC |
5a4d098 | Steven Johnson | 30 July 2020, 17:00:06 UTC | Remove RoundingMode.h It is never included anywhere. | 30 July 2020, 17:00:06 UTC |
d63a800 | Andrew Adams | 30 July 2020, 16:36:57 UTC | Merge pull request #5148 from halide/abadams/fix_autoschedule_feature_transposition Fix transposed variable names | 30 July 2020, 16:36:57 UTC |
7dc9790 | Volodymyr Kysenko | 30 July 2020, 06:36:38 UTC | Merge pull request #5149 from halide/vksnk/c-no-asserts Fix 'unused variable' warning in the code generated by CodeGen_C with 'no_asserts' feature | 30 July 2020, 06:36:38 UTC |
f5efe01 | Volodymyr Kysenko | 30 July 2020, 00:53:32 UTC | Updated test | 30 July 2020, 00:53:32 UTC |
b481633 | Volodymyr Kysenko | 29 July 2020, 22:38:27 UTC | Remove create_assertion(string, string) | 29 July 2020, 22:38:27 UTC |
dcc56b7 | Volodymyr Kysenko | 29 July 2020, 21:29:43 UTC | Fix 'unused variable' warning in the code generated by CodeGen_C with 'no_asserts' feature | 29 July 2020, 21:33:43 UTC |
0e361d4 | Andrew Adams | 29 July 2020, 18:19:45 UTC | Fix transposed variable names | 29 July 2020, 18:19:45 UTC |
83b452f | Steven Johnson | 29 July 2020, 17:24:31 UTC | Merge pull request #5097 from halide/srj-wabt WebAssembly Updates | 29 July 2020, 17:24:31 UTC |
419c2dd | Steven Johnson | 28 July 2020, 22:12:36 UTC | Merge branch 'master' into srj-wabt | 28 July 2020, 22:12:36 UTC |
cc281d9 | Volodymyr Kysenko | 28 July 2020, 21:42:59 UTC | Merge pull request #5142 from halide/vksnk/frees-are-free-after-all-2 Handle Allocate without matching Free in the CodeGen_C | 28 July 2020, 21:42:59 UTC |
aa3041f | Alex Reinking | 28 July 2020, 17:21:13 UTC | Merge branch 'master' into srj-wabt | 28 July 2020, 17:21:13 UTC |
3d7854d | Alex Reinking | 28 July 2020, 17:12:00 UTC | fix option name in dependent option | 28 July 2020, 17:12:00 UTC |
0bb6172 | Steven Johnson | 28 July 2020, 17:07:47 UTC | Merge pull request #5137 from halide/tests-env Set HL_{JIT_}TARGET from CMake Halide_TARGET variable. | 28 July 2020, 17:07:47 UTC |
9db0d79 | Alex Reinking | 28 July 2020, 16:57:46 UTC | Fix case when LLVM reports WebAssembly capability, but liblld is not installed | 28 July 2020, 16:57:46 UTC |
f923ef7 | Steven Johnson | 28 July 2020, 16:40:47 UTC | Merge pull request #5143 from halide/vksnk/update-alloc-doc Update comment description for the Allocate node | 28 July 2020, 16:40:47 UTC |
fae67fc | Volodymyr Kysenko | 28 July 2020, 16:16:27 UTC | Update comment | 28 July 2020, 16:16:27 UTC |
1cd8bde | Volodymyr Kysenko | 28 July 2020, 15:00:51 UTC | make format | 28 July 2020, 15:00:51 UTC |
43c8f8d | Volodymyr Kysenko | 28 July 2020, 06:20:14 UTC | Update comment description for the Allocate node | 28 July 2020, 06:20:14 UTC |
3c977d6 | Volodymyr Kysenko | 28 July 2020, 06:02:53 UTC | Handle Allocate without matching Free in the CodeGen_C | 28 July 2020, 06:06:38 UTC |
f3b2abd | Andrew Adams | 26 July 2020, 22:08:24 UTC | Merge pull request #5139 from halide/abadams/no_asserts Don't codegen asserts in blocks when noasserts is on | 26 July 2020, 22:08:24 UTC |
3d4f56a | Andrew Adams | 25 July 2020, 19:00:56 UTC | Don't codegen asserts in blocks when noasserts is on Fixes #5138 | 25 July 2020, 19:00:56 UTC |
baa5e04 | Alex Reinking | 24 July 2020, 19:14:13 UTC | Set HL_{JIT_}TARGET from Halide_TARGET in tests. | 24 July 2020, 19:14:13 UTC |
30f2f3c | Alex Reinking | 24 July 2020, 19:13:15 UTC | Drop APPS_INTERNAL because it is confusing. | 24 July 2020, 19:13:15 UTC |
4232fe2 | Steven Johnson | 24 July 2020, 17:48:30 UTC | fixes | 24 July 2020, 17:48:30 UTC |
ef3724b | Steven Johnson | 24 July 2020, 17:47:10 UTC | Merge branch 'master' into srj-wabt | 24 July 2020, 17:47:10 UTC |
e9084ae | Steven Johnson | 24 July 2020, 17:33:29 UTC | Merge pull request #5124 from halide/srj-halide-target Rename the CMake var HL_TARGET -> Halide_TARGET, to disambiguate from the env var of the same name | 24 July 2020, 17:33:29 UTC |
a51673e | Steven Johnson | 23 July 2020, 23:57:15 UTC | Tweaks | 23 July 2020, 23:57:15 UTC |
e61d821 | Steven Johnson | 23 July 2020, 23:52:42 UTC | Update README_webassembly.md | 23 July 2020, 23:52:42 UTC |
18001d2 | Steven Johnson | 23 July 2020, 23:50:51 UTC | HALIDE_ -> Halide_ | 23 July 2020, 23:50:51 UTC |