462a145 | Dan Palermo | 10 January 2019, 03:06:45 UTC | Added support for halide_hexagon_set_thread_priority() - Added halide_hexagon_set_thread_priority IDL API to set the priority of threads created in the Halide runtime - Works similarly to halide_hexagon_set_performance[_mode] - A user would call halide_hexagon_set_thread_priority from the host app before dispatching a pipeline // Increase the default priority of Halide Hexagon threads halide_hexagon_set_thread_priority(NULL, 50); - Threads will be created with the previously specified priority - If not called, the default priority will be used - Compatible with old runtimes - priority won't be set w/ old - Set qurt_thread_pool.cpp priority default to 100 (to match master thread) - Limit priority range from 1 to 255 - Also set priority on existing master thread - Update QURT_THREAD_ATTR_PRIORITY_DEFAULT to match v62 qthread.h - Added more documentation to HalideRuntimeHexagonHost.h /** Set the priority for Hexagon threads. * - Valid priority values range from 1 to 255 * - Smaller number for higher priority * - The highest priority for a user thread is 1, 0 reserved for OS usage * - If not set, Halide thread priority will default to 100 * - This should be called before running a pipeline. */ // @{ extern int halide_hexagon_set_thread_priority(void *user_context, int priority); // @} - Added mmap_dlsym_libs that will search the list of loaded libs as 8996/sdm820 was failing to find halide_set_default_thread_priority using mmap_dlsym - For dlopenbuf platforms, using halide_get_symbol instead of dlsym directly in order to pick up Dillon's fallback technique of checking RTLD_SELF & RTLD_DEFAULT | 10 January 2019, 03:06:45 UTC |
402831d | Steven Johnson | 09 January 2019, 01:47:43 UTC | Merge pull request #3564 from halide/srj-llvm-fix Update for trunk LLVM | 09 January 2019, 01:47:43 UTC |
6e66b66 | Steven Johnson | 08 January 2019, 23:41:50 UTC | Update for trunk LLVM | 08 January 2019, 23:41:50 UTC |
b3065d5 | Steven Johnson | 07 January 2019, 16:46:06 UTC | Merge pull request #3559 from halide/srj-mulbounds Improve Bounds calc for uint16 Mul | 07 January 2019, 16:46:06 UTC |
f7a08ea | Andrew Adams | 07 January 2019, 05:01:39 UTC | Merge pull request #3556 from halide/fix_pseudostack_leak Fix a memory leak with conditional pseudostack slots | 07 January 2019, 05:01:39 UTC |
fd0ecc2 | Steven Johnson | 07 January 2019, 03:00:28 UTC | Fix piconit | 07 January 2019, 03:00:28 UTC |
e8c2e23 | Steven Johnson | 07 January 2019, 02:24:33 UTC | Improve Bounds calc for uint16 Mul For unsigned int multiplies in the uint16 range, we would do the overflow checking with signed integers, thus (incorrectly) assuming the bounds were neg-inf and pos-inf; changing to use unsigned math for this check allows us to correctly infer the finite bounds. | 07 January 2019, 02:24:33 UTC |
2541213 | Andrew Adams | 06 January 2019, 22:10:38 UTC | Merge pull request #3524 from halide/reset_unique_counter Reset unique name counters before the compilation of each target. | 06 January 2019, 22:10:38 UTC |
2efebd7 | Steven Johnson | 05 January 2019, 19:54:39 UTC | Merge pull request #3557 from halide/srj-err halide_error_constraints_make_required_region_smaller appears to be wrong | 05 January 2019, 19:54:39 UTC |
642db09 | Steven Johnson | 05 January 2019, 00:53:43 UTC | halide_error_constraints_make_required_region_smaller appears to be wrong constrained_max appears to be calculated incorrectly. also, drive-by fix to emit the dimension with the error message, | 05 January 2019, 00:53:43 UTC |
5c562b5 | Jing Pu | 04 January 2019, 23:40:18 UTC | change the counter back to std::atomic. | 04 January 2019, 23:40:18 UTC |
730dde8 | Andrew Adams | 04 January 2019, 23:00:17 UTC | Fix a memory leak with conditional pseudostack slots | 04 January 2019, 23:00:17 UTC |
5c1a5fa | Shoaib Kamil | 04 January 2019, 22:10:33 UTC | Merge pull request #3555 from halide/kamil/rungen_readme Clarify where registration files come from | 04 January 2019, 22:10:33 UTC |
f1ba855 | Shoaib Kamil | 04 January 2019, 21:37:03 UTC | Clarify where registration files come from | 04 January 2019, 21:37:03 UTC |
46938ac | Jing Pu | 03 January 2019, 21:27:03 UTC | Move reset before creating the generator instance. | 03 January 2019, 21:27:03 UTC |
1e58d2a | Steven Johnson | 03 January 2019, 18:10:06 UTC | Merge pull request #3551 from halide/srj-warn Augment warning/error flags for apps/ | 03 January 2019, 18:10:06 UTC |
294c78c | Steven Johnson | 02 January 2019, 18:37:45 UTC | Augment warning/error flags for apps/ Augment default CXXFLAGS in apps/Makefile.inc to match those in the main Makefile, cleaning errors as needed. | 02 January 2019, 18:51:09 UTC |
6bf4fdc | Steven Johnson | 02 January 2019, 17:43:13 UTC | Merge pull request #3547 from halide/srj-met CodeGen_LLVM::embed_constant_expr should simplify non-const exprs (Issue #3538) | 02 January 2019, 17:43:13 UTC |
53154f5 | Steven Johnson | 30 December 2018, 19:58:26 UTC | Merge pull request #3548 from LazyDodo/master Fix bad stubs when generator has no inputs. | 30 December 2018, 19:58:26 UTC |
d02acb4 | Ray Molenkamp | 29 December 2018, 19:23:37 UTC | Fix bad stubs when generator has no inputs. Issue outlined in #3546 issue 1. | 29 December 2018, 19:23:37 UTC |
c46857e | Steven Johnson | 29 December 2018, 01:20:30 UTC | CodeGen_LLVM::embed_constant_expr should simplify non-const exprs (Issue #3538) Any Expr that can be constant-folded should be at that point. | 29 December 2018, 01:20:30 UTC |
deaa629 | Steven Johnson | 28 December 2018, 17:44:46 UTC | Merge pull request #3542 from halide/srj-est Add a way to set estimates for Input<scalar[]> | 28 December 2018, 17:44:46 UTC |
a8a75e3 | Steven Johnson | 28 December 2018, 17:44:20 UTC | Merge pull request #3543 from halide/srj-ucon Special-case __user_context in RunGen.h for --estimates | 28 December 2018, 17:44:20 UTC |
61e6ab2 | Steven Johnson | 28 December 2018, 02:54:37 UTC | Special-case __user_context in RunGen.h for --estimates | 28 December 2018, 02:54:37 UTC |
4b9c64a | Steven Johnson | 28 December 2018, 01:41:34 UTC | Special-case bool params | 28 December 2018, 01:41:34 UTC |
df68a2c | Steven Johnson | 28 December 2018, 01:09:50 UTC | Add a way to set estimates for Input<scalar[]> There wasn't a way to do this before. Oops. | 28 December 2018, 01:09:50 UTC |
dd54192 | Andrew Adams | 27 December 2018, 23:55:13 UTC | Merge pull request #3541 from mbrukman/readme-cleanups Add code formatting and other cleanups. [skip ci] | 27 December 2018, 23:55:13 UTC |
bee3648 | Misha Brukman | 27 December 2018, 23:44:59 UTC | Add code formatting and other cleanups. [skip ci] Additional cleanups: * added a top-level heading for the project name to make it stand out in the README * fixed capitalization when referring to project names rather than their binary or directory names * replaced / (forward slash) with \ (backward slash) for sample Windows paths, and put them in code formatting as well * removed escaping of underscores since code formatting avoids Markdown interpretation | 27 December 2018, 23:49:22 UTC |
8a08a70 | Steven Johnson | 27 December 2018, 01:35:09 UTC | Merge pull request #3536 from halide/srj-pred Disable predicated store/load on x86 (Issue #3534) | 27 December 2018, 01:35:09 UTC |
04ebe92 | Steven Johnson | 27 December 2018, 00:33:15 UTC | Disable predicated store/load on x86 (Issue #3534) | 27 December 2018, 00:33:15 UTC |
b2b7f9c | Steven Johnson | 26 December 2018, 20:22:51 UTC | Merge pull request #3531 from halide/srj-regstub Remove RunGenStubs; add registration as a Generator output type | 26 December 2018, 20:22:51 UTC |
15c506c | Steven Johnson | 21 December 2018, 23:55:33 UTC | More fixes | 21 December 2018, 23:55:33 UTC |
4c47f33 | Steven Johnson | 21 December 2018, 23:42:46 UTC | Add missing suffix | 21 December 2018, 23:42:46 UTC |
2511b73 | Jing Pu | 21 December 2018, 23:33:26 UTC | change counters to thread_local. | 21 December 2018, 23:33:26 UTC |
74d65f5 | Steven Johnson | 21 December 2018, 23:12:30 UTC | Fix order | 21 December 2018, 23:12:30 UTC |
a7164a7 | Steven Johnson | 21 December 2018, 23:01:57 UTC | Remove RunGenStubs; add registration as a Generator output type This simplifies some use cases with multiple-registration for RunGen when used in conjunction with Bazel/Blaze. | 21 December 2018, 23:01:57 UTC |
0500f6f | Steven Johnson | 21 December 2018, 17:44:22 UTC | Merge pull request #3521 from halide/srj-regmulti Rework RunGen registration model | 21 December 2018, 17:44:22 UTC |
c24f924 | Andrew Adams | 21 December 2018, 00:42:26 UTC | Merge pull request #3525 from halide/abadams-patch-1 Update README.md | 21 December 2018, 00:42:26 UTC |
8de8386 | Andrew Adams | 21 December 2018, 00:37:37 UTC | Update README.md | 21 December 2018, 00:37:37 UTC |
4da6ae8 | Jing Pu | 20 December 2018, 23:03:51 UTC | Reset unique name counters before the compilation of each target. | 20 December 2018, 23:03:51 UTC |
a4b1601 | Andrew Adams | 20 December 2018, 20:52:34 UTC | Merge pull request #3522 from LazyDodo/master Fix HalideTraceViz on windows. | 20 December 2018, 20:52:34 UTC |
5a6459d | Ray Molenkamp | 20 December 2018, 19:48:22 UTC | fix whitespace | 20 December 2018, 19:48:22 UTC |
b7ddb59 | Ray Molenkamp | 20 December 2018, 19:46:36 UTC | Fix HalideTraceViz on windows. stdin and stdout are by default opened as text and need switching to binary before reading/writing binary data. | 20 December 2018, 19:46:36 UTC |
d25cd87 | Steven Johnson | 20 December 2018, 17:51:37 UTC | Fix | 20 December 2018, 17:51:37 UTC |
640dad7 | Steven Johnson | 20 December 2018, 01:18:29 UTC | Change link order | 20 December 2018, 01:18:29 UTC |
f2e5213 | Steven Johnson | 20 December 2018, 01:12:21 UTC | Tweak CMake rules | 20 December 2018, 01:12:21 UTC |
9203b1a | Steven Johnson | 20 December 2018, 01:03:03 UTC | Rework RunGen registration model Expand to allow for multiple filters to link against RunGen at once, selecting via a flag; this is of limited use for routine purposes, but should give us greater flexibility when experimenting with multiple auto-scheduler models. | 20 December 2018, 01:03:03 UTC |
0602b74 | Steven Johnson | 19 December 2018, 21:52:23 UTC | Merge pull request #3517 from minrk/python-cxxflags add $CXXFLAGS to $CCFLAGS in Python bindings | 19 December 2018, 21:52:23 UTC |
57d72e1 | Min RK | 19 December 2018, 14:36:57 UTC | respect $CXXFLAGS in Python bindings | 19 December 2018, 14:36:57 UTC |
8a1cc09 | Steven Johnson | 18 December 2018, 23:35:08 UTC | Merge pull request #3512 from halide/srj-bufconst Improve Buffer conversion checks when assertions are disabled | 18 December 2018, 23:35:08 UTC |
172078f | Steven Johnson | 18 December 2018, 19:28:56 UTC | Improve Buffer conversion checks when assertions are disabled assert_can_convert_from() was a complete no-op for builds that disable assertions, but there are meaningful compile-time checks that can (and should) be done even in these situations. | 18 December 2018, 19:28:56 UTC |
c3c34a0 | Andrew Adams | 17 December 2018, 23:29:32 UTC | Merge pull request #3505 from halide/bug_device_free Fix an issue that device frees are not injected. | 17 December 2018, 23:29:32 UTC |
76a8c27 | Jing Pu | 17 December 2018, 21:46:11 UTC | force a builder rerun. | 17 December 2018, 21:46:11 UTC |
3f26cbc | Jing Pu | 14 December 2018, 21:46:57 UTC | Merge branch 'master' into bug_device_free | 14 December 2018, 21:46:57 UTC |
d46b547 | Steven Johnson | 14 December 2018, 21:11:02 UTC | Merge pull request #3502 from halide/fix_rtti_flag Fix no-rtti flag detection | 14 December 2018, 21:11:02 UTC |
2bb0577 | Andrew Adams | 14 December 2018, 17:31:52 UTC | Merge pull request #3500 from halide/fix_sgemm_schedule Reschedule test/performance/matrix_multiplication | 14 December 2018, 17:31:52 UTC |
d6129e7 | Jing Pu | 14 December 2018, 01:49:14 UTC | remove asserts. | 14 December 2018, 01:49:14 UTC |
f6b15d7 | Jing Pu | 14 December 2018, 01:43:18 UTC | move finding last use into a separate pass. | 14 December 2018, 01:43:18 UTC |
f8c7d7d | Jing Pu | 13 December 2018, 23:23:35 UTC | Fix the bug that device frees are not injected. | 13 December 2018, 23:23:35 UTC |
8259e26 | Andrew Adams | 13 December 2018, 22:58:59 UTC | Fix no-rtti flag detection | 13 December 2018, 22:58:59 UTC |
d1160ae | Zalman Stern | 13 December 2018, 22:47:51 UTC | Merge pull request #3493 from benoitsteiner/master Added a #ifdef to use the large llvm CodeModel | 13 December 2018, 22:47:51 UTC |
16ef557 | Andrew Adams | 13 December 2018, 21:39:49 UTC | Reschedule test/performance/matrix_multiplication | 13 December 2018, 21:39:49 UTC |
9552e1a | Benoit Steiner | 12 December 2018, 18:31:36 UTC | Improved code formatting | 12 December 2018, 18:31:36 UTC |
32d4f80 | Benoit Steiner | 12 December 2018, 18:28:58 UTC | Added a CMake option to control which codemodel will be used | 12 December 2018, 18:28:58 UTC |
558fcf5 | Benoit Steiner | 12 December 2018, 18:19:34 UTC | Added a #ifdef to use the large llvm CodeModel instead of the default one (Small)e | 12 December 2018, 18:19:34 UTC |
9cc042c | Steven Johnson | 11 December 2018, 22:12:13 UTC | Merge pull request #3487 from halide/srj-dlopen Don't attempt to dlopen() empty strings | 11 December 2018, 22:12:13 UTC |
d0d6daf | Steven Johnson | 11 December 2018, 19:33:54 UTC | define NOMINMAX before windows.h | 11 December 2018, 19:33:54 UTC |
9d50c61 | Steven Johnson | 11 December 2018, 19:08:54 UTC | Don't attempt to dlopen() empty strings Apparently this is OK on Linux, but definitely not on OSX or elsewhere. | 11 December 2018, 19:08:54 UTC |
995d799 | Steven Johnson | 11 December 2018, 17:51:13 UTC | Merge pull request #3485 from halide/generator_plugins Add the ability for generators to load plugins | 11 December 2018, 17:51:13 UTC |
28a575a | Steven Johnson | 11 December 2018, 17:49:32 UTC | Merge pull request #3481 from halide/srj-rtti Build apps with same RTTI mode as libHalide | 11 December 2018, 17:49:32 UTC |
65c0400 | Steven Johnson | 10 December 2018, 17:30:38 UTC | Merge pull request #3475 from halide/conditional_unroll Add a temporary means to unroll loops only when they turn out to be constant-sized | 10 December 2018, 17:30:38 UTC |
4851546 | Steven Johnson | 10 December 2018, 17:19:40 UTC | Merge branch 'master' into srj-rtti | 10 December 2018, 17:19:40 UTC |
7c331d6 | Steven Johnson | 10 December 2018, 17:18:54 UTC | Merge pull request #3482 from halide/srj-ubsan Fix some UBSan warnings | 10 December 2018, 17:18:54 UTC |
dc4cde8 | Andrew Adams | 10 December 2018, 17:17:37 UTC | Add the ability for generators to load plugins | 10 December 2018, 17:17:37 UTC |
f169f50 | Steven Johnson | 09 December 2018, 21:48:45 UTC | Merge pull request #3483 from halide/srj-scopey Disallow copy ctor in various Scope utilities | 09 December 2018, 21:48:45 UTC |
205340a | Steven Johnson | 08 December 2018, 02:18:53 UTC | Disallow copy ctor in various Scope utilities Classes like ScopedFact will misbehave if they are accidentally copied (rather than moved); delete the copy ctor and default the move ctor. (Apparently VS2015 is copying ScopedFact when it could be moving it, which seems to be the cause of the debug-only crash; not sure if that's legal for C++11 or not, but this is a good fix in any event.) | 08 December 2018, 02:18:53 UTC |
5cb1bd8 | Steven Johnson | 08 December 2018, 00:07:04 UTC | Merge branch 'master' into conditional_unroll | 08 December 2018, 00:07:04 UTC |
18e6816 | Steven Johnson | 08 December 2018, 00:06:19 UTC | Merge pull request #3478 from halide/srj-vmla temporarily disable vmla.f32 and vmls.f32 (#3477) | 08 December 2018, 00:06:19 UTC |
4d57021 | Steven Johnson | 08 December 2018, 00:03:42 UTC | Fix some UBSan warnings Built and tested with -fsanitize=undefined; fixed a handful of technically-UB-but-no-big-deal issues reported: - misaligned loads - incorrect downcasting | 08 December 2018, 00:03:42 UTC |
f716264 | Steven Johnson | 07 December 2018, 23:50:34 UTC | Build apps with same RTTI mode as libHalide - Emit a flag into halide_config.xxx to indicate whether Halide was built with RTTI or not - modify apps/ to only use -fno-rtti when Halide is build with -fno-rtti (This greatly reduces various false positives when running with some sanitizers) | 07 December 2018, 23:50:34 UTC |
d45e63e | Andrew Adams | 07 December 2018, 22:11:21 UTC | Fix test for windows | 07 December 2018, 22:11:21 UTC |
d831da6 | Andrew Adams | 07 December 2018, 22:07:22 UTC | Add tracking issue | 07 December 2018, 22:07:22 UTC |
476ca87 | Andrew Adams | 07 December 2018, 22:05:27 UTC | Merge pull request #3476 from halide/custom_autoschedulers Make it possible to install an autoscheduler as a plugin | 07 December 2018, 22:05:27 UTC |
2ace34d | Andrew Adams | 07 December 2018, 18:27:18 UTC | Merge pull request #3474 from halide/backport_autosched_fixes A grab-bag of minor tweaks from the autoscheduler branch | 07 December 2018, 18:27:18 UTC |
d72e2dd | Steven Johnson | 07 December 2018, 18:25:01 UTC | temporarily disable vmla.f32 and vmls.f32 (#3477) | 07 December 2018, 18:25:01 UTC |
9d1a838 | Andrew Adams | 07 December 2018, 17:36:21 UTC | Merge pull request #3468 from halide/restrict_fused_groups Restrict fused groups to require either all or none of the funcs to be used. | 07 December 2018, 17:36:21 UTC |
2386fed | Zalman Stern | 07 December 2018, 07:31:29 UTC | Merge pull request #3472 from halide/srj-meta Emit halide_scalar_value_t with full storage (Issue #3470) | 07 December 2018, 07:31:29 UTC |
14aa85d | Andrew Adams | 06 December 2018, 23:25:26 UTC | Make it possible to install an autoscheduler as a plugin | 06 December 2018, 23:25:26 UTC |
596a2ca | Andrew Adams | 06 December 2018, 22:53:04 UTC | Add a temporary means to unroll loops only when they turn out to be constant-sized | 06 December 2018, 22:53:04 UTC |
5f736c0 | Andrew Adams | 06 December 2018, 22:13:15 UTC | Allow for unused input images | 06 December 2018, 22:13:15 UTC |
54a3972 | Shoaib Kamil | 06 December 2018, 22:08:59 UTC | Merge pull request #3473 from halide/kamil/remove_stale_runtime_comment Remove stale comment about halide_set_num_threads() | 06 December 2018, 22:08:59 UTC |
2aec980 | Shoaib Kamil | 06 December 2018, 22:06:56 UTC | Remove stale comment about halide_set_num_threads() | 06 December 2018, 22:06:56 UTC |
8a81e25 | Andrew Adams | 06 December 2018, 22:02:54 UTC | More helpful error message for pipeline loops | 06 December 2018, 22:02:54 UTC |
4667456 | Andrew Adams | 06 December 2018, 22:02:05 UTC | Add printer for tail strategies | 06 December 2018, 22:02:05 UTC |
9a0b5cb | Andrew Adams | 06 December 2018, 22:01:32 UTC | Forward missing method from Stage to Func | 06 December 2018, 22:01:32 UTC |
e54d6a6 | Andrew Adams | 06 December 2018, 22:00:57 UTC | negative-sized allocations are ok if unused | 06 December 2018, 22:00:57 UTC |
e13eb67 | Andrew Adams | 06 December 2018, 22:00:13 UTC | Add missing include | 06 December 2018, 22:00:13 UTC |
90eb020 | Andrew Adams | 06 December 2018, 21:59:45 UTC | Print correct expression in Autosched error message | 06 December 2018, 21:59:45 UTC |
798de3d | Steven Johnson | 06 December 2018, 21:46:42 UTC | Use Type::bytes() | 06 December 2018, 21:46:42 UTC |
c840203 | Steven Johnson | 06 December 2018, 21:37:55 UTC | Round up size so that UInt(1) -> 1 byte | 06 December 2018, 21:37:55 UTC |