sort by:
Revision Author Date Message Commit Date
46e9b3f Bring back broadcasting specialization. 18 November 2020, 07:26:21 UTC
7f5e139 Add padding to bias. 18 November 2020, 07:26:11 UTC
33cfdfd Fix compare_vs_tflite 18 November 2020, 06:54:22 UTC
8e782e1 Add --working_set option. 18 November 2020, 01:52:47 UTC
f286b71 Add operator/ to Interval 18 November 2020, 01:39:37 UTC
3385c59 CropShape -> Box. 18 November 2020, 01:24:21 UTC
c856a38 Merge branch 'interpret_nn' of https://github.com/halide/Halide into interpret_nn 18 November 2020, 01:19:32 UTC
bd9731b Add Interval helper to replace std::pair<int, int> 18 November 2020, 01:17:16 UTC
a66a10a Add UInt64 type support 18 November 2020, 00:49:46 UTC
cd633b7 Don't mess with subgraph name -- it's debug-only and sometimes bogus 17 November 2020, 23:54:24 UTC
81d2ba0 Merge branch 'interpret_nn' of https://github.com/halide/Halide into interpret_nn 17 November 2020, 20:04:23 UTC
2fba24e Sloppy bounds inference fixes. 17 November 2020, 19:21:35 UTC
fd7f3a7 Fuse adjacent ops. 17 November 2020, 19:21:15 UTC
5b1f43a Merge branch 'master' into interpret_nn 17 November 2020, 18:41:06 UTC
5137369 Update Android NDK support in apps/ (#5454) * Update Android NDK support in apps/ The Makefile help for apps/ assumed a fairly old version of the Android NDK (~2017). This updates to assume r19 or later: - No need to do `make-standalone-toolchain` anymore - Clang instead of GCC - assume static linking of libc++ is the right default I am limited in my ability to test on-device here (I don't have a device that will let me test HVX easily). apps/blur works fine onmy P2XL after this change, though. Also, drive-by fix to apps/simd_op_check to remove hvx_64 references. 17 November 2020, 18:38:06 UTC
bf94102 New scheduling algorithm kind of works (still disabled). 17 November 2020, 18:29:23 UTC
e2cfcae Add some helpers. 17 November 2020, 06:11:14 UTC
a295ba2 Remove redundant parameter. 17 November 2020, 01:31:40 UTC
f4053c2 Merge branch 'master' into interpret_nn 17 November 2020, 00:00:42 UTC
f60efd8 Add run_benchmark_on_device.sh 17 November 2020, 00:00:38 UTC
c0ca4ff Revert C++11 usage in hexagon_remote (#5446) * Revert C++11 usage in hexagon_remote Building hexagon_remote with C++11 ends up inserting some unwanted C++11-related symbols (__gxx_personality_v0) that aren't present. This reverts those changes, and modifies HalideRuntime.h so that most of the `__cplusplus` checks are now `__cplusplus >= 201103L` (i.e., C++ convenience features only exist when compiled under C++11 or later). * Appease MSVC idiocy 16 November 2020, 20:34:33 UTC
a592c45 Merge branch 'master' into interpret_nn 16 November 2020, 18:00:46 UTC
e3fc746 Fix typo in comment: jit_compile -> compile_jit (#5453) 16 November 2020, 17:34:31 UTC
e5ff1b6 Revert "Push lets near their uses (whenever possible) while CSEing += (#5387)" (#5448) This reverts commit d5e425ee29f165b9de3527a267837093f99e59be. 16 November 2020, 17:33:32 UTC
e090687 Update compare_vs_tflite.cpp 13 November 2020, 23:29:15 UTC
e586441 Fix AveragePool + MaxPool 13 November 2020, 23:15:51 UTC
233fd32 Fix AddOp 13 November 2020, 23:07:12 UTC
5375536 Update ops.cpp 13 November 2020, 22:26:55 UTC
ecf6e66 Clean up FillWithRandom 13 November 2020, 22:21:48 UTC
b75aef0 Update compare_vs_tflite.cpp 13 November 2020, 22:08:50 UTC
b212af2 Merge branch 'interpret_nn' of https://github.com/halide/Halide into interpret_nn 13 November 2020, 02:39:17 UTC
da7e6aa Interpreter now owns Model objects, so Tensor can be stateful. 13 November 2020, 02:38:55 UTC
bd8b276 Fix DepthwiseConv2D 13 November 2020, 02:18:07 UTC
f5e4332 Init stride in ParseTensor rather than Allocate() 13 November 2020, 00:28:10 UTC
caaa495 Fix minor bug in tflite_exploder 13 November 2020, 00:27:13 UTC
8ad70b3 Merge branch 'master' of https://github.com/halide/Halide into interpret_nn 12 November 2020, 23:32:05 UTC
fca1b44 update readme.md for hexagon (#5444) 12 November 2020, 23:24:18 UTC
37a1d13 Minor formatting 12 November 2020, 02:01:03 UTC
62bf0f2 Add --verbose 11 November 2020, 23:14:21 UTC
35da402 default --threads to 1 11 November 2020, 22:58:25 UTC
4c6d1a7 formatting fixes 11 November 2020, 22:58:05 UTC
86ef56a Update compare_vs_tflite.cpp 11 November 2020, 22:42:02 UTC
d76bbd2 Various changes: - Only show a few diffs - limit range for int random - add `-threads`, set threads for tflite and halide 11 November 2020, 22:40:37 UTC
acdad1b Improve random-seed handling for inputs 11 November 2020, 22:10:59 UTC
cd41e7a Use halide_benchmark for timing 11 November 2020, 22:03:53 UTC
813d231 handle float16 11 November 2020, 19:52:10 UTC
94c3304 Mark all targets .SECONDARY 11 November 2020, 19:51:06 UTC
846e971 logging tweaks in compare_vs_tflite 11 November 2020, 19:26:19 UTC
181c974 Update VectorizeLoops.cpp 11 November 2020, 18:59:48 UTC
921f3cc Merge branch 'master' into interpret_nn 11 November 2020, 18:55:53 UTC
c8ab278 add conditions for one-sided bounds to LT/GT, LE/GE, EQ, NEQ (#5438) add conditions for one-sided bounds to LT/GT, LE/GE, EQ, NEQ 11 November 2020, 18:41:34 UTC
eac39c7 is_zero -> is_const_zero (#5436) is_zero -> is_const_zero 11 November 2020, 17:25:31 UTC
6a11b14 Add compare_vs_tflite 11 November 2020, 02:03:58 UTC
78a8b46 Add compare_vs_tflite 11 November 2020, 02:00:15 UTC
a21fc7c Make sure Tensor::Data<void>() sets type 11 November 2020, 01:55:48 UTC
0853b06 Update interpreter.cpp 11 November 2020, 01:48:23 UTC
8259941 Fix bonehead mistake 11 November 2020, 01:45:53 UTC
c9d635a Add ModelInterpreter::Inputs() 11 November 2020, 01:29:50 UTC
d21febf Allow IsType<void>() -> true 11 November 2020, 00:14:53 UTC
f373770 Add 'verbose' option to ScheduleOptions 10 November 2020, 23:41:17 UTC
f827283 Add ModelInterpreter::Outputs(), implement GetTensor() 10 November 2020, 23:02:20 UTC
364d1da Add copy/move ctors to all base classes Ensure that things like Tensor can't be accidentally copied 10 November 2020, 22:19:48 UTC
dd768c0 Merge branch 'master' into interpret_nn 10 November 2020, 21:37:15 UTC
d53b9ef Make simd_op_check.h more overridable (#5442) 10 November 2020, 18:11:14 UTC
f524111 FIx for trunk LLVM (#5435) 05 November 2020, 19:55:10 UTC
8b333bc is_zero is a compile-time check if it's the constant zero (#5433) I think for this code we want a runtime check if it's zero or not. 05 November 2020, 17:10:44 UTC
c45124f Use new flag for clang tidy behavior in cmake (#5428) * Use new flag for clang tidy 04 November 2020, 17:35:32 UTC
ece39b6 Upgrade PyBind11 version in CMake to 2.5 (#5427) v2.4.3 can generate a lot of compiler warnings under C++17; v2.5 fixes there. Note 1: I am unsure about the issues with keeping in sync with Ubuntu 20.04; tagging @alexreinking for comments Note 2: the current version of PyBind11 is actually v2.6, but it has many more changes and upgrading looks nontrivial; deliberately holding off on that upgrade for now. 04 November 2020, 00:12:15 UTC
cd3f1d8 Fix various clang-tidy issues (#5426) * Fix various clang-tidy issues For some reason, these are only getting flagged under C++17 builds, but they are legit (minor) issues we want to fix. 03 November 2020, 01:31:02 UTC
69643ce Fix for trunk LLVM (#5425) 02 November 2020, 23:40:01 UTC
2c488d7 tfexploder: append the builtin op name to the output file also, drive-by removal of unused var in tflite_parser 02 November 2020, 19:04:43 UTC
883bab8 Gather deps for interpreter into a single var 02 November 2020, 18:25:47 UTC
3381612 Add a from-buffer call to tfliteparser (avoid need for callers to know tflite internals) 02 November 2020, 18:08:33 UTC
61222a1 Disable CMake builds files for interpret_nn 02 November 2020, 17:55:09 UTC
982a851 Merge branch 'master' into interpret_nn 02 November 2020, 17:24:25 UTC
025f054 Remove superfluous boundary condition in resize (#5414) * Remove superfluous boundary condition in resize and tweak schedule. ~10-15% faster 30 October 2020, 17:12:05 UTC
6994a15 fix GCC -> GNU in generator expressions (#5419) 30 October 2020, 16:20:38 UTC
acb818b Add missing operator^ overloads in CppVector (Issue #5415) (#5416) 30 October 2020, 16:18:39 UTC
91b88b5 Don't declare round/roundf for multi-threaded MSVC builds (Issue #5403) (#5417) 30 October 2020, 16:18:09 UTC
feb81a2 doxygen wasn't finding the runtime (#5410) 28 October 2020, 21:08:39 UTC
3fd654f loosen preconditions on div by single point in Bounds.cpp (#5407) Loosen preconditions on div by single point for integers 28 October 2020, 21:07:57 UTC
d5e425e Push lets near their uses (whenever possible) while CSEing += (#5387) * Push lets near their uses (whenever possible) while CSEing += The CSE pass tries to do CSE jointly on the index and value of store nodes. This is to stop: f[x] = f[x] + y from turning into f[x] = f[z] + y This is because the two 'x' indices are not CSE'd together. However, one problem with jointly CSEing the store index and vaues is that after CSEing, the pass puts the lets (the values that were CSE'd) before the new store. That is we get, let(t0...) let(t1...) f[t0] = f[t0] + function_of(t1) Suppose however, there was nothing to CSE between the store index and value, that is the index was unchanged after CSE. In that case, moving lets before the store puts the lets too far away from their uses. This is ok except, there are passes like LoopCarry that are beneficial when they are able to see a long continuous block of stores (eg. when unrolling). But now, they'll see a long sequence of LetStmts. Instead, if the index was unchanged after CSE, we should build up the stores as f[x] = let(t0.. in (value)); This way LoopCarry is given a chance to see a series of stores. Handling this in CSE, means the LoopCarry pass need not be complicated. Change-Id: Iae19e1f69a6b38f3224a64b0c4781533e3862970 * In CSE, when we bundle together the store index and store value, create LetStmts only for CSE'd values that are needed by the store index. The rest can be Let expressions around the store value. * Formatting fix for printing the IR after CSE * Be smarter about pushing lets near their uses when CSEing += In a previous patch, in the code that handles CSEing += store operations, we weren't general enough; we only handled the specific case when the store index remained unchanged when CSE'd together with the store value. This patch is more general. Even if the store index changes, only the use-def chains created from the values used by the new store index that end in defs created by CSE are retained as Let stmts around the new store. Others go as Let expressions around the store value. * Fix some formatting issues exposed by clang-format and remove the inclusion of ExprUsesVar.h in CSE.cpp because it is not used anymore. * Use override in GetVarsUsed in CSE.cpp 28 October 2020, 20:19:40 UTC
7896513 Update IntrusivePtr.h 28 October 2020, 00:38:57 UTC
6e0b499 Update IntrusivePtr.h 28 October 2020, 00:38:57 UTC
e96a0e9 Update IntrusivePtr.h 28 October 2020, 00:38:57 UTC
116a3f0 More clang-tidy cleanup 28 October 2020, 00:38:57 UTC
864955e web assembly threads + demo app (#5395) Add webassembly demo app and enable webassembly in the makefile. 26 October 2020, 23:23:46 UTC
deb10c6 Add -fno-threadsafe-statics 26 October 2020, 17:12:50 UTC
f853c89 Upgrade hexagon_remote/Makefile to use C++11 HalideRuntime.h now requires at least C++11 (for C++ files), so ensure that we pass `-std=c++11` for all those when building the remote. 26 October 2020, 17:12:50 UTC
9ea017c Fix transitive dependencies 26 October 2020, 16:47:38 UTC
a0b5984 Refactor various things into app_util.h 24 October 2020, 00:40:07 UTC
be59cc4 Split interpret_nn.h/cpp to model.h/cpp and interpreter.h/cpp. 23 October 2020, 20:58:06 UTC
7142fa7 Add Stage.gpu_lanes to Python bindings. 23 October 2020, 17:19:50 UTC
fef108b Upgrade WABT version to 1.0.19 23 October 2020, 16:56:44 UTC
33e67ba Add tflite_exploder 23 October 2020, 00:19:46 UTC
2d1aee9 Add a test case for multiple argument memoize_tag to demonstrate usage. (#5393) * Add a test case for multiple argument memoize_tag usage to demonstrate how it works. * Fix formatting typo. 22 October 2020, 20:39:39 UTC
a86b66e Merge branch 'master' into interpret_nn 22 October 2020, 20:27:58 UTC
475c7a0 Use locally declared type. 22 October 2020, 16:35:54 UTC
4f072bc Fix bounds resulting in vector types. 22 October 2020, 16:35:54 UTC
2da26cf Merge branch 'interpret_nn' of https://github.com/halide/Halide into interpret_nn 22 October 2020, 04:46:26 UTC
back to top