en Copyright 2015 Java http://172.16.0.5:8080/history/llvm-project-15.0.7/openmp/libomptarget/plugins/exports#410bfa00 [openmp] Introduce optional plugin init/deinit functionsWill allow plugins to migrate away from using global variables tomanage lifetime, which will fix a segfault discovered in relation to D127432Reviewed By: jhuber6Differential Revision: https://reviews.llvm.org/D130712(cherry picked from commit 1f9d3974e444f95ddb600a6964ed14ded559e89c) List of files: /llvm-project-15.0.7/openmp/libomptarget/plugins/exports Thu, 28 Jul 2022 15:21:36 +0000 Jon Chesterfield <jonathanchesterfield@gmail.com> http://172.16.0.5:8080/history/llvm-project-15.0.7/openmp/libomptarget/plugins/exports#e01ce4e8 [Libomptarget] Add checks for CUDA subarchitecture using new infoThis patch extends the `is_valid_binary` routine to also check if thebinary's architecture string matches the one parsed from the runtime.This should allow us to only use the binary whose compute capabilitymatches, allowing us to support basic multi-architecture binaries forCUDA.Depends on D127432Reviewed By: jdoerfertDifferential Revision: https://reviews.llvm.org/D127505 List of files: /llvm-project-15.0.7/openmp/libomptarget/plugins/exports Fri, 10 Jun 2022 13:37:21 +0000 Joseph Huber <jhuber6@vols.utk.edu> http://172.16.0.5:8080/history/llvm-project-15.0.7/openmp/libomptarget/plugins/exports#10aa83ff [OpenMP] Allow to explicitly deinitialize device resourcesThere are two problems this patch tries to address:1) We currently free resources in a random order wrt. plugin and libomptarget destruction. This patch should ensure the CUDA plugin is less fragile if something during the deinitialization goes wrong.2) We need to support (hard) pause runtime calls eventually. This patch allows us to free all associated resources, though we cannot reinitialize the device yet.Follow up patch will associate one event pool per device/context.Differential Revision: https://reviews.llvm.org/D120089 List of files: /llvm-project-15.0.7/openmp/libomptarget/plugins/exports Thu, 17 Feb 2022 22:04:12 +0000 Johannes Doerfert <johannes@jdoerfert.de> http://172.16.0.5:8080/history/llvm-project-15.0.7/openmp/libomptarget/plugins/exports#f44e41af Runtime for Interop directiveThis implements the runtime portion of the interop directive.It expects the frontend and IRBuilder portions to be in placefor proper execution. It currently works only for GPUsand has several TODOs that should be addressed going forward.Reviewed By: RaviNarayanaswamyDifferential Revision: https://reviews.llvm.org/D106674 List of files: /llvm-project-15.0.7/openmp/libomptarget/plugins/exports Thu, 27 Jan 2022 20:15:25 +0000 Sri Hari Krishna Narayanan <snarayan@mcs.anl.gov> http://172.16.0.5:8080/history/llvm-project-15.0.7/openmp/libomptarget/plugins/exports#29df4ab3 [OpenMP][Offloading] Add support for event related interfacesThis patch adds the support form event related interfaces, which will be usedlater to fix data race. See D104418 for more details.Reviewed By: jdoerfert, ye-luoDifferential Revision: https://reviews.llvm.org/D108528 List of files: /llvm-project-15.0.7/openmp/libomptarget/plugins/exports Sat, 28 Aug 2021 20:24:06 +0000 Shilei Tian <tianshilei1992@gmail.com> http://172.16.0.5:8080/history/llvm-project-15.0.7/openmp/libomptarget/plugins/exports#d2f85d09 [OpenMP][Libomptarget] Adding `print_device_info` to RTL and `omptarget`This patch introduces a function in the device's plugin to print thedevice information. This patch relates to another patch that introducesa CLI tool to obtain the device information from the omplibrary directly.It is inspired by PGI's pgaccelinfo.The modifications are as follows:1. Introduce the optional `void __tgt_rtl_print_device_info(RTLdevID)` function into the RTL.2. Introduce the `bool __tgt_print_device_info(devID)` function into `omptarget` interface. Returns false if the RTL is not implemented3. Added `bool printDeviceInfo(RTLDevID)` to the `DeviceTy`4. Implement the `__tgt_rtl_print_device_info` for CUDA. Added additional CUDA Runtime calls.Reviewed By: jdoerfertDifferential Revision: https://reviews.llvm.org/D106751 List of files: /llvm-project-15.0.7/openmp/libomptarget/plugins/exports Wed, 28 Jul 2021 01:47:40 +0000 Jose M Monsalve Diaz <jmonsalvediaz@anl.gov> http://172.16.0.5:8080/history/llvm-project-15.0.7/openmp/libomptarget/plugins/exports#2b6f2008 [OpenMP] Add function for setting LIBOMPTARGET_INFO at runtimeSummary:This patch adds a new runtime function __tgt_set_info_flag that allows theuser to set the information level at runtime without using the environmentvariable. Using this will require an extern function, but will eventually beadded into an auxilliary library for OpenMP support functions.This patch required moving the current InfoLevel to a global variable which mustbe instantiated by each plugin.Reviewed By: jdoerfertDifferential Revision: https://reviews.llvm.org/D100774 List of files: /llvm-project-15.0.7/openmp/libomptarget/plugins/exports Wed, 21 Apr 2021 21:31:09 +0000 Joseph Huber <jhuber6@vols.utk.edu> http://172.16.0.5:8080/history/llvm-project-15.0.7/openmp/libomptarget/plugins/exports#542d9c21 [libomptarget] Load images in order of registrationThis makes sure that images are loaded in the order in which they are registered with libomptarget.If a target can load multiple images and these images depend on each other (for example if one image contains the programs target regions and one image contains library code), then the order in which images are loaded can be important for symbol resolution (for example, in the VE plugin).In this case: because the same code exist in the host binaries, the order in which the host linker loads them (which is also the order in which images are registered with libomptarget) is the order in which the images have to be loaded onto the device.Reviewed By: JonChesterfieldDifferential Revision: https://reviews.llvm.org/D95530 List of files: /llvm-project-15.0.7/openmp/libomptarget/plugins/exports Wed, 24 Feb 2021 17:14:35 +0000 Manoel Roemmer <manoel.roemmer@rwth-aachen.de> http://172.16.0.5:8080/history/llvm-project-15.0.7/openmp/libomptarget/plugins/exports#ec8f4a38 [OpenMP][Libomptarget] Introduce Remote Offloading PluginThis introduces a remote offloading plugin for libomptarget. Thisimplementation relies on gRPC and protobuf, so this library will onlybuild if both libraries are available on the system. The correspondingserver is compiled to `openmp-offloading-server`.This is a large change, but the only way to split this up is into RTL/serverbut I fear that could introduce an inconsistency amongst them.Ideally, tests for this should be added to the current ones that but that isproblematic for at least one reason. Given that libomptarget registers pluginon a first-come-first-serve basis, if we wanted to offload onto a local x86through a different process, then we'd have to either re-order the plugin listin `rtl.cpp` (which is what I did locally for testing) or find a bettersolution for runtime plugin registration in libomptarget.Differential Revision: https://reviews.llvm.org/D95314 List of files: /llvm-project-15.0.7/openmp/libomptarget/plugins/exports Sat, 23 Jan 2021 17:23:50 +0000 Atmn Patel <atmndp@gmail.com> http://172.16.0.5:8080/history/llvm-project-15.0.7/openmp/libomptarget/plugins/exports#a014fbbc [OpenMP] Improve D2D memcpy to use more efficient driver APISummary:In current implementation, D2D memcpy is first to copy data back to host and thencopy from host to device. This is very efficient if the device supports D2Dmemcpy, like CUDA.In this patch, D2D memcpy will first try to use native supported driver API. Ifit fails, fall back to original way. It is worth noting that D2D memcpy in thisscenerio contains two ideas:- Same devices: this is the D2D memcpy in the CUDA context.- Different devices: this is the PeerToPeer memcpy in the CUDA context.My implementation merges this two parts. It chooses the best API according tothe source device and destination device.Reviewers: jdoerfert, AndreyChurbanov, grokosReviewed By: jdoerfertSubscribers: yaxunl, guansong, sstefan1, openmp-commitsTags: #openmpDifferential Revision: https://reviews.llvm.org/D80649 List of files: /llvm-project-15.0.7/openmp/libomptarget/plugins/exports Thu, 04 Jun 2020 20:58:37 +0000 Shilei Tian <tianshilei1992@gmail.com> http://172.16.0.5:8080/history/llvm-project-15.0.7/openmp/libomptarget/plugins/exports#03ff643d [OpenMP] Put old APIs back and added new _async series for backward compatibilitySummary: According to comments on bi-weekly meeting, this patch put back old APIs and added new `_async` seriesReviewers: jdoerfertReviewed By: jdoerfertSubscribers: yaxunl, guansong, openmp-commitsTags: #openmpDifferential Revision: https://reviews.llvm.org/D77822 List of files: /llvm-project-15.0.7/openmp/libomptarget/plugins/exports Fri, 10 Apr 2020 02:40:30 +0000 Shilei Tian <tianshilei1992@gmail.com> http://172.16.0.5:8080/history/llvm-project-15.0.7/openmp/libomptarget/plugins/exports#32ed2927 [OpenMP] Optimized stream selection by scheduling data mapping for the same target region into a same streamSummary:This patch introduces two things for offloading:1. Asynchronous data transferring: those functions are suffix with `_async`. They have one more argument compared with their synchronous counterparts: `__tgt_async_info*`, which is a new struct that only has one field, `void *Identifier`. This struct is for information exchange between different asynchronous operations. It can be used for stream selection, like in this case, or operation synchronization, which is also used. We may expect more usages in the future.2. Optimization of stream selection for data mapping. Previous implementation was using asynchronous device memory transfer but synchronizing after each memory transfer. Actually, if we say kernel A needs four memory copy to device and two memory copy back to host, then we can schedule these seven operations (four H2D, two D2H, and one kernel launch) into a same stream and just need synchronization after memory copy from device to host. In this way, we can save a huge overhead compared with synchronization after each operation.Reviewers: jdoerfert, ye-luoReviewed By: jdoerfertSubscribers: yaxunl, lildmh, guansong, openmp-commitsTags: #openmpDifferential Revision: https://reviews.llvm.org/D77005 List of files: /llvm-project-15.0.7/openmp/libomptarget/plugins/exports Tue, 07 Apr 2020 18:51:56 +0000 Shilei Tian <tianshilei1992@gmail.com> http://172.16.0.5:8080/history/llvm-project-15.0.7/openmp/libomptarget/plugins/exports#c5fe030c [OpenMP][libomptarget] Enable usage of unified memory for declare target link variablesSummary: This patch enables the usage of a host variable on the device for declare target link variables when unified memory is available.Reviewers: ABataev, caomhin, grokosReviewed By: grokosSubscribers: Hahnfeld, guansong, jdoerfert, openmp-commitsTags: #openmpDifferential Revision: https://reviews.llvm.org/D60884llvm-svn: 362505 List of files: /llvm-project-15.0.7/openmp/libomptarget/plugins/exports Tue, 04 Jun 2019 15:05:53 +0000 Gheorghe-Teodor Bercea <gheorghe-teod.bercea@ibm.com> http://172.16.0.5:8080/history/llvm-project-15.0.7/openmp/libomptarget/plugins/exports#3de4cd12 [OpenMP] Initial implementation of OpenMP offloading library - libomptarget plugins.This is the patch upstreaming the plugins part of libomptarget (CUDA, generic-elf-64).Differential Revision: https://reviews.llvm.org/D14253llvm-svn: 293724 List of files: /llvm-project-15.0.7/openmp/libomptarget/plugins/exports Wed, 01 Feb 2017 00:14:41 +0000 George Rokos <grokos@us.ibm.com>