mirrors / dmlc--dgl

文件历史

4,090 提交 34 分支 42 Git 标签
v2.2.1
提交图

542 次代码提交

作者 SHA1 备注 提交日期
Muhammed Fatih BALIN 7de2e51b5e [CUDA] Remove unused headers for CCCL 2.4 compat (#7329) 2024-04-20 13:34:28 +08:00
Triston a4e19691d5 [Determinism] Enable environment var to use cusparse spmm deterministic algorithm (#7310) 
Co-authored-by: Hongzhi (Steve), Chen <chenhongzhi.nkcs@gmail.com>
 2024-04-19 09:00:40 +08:00
Muhammed Fatih BALIN 2ff3006c35 [CUDA][Bug] CSR transpose bug in CUDA 12 (#7295) 2024-04-12 02:54:13 -04:00
Muhammed Fatih BALIN 73e01d6daa [CUDA] Update CCCL to 2.3.0 (#7171) 2024-02-29 00:37:00 -05:00
Muhammed Fatih BALIN 5185c522b0 [Misc] Eliminate compile warnings (#6685) 2023-12-04 21:47:28 -05:00
Muhammed Fatih BALIN 5e78e070f4 [Misc] Fix signed unsigned comparison warning (#6602) 2023-11-23 08:17:16 +08:00
Muhammed Fatih BALIN bfde142239 [CUDA] Fix issue about integer overflow (#6586) 2023-11-22 13:01:47 +08:00
Songqing Zhang d932141471 [Misc] Fix compile warning (#6342) 2023-09-20 09:05:02 +08:00
Hugo MacDermott-Opeskin 4a42027d4a [Build] Add CMake changes from conda-forge build (#6189) 
Co-authored-by: Hongzhi (Steve), Chen <chenhongzhi.nkcs@gmail.com>
 2023-09-01 11:21:03 +08:00
Hongzhi (Steve), Chen ceb2572475 [Misc] re-enable destructor method of WorkspacePool. (#6173) 
Co-authored-by: Ubuntu <ubuntu@ip-172-31-28-63.ap-northeast-1.compute.internal>
 2023-08-18 12:32:12 +08:00
Songqing Zhang 5a41741485 [Bug] fix memory leak in cpp tests (#6141) 2023-08-17 16:10:52 +08:00
Xin Yao 08d18a47f2 [Build] Fix bf16/fp16 building issues for CUDA 12.2 (#6074) 
Signed-off-by: Xin Yao <xiny@nvidia.com>
 2023-08-14 10:12:41 +08:00
Chang Liu 88964a8253 [Bugfix] Fix cusparseCreateCsr format for cuda12 (#6121) 2023-08-10 07:35:32 -07:00
Songqing Zhang 6f28e1adb7 [misc] Fix build warnings (#6037) 
Co-authored-by: Hongzhi (Steve), Chen <chenhongzhi.nkcs@gmail.com>
 2023-07-26 14:38:03 +08:00
Muhammed Fatih BALIN 69a532c1ab [Feature] Gpu cache for node and edge data (#4341) 
Co-authored-by: xiny <xiny@nvidia.com>
 2023-07-24 13:17:10 +08:00
AdamGrabowski 4135b1bd5b [Performance] Fused sampling with compaction (#5924) 
Co-authored-by: Hesham Mostafa <hesham.mostafa@intel.com>
 2023-07-20 09:38:22 +08:00
Muhammed Fatih BALIN d3bd4c6112 [Feature] Adding kappa feature for labor (Cooperative Minibatching) (#6006) 
Co-authored-by: Hongzhi (Steve), Chen <chenhongzhi.nkcs@gmail.com>
 2023-07-19 10:34:28 +08:00
Muhammed Fatih BALIN f5f7e08e10 [Performance][CUDA] Sorting for indices for UVM code path. (#5882) 2023-07-14 14:12:30 +08:00
Muhammed Fatih BALIN 83115794c2 [Performance][CUDA] Faster CSRToCOO (#5648) 
Co-authored-by: Hongzhi (Steve), Chen <chenhongzhi.nkcs@gmail.com>
 2023-07-14 12:57:38 +08:00
Muhammed Fatih BALIN dc06060bb3 [CUDA] fixing partition bug (#6001) 2023-07-14 12:56:24 +08:00
Muhammed Fatih BALIN c3aea1b62c [Performance][CUDA] Labor UVA optimization (#5885) 
Co-authored-by: Xin Yao <xiny@nvidia.com>
Co-authored-by: Rhett Ying <85214957+Rhett-Ying@users.noreply.github.com>
 2023-07-13 19:56:31 +08:00
Ilia Taraban 62351d8f1b [Misc] Disable LibXSMM SpMM Cmp for AVX2 platforms (#5962) 2023-07-10 13:36:56 -07:00
caojy1998 55af15d4a9 [BugFix]Use Python API for relabel_nodes (#5937) 
Co-authored-by: Ubuntu <ubuntu@ip-172-31-6-31.ap-northeast-1.compute.internal>
 2023-07-04 20:44:59 +08:00
Rhett Ying d22049e879 [TensorpipeDeprecation] remove unused arguments from wait_for_senders() (#5917) 2023-06-28 16:32:29 +08:00
Rhett Ying 5542558433 [DistDGL] remove RPCSender/Receiver base classes (#5879) 2023-06-16 16:35:59 +08:00
Rhett Ying 4e5780e3e3 [DistDGL] remove unused rpc related files (#5878) 2023-06-16 13:40:11 +08:00
Rhett Ying 66c04855da [DistDGL] remove tensorpipe from third_party and no build (#5876) 2023-06-15 17:31:05 +08:00
keli-wen ae17185a65 [Sparse] Update code and add unittest for formats (#5859) 2023-06-13 15:46:34 +08:00
Rhett Ying 72ec1c95c6 [DistDGL] remove tensorpipe cpp code (#5850) 
Co-authored-by: Hongzhi (Steve), Chen <chenhongzhi.nkcs@gmail.com>
 2023-06-12 10:50:58 +08:00
Hongzhi (Steve), Chen 9ff56d2098 [Cleanup] Remove featgraph and unused TVM dependency. (#5767) 
Co-authored-by: Ubuntu <ubuntu@ip-172-31-28-63.ap-northeast-1.compute.internal>
 2023-06-02 15:14:31 +08:00
Andrzej Kotłowski 6862e37259 Upgrade libxsmm (#5725) 
Co-authored-by: Rhett Ying <85214957+Rhett-Ying@users.noreply.github.com>
 2023-05-24 11:25:57 +08:00
nv-dlasalle 46af76c30e [Performance Improvement] Make GPU sampling and to_block use pinned memory to decrease required synchronization (#5685) 2023-05-17 16:18:35 -07:00
Ilia Taraban ff9573c410 [Misc] Disable BF16 LibXSMM SpMM for AVX2 platforms (#5677) 2023-05-10 22:05:57 +02:00
Andrzej Kotłowski e0d2250efa [Performance] Improve COOToCSR implementation (#5508) 
Co-authored-by: Hongzhi (Steve), Chen <chenhongzhi.nkcs@gmail.com>
 2023-05-11 01:06:39 +08:00
Ilia Taraban 78ecd5086b [Fix] fix libxsmm build issues on Mac OS (#5626) 2023-04-28 21:55:51 +08:00
Ilia Taraban 8ecbfa57ae [Fix] restore SpMMSumCsrNaive function for float and double (#5615) 2023-04-27 07:59:49 +08:00
Ilia Taraban 9cb669391f [Misc] Add libxsmm_config.h only when LIBXSMM is enabled (#5577) 2023-04-21 10:15:43 +08:00
Xin Yao bea5c78b3f [Fix] Remove curand host functions (#5552) 2023-04-17 10:37:46 +08:00
Chang Liu 62b5f50a5c [Feature] Import PyTorch's pin_memory() method for DGL graph structure (#5366) 2023-04-10 19:21:32 -07:00
Tianqi Zhang (张天启) 27b008b9b1 [BugFix] Fix unstable behavior of beuteforce-sharemem KNN (#5515) 2023-04-10 22:26:10 +08:00
peizhou001 c51cc82e4b [Enhancement]Set default graph dataloader thread number (#5479) 
Co-authored-by: Ubuntu <ubuntu@ip-172-31-16-19.ap-northeast-1.compute.internal>
 2023-04-10 17:52:24 +08:00
Ilia Taraban acb4eb7ecd [Feature] Add bfloat16 support for CPU (#5497) 
Co-authored-by: Hongzhi (Steve), Chen <chenhongzhi.nkcs@gmail.com>
 2023-04-06 11:16:41 +02:00
Xin Yao 170203ae2b [Performance] Creating out buffers for segment_mm|sddmm via torch.empty() (#5462) 
* update for segmentMM

* update for sddmm

* fix a bug
 2023-03-23 17:41:07 +08:00
Hongzhi (Steve), Chen 4cf5f682a6 [Misc] Polish metapath_randomwalk.h (#5471) 
Co-authored-by: Ubuntu <ubuntu@ip-172-31-28-63.ap-northeast-1.compute.internal>
 2023-03-20 22:17:39 +08:00
Daniil Sizov 87fb7ed05b [Config] Enable libxsmm by default for AVX cpu (#5165) 
* Enable AVX by default

* Fix linting errors

* Fix win64 build (libxsmm not linked)

Libxsmm on Win64 is not linked, should be disabled by default

* Fix clang format issues

* Change lower supported cpu version to LIBXSMM_X86_AVX2

Change lower supported cpu version to LIBXSMM_X86_AVX2 to address https://github.com/dmlc/dgl/issues/3459 issue

* Fix unit test

Remove assumption that libxsmm is enabled in the config by default (only true for intel CPUs with AVX2 instructions)

---------

Co-authored-by: Ubuntu <ubuntu@ip-172-31-15-137.us-west-2.compute.internal>
Co-authored-by: Quan (Andy) Gan <coin2028@hotmail.com>
 2023-03-16 02:05:22 +08:00
Xin Yao 8d5d8962ad [Refactor] Replace third_party/nccl with PyTorch's NCCL backend (#4989) 
* expose GeneratePermutation

* add sparse_all_to_all_push

* add sparse_all_to_all_pull

* add unit test

* handle world_size=1

* remove python nccl wrapper

* remove the nccl dependency

* use pinned memory to speedup D2H copy

* fix lint

* resolve comments

* fix lint

* fix ut

* resolve comments
 2023-03-08 12:59:10 +08:00
Rhett Ying b1ec112eeb Fix compile error on ubuntu22.04_g++11.3.0 (#5434) 2023-03-08 09:49:23 +08:00
peizhou001 5854ef5ec7 [Enhancement]Speed up ToBlockCPU with concurrent id hash map (#5297) 2023-03-07 09:59:48 +08:00
Kacper Pietkun 308bd6f5b2 removed pragma omp for (#5334) 2023-03-01 11:48:04 +08:00
peizhou001 11d12f3cca [Refactor] Extract common code in gpu and cpu ToBLock (#5305) 2023-02-27 18:04:05 +08:00