Граф коммитов

30 Коммитов

Автор SHA1 Сообщение Дата
ghostplant 0246d68184 update compute compate list 2023-08-04 06:13:08 +08:00
ghostplant d592a86c6b add custom defines & atomic macro 2023-07-01 15:32:15 +08:00
ghostplant 3abbbf2358 deal with persistent device context 2023-06-13 23:40:28 +08:00
ghostplant 30308c6db7 aligning memsize for GPU memalloc 2023-04-07 23:41:23 +08:00
ghostplant 3df7c2f0d2 add -ldl to cuda 2023-04-07 20:15:39 +08:00
ghostplant 5b8ec8b63f make moduleLoad compatible for src & bin 2023-02-24 19:05:42 +08:00
Wei CUI cf369c9fcf update ROCm dockerfile to 5.3 2022-11-04 02:00:34 +08:00
Wei CUI ea21104493 compile gfx code by source hint 2022-10-13 15:14:18 +08:00
ghostplant 44679fab4c fix corner cases 2022-09-07 23:05:56 +08:00
Wei CUI 94143f6b5e support CBV for hlsl_win64 & hlsl_xbox 2022-08-18 16:37:01 +08:00
ghostplant 4ecbcf80f8 handle cumemalloc size of zero 2022-08-16 14:46:53 +08:00
ghostplant a085e945d1 allow compilation for CUDA_ALL_ARCH 2022-07-31 00:03:45 +08:00
ghostplant d484dcddeb update args parsing of booting threads 2022-07-23 14:50:12 +08:00
Wei CUI 4545fa31d2 upgrade hlsl runtime to v0.3 2022-07-08 19:00:00 +08:00
ghostplant 9658b1cca3 refine execute_module.hpp 2022-07-05 21:50:46 +08:00
ghostplant dde0726412 update clang-format option & compile time 2022-05-08 18:17:55 +08:00
Wei CUI 82c5bd8dd3 split moduleCompile and moduleLoad 2022-04-25 14:29:15 +08:00
ghostplant b04370b31d update vamap_args 2022-03-25 00:53:44 +08:00
Wei CUI 7d7566d330 fix bug when single tensor memsize >= 4GB 2022-03-22 00:06:50 +08:00
Wei CUI 95c832d5ee setup codegen flags for vamap 2022-03-18 21:20:32 +08:00
Wei CUI 8d37c4c093 enhance runtime checking of python version & cuda 2021-08-23 15:18:47 +08:00
ghostplant 06a6bfae57
support mcpu tuning in pytorch JIT plugin (#255) 2021-04-29 22:30:11 +08:00
ghostplant 356ca4a474
Backend stream (#242)
* include stream in backend interfaces

* synchronize stream for win64 backends
2021-04-18 15:53:22 +08:00
ghostplant 1100a3d806
move tempfile/syscall to ab_utils (#188) 2021-03-02 16:47:47 +08:00
ghostplant 7737b7e6db
update tf/pytorch plugin to support graph execution (#180) 2021-02-26 23:18:42 +08:00
ghostplant 505698739f
add missing Dockerfiles for win64 backends (#179) 2021-02-26 14:15:05 +08:00
ghostplant 68e6d04ccb
rename c-hlsl to c-hlsl_win64 (#176)
* rename c-hlsl to c-hlsl_win64
2021-02-26 10:10:26 +08:00
ghostplant 5fff4f2bc9
add foreground timeout monitoring (#174) 2021-02-25 12:34:48 +08:00
ghostplant af453be434
Upgrade Next Format (#170)
* upgrade next format
2021-02-23 23:28:39 +08:00
ghostplant 8307b33e85
Upgrade Evaluate Format (#165)
* upgrade exxcution format

* proper refactor

* split backend interfaces to headers

* part refactor: cancel prebuild; trans src v2; incomplete backends

* moving eval_client to graph_evaluator

* refactor backend C interface

* init C backend implementation for c-mcpu

* finish upgrade for scpu/mcpu/sycl/cuda/rocm

* finish upgrade for c-ocl

* finish partial upgrade for c-gc

* finish partial upgrade for c-hlsl

* add codehub folder

* remove frameworks/antares

* synchronize JIT plugin parser
2021-02-21 14:15:27 +08:00