6fefd8821a
Attempted to allow GPU acceleration on MacOS with neighbor builds on the device by enforcing the old neighbor list code path (will revisit)
2023-01-25 10:42:55 -06:00
722e583b59
use available introspection API to get accumulator data type. update name of flag.
2023-01-25 05:22:49 -05:00
adf43d7fee
Fixed the issues with some OpenCL implementation to avoid errors casting changing the pointer address spaces
2023-01-25 00:02:25 -06:00
64b5ad8966
Merge branch 'develop' into amoeba-gpu
2023-01-24 20:22:59 -05:00
aaa918cbe7
Fixed bugs with access mode on the host side of thetai[1-3]
2023-01-24 17:05:48 -06:00
5014e04341
Removed commented out code, ensured that ic_kspace is not nullptr when call precompute_kspace for hippo/gpu
2023-01-24 08:40:08 -06:00
27da716852
getting out of the insurance business :-)
2023-01-23 16:45:41 -05:00
8e79e2efa5
More cleanup, fixed bugs with hippo fphi kernels for mixed precision
2023-01-23 00:18:42 -06:00
658328dd9d
Added a note in the amoeba doc page on the not-yet resolved issue with integrated GPUs, removed commented out and debugging stuffs in the AM/HP kernels
2023-01-22 17:24:15 -06:00
8eb722a32a
Enforced synchronous host-device transfers for cgrid_brick and fdip arrays
2023-01-19 13:22:27 -06:00
03ab42fd52
correct calling sequence for matching argument types
2023-01-19 08:57:24 -05:00
4244d2e6cd
silence compiler warnings about unused parameters and variables
2023-01-19 08:56:54 -05:00
eddd3d6f25
Fixed a bug with extra being nullptr when _host_view is true: always allocate extra
...
(Note that BaseAmoeba has its own cast_extra_data() that doesn't know if extra is allocated properly, it is the case when _host_view is false for dedicated GPUs for example)
2023-01-18 20:04:45 -06:00
f86375c992
Attempted to ensure that extra gets allocated in the exactly same way as other added fields (charge, quat and vel)
2023-01-17 09:47:09 -06:00
71931d1d44
Cleaned up, and added missing zero timers for extra fields transfers
2023-01-17 09:39:03 -06:00
973b46a907
Attempted to resolve the memory access runtime errors when acquiring single and mixed precision arrays from the GPU lib
2023-01-16 10:12:42 -06:00
9dc0369cee
Attempted to resolve the address space change issue when casting for OpenCL 2.0 (ref: https://www.intel.com/content/www/us/en/developer/articles/technical/the-generic-address-space-in-opencl-20.html#06_address_space_casting )
2023-01-15 23:28:48 -06:00
c9ae41246d
Ran the four make commands in the src folder: make fix-whitespace; make fix-homepage; make fix-errordocs; make fix-permissions
2023-01-15 16:05:36 -06:00
212da7f109
Merge branch 'develop' into amoeba-gpu
2023-01-14 18:36:26 -06:00
5cbe303af4
Merge branch 'develop' into collected-small-changes
2023-01-04 07:28:03 -05:00
d9abc3fcc0
update CUDA Toolkit / GPU compatibility lists and GPU package compilation settings
2023-01-03 11:56:44 -05:00
396d577f40
port DPD exclusions corrections to GPU package
2023-01-02 12:04:10 -05:00
8af77c690c
Merge branch 'develop' into amoeba-gpu
2022-12-14 13:16:41 -06:00
959b9c220f
Cleaned up unused member functions and hd_balancer calls
2022-11-07 15:49:37 -06:00
a3cc0e8432
Reverted the block size tuning, which caused bugs for low atom counts (will revisit later)
2022-11-04 13:45:59 -05:00
2f1f7ee0fa
Cleaned up code
2022-11-03 23:45:40 -05:00
e5a808fb8d
apply correct platform selection for OpenCL context
2022-11-01 04:05:57 -04:00
80a141d9c8
silence compiler warnings
2022-11-01 03:38:08 -04:00
ad54268544
silence compiler warning
2022-10-19 14:31:21 -04:00
9d081a5916
more adjustments for bogus timer results on Intel OpenCL
2022-10-19 07:39:56 -04:00
f867adc541
GPU Package fix where timing disable could result in event/marker destruction before completion on accelerator during initialization.
2022-10-19 02:16:29 -04:00
51c6eddd0d
Fix to make the property list empty for command queues when timing disabled.
2022-10-19 02:15:39 -04:00
7c9666798e
whitespace
2022-10-08 09:34:20 -04:00
7551c0a3ca
GPU Package: Documenting some additional preprocessor flags, updating oneapi Makefile.
2022-10-07 22:44:21 -07:00
00f46120c7
Removed max_cus() from Device, used device->gpu->cus() instead
2022-10-07 15:50:30 -05:00
5a98a38e24
GPU Package: Switching to parallel GPU initialization / JIT compilation.
2022-10-07 13:25:14 -07:00
f715f174bb
GPU Package: Print OCL platform name to screen when multiple platforms
2022-10-06 21:40:42 -07:00
a6a39d47e1
Fixing potential issues with automatic splitting of accelerators for NUMA.
2022-10-06 20:48:02 -07:00
e9f39f85d2
Fixing issue where shared main memory property only set for NVIDIA devices.
2022-10-06 20:05:33 -07:00
6b9e83fe20
Added timing for the induced dipole spreading part, computed the block size to ensure all the CUs are occupied by the fphi_uind and fphi_mpole kernels
2022-10-06 15:03:58 -05:00
2ef6a59c0a
Merge branch 'develop' into amoeba-gpu
2022-10-01 00:38:24 -05:00
9a1f23a079
Cosmetic changes and cleanup
2022-09-30 17:32:25 -05:00
1d75ca3b20
Moved precompute() out of the terms in amoeba and hippo, to be involed in the first term in a time step: multipole for amoeba and repulsion for hippo
2022-09-30 16:31:13 -05:00
fb675028b9
whitespace
2022-09-29 02:42:11 -04:00
71464d8314
GPU Package: Fixing logic in OpenCL backend that could result in unnecessary device allocations.
2022-09-28 22:30:09 -07:00
6e34d21b24
GPU Package: Switching back to timer disabling with multiple MPI tasks per GPU. Logic added to prevent mem leak.
2022-09-28 21:02:16 -07:00
e6d2582642
Updated fphi_mpole, renamed precompute_induce to precompute_kspace
2022-09-28 15:08:18 -05:00
166701f13a
Fixed missing commas in the argument list of the macros in amoeba and hippo cu files, added amoeba_convolution_gpu.cpp and .h to the source file list in GPU.cmake
2022-09-23 11:53:09 -05:00
785131932c
Added fphi_mpole in amoeba/gpu, fixed a bug in the kernel when indexing grid
2022-09-20 13:58:17 -05:00
356c46c913
Replaced mem allocation/deallocation inside moduli() with using member variables and mem resize if needed
2022-09-18 16:28:30 -05:00
