|
|
7e92809288
|
Merge pull request #3069 from Vsevak/fix-hip-ffast-math
Fix HIP Makefile under lib/gpu
|
2022-01-03 11:37:55 -05:00 |
|
|
|
14e5474174
|
restore obsolete compilation settings similar to parallel makefile
|
2021-12-27 20:31:42 -05:00 |
|
|
|
b781410f92
|
Delete fast-math flag from Makefile.hip for AMD platforms
|
2021-12-28 03:11:02 +03:00 |
|
|
|
5594a38bb7
|
replace explicit Makefile.mpi with symbolic link
|
2021-12-27 10:47:23 -05:00 |
|
|
|
6357f19260
|
Added back Makefile.mpi in lib/gpu/ to be consistent with documentation; updated Makefile.*; and removed the unnecessary Makefile.turing
|
2021-12-27 00:14:04 -06:00 |
|
|
|
ede188652b
|
update a few GPU kernels so they can be compiled on GPUs without double precisions support
|
2021-10-21 07:33:00 -04:00 |
|
|
|
87b63f768f
|
Only check for GPU double precision support if a GPU is present
|
2021-10-18 12:15:05 -04:00 |
|
|
|
e990a1cf61
|
remove ambiguity between "double_precision" class member variable and function
|
2021-10-16 21:07:04 -04:00 |
|
|
|
afad3f42d5
|
Report only compatible GPU, i.e. no GPU if mixed/double precision is requested by the hardware does not support it
|
2021-10-13 21:15:16 -04:00 |
|
|
|
dd6e3c1acc
|
avoid variable length array and signed vs. unsigned warnings
|
2021-10-08 20:07:20 -04:00 |
|
|
|
79fbbd4f33
|
Cleaned up the API of amoeba and hippo to remove unncessary arguments
|
2021-10-04 14:40:58 -05:00 |
|
|
|
0f0f6a51de
|
Renamed sp_polar to sp_amoeba, and replaced special_wscale with special_hal for amoeba
|
2021-10-02 16:02:44 -05:00 |
|
|
|
5a6426bf96
|
Only transfer data arrays that are needed in each kernel
|
2021-10-02 00:56:15 -05:00 |
|
|
|
f4d3d3a2b5
|
Gradually cleaned up and removed redundancy in amoeba and hippo
|
2021-10-02 00:09:53 -05:00 |
|
|
|
f126f785a4
|
Removed duplicates in the amoeba kernels
|
2021-10-01 10:19:17 -05:00 |
|
|
|
3328ac0df2
|
Attempted to remove some redundancy in data transfers in the amoeba kernels; keeping HIPPO independent of AMOEBA for now
|
2021-10-01 09:58:21 -05:00 |
|
|
|
ad9d45639e
|
Fixed bugs with damprep where ucl_powr in mixed precision failed with a negative single-reprecision base
|
2021-09-29 12:32:08 -05:00 |
|
|
|
01381b7f54
|
Fixed bugs in the repulsion kernel, now working correctly with the double precision mode
|
2021-09-29 11:57:25 -05:00 |
|
|
|
4be44c386f
|
Added necessary arguments to the hippo repulsion kernel
|
2021-09-29 09:40:33 -05:00 |
|
|
|
17edd797a7
|
Adding API for the repulsion term to hippo/gpu
|
2021-09-28 23:42:04 -05:00 |
|
|
|
b95508125b
|
Adding the repulsion kernel for hippo
|
2021-09-28 23:24:34 -05:00 |
|
|
|
6286a119b3
|
Removed precompute() in hippo
|
2021-09-28 23:12:07 -05:00 |
|
|
|
98a2b67292
|
Changed to the API of BaseAmoeba to reduce duplicates in hippo
|
2021-09-28 17:39:55 -05:00 |
|
|
|
b874feb127
|
Removed trailing spaces
|
2021-09-28 17:28:33 -05:00 |
|
|
|
bf88ab77fa
|
Cleaned up unused variables in kernel (to be continued)
|
2021-09-28 15:06:30 -05:00 |
|
|
|
e80eea56ba
|
Added udirect2b and umutual2b for hippo
|
2021-09-28 14:59:39 -05:00 |
|
|
|
8d54547bc0
|
Commented out debugging commands in the hippo kernels, added (numtyp) to numerics in hippo_extra, replaced fabs with explicit func
|
2021-09-28 00:50:33 -05:00 |
|
|
|
d27836952a
|
Fixed a bug in neighbor.cpp to make special_flag consistent between amoeba and hippo (to be 2 instead of 0), that caused missing neighbors with hippo
|
2021-09-27 16:12:49 -05:00 |
|
|
|
2efd841a7e
|
Trying to find the difference in the neighbor list build in hippo vs amoeba
|
2021-09-27 11:35:35 -05:00 |
|
|
|
7437c98628
|
Fixed bugs in the polar real kernel in hippo, getting closer..
|
2021-09-26 09:11:09 -05:00 |
|
|
|
5193dcf8c5
|
Working on the polar real-space term of hippo
|
2021-09-26 00:56:29 -05:00 |
|
|
|
edbed9c9c9
|
Fixed bugs in HippoT::compute_dispersion_real and compute_multipole_real to ensure that answers only get copied back from device in the last kernel activated.
|
2021-09-26 00:13:40 -05:00 |
|
|
|
9da8c932ab
|
make check more obvious
|
2021-09-25 21:33:10 -04:00 |
|
|
|
f8bc091cb8
|
Kept working on the multipole real-space term of hippo
|
2021-09-25 13:17:06 -05:00 |
|
|
|
78ef0d631f
|
Working on the multipole real-space term of hippo
|
2021-09-25 12:25:34 -05:00 |
|
|
|
530912a930
|
detect double precision support according to OpenCL specs (1.2 and later)
|
2021-09-25 07:20:52 -04:00 |
|
|
|
e77df80ce2
|
Working hippo multipole real-space term, added helper functions in a separate file
|
2021-09-24 16:44:43 -05:00 |
|
|
|
42dca75225
|
add check and suitable error message when fp64 is required but not available
|
2021-09-24 12:17:58 -04:00 |
|
|
|
ad8164dfc0
|
Fixed bugs in the dispersion real-space term for hippo. NOTE: CPU version filter out neighbors with zero special_disp
|
2021-09-24 00:21:25 -05:00 |
|
|
|
17ba0d5804
|
possible workaround for some GPU package neighbor list issue
|
2021-09-22 21:47:32 -04:00 |
|
|
|
2428f1f4d5
|
Updated hippo kernels
|
2021-09-22 11:44:41 -05:00 |
|
|
|
bebef18495
|
Cleaned up and minor changes
|
2021-09-21 23:46:21 -05:00 |
|
|
|
d77d5b7f0a
|
Added classes for hippo/gpu, refactored BaseAmoeba and made room for the dispersion real-space term in hippo
|
2021-09-21 15:40:06 -05:00 |
|
|
|
a2fd784034
|
Added the dispersion real space term, which is for HIPPO.
|
2021-09-21 10:55:38 -05:00 |
|
|
|
afccf1933f
|
correctly specify the destructor function name.
|
2021-09-20 23:40:14 -04:00 |
|
|
|
42034bd1c9
|
Fixed bugs for undefined tagint and ucl_powr ambiguity in kernels for OpenCL builds
|
2021-09-20 12:48:29 -05:00 |
|
|
|
4e88cd158e
|
Fixed bugs with _tep and _fieldp to allow mixed-precision builds, being defensive with acctyp for these variables
|
2021-09-20 11:38:50 -05:00 |
|
|
|
0228867d8e
|
Added the dispersion real space kernel and transfer special coeffs to the device
|
2021-09-19 23:40:43 -05:00 |
|
|
|
1166845fcf
|
Prepared data structure for the dispersion real-space term
|
2021-09-18 10:22:22 -05:00 |
|
|
|
5d801e985f
|
More cleanup
|
2021-09-17 23:24:23 -05:00 |
|
