Ye Luo
|
7b3a09e4f1
|
Move gen_test_params.py under tests/
|
2021-12-22 10:54:03 -06:00 |
Ye Luo
|
d02aa864f4
|
Restore pure makefile build.
|
2021-12-21 19:01:32 -06:00 |
Ye Luo
|
be33e614ce
|
Move FFTXlib source files under src.
|
2021-12-21 18:35:10 -06:00 |
Pietro Bonfa
|
f9f2609873
|
Fixed Makefile
|
2021-10-22 10:36:01 +02:00 |
Pietro Bonfa
|
3ed9a958d3
|
Oops, I forgot it starts from start (see changes to understand)! Also removed useless copy.
|
2021-10-22 10:25:44 +02:00 |
Pietro Bonfa
|
0ae2ec1c49
|
Check all g vectors in all tests, fixed accuracy loss, removed unused variables
|
2021-10-20 12:54:14 +02:00 |
Pietro Bonfa
|
00ae01e896
|
Comments
|
2021-10-20 09:57:41 +02:00 |
Ye Luo
|
41aef76228
|
Fix CMake.
|
2021-10-19 09:40:09 -05:00 |
Pietro Bonfa
|
f6610c6303
|
Removed useless computation of global g offset
|
2021-10-19 16:29:18 +02:00 |
Pietro Bonfa
|
8df478ebf7
|
Not only gamma! (or a bunch of other vectors on the various procs)
|
2021-10-19 14:27:27 +02:00 |
Pietro Bonfa
|
eb247841cc
|
Populate g vectors correctly in invfft
|
2021-10-19 13:59:09 +02:00 |
Pietro Bonfa
|
7bbf02fbf4
|
Adding g-vector map generation
|
2021-10-19 08:34:30 +02:00 |
Pietro Bonfa
|
c336ed75bf
|
Fixed data initialization
|
2021-09-20 10:13:23 +02:00 |
Pietro Bonfa
|
16277ca0bb
|
Do not test fft_scatter when there's nothing to scatter
|
2021-09-09 12:20:45 +02:00 |
Pietro Bonfa
|
4ab72a0077
|
Fixing tests
|
2021-09-08 17:22:31 +02:00 |
Victor Yu
|
5c9c38fbd5
|
Add include 'mpif.h' to FFTXlib/tests
It's used when not using the MPI module.
|
2021-09-03 10:41:58 -04:00 |
Ye Luo
|
ade26bea7f
|
Connect FFTXlib/tests to CMake/CTest
|
2021-09-03 09:26:54 -04:00 |
Victor Yu
|
c3e3bbad5b
|
Update FFTXlib/tests
Should compile and run again.
|
2021-09-03 09:26:54 -04:00 |
Pietro Bonfa
|
2412833213
|
Fixed bug in many_cft3s_gpu for 'Rho' case. Activated batched Rho fft together with tests in test_fwinv_gpu.
|
2018-11-05 15:23:18 +01:00 |
Pietro Bonfa
|
f472f9b964
|
Fixed test_fwinv_gpu.f90. But device selection still missing
|
2018-10-26 15:09:59 +02:00 |
Pietro Bonfa
|
34e9ae5b78
|
Almost there but not quite yet (fft_scatter_many_yz still failing for large MPI pools). Device selection still missing.
|
2018-10-26 14:57:14 +02:00 |
Pietro Bonfa
|
ddfe4181ad
|
Only zero out what needed in fft_scatter_yz_gpu. Test partially fixed (few trivial things left for tomorrow)
|
2018-10-25 18:52:38 +02:00 |
Pietro Bonfa
|
31f7be881e
|
nx/=ldx (and equivalent in other directions) is not supported by all cfft3d implementations. Reverted to correct syntax and added a worning.
|
2018-06-11 12:14:15 +02:00 |
Pietro Bonfa
|
6055c6ab17
|
A few subtle bugs in 3dfft: wrong data order and partial scaling when leading dimension is greater than data dimension, mess with optional input and save statement (always set the value of optional input!). Test updated accordingly.
|
2018-06-08 20:22:37 +02:00 |
Pietro Bonfa
|
359ae50eb6
|
Corrected subtle error in test_fft_scatter_mod_gpu and added implementation of howmany for parallel GPU driver. This commit is a WIP and will be used to further test the changes in the gpu-develop branch.
|
2018-05-08 10:52:38 +02:00 |
Pietro Bonfa
|
df8dece2e1
|
Removed fake error when checking against a CPU implementation that does not support howmany
|
2018-05-08 10:49:49 +02:00 |
Pietro Bonfa
|
3465554cda
|
Naive implementation of cuda stream support. Non blocking scatter version seems to work fine. Added in-place FFT to avoid a memory copy. Massive optimization still needed.
|
2018-04-26 17:39:43 +02:00 |
Pietro Bonfa
|
9a7b1c1099
|
Added in-place fft for cuda to avoid mem copy
|
2018-04-23 17:20:17 +02:00 |
Pietro Bonfa
|
0c43bdffab
|
Added initial test for CUDA/CUDA Fortran based FFT interfaces.
|
2018-03-21 16:31:26 +01:00 |
Pietro Bonfa
|
ab12a2be6e
|
Added test for scatter_mod_gpu
|
2018-03-19 15:28:59 +01:00 |
Pietro Bonfa
|
bdc1830983
|
Initial draft for unit testing system of FFTXlib
|
2018-03-16 16:34:19 +01:00 |