Pietro Delugas
df682b3e24
Merge branch 'ks_acc' into 'develop'
...
Davidson ported with OpenACC, CPU and GPU versions merged into one
See merge request QEF/q-e!1826
2022-06-20 07:41:25 +00:00
Ivan Carnimeo
6d6387950c
Clean up
2022-06-13 18:40:31 +02:00
Pietro Delugas
5c1e65933e
Merge remote-tracking branch 'official_qe/develop' into rism-devel
2022-05-25 16:03:00 +02:00
Ivan Carnimeo
d5a343bf1c
Some unnecessary precaution (with a bad impact on performance) removed
2022-05-09 18:40:32 +02:00
Ivan Carnimeo
0546abf585
Fix cmake compilation
2022-05-06 15:11:58 +02:00
Ivan Carnimeo
b35c39b070
Some cleanup
2022-05-06 11:18:58 +02:00
Ivan Carnimeo
687c65d612
Protect the CPU case for non-contiguos mp_sum to avoid RAM waste
2022-05-06 08:11:35 +02:00
Ivan Carnimeo
31d427bb87
This might be a compact way to perform mp_sum
...
passing non-contiguous device arrays
2022-05-05 15:38:57 +02:00
Ivan Carnimeo
5fd801fe03
regterg to OpenACC (11):
...
- pinned_buffer not necessary here
2022-05-05 12:15:38 +02:00
Ivan Carnimeo
3522ea3cd2
regterg to OpenACC (10):
...
- regterg_gpu bypassed, regterg now used for both CPU and GPU execution
2022-05-02 17:16:39 +02:00
Ivan Carnimeo
0db163b694
regterg to OpenACC (9):
...
- KScudaDGER removed, MYDGER (UtilXlib) used instead
2022-05-02 16:35:12 +02:00
Ivan Carnimeo
f5bbc9d45d
regterg to OpenACC (8):
...
- MYDDOT_VECTOR_GPU used for dot product (cfr cegterg)
2022-05-02 16:25:47 +02:00
Ivan Carnimeo
5520356462
regterg to OpenACC (7):
...
- CUDA Fortran (USE cudafor) removed
- cublasDgemm --> DGEMM
2022-05-02 15:53:24 +02:00
Ivan Carnimeo
7913cc4d32
regterg to OpenACC (6):
...
- reorder_evals_revecs removed
2022-05-02 15:44:56 +02:00
Ivan Carnimeo
318f4ebcdb
regterg to OpenACC (5):
...
- e_host, ew_host removed
2022-05-02 12:55:39 +02:00
Ivan Carnimeo
7d7d34a667
regterg to OpenACC (4):
...
- hr_d, sr_d, vr_d, ew_d --> hr, sr, vr, ew (done)
2022-05-02 12:49:15 +02:00
Ivan Carnimeo
b69e192bf7
regterg to OpenACC (3):
...
- hr_d, sr_d, vr_d, ew_d --> hr, sr, vr, ew (to be finished)
2022-05-02 11:11:42 +02:00
Ivan Carnimeo
69f6cf85c9
regterg to OpenACC (2):
...
- some cleanup
2022-04-29 10:21:02 +02:00
Ivan Carnimeo
b6d60091b4
regterg to OpenACC (1):
...
- psi_d, hpsi_d, spsi_d --> psi, hpsi, spsi (OpenACC)
- pinned_buffer added for non-contiguous mp_sum (cfr cegterg)
2022-04-29 10:12:57 +02:00
Ivan Carnimeo
bbc6627aa2
cegterg to OpenACC (13):
...
- some cleanup
2022-04-29 08:07:31 +02:00
Ivan Carnimeo
ef9f59d4d0
cegterg to OpenACC (12):
...
- small fixes for CPU version (this fixes the WARNING at commit (11))
2022-04-28 21:45:21 +02:00
Ivan Carnimeo
9d9b0b9acb
cegterg to OpenACC (11):
...
- cegterg_gpu merged with cegterg
- cegterg_gpu bypassed, only cegterg is used for both CPU and GPU
WARNING: this commit has been tested only on GPU and needs to be tested also on CPU
2022-04-28 21:16:48 +02:00
Ivan Carnimeo
b40b5c3d89
cegterg to OpenACC (10):
...
- e_host, ew_host removed
- WHERE/ELSEWHERE changed to a simple loop with OpenACC
2022-04-28 15:57:30 +02:00
Ivan Carnimeo
ff4c3a70c7
cegterg to OpenACC (9):
...
- some cleanup
2022-04-27 18:12:49 +02:00
Ivan Carnimeo
da21542acc
cegterg to OpenACC (8):
...
- cudafor removed (evc_d, e_d --> deviceptr)
2022-04-27 18:05:53 +02:00
Ivan Carnimeo
61163cfa42
cegterg to OpenACC (7):
...
- reorder_evals_cevecs removed
2022-04-27 16:29:58 +02:00
Ivan Carnimeo
4bdcca1f20
cegterg to OpenACC (6):
...
- reorder_evals_cevecs skipped and the OpenACC-ized CPU code is used instead
- reorder_evals_cevecs to OpenACC (reorder_evals_cevecs could be removed,
I leave it in this commit only to save the OpenACC version,
in case any bugs come out in the future)
2022-04-27 16:18:33 +02:00
Ivan Carnimeo
ccd60211cd
cegterg to OpenACC (5):
...
- variables with local scope declared as device_resident
2022-04-27 14:34:38 +02:00
Ivan Carnimeo
d2aa90b8c2
cegterg to OpenACC (4):
...
some cleanup
2022-04-22 18:01:59 +02:00
Ivan Carnimeo
32bc9d529c
cegterg to OpenACC (3):
...
ew_d --> ew
2022-04-22 17:51:17 +02:00
Ivan Carnimeo
fedbe658b6
cegterg to OpenACC (2):
...
hc_d, sc_d, vc_d --> hc, sc, vc
2022-04-22 17:36:38 +02:00
Ivan Carnimeo
d684c50deb
cegterg to OpenACC (1):
...
psi_d, hpsi_d, spsi_d --> psi, hpsi, spsi
MYDDOT (host) --> MYDDOT_VECTOR_GPU (device)
2022-04-22 16:24:59 +02:00
Ye Luo
18a0fd5843
Remove obsolete ELPA include.
2022-01-18 00:06:20 -06:00
Minoru Otani
ac536f1dd1
fix old RMM file path
2021-12-29 16:08:39 +09:00
Satomichi Nishihara
82d04cc156
fix Makefile
2021-12-29 16:07:11 +09:00
Satomichi Nishihara
0d7d749371
merge RISM-code
2021-12-29 16:07:10 +09:00
Paolo Giannozzi
5d74f827c4
Remove make.depend files from git
2021-10-24 21:29:29 +02:00
Pietro Delugas
6179ad15a7
fixing one more failing case for blocked Gram-Schmidt gpu version
2021-10-20 12:10:36 +02:00
Pietro Delugas
3b65738662
fixing one more bound issue in blocked Gram-Schmidt for GPUs
2021-10-18 10:13:33 +02:00
Pietro Delugas
082dab30fb
fix_issue_in_rmm_diagonalization
2021-10-15 18:39:10 +02:00
Pietro Delugas
152a6816a5
fixing gpu case
2021-09-26 22:21:52 +02:00
Pietro Delugas
2af7d4fa07
moving gram schmidt and rotate routined to dense
...
* complete the move of various files from in KS_Solvers/DENSE
* define an interface for rotate routines and an header file
* still to do unify the many rotate routines in the KS_Solvers/DENSE
2021-09-26 20:07:50 +02:00
Elena De Paoli
38b47df02f
Move rotate_xpsi and rotate_xpsi from PW/src into KS_Solver/DENSE
2021-09-23 16:53:13 +02:00
Elena De Paoli
e17fb5521f
"Tested version of gran_schimdt gamma and kappa gpu"
...
fixing dependencies in CMakeLists.txt in KS_SOLVERS
2021-09-23 08:58:23 +02:00
Elena De Paoli
2274183be9
Add gram_schimdt gpu with devxlib
2021-09-22 23:53:13 +02:00
Pietro Delugas
1068d3a7b6
removal of calls to zdotc for KS_SOLVERS/RMM
2021-08-12 21:46:01 +02:00
Pietro Delugas
3c1b5dae41
fixing compilation after alignement
2021-08-12 20:08:35 +02:00
Pietro Delugas
37a135b31d
upgrading CMakeLists files for RMM-DIIS
2021-08-12 20:08:34 +02:00
Pietro Delugas
b102d55c8f
adding a missing attributes(device) statement in gram_schmidt_k_gpu
2021-08-12 20:08:34 +02:00
Pietro Delugas
615f5d9164
Applying Andrea Ferretti patch
2021-08-12 20:08:34 +02:00