59bd1fe21b
Fix for 'perm' and 'local' not being set for hand-unrolled external-site Dslash, which caused incorrect behavior of G-parity kernel
2017-08-29 13:07:37 -07:00
4e907fef2c
Merge remote-tracking branch 'grid/develop' into feature/arm-neon
2017-08-29 17:47:36 +02:00
67888b657f
Merge branch 'gparity-handunroll' of https://github.com/giltirn/Grid into gparity-handunroll
2017-08-29 09:52:05 -04:00
74af885d4e
Removed some no-longer-needed associated with G-parity hand unrolled kernel
2017-08-29 09:50:37 -04:00
d36d2fb40d
Added ability to override default Ls in Benchmark_dwf
2017-08-28 06:53:56 -07:00
5b9267e88d
Cleaner comms benchmark treatment for one node runs
2017-08-27 18:24:48 -04:00
15fd4003ef
Improving presentation of results
2017-08-27 13:46:02 +01:00
4b4c2a715b
fcntl.h needed
2017-08-26 11:38:04 +01:00
54a5e6c1d0
Check if we get huge pages on linux. Larry Meadows piece of magic.
2017-08-25 22:36:08 +01:00
73aeca7dea
Merge branch 'feature/multi-communicator' into develop
2017-08-25 21:55:09 +01:00
ad89abb018
Fix
2017-08-25 20:43:37 +01:00
80c5bce5bb
Merge branch 'develop' into feature/multi-communicator
2017-08-25 20:21:26 +01:00
f68b5de9c8
No compile fix on Clang
2017-08-25 19:35:21 +01:00
d0f3d525d5
Optimal block size for KNL
2017-08-25 19:33:54 +01:00
f365a83fae
In G-parity unrolled kernel, replaced calls to permute and exchange with run-time-evaluated permute type with explicit calls to appropriate underlying functions
2017-08-25 14:24:11 -04:00
3a58217405
Updated
2017-08-25 14:29:53 +01:00
c289699d9a
updated from cambridge mpi3 shakeout
2017-08-25 11:41:01 +01:00
c3b1263e75
Benchmark prep
2017-08-25 09:25:54 +01:00
34a9aeb331
Reduced number of if-statement evaluations in G-parity unrolled kernel
2017-08-24 13:53:50 -07:00
102ea9ae66
CI update
2017-08-24 18:17:09 +01:00
5fa386ddc9
FFT test compile fixed
2017-08-24 10:17:52 +01:00
edabb3577f
Imported Benchmark_gparity
2017-08-23 16:54:06 -04:00
ce5df177ee
Removed superfluous implementation of G-parity twist for hand-unrolled kernel from GparityWilsonImpl
2017-08-23 15:05:22 -04:00
a0bb8e5b46
Added hand-unrolled kernel implementations of all the other dslash precision / comms precision combinations with G-parity
2017-08-23 14:44:40 -04:00
46f88e6d72
G-parity hand-unrolled intrinsics twist now uses one less permute and one less temporary
2017-08-23 13:21:10 -04:00
dd8f1ea189
Vectorized Mobius EOFA Dperp + shift operation
2017-08-23 13:17:26 -04:00
b61835c1a5
Added inplace version of intrinsic G-parity twist to hand-unrolled kernel
2017-08-23 12:33:48 -04:00
d9cd4f0273
Staggered multinode block cg debugged. Missing global sum.
...
Code stalls and resumes on KNL at cambridge. Curious.
CG iterations 23ms each, then 3200 ms pauses. Mean bandwidth reports
as 200MB/s. Comms dominant in the report. However, the time behaviour suggests it
is *bursty*.... Could be swap to disk?
2017-08-23 15:07:18 +01:00
459f70e8d4
Check-in of working Mobius EOFA class and tests
2017-08-22 22:38:30 -04:00
061e48fd73
Replaced slow unpack-repack in G-parity BC twist with intrinsics version
2017-08-22 18:12:12 -04:00
ab50145001
Implemented first, unoptimized version of hand-unrolled G-parity kernels
...
Improved Test_gparity
2017-08-22 17:12:25 -04:00
b49bec0cec
MAP_HUGETLB portability fix
2017-08-20 03:08:54 +01:00
ae56e556c6
finalise issue on new OPA revert
2017-08-20 02:53:12 +01:00
1cdf999668
Moving multicommunicator into mpi3 also for threading
2017-08-20 02:39:10 +01:00
11062fb686
Comms none fail fix
2017-08-20 01:37:07 +01:00
383ca7d392
Switch off comms for now until feature/multi-communicator is merged
2017-08-20 01:27:48 +01:00
a446d95c33
Trying to pass TeamCity and Travis
2017-08-20 01:10:50 +01:00
be66e7dd95
Merge branch 'develop' into feature/multi-communicator
2017-08-19 23:12:38 +01:00
6d0d064a6c
Update TODO
2017-08-19 23:11:30 +01:00
bfef525ed2
New benchmark prep
2017-08-19 23:10:12 +01:00
0b0cf62193
Fix mpi 3 interface change
2017-08-19 13:18:50 -04:00
7d88198387
Merge branch 'develop' into feature/multi-communicator
2017-08-19 13:03:35 -04:00
2f619482b8
Enable blocking stencil send
2017-08-19 12:53:59 -04:00
d6472eda8d
Use mmap
2017-08-19 12:53:18 -04:00
9e658de238
Use Vector
2017-08-19 12:52:44 -04:00
bcefdd7c4e
Align both allocator calls to 2MB
2017-08-19 12:49:02 -04:00
9d45fca8bc
Implement MobiusEOFAFermioncache.cc
2017-08-17 23:45:36 -04:00
ac9e6b63c0
More re-import of Mobius EOFA
2017-08-17 19:28:53 -04:00
e140b3f802
Beginning to re-import Mobius EOFA
2017-08-16 23:36:23 -04:00
d9d3d30cc7
Minor clean-up
2017-08-16 20:57:51 -04:00
