portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-09-20 09:15:38 +01:00
Code Releases Activity
5,244 Commits 86 Branches 14 Tags 74 MiB
6a3b09cf02
Commit Graph

264 Commits

Author SHA1 Message Date
Peter Boyle
e279b2be29 Merge develop 2019-08-14 23:01:59 +01:00
Peter Boyle
48e6efc7c9 Merge branch 'develop' into feature/gpu-port 
Conflicts:
	Grid/qcd/action/fermion/WilsonKernelsAsm.cc
	Grid/qcd/action/fermion/implementation/ImprovedStaggeredFermionImplementation.h
	Grid/qcd/action/fermion/implementation/StaggeredKernelsAsm.h
	benchmarks/Benchmark_comms.cc
 2019-08-14 18:56:54 +01:00
Peter Boyle
263dcbabab Simplify the comms benchmark 2019-07-30 22:51:04 +01:00
Peter Boyle
d85dcc72df Multinode fix 2019-07-20 07:13:28 +01:00
Peter Boyle
0561c2edeb Benchmarks modified for new GPU constructs 2019-06-15 12:52:56 +01:00
Peter Boyle
3e41b1055c Remove Gpu only kernels. 2019-06-09 11:20:01 +01:00
Peter Boyle
da8d87e9da Cuda switch off 2019-06-08 17:11:38 +01:00
Peter Boyle
6d77941990 Drop the 5D vec actions 2019-06-08 13:38:05 +01:00
Peter Boyle
47c063f984 Remove Ls Vec cases from benchmarks 2019-06-04 20:45:35 +01:00
Peter Boyle
ee6f96d85c
Merge pull request #210 from grid-test-organisation/feature/gpu-port-develop 
Cayley fermion functions for GPUs
 2019-05-18 19:06:20 +01:00
Peter Boyle
4e9df9e93c GPU patches 2019-05-18 17:43:11 +01:00
gfilaci
e3c56fd9b3 CayleyZeroCounters before benchmark loop 2019-05-13 15:52:00 +01:00
gfilaci
d9438627d9 M5D benchmark without vector copy overhead 2019-05-02 11:10:57 +01:00
gfilaci
6da9aa9971 replace std::vector with Vector in benchmark 2019-05-02 10:56:22 +01:00
gfilaci
b52fa38f8c seed initialisation of RNG5 2019-05-02 10:36:09 +01:00
Peter Boyle
c43a2b599a GPU support 2019-01-01 15:07:29 +00:00
Peter Boyle
b57a4d32aa Merge branch 'develop' into feature/gpu-port 2018-12-13 05:11:34 +00:00
Antonin Portelli
0ba3d469c7 Benchmark IO in single and double precision 2018-10-17 20:27:34 +01:00
Antonin Portelli
291bc2a1f0 IO benchmark on a list of directories 2018-10-15 17:25:08 +01:00
Peter Boyle
adbdc4e65b Half comms not working on GPU yet, so disable. 2018-09-11 05:15:22 +01:00
Peter Boyle
f4bfeb835d Drop back to smaller Ls 2018-09-09 14:25:06 +01:00
Antonin Portelli
a15a2dfd29 Merge branch 'develop' into feature/hadrons 2018-08-10 16:08:22 +01:00
paboyle
27cdb79063 Sha used to seed from a unique string 2018-08-10 15:11:01 +01:00
Peter Boyle
00b92a91b5 Optimising 2018-07-28 23:46:22 +01:00
paboyle
65533741f7 7 moms 2018-07-28 16:17:47 +01:00
Peter Boyle
131a6785d4
Merge branch 'feature/hadrons-a2a' into feature/hadrons-a2a 2018-07-27 23:03:42 +01:00
paboyle
44f4f5c8e2 Momentum loop 2018-07-27 23:00:16 +01:00
fionnoh
2679df034f Changes to meson field benchmark. Now includes the gammas in the final part of the naive method, both methods compute 
lhs^dag*Gamma*rhs (previously Gamma*lhs^dag*rhs), and checks results.
 2018-07-27 18:31:10 +01:00
paboyle
71e1006ba8 Updated meson field benchmark for dirac structures 2018-07-26 09:09:29 +01:00
fionnoh
24128ff109 Changes needed for MF benchmark to work with comms correctly 2018-07-23 15:51:37 +01:00
Peter Boyle
21a1710b43 Verbose vector length 2018-07-23 06:08:39 -04:00
paboyle
ec9939c1ba Test for faster implementation of meson field inner loop 
This should be possible to cache block at outer levels, global sum across nodes not performed
and deferred to caller to block them all into a big all reduce.
Nc=3 and Fermion is hard coded in an ugly way. We might think about benchmarking whether
a product without the conjugate should be made available by Grid.

It is not clear whether the explicit unroll, or the performing of conjugate on left once
was the real source of the speed up.

Gives 70-80 GF/s on my laptop (single) half that double, and 70GB/s to cache.

This is competitive with dslash and a reasonable stopping point for the optimisation. If necessary we can revisit.
 2018-07-10 12:38:51 +01:00
Peter Boyle
4b04ae3611 Printing improvement 2018-07-05 06:59:38 -04:00
Peter Boyle
2f776d51c6 Gpu specific benchmark saturates memory. Can enhance Grid to do this for expressions, 
but a bitof (known) work.
 2018-07-05 06:58:37 -04:00
paboyle
25becc9324 GPU tweaks for benchmarking; really necessary? 2018-06-13 20:26:07 +01:00
Peter Boyle
eb921041d0 Perf count control 2018-05-12 17:57:32 -04:00
Antonin Portelli
bfbf2f1fa0 no threaded stencil benchmark if OpenMP is not supported 2018-05-03 16:20:01 +01:00
Dr Peter Boyle
1dddd17e3c Benchmark improvements from tesseract 2018-04-27 11:44:46 +01:00
Peter Boyle
fa0d8feff4 Performance of CovariantCshift now non-embarrassing. 2018-04-26 17:56:27 +01:00
Peter Boyle
05b44aef6b Merge branch 'develop' of https://github.com/paboyle/Grid into develop 
Conflicts:
	benchmarks/Benchmark_su3.cc
 2018-04-26 15:38:49 +01:00
Peter Boyle
91a0a3f820 Improvement 2018-04-26 14:48:35 +01:00
Peter Boyle
8f44c799a6 Saving the benchmarking tests for Cshift 2018-04-26 14:48:03 +01:00
Guido Cossu
43f5a0df50 More timers in the integrator 2018-04-26 12:01:56 +09:00
paboyle
2baf193031 Merge branch 'develop' of https://github.com/paboyle/Grid into develop 2018-04-25 00:14:03 +01:00
paboyle
362ba0443a Cshift updates 2018-04-25 00:12:11 +01:00
Guido Cossu
c5b9147b53 Correction of a minor bug in the su3 benchmark 2018-04-24 08:03:57 -07:00
Guido Cossu
a1be533329 Corrected Flop count in Benchmark su3 and expanded the Wilson flow output 2018-04-24 01:19:53 -07:00
paboyle
b5510427f9 physical fermion interface, cshift benchmark in SU3. 2018-04-18 01:43:29 +01:00
paboyle
276f113f28 IO uses master boss node for metadata. 2018-03-30 16:17:05 +01:00
paboyle
ab6afd18ac Still compile if no LIME 2018-03-30 13:39:20 +01:00