1
0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-10 07:55:35 +00:00
Commit Graph

7266 Commits

Author SHA1 Message Date
Peter Boyle
3f2fd49db4 Merge branch 'develop' of https://github.com/paboyle/Grid into develop 2023-03-27 17:29:54 -07:00
Peter Boyle
0efa107cb6 Script update 2023-03-27 17:29:43 -07:00
Peter Boyle
8feedb4f6f Include files moved 2023-03-27 17:29:21 -07:00
Peter Boyle
05e562e3d7 Move the copy synch out to stencil and do one per call instead of one per packet 2023-03-27 17:28:38 -07:00
Peter Boyle
dd3bbb8fa2 MOve the synchronise out to the stencil so one call instead of one call per packet 2023-03-27 17:27:45 -07:00
Peter Boyle
2fbcf13c46 SYCL fix 2023-03-27 14:25:14 -07:00
Peter Boyle
4ea48ef0c4
Merge pull request #419 from lehner/feature/gpt
Separate rankSum from sum
2023-03-24 15:42:16 -04:00
Peter Boyle
5c85774ee3 Merge branch 'feature/dirichlet' of https://github.com/paboyle/Grid into feature/dirichlet 2023-03-24 15:40:57 -04:00
Peter Boyle
d8a9a745d8 stream synchronise 2023-03-24 15:40:30 -04:00
Peter Boyle
dcf172da3b
Merge pull request #415 from paboyle/feature/block_lanczos22
Feature/block lanczos22
2023-03-24 12:08:16 -04:00
Peter Boyle
d57ed25071
Merge branch 'feature/dirichlet' into feature/block_lanczos22 2023-03-24 12:08:09 -04:00
Peter Boyle
546be724e7
Merge pull request #421 from UniOfLeicester/feature/accel_Copy_plane
Populate the Cshift_table in the GPU
2023-03-24 12:04:06 -04:00
Peter Boyle
8a1b9073f9 Mshift update 2023-03-23 15:39:30 -04:00
Peter Boyle
1a7114d4b9 Temporary algorithm while sorting out mixed prec 2023-03-23 15:38:35 -04:00
Peter Boyle
3f385f717c Merge branch 'feature/dirichlet' of https://github.com/paboyle/Grid into feature/dirichlet
Conflicts:
	systems/PVC/benchmarks/run-2tile-mpi.sh
	systems/PVC/config-command
2023-03-23 14:52:53 -04:00
Peter Boyle
481bbaf1fc Interface to query memory use 2023-03-23 12:55:31 -04:00
Peter Boyle
281488611a WriteDiscard on construct 2023-03-23 10:28:50 -04:00
Peter Boyle
c180a52518 Merge branch 'feature/dirichlet' of https://www.github.com/paboyle/Grid into feature/dirichlet 2023-03-23 10:28:01 -04:00
Peter Boyle
90130e25e9 TODO list 2023-03-23 10:27:02 -04:00
Peter Boyle
23298acb81
Merge pull request #424 from giltirn/feature/dirichlet-precchange
Precision change implementation
2023-03-22 23:04:52 -04:00
Peter Boyle
52384e34cf Discard on construct 2023-03-22 19:40:32 -04:00
Peter Boyle
d0bb033ea2 Device resident GPU block buffer instead of UVM as hit likely UVM
bug. Code worked on CUDA 11.4 but fails on later drivers (certainly 530.30.02, but need to
find the perlmutter driver version).
2023-03-22 19:07:32 -04:00
Peter Boyle
c6621806ca Compiling on laptop and running 2023-03-21 17:27:09 -04:00
Peter Boyle
0b6f0f6d2f Merge branch 'feature/dirichlet' of https://www.github.com/paboyle/Grid into feature/dirichlet 2023-03-21 16:06:55 -04:00
Peter Boyle
b5b759df73 Merge branch 'develop' into feature/dirichlet 2023-03-21 16:05:46 -04:00
Peter Boyle
7db8dd7a95 Merge branch 'feature/dirichlet' of https://github.com/paboyle/Grid into feature/dirichlet 2023-03-21 16:04:27 -04:00
Peter Boyle
8b43be39c0 Config command 2023-03-21 16:00:52 -04:00
Peter Boyle
f17f879206 Test update 2023-03-21 15:59:29 -04:00
Peter Boyle
68428fceab Integrator update 2023-03-21 15:58:49 -04:00
Peter Boyle
4135f2dcd1 Compressor 2023-03-21 15:41:41 -04:00
Peter Boyle
c5bdf61215 AUdit fix 2023-03-21 15:38:39 -04:00
Peter Boyle
88e218e8ee Stencil updates 2023-03-21 15:37:58 -04:00
Peter Boyle
0f2b786436 Vector -> vector 2023-03-21 15:36:11 -04:00
Peter Boyle
e1c326558a COmms improvements 2023-03-21 08:53:56 -07:00
Peter Boyle
bae0f8ea99
Merge pull request #425 from rrhodgson/feature/CacheLogging
Huge Cache
2023-03-21 08:59:08 -04:00
Peter Boyle
bbbcd36ae5
Merge pull request #426 from rrhodgson/feature/LCDeflation
Batched Local Coherence Tools
2023-03-21 08:58:40 -04:00
Peter Boyle
39c0815d9e WriteDiscard 2023-03-21 08:57:29 -04:00
Peter Boyle
a997d24743 Remove nofma 2023-03-14 12:10:31 -07:00
Peter Boyle
861e5d7f4c SYCL version update. Why do they keep making incompatible changes 2023-03-14 12:10:02 -07:00
Peter Boyle
14cc142a14 Warning remove 2023-03-14 12:09:26 -07:00
Peter Boyle
f36b87deb5 syscall fix 2023-03-14 12:09:00 -07:00
Peter Boyle
eeb6e0a6e3 Renable cache blocking and efficient UPI type SHM comms 2023-03-14 09:10:27 -07:00
Peter Boyle
cad5b187dd Cleanup 2023-03-14 09:08:16 -07:00
Peter Boyle
87697eb07e SHared compile 2023-03-14 09:07:36 -07:00
a3e935c902 Batched block project/promote size checks 2023-02-27 11:38:16 +00:00
7731c7db8e Add huge cache type and allow Ncache==0 2023-02-26 14:15:28 +00:00
ff97340324 Expose cached bytes 2023-02-26 12:22:45 +00:00
Christopher Kelly
83d86943db Fixed compile bug in MemoryManagerShared caused by Audit function not being passed a string 2023-02-23 13:09:45 -05:00
Christopher Kelly
e82cf1d311 Further prec-change improvements
Mixed prec CG algorithm has been modified to precompute precision change workspaces

As the original Test_dwf_mixedcg_prec has been coopted to do a performance stability and reproducibility test, requiring the single-prec CG to be run 200 times, I have created a new version of Test_dwf_mixedcg_prec in the solver subdirectory that just does the mixed vs double CG test
2023-02-23 09:45:29 -05:00
Christopher Kelly
1db58a8acc Precision change improvements
Added a new, much faster implementation of precision change that uses (optionally) a precomputed workspace containing pointer offsets that is device resident, such that all lattice copying occurs only on the device and no host<->device transfer is required, other than the pointer table. It also avoids the need to unpack and repack the fields using explicit lane copying. When this new precisionChange is called without a workspace, one will be computed on-the-fly; however it is still considerably faster than the original implementation.

In the special case of using double2 and when the Grids are the same, calls to the new precisionChange will automatically use precisionChangeFast, such that there is a single API call for all precision changes.

Reliable update and mixed-prec multishift have been modified to precompute precision change workspaces

Renamed the original precisionChange as precisionChangeOrig

Fixed incorrect pointer offset bug in copyLane

Added a test and a benchmark for precisionChange

Added a test for reliable update CG
2023-02-21 10:52:42 -05:00