portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2026-07-18 16:13:28 +01:00
Code Releases Activity
5,771 Commits 95 Branches 14 Tags
e4784042915d5f622fefa7e7f4189c11943ae11f
Commit Graph
100 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
Peter Boyle e478404291 Tuned up significantly on GPU, but another 10x in coarse space required 2019-12-17 05:03:25 -05:00
Peter Boyle 9aafd20468 Simple block project promote runs faster on GPU 2019-12-17 05:01:39 -05:00
Peter Boyle 9e15474999 Accelerator loop attempt at speed up 2019-12-14 05:28:16 -05:00
Peter Boyle 152b525a4d Typo fix 2019-12-13 22:44:42 -05:00
Peter Boyle d18994eddc offload more of mgrid to GPU 2019-12-13 22:08:11 -05:00
Peter Boyle 736b19485e Faster set up and some dead code ifdef'ed out 2019-12-13 21:30:48 -05:00
Peter Boyle 5bfd1470ad Merge branch 'develop' into feature/hdcr 2019-12-10 21:51:06 -05:00
Peter Boyle 6957b0b58a Merge branch 'develop' of https://github.com/paboyle/Grid into develop 2019-12-10 21:50:42 -05:00
Peter Boyle d73f0b8618 Verbose for temporary debug 2019-12-10 21:50:06 -05:00
Peter Boyle 0b3a3562c3 Some MPI (summit) create sigusr2, so trap that 2019-12-10 21:49:12 -05:00
Peter Boyle 710fee5d26 Subspace setup testing code 
and timing verbose
 2019-12-10 21:48:42 -05:00
Peter Boyle bab0bf2e93 Merge branch 'develop' into feature/hdcr 2019-12-10 21:47:41 -05:00
Peter Boyle f2a4f13111 Must offload the Coarsened matrix if Stencil buffers are device resident 2019-12-10 19:32:12 -05:00
Peter Boyle edd1c924eb Merge branch 'develop' of https://github.com/paboyle/Grid into develop 2019-12-09 03:53:01 -05:00
Peter Boyle 9b6b0caa55 Junk commit fix 2019-12-09 03:01:58 -05:00
Peter Boyle 2a48617ac5 Merge branch 'develop' of https://github.com/paboyle/Grid into develop 2019-12-09 03:00:00 -05:00
Peter Boyle 876d9c957c QMR 2019-12-09 02:59:49 -05:00
Peter Boyle 295e535f93 QMR 2019-12-09 02:59:35 -05:00
Peter Boyle 58a31f0763 QMR implemented, preserve even if not used much 2019-12-09 02:59:13 -05:00
Peter Boyle 3d2fe80780 Temporary size depends on checkerboard/uncheckerboard. The Mdir cares 2019-12-09 02:58:24 -05:00
Peter Boyle e43fce1083 Clean up and simplify a little. 2019-12-09 02:55:45 -05:00
Peter Boyle 0dfdf80407 Logging 2019-12-09 02:54:52 -05:00
Peter Boyle 2912071f83 Add non hermitian operator 2019-12-09 02:51:53 -05:00
Peter Boyle 26605ef387 HDCR back to working 2019-12-09 02:51:01 -05:00
Peter Boyle 997790ad24 Allow subspace setup to no converge 2019-11-26 14:04:28 -05:00
Peter Boyle 900d6fad21 fp16 mandatory. Use SFW is not available as hdw 2019-11-26 13:26:43 -05:00
Peter Boyle d1a89af8c9 Change to reporting 2019-11-22 10:49:10 -05:00
Peter Boyle d91ba1f6cc NVCC timer support 2019-11-21 20:11:19 +00:00
Peter Boyle f4d27e7090 Merge branch 'develop' of https://github.com/paboyle/Grid into develop 2019-11-21 20:09:31 +00:00
Peter Boyle feb1ff3494 Fix nocompilee under NVCC 2019-11-21 20:03:39 +00:00
Peter Boyle 8ef6175acc Merge branch 'develop' of https://github.com/paboyle/Grid into develop 2019-11-21 15:02:21 -05:00
Peter Boyle e4399e3ee1 --enable-summit flag 2019-11-21 15:02:10 -05:00
Peter Boyle 98ea67b636 IBM summit optimisation. Synchronise in node is still btweeen 2 halves of AC922, so could 
be a little faster
 2019-11-21 15:00:46 -05:00
Peter Boyle ac614cbc53 Merge branch 'develop' of https://github.com/paboyle/Grid into develop 2019-10-31 11:46:43 -04:00
Peter Boyle ec8e060ec7 Summit jsrun GPU mapping updates. Conffigure with --enable-jsrun 2019-10-31 11:46:09 -04:00
Peter Boyle b473405652 Tensor ambiguous fix 2019-08-29 09:36:41 -05:00
Peter Boyle 9b7a6d197f Fix for GCC preprocessor/pragma handling bug 2019-08-23 14:37:46 +01:00
Peter Boyle 59cd7f3b70 Trips up NVCC for reasons I dont understand on summit 2019-08-23 06:03:49 -04:00
Peter Boyle 28d6be2a4e Fix GCC complaint 2019-08-22 18:56:37 +01:00
Peter Boyle 95f66cc93c Merge branch 'feature/gpu-port' into develop 2019-08-15 02:19:31 +01:00
Peter Boyle 12eb2a6a34 Zero() change 2019-08-15 01:43:00 +01:00
Peter Boyle 7c8902b04f Merge branch 'develop' into feature/gpu-port 2019-08-15 01:33:07 +01:00
Peter Boyle 4278caa030 Force a couple of things to compile on NVCC 2019-08-15 01:32:03 +01:00
Peter Boyle be37dfb6f8 Remove debug code 2019-08-15 01:31:40 +01:00
Peter Boyle 5e8437029f nvcc error suppress 2019-08-15 01:31:12 +01:00
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 3e49dc8a67 Reduction finished and hopefully fixes CI regression fail on single precisoin and force 2019-08-14 15:18:34 +01:00
Peter Boyle 96ac56cace Double precision variants for summation accuracy 2019-08-14 13:08:01 +01:00
Peter Boyle 2b037e3daa Update todo list 2019-08-14 13:07:26 +01:00
Peter Boyle 2d2de7aede Freeze the seed 2019-08-14 13:07:11 +01:00
Peter Boyle ce97638bac Think the reduction is now sorted and cleaned up 2019-08-11 11:09:01 +01:00
Peter Boyle 53e3ab4131 Fix force term 2019-08-11 11:06:13 +01:00
Peter Boyle 9cd33a7b9c Printing improvement 2019-07-31 08:01:24 +01:00
Peter Boyle 639dc1ab21 GPU reduction fix and also exit backtrace option 2019-07-31 01:23:23 +01:00
Peter Boyle 9117f61109 GPU friendly 2019-07-31 01:22:54 +01:00
Peter Boyle bca36d9bc3 Merge branch 'develop' of https://github.com/paboyle/Grid into develop 2019-07-30 22:51:23 +01:00
Peter Boyle 263dcbabab Simplify the comms benchmark 2019-07-30 22:51:04 +01:00
Peter Boyle 9dad7a0094 Reproducible reduction and axpy_norm offload from Gianluca. 
Hopefully get CG running entirely on GPU
 2019-07-30 00:14:12 +01:00
Peter Boyle 1282e1067f Do the force term on the accelerator too. Needed particularly because comms buffers 
are device memory.
 2019-07-29 22:58:35 +01:00
Peter Boyle 275c1c920f More info dump on error from CUDA 2019-07-26 12:18:53 +01:00
Peter Boyle fe700a183a Getting HMC to run 2019-07-26 12:18:29 +01:00
Peter Boyle 34108296cd Merge branch 'develop' into feature/gpu-port 
Conflicts:
	Grid/simd/Grid_avx512.h
 2019-07-20 17:05:35 +01:00
Peter Boyle 76c704b84b Intrinsics for CLANG are now fixed in v6 2019-07-20 16:52:24 +01:00
Peter Boyle ce255ec359 Relocate to fix build failure for comms none 2019-07-20 16:37:03 +01:00
Peter Boyle 1c096626cb Hypercube defaults to on if HPE detected, but override to off possible 2019-07-20 16:06:16 +01:00
Peter Boyle ce8b247426 Compiles 2019-07-20 15:16:02 +01:00
Peter Boyle 80481f81be Constructor typo 2019-07-20 09:58:24 +01:00
Peter Boyle d85dcc72df Multinode fix 2019-07-20 07:13:28 +01:00
Peter Boyle 3fedcd6d52 Compiles 2019-07-20 07:12:44 +01:00
Peter Boyle 25ba4c5f80 Merge branch 'develop' into feature/gpu-port 
Conflicts:
	HMC/Mobius2p1fEOFA.cc
	tests/forces/Test_rect_force.cc
 2019-07-19 11:01:55 +01:00
Peter Boyle 671bcbcccb Merge branch 'develop' of https://github.com/paboyle/Grid into develop 2019-07-19 10:48:22 +01:00
Peter Boyle ff325376cb Fix single precision deriv test fail 2019-07-19 10:47:44 +01:00
Peter Boyle 9e926e3fc5 Build fix in develop 2019-07-19 10:01:52 +01:00
Peter Boyle 775eaee199 Fix for suspected Intel 2018.1 compiler bug under O3 2019-07-19 07:57:34 +01:00
Peter Boyle 0fd2827d5d Fix fail in single 2019-07-19 05:28:26 +01:00
Peter Boyle bdd79f9ef8 TODO update 2019-07-18 22:04:28 +01:00
Peter Boyle 0695f8cec2 Single precision compile fix. Soon deprecate single precision 2019-07-18 22:02:31 +01:00
Peter Boyle 9fa705c5a0 comma fix 2019-07-18 21:38:11 +01:00
Peter Boyle 331f5a53dc New header 2019-07-18 14:51:09 +01:00
Peter Boyle a23dc295ac Remove compiler errors and warnings 2019-07-18 14:47:02 +01:00
Peter Boyle 08904f830e Merge develop 2019-07-16 11:59:56 +01:00
Peter Boyle fa9cd50c5b Merge branch 'develop' into feature/gpu-port 2019-07-16 11:55:17 +01:00
Peter Boyle 7c11525d1a Local stencil for complex wilson loops etc 2019-07-14 14:05:09 +01:00
Peter Boyle 42c1dbb1d1 General local stencil first cut for Patrick force term 2019-07-14 14:04:28 +01:00
Peter Boyle 6179acfda0 Put back a call that was required 2019-07-14 13:59:54 +01:00
Peter Boyle 07601ac1f5 Replace instantiation of Gparity 2019-07-12 17:18:12 +01:00
Peter Boyle 705a8098b2 Merge branch 'feature/gpu-port' of https://github.com/paboyle/Grid into feature/gpu-port 
Conflicts:
	Grid/stencil/Stencil.h
 2019-07-12 17:14:11 +01:00
Peter Boyle a29b43d755 Stencil comms cleaner 2019-07-12 17:12:25 +01:00
Peter Boyle 368c8369ce Merge branch 'feature/gpu-port' of https://github.com/paboyle/Grid into feature/gpu-port 2019-07-12 17:11:29 +01:00
Peter Boyle c0d89a2dbb TODO updates 2019-07-12 17:11:15 +01:00
Peter Boyle 78ebd93281 Cuda 9.1 happy 2019-07-12 17:11:00 +01:00
Peter Boyle 3d58daf70f Safety check 2019-07-12 17:10:35 +01:00
Peter Boyle bd155ca5c0 Overlap comms with comput now supported 2019-07-12 09:09:40 +01:00
Peter Boyle 91e2cf9b40 All axes can be used for comms now 2019-07-12 09:08:26 +01:00
Peter Boyle 3cc9947731 Better welcome printing 2019-07-12 06:47:51 +01:00
Peter Boyle f15eeb0283 localise scope of variables declared in macro 2019-07-12 06:47:01 +01:00
Peter Boyle 0996ba9396 Pretty messaging 2019-07-12 06:45:31 +01:00
Peter Boyle 966a203dcb Interactions with GPU compilation 2019-07-11 03:16:17 +01:00
Peter Boyle 44170cc15f Initialise CUDA device prior to entering MPI. 
This may or may not interact with Summit which configures MPI - CUDA mapping with jsrun.
TBD
Cases of OpenMPI and MVAPICH are covered, and default to cudaSetDevice(0) otherwise
 2019-07-11 03:14:23 +01:00