53 Commits

Author	SHA1	Message	Date
Peter Boyle	417dbfa257	Fix	2021-08-10 08:55:35 -07:00
peterx.a.boyle	1eda4d8e0b	Merge branch 'develop' of https://github.com/paboyle/Grid into develop	2021-08-10 05:41:18 -07:00
peterx.a.boyle	50181f16e5	Level 0 IPC set up	2021-08-10 05:35:15 -07:00
Peter Boyle	80ac2a73ca	Check is wrong (HtoD / DtoH)	2021-08-05 18:33:20 -04:00
Peter Boyle	29a22ae603	Simpler SYCL setup	2021-06-22 17:57:20 +00:00
Peter Boyle	6cd9224dd7	SYCL comms buffer allocate	2021-06-16 17:10:55 +00:00
Christoph Lehner	9c9566b9c9	Merge pull request #23 from paboyle/develop ... Sync	2021-03-01 12:33:51 +01:00
Peter Boyle	cd99edcc5f	maxLocalNorm2()	2021-02-04 18:25:49 -05:00
Christoph Lehner	4705aa541d	Allow user to configure ShmDims via environment variables	2021-02-04 14:25:55 +01:00
Peter Boyle	cf76741ec6	Intel DPCPP Gold happy now (compiles all, runs Benchmark_dwf_fp32 )	2020-12-03 03:47:11 -08:00
Peter Boyle	aa8aba6543	--shm-force-mpi	2020-11-16 20:15:50 -05:00
Peter Boyle	13df14f96e	Switch off SHM paths with --disable-shm	2020-11-16 18:07:15 -05:00
Peter Boyle	18ef8056ec	Hide Shared Memory	2020-11-13 04:10:40 +01:00
Peter Boyle	d05ce01809	TOFU behaviour now optional THREAD_MULTIPLE or THREAD_SERIALIZED	2020-11-13 03:52:19 +01:00
Michael Marshall	b71a081cba	Asynchronous calls removed - reflect this in Communicator_none.cc ... (Opportunistic doc update - OpenMP support on Mac OS)	2020-09-21 09:33:23 +01:00
Peter Boyle	c48909590b	MPI asynch call removal	2020-09-17 20:47:32 +01:00
Peter Boyle	446ef40570	HIP IPC	2020-09-17 20:31:46 +01:00
Peter Boyle	a8309638d4	UVM check in MPI calls	2020-09-03 20:29:26 -04:00
Peter Boyle	0c3095e173	Comms buffers to device memory	2020-09-03 15:45:35 -04:00
Christoph Lehner	06007db3d9	true shm_none implementation with GPUs that disables the use of device shared memory for the stencils	2020-08-14 18:37:00 +02:00
Christoph Lehner	3abe09025a	when using SHM_NONE allow multiple ranks per node but without using shared memory	2020-08-06 14:42:38 +02:00
Christoph Lehner	197612bc7a	fast cpu basisRotate and other small cleanups	2020-07-30 07:08:54 -04:00
nmeyer-ur	8726e94ea7	merge upstream develop	2020-07-07 20:26:47 +02:00
nmeyer-ur	1635c263ee	disable TOFU by default	2020-06-30 19:27:08 +02:00
nmeyer-ur	465856331a	switch back to serialized; wrong results on single too	2020-06-15 15:39:39 +02:00
nmeyer-ur	cc958aa9ed	switch back to standard MPI_init due to wrong results in Benchmark_wilson using comms-overlap	2020-06-15 14:21:38 +02:00
Christoph Lehner	3dccd7aa2c	Catch edge case in SharedMemoryMPI::GetShmDims; Change default units to consistent MB in init args; Want last element not past last element in MemoryManagerCache.cc	2020-06-14 13:26:01 -04:00
Christoph Lehner	7974acff54	merged sycl to feature-gpt	2020-06-12 06:49:38 -04:00
Peter Boyle	cdf0a04fc5	Merge branch 'develop' into sycl	2020-06-09 04:00:12 -04:00
Christoph Lehner	9fcb47ee63	Explicit error message instead of infinite loop in GlobalSharedMemory::GetShmDims	2020-06-02 07:44:38 -04:00
nmeyer-ur	4fedd8d29f	switch to MPI_THREAD_SERIALIZED instead of SINGLE	2020-05-27 14:08:34 +02:00
nmeyer-ur	9a86059761	symmetrize VLA and fixed size build messages	2020-05-20 20:05:42 +02:00
nmeyer-ur	b780b7b7a0	guard prevents multiple TOFU messages	2020-05-20 19:20:59 +02:00
nmeyer-ur	fc2e9850d3	temporarily enable TOFU by default when using A64FX or A64FXFIXEDSIZE	2020-05-11 13:25:02 +02:00
nmeyer-ur	ffaaed679e	MPI_THREAD_SINGLE hack for Fugaku, enabled by -DTOFU	2020-05-11 13:21:39 +02:00
Peter Boyle	ea08f193e7	Allocator cache spliit into large/small pools	2020-05-10 05:24:26 -04:00
Peter Boyle	28a1fcaaff	First compile against SYCL	2020-05-05 11:13:27 -07:00
Christoph Lehner	856d168e41	global sum over vectors of uint64_t	2020-03-29 07:56:05 -04:00
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	ec8e060ec7	Summit jsrun GPU mapping updates. Conffigure with --enable-jsrun	2019-10-31 11:46:09 -04:00
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	fa9cd50c5b	Merge branch 'develop' into feature/gpu-port	2019-07-16 11:55:17 +01:00
Peter Boyle	0996ba9396	Pretty messaging	2019-07-12 06:45:31 +01:00
Peter Boyle	2095c12eac	Make detection of HPE 8600 automatic	2019-05-22 09:54:21 +01:00
Peter Boyle	170ba4e619	Ensure different MPI ranks use different GPUs. The mapping works on Tesseract.	2019-04-28 07:32:30 +01:00
Peter Boyle	bc14e86812	Simple check	2019-04-17 12:07:42 +01:00
Peter Boyle	780a67844e	Simple checks	2019-04-17 12:07:17 +01:00
Antonin Portelli	f80c548365	quieter initialisation	2019-02-10 20:47:35 +00:00
Peter Boyle	b57a4d32aa	Merge branch 'develop' into feature/gpu-port	2018-12-13 05:11:34 +00:00