Peter Boyle
2264afda84
Allreduce benchmarking
2026-08-27 22:11:06 -04:00
Peter Boyle
792d2a8258
Masked/orthog dim CartesianRingAllReduce and CartesianRingAllGather
2026-08-27 17:10:49 -04:00
Peter Boyle
7abc19dc03
Allow subsets of dimensions for spatial only sums
2026-08-27 17:03:14 -04:00
Peter Boyle
074bddd48b
Probing environment effects on dense inversion time
2026-08-27 13:56:02 -04:00
Peter Boyle
6cea81bc2c
More probing why test harness vs. slate was a little faster running Grid's dense inverse than normal environment
2026-08-27 13:42:18 -04:00
Peter Boyle
68b9fa86d7
Faster leaf inverse
2026-08-27 12:53:26 -04:00
Peter Boyle
d44964d923
Faster dense inverse, I hope
2026-08-27 12:13:46 -04:00
Peter Boyle
52998539bc
Bigger leaf operations should spead up distributed inverse
2026-08-27 12:08:37 -04:00
Peter Boyle
88acdb8172
More perf diagnostics
2026-08-27 11:22:35 -04:00
Peter Boyle
0857f3aa82
Mor ecomms probing
2026-08-27 09:26:24 -04:00
Peter Boyle and Claude Fable 5
0979371c45
SUMMA telemetry: ring time histogram by message size
...
Per log2-size bucket: messages, GB, seconds inside SendToRecvFrom, GB/s,
% of ring time. Decomposes the 2 GB/s/rank average (probe: 11-20 GB/s
at 8 MB) into latency-bound small messages vs slow large ones vs wait.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com >
Claude-Session: https://claude.ai/code/session_01RAhdQHrkzKmzfxpxW1whLn
2026-08-27 02:02:30 -04:00
Peter Boyle and Claude Fable 5
96071aa492
Dense allgather: stage/scatter x by global-site order, not rank-major block
...
DENSE_DEVICE_SUM=4 assumed each rank's rows of x were the contiguous
[me*nrows,...) block; x and the slab columns are in global-site order
(hX[myGsite*nbasis+b]), the contiguous block is the rank-major index.
Coincide on one rank only -> Frontier VERIFY 0.9965. Stage via myGsite,
scatter through the inverse of BuildRankMajorMap. 4-rank laptop VERIFY
7.38e-07, identical to the allreduce path.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com >
Claude-Session: https://claude.ai/code/session_01RAhdQHrkzKmzfxpxW1whLn
2026-08-27 02:02:01 -04:00
Peter Boyle
e5f1553232
Better reduction
2026-08-26 23:36:31 -04:00
Peter Boyle
06c6011229
Timestamp improvement
2026-08-26 21:56:04 -04:00
Peter Boyle
cbe97cc40d
Smoother search improvement
2026-08-26 21:37:36 -04:00
Peter Boyle
dc1ae3185a
Ring allgather too. Let's nail the Dense CoarseCoarse
2026-08-26 21:05:36 -04:00
Peter Boyle
8e4a304626
Less aggressive L5
2026-08-26 20:17:19 -04:00
Peter Boyle
5372dce0c2
Rerecord, not fade away !
2026-08-26 20:15:14 -04:00
Peter Boyle
bc67e2c584
abs thrust issue
2026-08-26 17:58:18 -04:00
Peter Boyle
e15099eb7b
Both changes to improve the buffer persistence in teh dense distributed inverse and to add verbosity to polynomial smoothers
2026-08-26 17:48:03 -04:00
Peter Boyle
5f8f396e9b
More jobs for testing Mgrid
2026-08-26 14:22:46 -04:00
Peter Boyle
841e59e8c0
GCR coefficient recording and playback
2026-08-26 14:22:00 -04:00
Peter Boyle
b55ca3d50d
Consolidate polynomial smoothers + logging time breakdown in Dense Coarsest invers
2026-08-26 14:20:50 -04:00
Peter Boyle
d7c9076228
Power method added
2026-08-26 06:54:20 -04:00
Peter Boyle
7a292bc4c6
Slate compile fix
2026-08-25 21:39:26 -04:00
Peter Boyle
9213f6e533
Improvements for mulktigrid. + Schur setup improvements
2026-08-25 21:37:12 -04:00
Peter Boyle
5685203b7b
Frontier jobs
2026-08-25 19:18:16 -04:00
Peter Boyle
22dda4a8ed
Update
2026-08-25 19:17:59 -04:00
Peter Boyle
eb22145f49
Updated slate
2026-08-25 17:27:37 -04:00
Peter Boyle
5d92f56bed
omp issues combined with hipcc
2026-08-25 15:45:49 -04:00
Peter Boyle
d42d1855c2
SLATE build updates
2026-08-25 15:33:07 -04:00
Peter Boyle
0a81a41729
Slate changes and warn suppression
2026-08-25 15:29:21 -04:00
Peter Boyle
229f4808c7
Ring all reduce first cut
2026-08-25 12:53:24 -04:00
Peter Boyle
167753aa3f
Adding ring all reduce for large vectors
2026-08-25 12:07:11 -04:00
Peter Boyle
dbbd5dba35
Potentially faster GCR linalg
2026-08-25 07:29:02 -04:00
Peter Boyle
33757c3c94
FOrmat warn avoidance
2026-08-24 20:32:54 -04:00
Peter Boyle
7438e56f94
More controls
2026-08-24 19:42:38 -04:00
Peter Boyle
47416230b1
Verbose on finalize improvement
2026-08-24 19:42:38 -04:00
Peter Boyle
6a854ebd0d
Print trace for teh GCR step
2026-08-24 19:42:38 -04:00
Peter Boyle
8c6bcddbab
Coefficent of GCR printing for poly recording
2026-08-24 19:40:27 -04:00
Peter Boyle
effa31e761
Drop redundant set zero's and optimise persistent history vectors for smoothers
2026-08-24 15:48:26 -04:00
Peter Boyle
7792d7a1a7
Verbose prettier
2026-08-24 11:00:16 -04:00
Peter Boyle
8261dc39d9
More tracing
2026-08-24 10:56:16 -04:00
Peter Boyle
9ed7049736
COMpile fixes
2026-08-23 22:58:06 -04:00
Peter Boyle
5a70066e5b
Claude build fixeds
2026-08-23 21:16:56 -04:00
Peter Boyle
6be8b478c3
Claude 2D block cyclic inverse work
2026-08-23 20:42:21 -04:00
Peter Boyle
34a1220035
Claude 2D block cyclic inverse
2026-08-23 20:41:46 -04:00
Peter Boyle
c97cfe9f5b
This fails
2026-08-23 12:14:58 -04:00
Peter Boyle
01858754ed
Unhappy test looking for work aroudn to avoid allgather
2026-08-23 12:14:23 -04:00
Peter Boyle
c388dd98d6
More verbose phasing on finish
2026-08-23 11:00:43 -04:00
Peter Boyle
38344e035f
Don't print Finalize until made collective
2026-08-23 10:56:36 -04:00
Peter Boyle
14d2f2c06e
Improvements
2026-08-23 02:38:23 -04:00
Peter Boyle
06fa2fc160
use shm buffers
2026-08-22 17:18:12 -04:00
Peter Boyle
a9c22d787a
Benchmark comms with the exact sequence in Stencil as it is now.
2026-08-22 17:14:10 -04:00
Peter Boyle
1e287b6b30
Changes to dense inverse AND improved timing on comms
2026-08-22 11:43:13 -04:00
Peter Boyle
152e9ecf3b
Bandwidth measures inside stencil
2026-08-22 10:51:44 -04:00
Peter Boyle
2e0c19c915
Allgather test
2026-08-21 21:25:18 -04:00
Peter Boyle
4d6931620b
Gather based distributed dense inverse
2026-08-21 12:34:01 -04:00
Peter Boyle
d62fc91996
Better ROCTX tracing
2026-08-20 11:46:33 -04:00
Peter Boyle
3ed05dbbb1
Verbose trasing
2026-08-20 11:46:33 -04:00
Peter Boyle
97c9178785
Ready for faster 3 level solve with dense coarse and BLAS on SRHS !
...
Exciting
2026-08-20 11:44:26 -04:00
Peter Boyle
2fde4f3e50
Trace hooks
2026-08-20 07:38:28 -04:00
Peter Boyle
b31547c903
no compile fix
2026-08-19 23:23:18 -04:00
Peter Boyle
2d56323012
No compile on Frontier
2026-08-19 23:06:38 -04:00
Peter Boyle
63fefb8ea5
new tests for lex lattice
2026-08-19 20:33:10 -04:00
Peter Boyle
e4ec5d0009
FOr pvdagm preconditioners if they work
2026-08-19 20:17:56 -04:00
Peter Boyle
a5e4ce3e48
Better comment
2026-08-19 20:15:21 -04:00
Peter Boyle
2900ce33b5
New files, including v2 multiRHS coarse op
2026-08-19 20:05:48 -04:00
Peter Boyle
ea5bf89955
Failed experiment in the style of Saul Cohens non-red black CG approach to mgrid.
2026-08-19 20:04:29 -04:00
Peter Boyle
109b77ea75
Drop D2 field types
2026-08-19 20:03:40 -04:00
Peter Boyle
333315569c
Benchmark lex lattices -- what is the benefit of Grid's SIMDization layout
2026-08-19 20:02:58 -04:00
Peter Boyle
995e149256
Scalar "vector" type to provide standard lexicographic local lattice layouts
2026-08-19 20:02:01 -04:00
Peter Boyle
971b0f3546
Deprecate doubled vector as much as possible.
...
Arm complex instructions on M3/M4 NEON v8.3 and simplify A64FX code paths/broaden.
Grid_vector_types and Simd.h mainly reorg and prep for sComplexD and sComplexF alternat Nsimd=1 types
2026-08-19 20:00:26 -04:00
Peter Boyle
ab9603b28d
Support for lexLattice in coarse space and eliminating layout changes in
...
multigrid
2026-08-19 19:58:35 -04:00
Peter Boyle
adfd5853cc
lex layout arrays support in fermion instantiations
2026-08-19 19:57:26 -04:00
Peter Boyle
cc54c90ddc
sComplex simd, also commit coarse space as lex lattice tests
2026-08-19 19:54:18 -04:00
Peter Boyle
6a4071cdc2
New lex lattice subdirs
2026-08-19 19:52:33 -04:00
Peter Boyle
e92be14df8
Recurse to lexLattice
2026-08-19 19:50:56 -04:00
Peter Boyle
e9be1a91b3
Speed up on mac arm with --enable-simd=NEONv8.3
2026-08-19 19:49:59 -04:00
Peter Boyle
7db3b96d2d
Deprecate D2 fields
2026-08-19 19:49:26 -04:00
Peter Boyle
429ad2a71a
sComplex and lexLattice prep work
2026-08-19 19:42:56 -04:00
Peter Boyle
0b1598e9ef
Annoying warn on printf fix
2026-08-19 19:41:43 -04:00
Peter Boyle
8d04b91932
Clean up whitespace. Gauge group needs a Gimpl or some other policy parameter.
...
Perhaps rename iGaugeGroupd and pass EITHER vComplexD, vComplexF OR sCompexD / sComplexF
Use a typedef to restore the original Gauge group
2026-08-19 19:39:42 -04:00
Peter Boyle
665094d94b
Lex gauge actions.
2026-08-19 19:38:48 -04:00
Peter Boyle
919d126af1
Remove D2 lattice fields; remove WilsonCompressor based half precision project; add sComplex lexLattice layouts
...
Fix staggered instantiations for new types
2026-08-19 19:37:29 -04:00
Peter Boyle
abd5df1729
Discourage use of D2 lattices -- it was low performance.
...
Introduce scalar lexLattice layout
2026-08-19 19:35:59 -04:00
Peter Boyle
3243ad91e8
Fix race condition during read after write due to missing barrier.
...
Enable both lexLattice (scalar order) and Lattice (vector or partial SOA order)
2026-08-19 19:34:54 -04:00
Peter Boyle
c10d451f4d
Enable interafaces with both coarse and fine being either single higher dim multiRHS lattice OR vector of lower dim fields
2026-08-19 19:33:55 -04:00
Peter Boyle
9c4a61850e
Features for better BLAS based coarse grids, for both single and multi RHS.
...
Use sComplex and lexLattice in future to avoid layout changes and copies
2026-08-19 19:32:36 -04:00
Peter Boyle
5e3fe24412
Recursive Schur complement based distributed matrix inverse by Claude after discussing plan and instructions from me
2026-08-19 19:31:46 -04:00
Peter Boyle
7b81ca6b52
Changes to make Nsimd=1 sComplex lexLattices functional for "where" statements
2026-08-19 19:30:34 -04:00
Peter Boyle
8c9ea4cde2
Json "" warning remove
2026-08-19 19:29:27 -04:00
Peter Boyle
0ac72cb6a3
Don't use host as it sucks.
2026-08-15 12:24:49 -04:00
Peter Boyle
c4b089cf15
More timing updates
2026-08-15 09:51:56 -04:00
Peter Boyle
85e1bbf4bc
Added gather benchmark to cross refer to the 'lazy' analogue of summing a vector with many zeroes in regions not owned by THIS rank
2026-08-15 09:51:12 -04:00
Peter Boyle
032f111c8b
Instrument as not faster
2026-08-15 07:42:18 -04:00
Peter Boyle
e38528888a
Submission scripts for distributed schur coarse grid inverse
2026-08-14 18:28:46 -04:00
Peter Boyle
7c9a6d354e
Updates for thrust
2026-08-14 18:21:45 -04:00
Peter Boyle
3ec90803ac
no compil patches
2026-08-14 18:21:13 -04:00
Peter Boyle
c22473f15d
First test of distributed schur recursive inverse
2026-08-14 17:53:35 -04:00
Peter Boyle
