Code

Venado optimizations#297

Code

Draft

mewall wants to merge 66 commits intomasterlanl/qmd-progress:masterfrom hackathonlanl/qmd-progress:hackathonCopy head branch name to clipboard

+30,635-124

Commits

Commits on Jan 14, 2025

• Add gpmdk for the Venado hackathon

  mewall

  committed
  9d12c58

  Copy full SHA for 9d12c58

• Add build script for hackathon

  mewall

  committed
  f2a3c08

  Copy full SHA for f2a3c08

• Update bml submodule

  mewall

  committed
  3efdf37

  Copy full SHA for 3efdf37

• Update bml submodule

  mewall

  committed
  8c1bcab

  Copy full SHA for 8c1bcab

• Add electrons.dat to latteTBparams

  mewall

  committed
  48b8022

  Copy full SHA for 48b8022

• Add MPI barrier to pin down performance issue

  mewall

  committed
  05314a5

  Copy full SHA for 05314a5

• Add venado build, env, and run scripts

  mewall

  committed
  ff1bd08

  Copy full SHA for ff1bd08

• Bug fix

  mewall

  committed
  7525e35

  Copy full SHA for 7525e35

• Add TrpCage example for gpmdk

  mewall

  committed
  17ad747

  Copy full SHA for 17ad747

• Fixed line truncations

  

  Christian Francisco Negre

  authored and

  mewall

  committed
  639473f

  Copy full SHA for 639473f

• Added sedacs partition and field induced forces

  

  Christian Francisco Negre

  authored and

  mewall

  committed
  61a5be8

  Copy full SHA for 61a5be8

• Added main fro field

  

  Christian Francisco Negre

  authored and

  mewall

  committed
  a61c7a6

  Copy full SHA for a61c7a6

• NVTX tags

  mewall

  committed
  f62b952

  Copy full SHA for f62b952

• Introducing nvtx tags and some optimizations

  mewall

  committed
  c1aaa69

  Copy full SHA for c1aaa69

• Debug resizing and resize two more arrays

  mewall

  committed
  e3f9a44

  Copy full SHA for e3f9a44

• Resize zqt array and use new bml_transpose_inplace to avoid allocation

  mewall

  committed
  2c63bed

  Copy full SHA for 2c63bed

• Move response into gpmdk in preparation for GPU kernel

  mewall

  committed
  a044282

  Copy full SHA for a044282

• Add some useful build scripts

  mewall

  committed
  dfeb3a1

  Copy full SHA for dfeb3a1

• Update bml submodule

  mewall

  committed
  1aebd59

  Copy full SHA for 1aebd59

• Reduce allocations in hcsf method

  mewall

  committed
  15594cf

  Copy full SHA for 15594cf

• Update build script

  mewall

  committed
  11695ec

  Copy full SHA for 11695ec

• Modifications to support the new bml_transpose Fortran API

  mewall

  committed
  f1d152c

  Copy full SHA for f1d152c

• Add nvtx tags for charges and thread charge calculation

  mewall

  committed
  6195172

  Copy full SHA for 6195172

• Add nvtx tag methods to gpmdk

  mewall

  committed
  2c72a40

  Copy full SHA for 2c72a40

• Debug cray build and work on omp offload

  mewall

  committed
  2c7afd7

  Copy full SHA for 2c7afd7

• Working omp offload for gpmdcov_response

  mewall

  committed
  0ae84e1

  Copy full SHA for 0ae84e1

• Bug fix

  mewall

  committed
  468bca1

  Copy full SHA for 468bca1

• Allocate smaller array for work in kernel

  mewall

  committed
  07f9dcb

  Copy full SHA for 07f9dcb

• Code decorations to investigate MPI imbalance

  mewall

  committed
  5edf99b

  Copy full SHA for 5edf99b

• Build updates

  mewall

  committed
  b941d17

  Copy full SHA for b941d17

• Beginning update of graph MPI

  mewall

  committed
  77f9065

  Copy full SHA for 77f9065

• Preliminary low-communication graph update introduced. Not tested.

  mewall

  committed
  ee06263

  Copy full SHA for ee06263

• Working subgraph graph update method with less MPI communication

  mewall

  committed
  df07623

  Copy full SHA for df07623

• Added new script

  

  Christian Francisco Negre

  authored and

  mewall

  committed
  a4a61d9

  Copy full SHA for a4a61d9

• added bch dslp script

  

  Christian Francisco Negre

  authored and

  mewall

  committed
  6f08dcd

  Copy full SHA for 6f08dcd

• Attempt to fix the hackathon branch

  mewall

  committed
  769d607

  Copy full SHA for 769d607

• Eliminate debug output of FORCESS

  mewall

  committed
  6a2cb5f

  Copy full SHA for 6a2cb5f

• Update bml submodule

  mewall

  committed
  6722bda

  Copy full SHA for 6722bda

• Update bml submodule and move build scripts to scripts/ dir

  mewall

  committed
  87cf0c8

  Copy full SHA for 87cf0c8

• breaking lines

  

  Christian Francisco Negre

  authored and

  mewall

  committed
  63828d5

  Copy full SHA for 63828d5

• fixing gpmd.py

  

  Christian Francisco Negre

  authored and

  mewall

  committed
  52e7fd6

  Copy full SHA for 52e7fd6

• Add more nvtx tags

  mewall

  committed
  23c3af7

  Copy full SHA for 23c3af7

• Added err_var

  

  Christian Francisco Negre

  authored and

  mewall

  committed
  69d937a

  Copy full SHA for 69d937a

• Workaround for Cray matmul bug in gpmdk

  mewall

  committed
  a4dc3a2

  Copy full SHA for a4dc3a2

• Report max time and rank for dH+dS

  mewall

  committed
  26f34d8

  Copy full SHA for 26f34d8

• Fix kernel bug. Better matmul fix.

  Show description for e2b232c
  mewall

  committed
  e2b232c

  Copy full SHA for e2b232c

• Clean up comments and begin working on offload of get_dH_or_dS_vect

  mewall

  committed
  fde467f

  Copy full SHA for fde467f

• Added protections against double alloc

  

  Christian Francisco Negre

  authored and

  mewall

  committed
  59a45c5

  Copy full SHA for 59a45c5

• Prepare for OMP offload optimization

  mewall

  committed
  49154ca

  Copy full SHA for 49154ca

• Fix bug in graph update

  mewall

  committed
  9fff9ab

  Copy full SHA for 9fff9ab

• Use MAGMA pointer in offload response kernel

  mewall

  committed
  0315f8c

  Copy full SHA for 0315f8c

• Working on openACC

  mewall

  committed
  9b5b4bd

  Copy full SHA for 9b5b4bd

• Working offload of response using magma pointer

  mewall

  committed
  829eb87

  Copy full SHA for 829eb87

• fixing compilation bug gfortran

  

  Christian Francisco Negre

  authored and

  mewall

  committed
  578dec1

  Copy full SHA for 578dec1

• Added voltage option

  

  Christian Francisco Negre

  authored and

  mewall

  committed
  bdc7776

  Copy full SHA for bdc7776

• Added dos

  

  Christian Francisco Negre

  authored and

  mewall

  committed
  7779af8

  Copy full SHA for 7779af8

• changes for coarse MD

  

  Christian Francisco Negre

  authored and

  mewall

  committed
  c5dc023

  Copy full SHA for c5dc023

• Update bml submodule

  mewall

  committed
  7b2ece1

  Copy full SHA for 7b2ece1

• Added formatting statements to prg_system_mod XYZ trajectory writing code to fix issue with XYZ coordinates being printed across two lines instead of on a single line

  

  rcorrigan

  authored and

  mewall

  committed
  bfd58e4

  Copy full SHA for bfd58e4

• Opts for nvidia build

  Show description for 8a35116
  mewall

  committed
  8a35116

  Copy full SHA for 8a35116

• NVIDIA opt get_hsmat

  Show description for 89f9eaa
  mewall

  committed
  89f9eaa

  Copy full SHA for 89f9eaa

• NVIDIA opt hsderivative

  Show description for 305dbc9
  mewall

  committed
  305dbc9

  Copy full SHA for 305dbc9

• OpenACC accelerated nonorthocoul

  Show description for cb20d3d
  mewall

  committed
  cb20d3d

  Copy full SHA for cb20d3d

• Add variables to select MD steps for nsys profiling

  mewall

  committed
  33cf90a

  Copy full SHA for 33cf90a

• Disable bisection in gpmd. Nonorthocoul opts.

  mewall

  committed
  0933417

  Copy full SHA for 0933417

Commits on Jan 23, 2025

• Fix merge error

  mewall

  committed
  03ef662

  Copy full SHA for 03ef662