Code

Gpu#3

Code

Open

jwscook wants to merge 23 commits intomainjwscook/MPIQR.jl:mainfrom gpujwscook/MPIQR.jl:gpuCopy head branch name to clipboard

+181-117

Commits

Commits on Sep 30, 2024

• similar instead of zeros etc

  jwscook

  committed
  e89afd5

  Copy full SHA for e89afd5

Commits on Oct 2, 2024

• runs but is wrong on gpu

  jwscook

  committed
  7314c63

  Copy full SHA for 7314c63

• update tests

  jwscook

  committed
  8836e1b

  Copy full SHA for 8836e1b

• works w/ 1 rank on GPU but is it slower?

  jwscook

  committed
  7438a8a

  Copy full SHA for 7438a8a

Commits on Oct 3, 2024

• dispatchable axpy!

  jwscook

  committed
  3da19d6

  Copy full SHA for 3da19d6

• better use of views

  jwscook

  committed
  7dd8fab

  Copy full SHA for 7dd8fab

Commits on Oct 4, 2024

• is gpu branch as fast as main?

  jwscook

  committed
  c196916

  Copy full SHA for c196916

Commits on Oct 5, 2024

• add a function workaround for lack of dotu

  jwscook

  committed
  ac7d3b8

  Copy full SHA for ac7d3b8

Commits on Oct 6, 2024

• try reduce allocs

  jwscook

  committed
  da13f99

  Copy full SHA for da13f99

• type instabilities in columnowver and iterator

  jwscook

  committed
  67cdee5

  Copy full SHA for 67cdee5

Commits on Oct 7, 2024

• still slow on gpu. time is in norm?!

  jwscook

  committed
  354cd14

  Copy full SHA for 354cd14

• gitignore

  jwscook

  committed
  c7b26a3

  Copy full SHA for c7b26a3

• more normal indices in recurse

  jwscook

  committed
  adf93d0

  Copy full SHA for adf93d0

Commits on Oct 8, 2024

• optimise blocks shennanigans

  jwscook

  committed
  3df04ae

  Copy full SHA for 3df04ae

Commits on Oct 19, 2024

• use sqrt and dot instead of norm

  jwscook

  committed
  4ffb437

  Copy full SHA for 4ffb437

Commits on Nov 6, 2024

• solve against multiple rhses

  jwscook

  committed
  6404544

  Copy full SHA for 6404544

Commits on Nov 9, 2024

• fix for abstractarray

  jwscook

  committed
  0224429

  Copy full SHA for 0224429

• fix dispatch on Base:*(::MPIQRMatrix ...

  jwscook

  committed
  2307b79

  Copy full SHA for 2307b79

Commits on Nov 14, 2024

• fix mul

  jwscook

  committed
  0287300

  Copy full SHA for 0287300

Commits on Jan 10, 2025

• allow arbitrary sized matrices regardless of blocksize and/or number of ranks

  jwscook

  committed
  adcb4ca

  Copy full SHA for adcb4ca

Commits on Jan 17, 2025

• same changes made as in main

  jwscook

  committed
  8bb79d8

  Copy full SHA for 8bb79d8

Commits on Jan 20, 2025

• better consistency of types in qr!

  jwscook

  committed
  aed8fd7

  Copy full SHA for aed8fd7

• allow broadcasted setindex by <: AbstractMatrix

  jwscook

  committed
  dc1984d

  Copy full SHA for dc1984d