Giter Club home page Giter Club logo

cusolverrf.jl's People

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar  avatar  avatar  avatar

cusolverrf.jl's Issues

Testset RFBacthedLowLevel factorization is failing with CUDA 12.3

On the branch master, the following testset fails with the following error:

CUDA runtime 12.3, artifact installation
CUDA driver 12.3
NVIDIA driver 535.104.12, originally for CUDA 12.2

CUDA libraries: 
- CUBLAS: 12.3.2
- CURAND: 10.3.4
- CUFFT: 11.0.11
- CUSOLVER: 11.5.3
- CUSPARSE: 12.1.3
- CUPTI: 21.0.0
- NVML: 12.0.0+535.104.12

Julia packages: 
- CUDA: 5.1.0
- CUDA_Driver_jll: 0.7.0+0
- CUDA_Runtime_jll: 0.10.0+1

Toolchain:
- Julia: 1.9.3
- LLVM: 14.0.6

1 device:
  0: NVIDIA A100 80GB PCIe (sm_80, 79.143 GiB / 80.000 GiB available)
RFBacthedLowLevel factorization: Error During Test at /home/montalex/git/CUSOLVERRF.jl/test/cusolverRF.jl:29
  Got exception outside of a @test
  ReadOnlyMemoryError()
  Stacktrace:
    [1] macro expansion
      @ ~/.julia/packages/CUDA/nIZkq/lib/cusolver/libcusolverRF.jl:236 [inlined]
    [2] #1195
      @ ~/.julia/packages/CUDA/nIZkq/lib/utils/call.jl:27 [inlined]
    [3] #1
      @ ~/.julia/packages/CUDA/nIZkq/lib/cusolver/libcusolver.jl:17 [inlined]
    [4] retry_reclaim(f::CUDA.CUSOLVER.var"#1#2"{CUDA.CUSOLVER.var"#1195#1196"{CUSOLVERRF.RfHandle}}, isfailed::Base.Fix2{typeof(in), Tuple{CUDA.CUSOLVER.cusolverStatus_t}})
      @ CUDA ~/.julia/packages/CUDA/nIZkq/src/pool.jl:359
    [5] check
      @ ~/.julia/packages/CUDA/nIZkq/lib/cusolver/libcusolver.jl:16 [inlined]
    [6] cusolverRfBatchAnalyze
      @ ~/.julia/packages/CUDA/nIZkq/lib/utils/call.jl:26 [inlined]
    [7] CUSOLVERRF.RFBatchedLowLevel(lu_host::CUSOLVERRF.RFSymbolicAnalysis{Float64, Int32}, batchsize::Int64; options::Base.Pairs{Symbol, Union{}, Tuple{}, NamedTuple{(), Tuple{}}})
      @ CUSOLVERRF ~/git/CUSOLVERRF.jl/src/rf_wrapper.jl:222
    [8] RFBatchedLowLevel
      @ ~/git/CUSOLVERRF.jl/src/rf_wrapper.jl:200 [inlined]
    [9] #RFBatchedLowLevel#5
      @ ~/git/CUSOLVERRF.jl/src/rf_wrapper.jl:197 [inlined]
   [10] CUSOLVERRF.RFBatchedLowLevel(A::CuSparseMatrixCSR{Float64, Int32}, batchsize::Int64)
      @ CUSOLVERRF ~/git/CUSOLVERRF.jl/src/rf_wrapper.jl:192
   [11] macro expansion
      @ ~/git/CUSOLVERRF.jl/test/cusolverRF.jl:39 [inlined]
   [12] macro expansion
      @ ~/Applications/julia/julia-1.9.3/share/julia/stdlib/v1.9/Test/src/Test.jl:1498 [inlined]
   [13] macro expansion
      @ ~/git/CUSOLVERRF.jl/test/cusolverRF.jl:30 [inlined]
   [14] macro expansion
      @ ~/Applications/julia/julia-1.9.3/share/julia/stdlib/v1.9/Test/src/Test.jl:1498 [inlined]
   [15] top-level scope
      @ ~/git/CUSOLVERRF.jl/test/cusolverRF.jl:4
   [16] include(fname::String)
      @ Base.MainInclude ./client.jl:478
   [17] top-level scope
      @ ~/git/CUSOLVERRF.jl/test/runtests.jl:15
   [18] include(fname::String)
      @ Base.MainInclude ./client.jl:478
   [19] top-level scope
      @ none:6
   [20] eval
      @ ./boot.jl:370 [inlined]
   [21] exec_options(opts::Base.JLOptions)
      @ Base ./client.jl:280
   [22] _start()
      @ Base ./client.jl:522
Test Summary:                     | Pass  Error  Total   Time
cusolverRF                        |    2      1      3  19.5s
  RFLowLevel factorization        |    2             2  15.4s
  RFBacthedLowLevel factorization |           1      1   1.9s
ERROR: LoadError: Some tests did not pass: 2 passed, 0 failed, 1 errored, 0 broken.
in expression starting at /home/montalex/git/CUSOLVERRF.jl/test/cusolverRF.jl:2
in expression starting at /home/montalex/git/CUSOLVERRF.jl/test/runtests.jl:15
ERROR: Package CUSOLVERRF errored during testing

CUSOLVERRF is broken with CUDA 12.0

CUDA 12.0 has made some change in cusparse, apparently impacting CUSOLVERRF.
Also, the functions used in the backsolve have been deprecated and removed in CUDA 12.0.

Performance regression from CUDA 11.8 to CUDA 12.1

We observed a performance regression due to the new API of CUSPARSE in CUDA 12. This is due to the analysis step being required before each backsolve. This was not the case with CUDA version < 12.

CUDA 12:

for (desc, info) in operations
CUSPARSE.cusparseSpSV_analysis(
CUSPARSE.handle(), s.transa, Ref{T}(alpha), desc, descX, descX, T, s.algo, info, s.buffer,
)

CUDA 11:

For now we detect the loaded CUDA version and use the faster approach of CUDA version < 12, where the analysis stage is only called once.

TagBot trigger issue

This issue is used to trigger TagBot; feel free to unsubscribe.

If you haven't already, you should update your TagBot.yml to include issue comment triggers.
Please see this post on Discourse for instructions and more details.

If you'd like for me to do this for you, comment TagBot fix on this issue.
I'll open a PR within a few hours, please be patient!

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.