[JOSS REVIEW] Use of deprecated function cusparseScsrmv in CUDA executor

[adam-m-jcbs]

This issue is part of the functionality aspect of a JOSS review (see #597 )

I have run into a new issue in building CUDA-enabled ginkgo while building. After a successful cmake, an attempt to build with make gives

[  1%] Building CXX object cuda/CMakeFiles/ginkgo_cuda.dir/base/executor.cpp.o
In file included from /home/ajacobs/Reporoot/ginkgo/cuda/base/executor.cpp:43:
/home/ajacobs/Reporoot/ginkgo/cuda/base/cusparse_bindings.hpp: In function ‘void gko::kernels::cuda::cusparse::spmv(cusparseHandle_t, cusparseOperation_t, gko::int32, gko::int32, gko::int32, const float*, cusparseMatDescr_t, const float*, const int32*, const int32*, const float*, const float*, float*)’:
/home/ajacobs/Reporoot/ginkgo/cuda/base/cusparse_bindings.hpp:118:33: error: ‘cusparseScsrmv’ was not declared in this scope; did you mean ‘cusparseSbsrmv’?
  118 | GKO_BIND_CUSPARSE32_SPMV(float, cusparseScsrmv);

I deploy the latest stable CUDA tookit and dev environment, so when I looked at my cusparse header files I confirmed that cusparseScsrmv is no longer provided and was deprecated as mentioned here (section 8.8.).

NVIDIA suggests "use cusparseCsrmvEx() instead."

In the meantime, I will explore workarounds (and welcome suggestions). As CUDA toolkits are made current, however, this will become an issue for more users/systems.

[upsj]

Thanks for letting us know! Can you provide the exact CUDA version you are using?

[adam-m-jcbs]

Lightning fast as usual @upsj ! I aspire!

Here is the result of deviceQuery:

/deviceQuery Starting...

 CUDA Device Query (Runtime API) version (CUDART static linking)

Detected 1 CUDA Capable device(s)

Device 0: "Quadro T2000"
  CUDA Driver Version / Runtime Version          11.0 / 11.0
  CUDA Capability Major/Minor version number:    7.5
  Total amount of global memory:                 3912 MBytes (4101898240 bytes)
  (16) Multiprocessors, ( 64) CUDA Cores/MP:     1024 CUDA Cores
  GPU Max Clock rate:                            1785 MHz (1.78 GHz)
  Memory Clock rate:                             4001 Mhz
  Memory Bus Width:                              128-bit
  L2 Cache Size:                                 1048576 bytes
  Maximum Texture Dimension Size (x,y,z)         1D=(131072), 2D=(131072, 65536), 3D=(16384, 16384, 16384)
  Maximum Layered 1D Texture Size, (num) layers  1D=(32768), 2048 layers
  Maximum Layered 2D Texture Size, (num) layers  2D=(32768, 32768), 2048 layers
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       49152 bytes
  Total number of registers available per block: 65536
  Warp size:                                     32
  Maximum number of threads per multiprocessor:  1024
  Maximum number of threads per block:           1024
  Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
  Max dimension size of a grid size    (x,y,z): (2147483647, 65535, 65535)
  Maximum memory pitch:                          2147483647 bytes
  Texture alignment:                             512 bytes
  Concurrent copy and kernel execution:          Yes with 3 copy engine(s)
  Run time limit on kernels:                     No
  Integrated GPU sharing Host Memory:            No
  Support host page-locked memory mapping:       Yes
  Alignment requirement for Surfaces:            Yes
  Device has ECC support:                        Disabled
  Device supports Unified Addressing (UVA):      Yes
  Device supports Managed Memory:                Yes
  Device supports Compute Preemption:            Yes
  Supports Cooperative Kernel Launch:            Yes
  Supports MultiDevice Co-op Kernel Launch:      Yes
  Device PCI Domain ID / Bus ID / location ID:   0 / 1 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 11.0, CUDA Runtime Version = 11.0, NumDevs = 1
Result = PASS

Arch's package database defaults to CUDA 11.0.2 (https://www.archlinux.org/packages/community/x86_64/cuda/).

I do suspect it's possible CUDA on my system went from a late 10.x into 11.0 between builds, causing cusparseScsrmv to disappear.

[thoasm]

Hello @adam-m-jcbs,
thank you for your feedback. I am assuming you are using CUDA 11 since it is the 'latest stable CUDA toolkit'.
In CUDA 11, a lot of functionality was removed (that we used in Ginkgo). We are currently working on getting it to work with CUDA 11 in the PR #603, but since there are so many changes (and new functions with not ideal documentation), it takes a while to make it work.

[thoasm]

If you are using ArchLinux, then yes, that actually happened to me as well recently.

[adam-m-jcbs]

Thanks for pointing that out, @thoasm . I missed that PR. For the purposes of the functionality review #597 I will downgrade to 10.x and evaluate what was submitted. This issue is more relevant for the future.