Incorrect scalar-sparse matrix multiplication #3010
Description
Describe the bug
The scalar-matrix multiplication returns a wrong matrix after updating to v5.9.6
To reproduce
The Minimal Working Example (MWE) for this bug:
using SparseArrays, LinearAlgebra
using CUDA
using CUDA.CUSPARSE
A = spdiagm(1 => sqrt.(1:20))
# A = 20 * sprand(20, 20, 0.2) # this surprisingly works
dA = CuSparseMatrixCSC(A)
B = 0.1 * A
dB = 0.1 * dA
SparseMatrixCSC(dB) - B |> norm # 0.447213595499958Manifest.toml
The file content was too big. I can upload it if necessary.
Expected behavior
A correct matrix
Version info
Details on Julia:
Julia Version 1.12.3
Commit 966d0af0fdf (2025-12-15 11:20 UTC)
Build Info:
Official https://julialang.org release
Platform Info:
OS: Linux (x86_64-linux-gnu)
CPU: 32 × 13th Gen Intel(R) Core(TM) i9-13900KF
WORD_SIZE: 64
LLVM: libLLVM-18.1.7 (ORCJIT, alderlake)
GC: Built with stock GC
Threads: 16 default, 1 interactive, 16 GC (on 32 virtual cores)
Environment:
JULIA_EDITOR = code
JULIA_VSCODE_REPL = 1
JULIA_NUM_THREADS = 16
Details on CUDA:
CUDA toolchain:
- runtime 13.0, artifact installation
- driver 580.95.5 for 13.0
- compiler 13.1
CUDA libraries:
- CUBLAS: 13.1.0
- CURAND: 10.4.0
- CUFFT: 12.0.0
- CUSOLVER: 12.0.4
- CUSPARSE: 12.6.3
- CUPTI: 2025.3.1 (API 13.0.1)
- NVML: 13.0.0+580.95.5
Julia packages:
- CUDA: 5.9.6
- GPUArrays: 11.3.2
- GPUCompiler: 1.7.5
- KernelAbstractions: 0.9.39
- CUDA_Driver_jll: 13.1.0+0
- CUDA_Compiler_jll: 0.4.0+0
- CUDA_Runtime_jll: 0.19.2+0
Toolchain:
- Julia: 1.12.3
- LLVM: 18.1.7
1 device:
0: NVIDIA GeForce RTX 4090 (sm_89, 20.744 GiB / 23.988 GiB available)