Closed
Description
Describe the bug
std
function from Statistics is type unstable when used on CuArray with the dims argument
To reproduce
The Minimal Working Example (MWE) for this bug:
using CUDA
a = rand(3, 10) |> cu
@code_warntype std(a) # Type Inference works
@code_warntype std(a, dims = 2) # Type Inference fails
Dependencies
Status `~/research/CUDATests/Project.toml`
[6e4b80f9] BenchmarkTools v0.5.0
[052768ef] CUDA v1.2.0 #master (https://github.com/JuliaGPU/CUDA.jl.git)
Verified on both the latest release and master
Version info
Details on Julia:
Julia Version 1.4.2
Commit 44fa15b150* (2020-05-23 18:35 UTC)
Platform Info:
OS: Linux (x86_64-pc-linux-gnu)
CPU: Intel(R) Xeon(R) CPU E5-2603 v4 @ 1.70GHz
WORD_SIZE: 64
LIBM: libopenlibm
LLVM: libLLVM-8.0.1 (ORCJIT, broadwell)
Environment:
JULIA_PKG_DEVDIR = /home/avik-pal/research
Details on CUDA:
CUDA toolkit 10.1.243, artifact installation
CUDA driver 10.1.0
NVIDIA driver 418.67.0
Libraries:
- CUBLAS: 10.2.1
- CURAND: 10.1.1
- CUFFT: 10.1.1
- CUSOLVER: 10.2.0
- CUSPARSE: 10.3.0
- CUPTI: 12.0.0
- NVML: 10.0.0+418.67
- CUDNN: 7.6.5 (for CUDA 10.1.0)
- CUTENSOR: 1.2.0 (for CUDA 10.1.0)
Toolchain:
- Julia: 1.4.2
- LLVM: 8.0.1
- PTX ISA support: 3.2, 4.0, 4.1, 4.2, 4.3, 5.0, 6.0, 6.1, 6.3
- Device support: sm_30, sm_32, sm_35, sm_37, sm_50, sm_52, sm_53, sm_60, sm_61, sm_62, sm_70, sm_72, sm_75
9 device(s):
- GeForce GTX 1080 Ti (sm_61, 10.905 GiB / 10.917 GiB available)
- Tesla P100-PCIE-16GB (sm_60, 15.890 GiB / 15.899 GiB available)
- GeForce GTX 1080 Ti (sm_61, 10.232 GiB / 10.917 GiB available)
- Tesla P100-PCIE-16GB (sm_60, 15.890 GiB / 15.899 GiB available)
- Tesla V100-PCIE-32GB (sm_70, 30.338 GiB / 31.719 GiB available)
- Tesla V100-PCIE-32GB (sm_70, 31.586 GiB / 31.719 GiB available)
- Tesla V100-PCIE-32GB (sm_70, 28.039 GiB / 31.719 GiB available)
- Tesla V100-PCIE-32GB (sm_70, 31.594 GiB / 31.719 GiB available)
- Tesla V100-PCIE-16GB (sm_70, 15.741 GiB / 15.752 GiB available)
Additional context
Add any other context about the problem here.