Hi there, I hope this is not a duplicate. If so, please fill free to shit this post. I’m relatively new to Julia. I currently working on a numerical project which I want to implement on the GPU. Therefore I’m using the Julia CUDA libraries. I want to create some initial date using my Nvidia-GPU ,…

for completeness, I’ll include a snippet which doesn’t crash :slight_smile: @inline function CUDAnative.exp(x::ComplexF64) scale = CUDAnative.exp( x.re ) return ComplexF64( scale * CUDAnative.cos(x.im), scale * CUDAnative.sin(x.im) ) end function kernel_initial_psi(a,N,k_in) j = (block…

Please quote your code, this will help us help you! [image] PSA: how to quote code with backticks Meta Discussion This is a short post on how to use backticks (` and ```) to quote code, so it is easy to read. This post can be linked to new users who are confus…

Thanks, this is what I was looking for. But the main problem still exists :frowning:

hiya, CUDAnative.exp just wraps the nvidia libdevice functions, and there is no complex version of this function. But you can define it! define @inline function CUDAnative.exp(x::ComplexF64) scale = CUDAnative.exp( x.re ) return ComplexF64( scale * CUDAnative.cos(x.im), scale * CUDAnativ…

Hi, Your answer makes some things clear now, first thanks for that. But when I define the new function CUDAnative.exp for the ComplexF64 data type I get a new error message that’s driving me a little bit nuts. I mean that is of course caused by my lack of experience with Julia, which is actually …

ah, ok you have more problems with your code you use k for both initial wave direction, and the index (blockIdx().y - 1) * blockDim().y + threadIdx().y I think this is the main problem you have a line blocks = ceil.(Int, (size(psi_int,3), size(psi_int,4)) ./ threads) but threads is not defi…

Maybe one last question concerning this problem, I tried your code and for N>10 (of course I edited threads) it works anymore and gives me the error message ERROR: LoadError: CUDA error: invalid argument (code #1, ERROR_INVALID_VALUE) Stacktrace: [1] #_cudacall#24(::Tuple{Int64,Int64,Int64}, ::Tu…

this issue isn’t specific to Julia: you’re perhaps hitting a limit on threads per block, which is 1024. 11^3 > 1024, so that is not allowed if you only have a single block. Personally, I like to just use one-dimensional block and thread groups, and have a little logic to partition what each block …

Base function in Cuda kernels

General Usage

ffevotte March 14, 2019, 6:09pm 2

Please quote your code, this will help us help you!

Topic		Replies	Views
Cuda kernel error New to Julia gpu	9	3084	December 9, 2019
Question about CUDA kernels GPU question	4	697	February 10, 2023
CUDA with IJulia results in unexpected errors GPU	7	783	June 18, 2021
Error in Cuda function : ERROR: LoadError: CuError(1, nothing) General Usage question , cudanative , cuda	8	1180	July 26, 2019
QuadGK don't work in CUDA kernel General Usage question , package , cudanative , cuda	6	1041	July 30, 2019

Base function in Cuda kernels

Related topics