gpu.js
gpu.js copied to clipboard
gpujs
Chain combineKernels?
When I try to pass a combinedKernel into combineKernels() to chain them, I got this:
Uncaught TypeError: arguments[i].setCanvas is not a function at GPU.combineKernels
I'm using GPU.js to run bellman Ford algorithm, code here: https://github.com/pan-long/SSSP-on-GPU/blob/master/bellman_ford_gpu.js
I have to have a loop to keep running bellmanFordKernel, it's now slower than Dijkstra running on GPU. I'm wondering is it because of performance penalty brought up by keep transferring data between CPU and GPU. Then I found out combineKernels, but it doesnt support chaining.
On Wed, Nov 15, 2017, 10:26 PM Robert Plummer [email protected] wrote:
Can we getore context, like example code?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/gpujs/gpu.js/issues/224#issuecomment-344608080, or mute the thread https://github.com/notifications/unsubscribe-auth/AD3olvFKMYsaPtJ-dLW009Vo-B6KJ3BNks5s2vSsgaJpZM4Qe9Ds .
I think you are dead right. We need to add this as a feature, so we can chain values through. Give me some time this week and I'll add support. We'll get this thing faster!
Can we have access to that code again? Or something simple that outlines the chaining being done? Clarity helps us help you! 😉
Hi, I have made it public. Thanks for your help!
On Sat, Nov 18, 2017, 10:05 AM Robert Plummer [email protected] wrote:
Can we have access to that code again? Or something simple that outlines the chaining being done? Clarity helps us help you! 😉
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/gpujs/gpu.js/issues/224#issuecomment-345410426, or mute the thread https://github.com/notifications/unsubscribe-auth/AD3oltX5oiYOhVceIskfoVqVrrOzenllks5s3jtlgaJpZM4Qe9Ds .
Been thinking about this over the weekend, I will have a solution soon, fyi.
This needs clearer direction. Can you give a non-working jsfiddle?
