rcurtin_irc changed the topic of #mlpack to: mlpack: a scalable machine learning library (https://www.mlpack.org/) -- channel logs: https://libera.irclog.whitequark.org/mlpack -- NOTE: messages sent here might not be seen by bridged users on matrix, gitter, or slack
<EshaanAgarwal[m]> * rcurtin zoq: I have sent link to my proposal. Actually it's still incomplete ( small parts are still left here and there) but I would love to hear out your suggestions and reviews on the rest of it.
<EshaanAgarwal[m]> Link to my proposal is following -
<EshaanAgarwal[m]> Can you please review it whenever you have free time ?
<EshaanAgarwal[m]> * rcurtin zoq: I have sent link to my proposal. Actually it's still incomplete ( small parts are still left here and there) but I would love to hear out your suggestions and reviews on the rest of it.
<EshaanAgarwal[m]> Can you please review it whenever you have free time ?
<zoq[m]> <GopiMTatiraju[m]> "Oh yupp, that makes sense, I was..." <- What I did for the `shuffle` implementation is to use `curandGenerateUniform` and implement a rowwise and colwise kernel. Haven't timed it yet, but I think it's okay.
<GopiMTatiraju[m]> Yea, thrust and cuRAND, these libraries can be very helpful for initial implementation...
<GopiMTatiraju[m]> We can time it and see what happens then...
<GopiMTatiraju[m]> If it works then good otherwise we might wanna try something else...
<zoq[m]> Yeah, I could even see for some sizes it doesn't make much sense to do it on the GPU.
<GopiMTatiraju[m]> How would you time each individual kernel?
<GopiMTatiraju[m]> Just compare CPU and GPU implementation's time?
<zoq[m]> Yeah, and also just manually swap out different kernel implementations.
<GopiMTatiraju[m]> For `solve()` also we can take help of `cuSOLVER`
<GopiMTatiraju[m]> * For `solve()` also we can take help of `cuSOLVER`
AshvithShetty[m] has quit [Quit: You have been kicked for being idle]
<GopiMTatiraju[m]> rcurtin zoq [here](https://gitlab.com/conradsnicta/bandicoot-code/-/blob/unstable/include/bandicoot_bits/cuda/kernels/twoway/equ_array_log_post.cu) in the implementation of `log()` we are calling another log(), this logs comes from where?
<rcurtin[m]> that should be the standard builtin implementation of `log()`; on the CPU this would come from `<cmath>` I think, but on the GPU it's provided by CUDA
<GopiMTatiraju[m]> Yea, there are many functions like `tan` `cos` which we can implement in the same way then, right?
<rcurtin[m]> should be, yes
<GopiMTatiraju[m]> okayyy...
<GopiMTatiraju[m]> Open Source work with a job becomes very hectic, its great that you guys are still putting in soo much efforts and time into this.
<EshaanAgarwal[m]> Just curious! but is there any upper bound on the length of proposal for GSoC ?
<EshaanAgarwal[m]> Hi !
<GopiMTatiraju[m]> I don't think so, but considering that mentors have soo much to do, I would say a to the point proposal(not very lengthy but covers all the crux points) would be better...
078AAHW1E has quit [Quit: You have been kicked for being idle]