07:00 <dkaratza[m]> ms: good morning, can you send me the link?

07:01 <ms[m]> dkaratza: yep, just a sec

07:01 <ms[m]> dkaratza: https://ethz.zoom.us/j/8965226326

07:42 <dkaratza[m]> ms: so, the thing is that i can remove the pages by just removing `include`, but I would like to prevent the files from being generated as well. For example, the files `cmake_toolchains.rst`, `cmake_variables.rst`

07:44 <dkaratza[m]> sorry, only `cmake_toolchains.rst`, not the variables

07:48 <ms[m]> dkaratza: gotcha, you can remove it here: https://github.com/STEllAR-GROUP/hpx/blob/b184a945572ce8b680ffb1623e29424096d0f4de/docs/CMakeLists.txt#L125-L157

07:48 <ms[m]> and this file can also go in that case: https://github.com/STEllAR-GROUP/hpx/blob/master/cmake/templates/cmake_toolchains.rst.in

07:56 <dkaratza[m]> ms: great, thank you

07:56 <dkaratza[m]> and `make -jN install` now becomes just `cmake --build`?

07:57 <ms[m]> it becomes `cmake --build <path to build dir> --target install`

07:57 <ms[m]> where `<path to build dir>` can probably be `.` if we assume that users are running the commands in the build directory

07:58 <dkaratza[m]> and for the tests it's `cmake --build tests`?

08:02 <ms[m]> for the tests just `ctest`

08:02 <ms[m]> to build them `cmake --build . --target tests`

08:02 <ms[m]> sorry btw, I didn't realize you did push your changes to the pr yesterday

08:02 <ms[m]> I'll have a closer look before the meeting

08:05 <dkaratza[m]> no problem

15:05 <gonidelis[m]> hkaiser: r u joining?

16:36 <jedi18[m]> hkaiser: Here are the graphs for the speedup relative to the seq execution https://github.com/Jedi18/scan_benchmarks

16:36 <jedi18[m]> These are the default chunk size measurements, I'll try varying the chunk size next

16:36 <hkaiser> nice!

16:37 <jedi18[m]> Ran these on a single node of buran btw

16:37 <hkaiser> nod, good work!

16:38 <hkaiser> how many cores did that use?

16:40 <jedi18[m]> hkaiser: I dind't specify the number of cores (just ran srun bin/...)

16:40 <jedi18[m]> Is there a way to check the number of cores?

16:40 <hkaiser> ok, so all of the cores were used

16:41 <hkaiser> you can do --hpx:print-bind to see how threads are being bound

16:41 <srinivasyadav227> jedi18[m]: it uses 48 cores default on buran

16:44 <jedi18[m]> Oh ok, thanks

16:47 <hkaiser> jedi18[m]: what type of elements did you use ? int? double?

16:52 <jedi18[m]> Int, and the it's the average of 50 iterations for each input size

17:14 <pedro_barbosa[m]> Hey guys, I was trying to adapt correct a problem I had in an example I have in HPXCL, I have a program that runs on a GPU and every iteration of the kernel I want to swap the input with the output, so I copy the output to the host, swap the values and copy it back to the GPU, however this has poor performance cause I have to copy the arrays every iteration, so my idea was to swap the two buffers on the argument array that I pass to

17:14 <pedro_barbosa[m]> the function. expecting the kernel to then use the output array as input and vice versa without having to copy everything back to the GPU. Can someone confirm if it is possible to do it like this?

17:15 <pedro_barbosa[m]> s/adapt//

17:15 <pedro_barbosa[m]> s/adapt//, s/had/have/

17:15 <pedro_barbosa[m]> s/adapt//, s/had/have/, s/I have//

17:16 <pedro_barbosa[m]> s/Hey guys, I was trying to adapt correct a problem I had in an example I have in HPXCL, I have a program that runs on a GPU and every iteration of the kernel I want to swap the input with the output, so I copy the output to the host, swap the values and copy it back to the GPU, however this has poor performance cause I have to copy the arrays every iteration, so my idea was to swap the two buffers on the argument array that I pass to

17:16 <pedro_barbosa[m]> copy it back to the GPU, however this has poor performance cause I have to copy the arrays every iteration, so my idea was to swap the two buffers on the argument array that I pass to the function, expecting the kernel to then use the output array as input and vice versa without having to copy everything back to the GPU. Can someone confirm if it is possible to do it like this?/

17:16 <pedro_barbosa[m]> the function. expecting the kernel to then use the output array as input and vice versa without having to copy everything back to the GPU. Can someone confirm if it is possible to do it like this?/Hey guys, I was trying to correct a problem I have in an example in HPXCL, I have a program that runs on a GPU and every iteration of the kernel I want to swap the input with the output, so I copy the output to the host, swap the values and