Skip to main content
33 events
when toggle format what by license comment
Nov 7, 2023 at 19:42 history edited Henners2002 CC BY-SA 4.0
edited title
Nov 7, 2023 at 19:36 history edited Henners2002 CC BY-SA 4.0
added 54 characters in body
Nov 7, 2023 at 19:30 answer added Henners2002 timeline score: 1
Nov 7, 2023 at 19:22 history edited Henners2002 CC BY-SA 4.0
added 1125 characters in body
Nov 7, 2023 at 1:45 answer added user555045 timeline score: 1
Nov 6, 2023 at 22:18 history edited Sᴀᴍ Onᴇᴌᴀ CC BY-SA 4.0
update image description
Nov 6, 2023 at 21:54 history edited greybeard CC BY-SA 4.0
replace rendered markup with the real thing - as a preformatted block. Interpreted, it looks - spaceous
Nov 6, 2023 at 20:11 comment added iSR5 @Henners2002 I just saw the results, I'm glad that helped through your journey.
Nov 6, 2023 at 20:03 history edited Henners2002
edited tags
Nov 6, 2023 at 19:53 comment added Henners2002 @iSR5 I have tested it it was 100% worth it you can check the thread for the results.
Nov 6, 2023 at 19:51 comment added iSR5 @Henners2002 honestly, I haven't used it with vectorized calculations, but you can test it out, and see if it's going to be worth it.
Nov 6, 2023 at 17:39 history edited Henners2002 CC BY-SA 4.0
edited title
Nov 6, 2023 at 14:58 history edited Henners2002 CC BY-SA 4.0
added 127 characters in body
Nov 6, 2023 at 12:58 answer added Henners2002 timeline score: 1
Nov 6, 2023 at 12:14 answer added Henners2002 timeline score: 1
Nov 6, 2023 at 10:58 comment added Henners2002 @iSR5 Is it possible to use the MemoryMarshall approach on vectorised calculations?
Nov 6, 2023 at 10:54 answer added Henners2002 timeline score: 3
Nov 4, 2023 at 14:12 comment added iSR5 using Span<T> would be a better approach, but you need to keep thread-safety in mind. There is also MemoryMarshal approach, but I don't think it would be worth as you might gain a very tiny speed over Span, unless you have a huge arrays, then it might worth a tray.
Nov 3, 2023 at 13:52 comment added Henners2002 @AlexanderPetrov Code has been corrected and Benchmarks updated. Do you have any performance enhancements I could try?
Nov 3, 2023 at 13:51 history edited Henners2002 CC BY-SA 4.0
added 380 characters in body
Nov 3, 2023 at 12:44 comment added Alexander Petrov Show a fully working example. Now your code does not copy values from testSpace to result at all, multiplying zeros in this way.
Nov 3, 2023 at 12:41 comment added Alexander Petrov Do size 999 or 1001 and see yourself.
Nov 3, 2023 at 12:38 comment added Henners2002 @AlexanderPetrov Yes it does. In the loop condition there is testSpace.Length - testSpace.Length % batchSize This ensures the the upper limit of the loop is always a multiple of batchSize
Nov 3, 2023 at 12:36 comment added Alexander Petrov Your code has a bug. It does not handle the tail remaining in the case when the array size is not a multiple of batchSize.
Nov 3, 2023 at 12:35 comment added Henners2002 @OlivierJacot-Descombes Yeah I think the TPL would be a good option of the test space was larger. Maybe If I played around with this and the degree of parallelism I could get better results.
Nov 3, 2023 at 12:31 comment added Henners2002 @PeterCsala I have included Benchmarks of five relevant methods
Nov 3, 2023 at 12:31 comment added Olivier Jacot-Descombes Note that tasks must have chunks to process that are big enough in order to compensate for the overhead involved in task management. Did you have a loop into Task Parallel Library (TPL)?
Nov 3, 2023 at 12:30 history edited Henners2002 CC BY-SA 4.0
added 171 characters in body
Nov 3, 2023 at 12:24 comment added Peter Csala Anything that extends the review context is more than welcome.
Nov 3, 2023 at 12:23 comment added Henners2002 @PeterCsala I ran a benchmark. I can post the results of that if you would like. I've compared it against LINQ, a regular for loop and a SIMD instruction with heap allocation
Nov 3, 2023 at 12:21 comment added Peter Csala How did you measure the performance of your code?
Nov 3, 2023 at 12:20 history edited Peter Csala CC BY-SA 4.0
deleted 37 characters in body
Nov 3, 2023 at 12:00 history asked Henners2002 CC BY-SA 4.0