Mir updates

Mar 30, 2020

9il

Mar 30, 2020

Yazan D

Mar 30, 2020

Mar 30, 2020

Mar 30, 2020

Apr 02, 2020

Apr 02, 2020

Apr 02, 2020

Apr 02, 2020

Apr 04, 2020

Docs and links have been updated for http://mir-algorithm.libmir.org, Mir Algorithm http://mir-core.libmir.org, Mir Core http://mir-random.libmir.org, Mir Random http://mir-optim.libmir.org, Mir Optim http://mir-cpuid.libmir.org, Mir CPUID http://mir-blas.libmir.org, Mir BLAS http://mir-lapack.libmir.org, Mir LAPACK http://mir.libmir.org, Mir Backport http://asdf.libmir.org, ASDF - JSON library I have started this a year ago and recently I got time to finish this work. All of the libraries above have: 1. Meson and dub support 2. Imports validation for documented examples 3. AWS S3 + Circle CI automated documentation updates Mir Optim was completely reworked and got Box Constrained QP Solver ASDF, mir-algorithm, mir-lapack, mir-random, mir-core have been extended --- This work has been sponsored by Symmetry Investments and Kaleidic Associates. http://symmetryinvestments.com/ https://github.com/kaleidicassociates

On Monday, 30 March 2020 at 06:33:13 UTC, 9il wrote: > Docs and links have been updated for > > http://mir-algorithm.libmir.org, Mir Algorithm > http://mir-core.libmir.org, Mir Core > http://mir-random.libmir.org, Mir Random > http://mir-optim.libmir.org, Mir Optim > http://mir-cpuid.libmir.org, Mir CPUID > http://mir-blas.libmir.org, Mir BLAS > http://mir-lapack.libmir.org, Mir LAPACK > http://mir.libmir.org, Mir Backport > http://asdf.libmir.org, ASDF - JSON library > > [...] Thanks for your work!

On Monday, 30 March 2020 at 06:33:13 UTC, 9il wrote: > [snip] Thanks, I like 'em. I noticed that the little icon in the tabs has changed from most of them. However, the mir random is unchanged from before. Also, on the mir.glas page, one of the lines says "matrix-vector operations %3 done, partially optimized for now". Another line says "l3 was moved to mir-glas", which is confusing because it should be at the mir-glas documentation page anyway.

On Monday, 30 March 2020 at 12:23:03 UTC, jmh530 wrote: > On Monday, 30 March 2020 at 06:33:13 UTC, 9il wrote: >> [snip] > > Thanks, I like 'em. > > I noticed that the little icon in the tabs has changed from most of them. However, the mir random is unchanged from before. Probably it is because of your browser cache, likely will be updated after a while. > Also, on the mir.glas page, one of the lines says "matrix-vector operations %3 done, partially optimized for now". Another line says "l3 was moved to mir-glas", which is confusing because it should be at the mir-glas documentation page anyway. We don't have documentation for mir-glas library, only for mir (backports) package, which has mir.glas package. I don't know what to do with mir-glas, it is too good to be forgotten, but I don't see a commercial perspective in it.

On Tue, Mar 31, 2020 at 12:15 AM 9il via Digitalmars-d-announce < digitalmars-d-announce@puremagic.com> wrote: > On Monday, 30 March 2020 at 12:23:03 UTC, jmh530 wrote: > > On Monday, 30 March 2020 at 06:33:13 UTC, 9il wrote: > >> [snip] > > > > Thanks, I like 'em. > > > > I noticed that the little icon in the tabs has changed from most of them. However, the mir random is unchanged from before. > > Probably it is because of your browser cache, likely will be updated after a while. > > > Also, on the mir.glas page, one of the lines says "matrix-vector operations %3 done, partially optimized for now". Another line says "l3 was moved to mir-glas", which is confusing because it should be at the mir-glas documentation page anyway. > > We don't have documentation for mir-glas library, only for mir (backports) package, which has mir.glas package. > > I don't know what to do with mir-glas, it is too good to be forgotten, but I don't see a commercial perspective in it. > Why not? Where does it fall short of being useful?

April 02, 2020

Re: Mir updates

Posted by 9il
in reply to Manu

Permalink

9il

Posted in reply to Manu

Permalink

On Thursday, 2 April 2020 at 06:20:22 UTC, Manu wrote:
> On Tue, Mar 31, 2020 at 12:15 AM 9il via Digitalmars-d-announce < digitalmars-d-announce@puremagic.com> wrote:
>
>> On Monday, 30 March 2020 at 12:23:03 UTC, jmh530 wrote:
>> > On Monday, 30 March 2020 at 06:33:13 UTC, 9il wrote:
>> >> [snip]
>> >
>> > Thanks, I like 'em.
>> >
>> > I noticed that the little icon in the tabs has changed from most of them. However, the mir random is unchanged from before.
>>
>> Probably it is because of your browser cache, likely will be updated after a while.
>>
>> > Also, on the mir.glas page, one of the lines says "matrix-vector operations %3 done, partially optimized for now". Another line says "l3 was moved to mir-glas", which is confusing because it should be at the mir-glas documentation page anyway.
>>
>> We don't have documentation for mir-glas library, only for mir (backports) package, which has mir.glas package.
>>
>> I don't know what to do with mir-glas, it is too good to be forgotten, but I don't see a commercial perspective in it.
>>
>
> Why not? Where does it fall short of being useful?

For my work, I use `mir-blas` - ndslice bindings to CBLAS API. It is faster to write binding rather than finish BLAS implementation.

It is useful. And I think it is very promising for Dlang promotion and can really involve new companies. I will be happy to finish it if a company sponsors the work. Companies like CPU/GPU vendors or companies that require slim and fast BLAS implementation will win from this work. However, I have no idea how to sell mir-glas, it is more complex to find a client rather than finish it.

On Thursday, 2 April 2020 at 13:16:46 UTC, 9il wrote: > [snip] > > For my work, I use `mir-blas` - ndslice bindings to CBLAS API. It is faster to write binding rather than finish BLAS implementation. > > It is useful. And I think it is very promising for Dlang promotion and can really involve new companies. I will be happy to finish it if a company sponsors the work. Companies like CPU/GPU vendors or companies that require slim and fast BLAS implementation will win from this work. However, I have no idea how to sell mir-glas, it is more complex to find a client rather than finish it. Can mir-glas be used as a replacement blas implementation for mir-lapack?

On Thursday, 2 April 2020 at 13:39:41 UTC, jmh530 wrote: > On Thursday, 2 April 2020 at 13:16:46 UTC, 9il wrote: >> [snip] >> >> For my work, I use `mir-blas` - ndslice bindings to CBLAS API. It is faster to write binding rather than finish BLAS implementation. >> >> It is useful. And I think it is very promising for Dlang promotion and can really involve new companies. I will be happy to finish it if a company sponsors the work. Companies like CPU/GPU vendors or companies that require slim and fast BLAS implementation will win from this work. However, I have no idea how to sell mir-glas, it is more complex to find a client rather than finish it. > > Can mir-glas be used as a replacement blas implementation for mir-lapack? Yes (when finished). However, mir-lapack would need to link with a LAPACK library. For example, OpenBLAS can be compiled with LAPACK as a solid library. We can do the same. Actually, when the library is finished it can be available in the dub registry as well as it should be packed for Linux and windows as a standalone BLAS/LAPACK library with common API.

Forums