[Beowulf] Not OT, but a quick link to an article on InsideHPC

C Bergström cbergstrom at pathscale.com
Fri Mar 24 08:56:33 PDT 2017


Total bullshit - The website has been down a few times, but they were
things like a server restart mishape and people errors. I'm letting
the site go down now so people realize I'm serious about this.

I generally have very little or nothing to do with Open64. That is/was
a semi-hostile group of people whom I didn't want much interaction
with at all. Open64 is/was dead and maybe you mistook my words as
PathScale when what I meant was that Open64 is dead to me and I wanted
nothing to do with them.
------------
Clarification on open source - We made EKOPath (Not ENZO) fully open
source from 2009-2012. It was done under the path64 naming to avoid
confusion with our EKOPath brand. The code was (I've pulled it)
available on github and anyone could download and build it. The amount
of people who cared seemed to be so small and the confusion in
answering questions just wasn't worth it. In 2012 we decided just to
switch back to a closed model for both EKOPath and ENZO.

I realize we may be quiet and not dancing around posting bs marketing,
but I can assure you me and the small army of engineers working on the
compiler are quite alive. At peak we had 25 people working on the
compiler and other connected things. (This is btw more than "old"
PathScale had in terms of compiler people)

Sorry if my tone is wrong, but I do appreciate the feedback.

Some things we've done
* OpenACC across CPU/GPU (ARMv8, Power8, AMD dGPU, AMD APU and NVIDIA)
* OpenMP4 ...

Tons of work on top of clang/llvm infrastructure to fix it's complete
deficiency in the "high" level optimization catagories. All the
multicore/GPU optimizations which are needed for autopar, autogpu and
OpenACC (which puts a lot of pressure on the compiler)

KNC hardware wasn't all that great, but we wrote a micro OS/firmware
that replaced the bloated thing from Intel and exposed an interface
similar to a GPU (open source)

We've got a semi-experimental IDE which tightly couples clang+compiler
backend for showing users what's being optimized and what's not..

In recent years we've done work for several national labs, AMD, Cavium
and others... (Not to mention other products we've done which are very
specific to some aerospace customers)...

In honesty - traditional EKOPath customers may be disappointed because
all this work focuses on what I think is value for the future and not
immediate value for Intel centric world. Actually in hindsight, I
should have just had our core business go battle against Intel on
common benchmarks. If we're a "SPEC" compiler then people would
actually care..


On Fri, Mar 24, 2017 at 11:40 PM, Prentice Bisbal <pbisbal at pppl.gov> wrote:
> Chris
>
> Don't take this the wrong way, but I thought Pathscale closed shop a long
> time ago. We license Pathscale here, and I haven't been able to access the
> pathscale.com website for months, so I assumed you closed shop a long time
> ago. Googling for information on the current state of Pathscale, I came
> across a post on a Open64 mailing list/forum that made it sound like
> Pathscale had tried to open source the compiler, failed, and went out of
> business. That post was from someone withing the company (you?). I can try
> to find the post in question if interested, but it would probably take some
> time.
>
> Prentice
>
>
> On 03/23/2017 05:27 PM, C Bergström wrote:
>>
>> Tiz the season for HPC software to die?
>>
>> https://www.hpcwire.com/2017/03/23/hpc-compiler-company-pathscale-seeks-life-raft/
>>
>> (sorry I don't mean to hijack your thread, but timing of both
>> announcements is quite overlapping)
>>
>> On Fri, Mar 24, 2017 at 5:16 AM, Joe Landman <joe.landman at gmail.com>
>> wrote:
>>>
>>> For those who I've not talked with yet ...
>>>
>>> http://insidehpc.com/2017/03/scalable-informatics-closes-shop/
>>>
>>>
>>> --
>>> Joe Landman
>>> e: joe.landman at gmail.com
>>> t: @hpcjoe
>>> c: +1 734 612 4615
>>> w: https://scalability.org
>>> _______________________________________________
>>> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin Computing
>>> To change your subscription (digest mode or unsubscribe) visit
>>> http://www.beowulf.org/mailman/listinfo/beowulf
>>
>> _______________________________________________
>> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin Computing
>> To change your subscription (digest mode or unsubscribe) visit
>> http://www.beowulf.org/mailman/listinfo/beowulf
>
>
> _______________________________________________
> Beowulf mailing list, Beowulf at beowulf.org sponsored by Penguin Computing
> To change your subscription (digest mode or unsubscribe) visit
> http://www.beowulf.org/mailman/listinfo/beowulf


More information about the Beowulf mailing list