info
Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
I'll be presenting at this year's LLVM US Developers' Meeting
559 views
Skip to first unread message
John Reagan
Sep 5, 2017, 4:42:32 PM9/5/17
to
As an FYI, I submitted an abstract on how we are using LLVM to port OpenVMS to x86 along with plans for leveraging clang as our C++ offering.
I submitted a "lightning talk" (approx 5 minutes each, no questions, no discussion, essentially an "elevator talk"). Lightning talks are to show creative uses of LLVM.
It received all positive remarks from the program committee. The names of the reviewers are hidden from me to prevent bias, but one of the reviewers knows about GEM and mentioned it by name in the comment (I did not mention GEM by name in the abstract).
This will give us additional exposure to the overall LLVM community.
The conference is at the San Jose Convention Center, Oct 18/19 2017.
Of course I'll be speaking for longer than 5 minutes at the OpenVMS Technical Bootcamp. I'll be giving an update on the compilers and some design highlights for dealing with MACRO-32 register linkages and BLISS LINKAGEs. For those of you that cannot attend, I'll write up something for the newsgroup as well so you won't feel left out. Our linker engineer is also presenting on the calling standard so there will be some overlap with him.
I submitted a "lightning talk" (approx 5 minutes each, no questions, no discussion, essentially an "elevator talk"). Lightning talks are to show creative uses of LLVM.
It received all positive remarks from the program committee. The names of the reviewers are hidden from me to prevent bias, but one of the reviewers knows about GEM and mentioned it by name in the comment (I did not mention GEM by name in the abstract).
This will give us additional exposure to the overall LLVM community.
The conference is at the San Jose Convention Center, Oct 18/19 2017.
Of course I'll be speaking for longer than 5 minutes at the OpenVMS Technical Bootcamp. I'll be giving an update on the compilers and some design highlights for dealing with MACRO-32 register linkages and BLISS LINKAGEs. For those of you that cannot attend, I'll write up something for the newsgroup as well so you won't feel left out. Our linker engineer is also presenting on the calling standard so there will be some overlap with him.
VAXman-
Sep 6, 2017, 8:51:15 AM9/6/17
to
In article <e5bbac8b-f842-4095...@googlegroups.com>, John Reagan <xyzz...@gmail.com> writes:
>{...snip...}
>Of course I'll be speaking for longer than 5 minutes at the OpenVMS Technic=
>al Bootcamp. I'll be giving an update on the compilers and some design hig=
>hlights for dealing with MACRO-32 register linkages and BLISS LINKAGEs. Fo=
>r those of you that cannot attend, I'll write up something for the newsgrou=
>p as well so you won't feel left out. Our linker engineer is also presenti=
--
VAXman- A Bored Certified VMS Kernel Mode Hacker VAXman(at)TMESIS(dot)ORG
I speak to machines with the voice of humanity.
>{...snip...}
>Of course I'll be speaking for longer than 5 minutes at the OpenVMS Technic=
>al Bootcamp. I'll be giving an update on the compilers and some design hig=
>hlights for dealing with MACRO-32 register linkages and BLISS LINKAGEs. Fo=
>r those of you that cannot attend, I'll write up something for the newsgrou=
>p as well so you won't feel left out. Our linker engineer is also presenti=
>ng on the calling standard so there will be some overlap with him.
:)
--
VAXman- A Bored Certified VMS Kernel Mode Hacker VAXman(at)TMESIS(dot)ORG
I speak to machines with the voice of humanity.
IanD
Sep 14, 2017, 1:37:16 AM9/14/17
to
On Wednesday, September 6, 2017 at 6:42:32 AM UTC+10, John Reagan wrote:
> As an FYI, I submitted an abstract on how we are using LLVM to port OpenVMS to x86 along with plans for leveraging clang as our C++ offering.
>
> I submitted a "lightning talk" (approx 5 minutes each, no questions, no discussion, essentially an "elevator talk"). Lightning talks are to show creative uses of LLVM.
>
> It received all positive remarks from the program committee. The names of the reviewers are hidden from me to prevent bias, but one of the reviewers knows about GEM and mentioned it by name in the comment (I did not mention GEM by name in the abstract).
>
> This will give us additional exposure to the overall LLVM community.
>
> The conference is at the San Jose Convention Center, Oct 18/19 2017.
>
In Aussie feedback style, Bloody Awesome!!!
> As an FYI, I submitted an abstract on how we are using LLVM to port OpenVMS to x86 along with plans for leveraging clang as our C++ offering.
>
> I submitted a "lightning talk" (approx 5 minutes each, no questions, no discussion, essentially an "elevator talk"). Lightning talks are to show creative uses of LLVM.
>
> It received all positive remarks from the program committee. The names of the reviewers are hidden from me to prevent bias, but one of the reviewers knows about GEM and mentioned it by name in the comment (I did not mention GEM by name in the abstract).
>
> This will give us additional exposure to the overall LLVM community.
>
> The conference is at the San Jose Convention Center, Oct 18/19 2017.
>
> Of course I'll be speaking for longer than 5 minutes at the OpenVMS Technical Bootcamp. I'll be giving an update on the compilers and some design highlights for dealing with MACRO-32 register linkages and BLISS LINKAGEs.
>>> For those of you that cannot attend, I'll write up something for the newsgroup as well so you won't feel left out. Our linker engineer is also presenting on the calling standard so there will be some overlap with him.
In a strange way, it makes one feel part of the overall process, even if it's only by indirect association
Thank you!
Neil Rieck
Sep 15, 2017, 7:36:54 AM9/15/17
to
Neil Rieck
Waterloo, Ontario, Canada.
http://neilrieck.net/OpenVMS-Programmers-Corner.html
John Reagan
Sep 15, 2017, 9:15:57 AM9/15/17
to
IanD
Oct 14, 2017, 12:15:47 PM10/14/17
to
Were any of those young'ins interested in finding out more about OpenVMS?
John Reagan
Oct 14, 2017, 12:37:09 PM10/14/17
to
IanD
Oct 14, 2017, 12:46:11 PM10/14/17
to
>>> The conference is at the San Jose Convention Center, Oct 18/19 2017
Time for new glasses!!! 8-(
John Reagan
Oct 22, 2017, 4:22:39 PM10/22/17
to
Here's a recap.
It is really difficult to jam it all into 5 minutes. I wrote out what I thought was a tight presentation and about 10 slides. I practiced it 'out loud' to better pace my speaking. First experiment? 9 minutes! Ugh.
I cut more and squashed some slides together. 7 minutes.
I cut more and removed "Open" from "VMS" from the occurrences and finally got it down to 5 minutes and change.
I was first in the session (right after a break so there were a few stragglers opening/closing the doors.
I read from my script since I was tight on time. If I did it "off the cuff", I wasn't sure I'd finish on time. However, reading the script and advancing the slides was a little awkward. I'm not sure I'd do that again.
My presentation was a little different than the rest. The others talked about USING a particular feature of LLVM or ADDING a new feature to LLVM. I was talking about retargeting frontends and mapping GEM to LLVM. I did discuss things we had (or will) change in LLVM and clang, but there is really nothing that is rocket science. Others have hooked their own frontends to LLVM in the past. The "interesting" part is that we're doing 6 frontends, RTL enhancements, and adapting clang for OpenVMS all at the same time.
A friend in the audience said that people were listening and seemed impressed. I only had a few questions later on.
Given the amount of information I presented, I think it might have been as a poster paper. The afternoon break on day #2 has about a dozen of posters and authors spread around the lobby area. It would have let interested folks read and absorb at their rate. However, I wouldn't have had a room full of captives.
It is really difficult to jam it all into 5 minutes. I wrote out what I thought was a tight presentation and about 10 slides. I practiced it 'out loud' to better pace my speaking. First experiment? 9 minutes! Ugh.
I cut more and squashed some slides together. 7 minutes.
I cut more and removed "Open" from "VMS" from the occurrences and finally got it down to 5 minutes and change.
I was first in the session (right after a break so there were a few stragglers opening/closing the doors.
I read from my script since I was tight on time. If I did it "off the cuff", I wasn't sure I'd finish on time. However, reading the script and advancing the slides was a little awkward. I'm not sure I'd do that again.
My presentation was a little different than the rest. The others talked about USING a particular feature of LLVM or ADDING a new feature to LLVM. I was talking about retargeting frontends and mapping GEM to LLVM. I did discuss things we had (or will) change in LLVM and clang, but there is really nothing that is rocket science. Others have hooked their own frontends to LLVM in the past. The "interesting" part is that we're doing 6 frontends, RTL enhancements, and adapting clang for OpenVMS all at the same time.
A friend in the audience said that people were listening and seemed impressed. I only had a few questions later on.
Given the amount of information I presented, I think it might have been as a poster paper. The afternoon break on day #2 has about a dozen of posters and authors spread around the lobby area. It would have let interested folks read and absorb at their rate. However, I wouldn't have had a room full of captives.
Simon Clubley
Oct 22, 2017, 7:55:33 PM10/22/17
to
On 2017-10-22, John Reagan <xyzz...@gmail.com> wrote:
>
> Given the amount of information I presented, I think it might have
> been as a poster paper. The afternoon break on day #2 has about a
> dozen of posters and authors spread around the lobby area. It would
> have let interested folks read and absorb at their rate. However, I
> wouldn't have had a room full of captives.
>
No, but the lobby area sounds like a much more informal environment so
>
> Given the amount of information I presented, I think it might have
> been as a poster paper. The afternoon break on day #2 has about a
> dozen of posters and authors spread around the lobby area. It would
> have let interested folks read and absorb at their rate. However, I
> wouldn't have had a room full of captives.
>
you could have a real discussion with people interested enough to read
the poster.
Of course, there's no reason why you can't still put a poster online
and then hand out the URL to anyone interested or include the URL in
any other presentations.
Simon.
--
Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
Microsoft: Bringing you 1980s technology to a 21st century world
IanD
Oct 27, 2017, 11:14:48 AM10/27/17
to
'...seemed impressed'?
Come on!!! Pat yourself on the back a little more.
You adopt LLVM as a tool, a tool that is probably in flight on many ways (looking at the change logs for it, off and on. It's rather beyond me anyhow)) and is sill being developed.
By your own statement, you adapt it for 6 different interface tasks all at once and then to top it all off, you give a presentation to a group of people who were probably involved in LLVM's development!
The fact that a group at that level didn't walk out or play with their phones says a lot
While at their level they may not be falling off their chair in amazement, I happen to think it's pretty dam impressive, at least to someone at my low level
It's probably going to take time to get known in those circles anyhow.
Reading between the lines and not knowing you other than what you post here, sounds like your into slightly more informal presentations than just script reading, hopefully there will be other opportunities to present again and then you might have more time to inject experiences learned versus a straight out technical presentation. Then again, I could be just taking crap and misreading things altogether...
At least the word is getting out that VMS/VSI is now on-board and running in the LLVM space
John Reagan
Oct 28, 2017, 10:19:45 AM10/28/17
to
On Friday, October 27, 2017 at 11:14:48 AM UTC-4, IanD wrote:
> On Monday, October 23, 2017 at 7:22:39 AM UTC+11, John Reagan wrote:
> > Here's a recap.
> >
> > It is really difficult to jam it all into 5 minutes. I wrote out what I thought was a tight presentation and about 10 slides. I practiced it 'out loud' to better pace my speaking. First experiment? 9 minutes! Ugh.
> >
> > I cut more and squashed some slides together. 7 minutes.
> >
> > I cut more and removed "Open" from "VMS" from the occurrences and finally got it down to 5 minutes and change.
> >
> > I was first in the session (right after a break so there were a few stragglers opening/closing the doors.
> >
> > I read from my script since I was tight on time. If I did it "off the cuff", I wasn't sure I'd finish on time. However, reading the script and advancing the slides was a little awkward. I'm not sure I'd do that again.
> >
> > My presentation was a little different than the rest. The others talked about USING a particular feature of LLVM or ADDING a new feature to LLVM. I was talking about retargeting frontends and mapping GEM to LLVM. I did discuss things we had (or will) change in LLVM and clang, but there is really nothing that is rocket science. Others have hooked their own frontends to LLVM in the past. The "interesting" part is that we're doing 6 frontends, RTL enhancements, and adapting clang for OpenVMS all at the same time.
> >
> > A friend in the audience said that people were listening and seemed impressed. I only had a few questions later on.
> >
> > Given the amount of information I presented, I think it might have been as a poster paper. The afternoon break on day #2 has about a dozen of posters and authors spread around the lobby area. It would have let interested folks read and absorb at their rate. However, I wouldn't have had a room full of captives.
>
>
> Thank you
>
> '...seemed impressed'?
>
> Come on!!! Pat yourself on the back a little more.
Thanks. It isn't just me. There is a small, but experienced team as well.
> On Monday, October 23, 2017 at 7:22:39 AM UTC+11, John Reagan wrote:
> > Here's a recap.
> >
> > It is really difficult to jam it all into 5 minutes. I wrote out what I thought was a tight presentation and about 10 slides. I practiced it 'out loud' to better pace my speaking. First experiment? 9 minutes! Ugh.
> >
> > I cut more and squashed some slides together. 7 minutes.
> >
> > I cut more and removed "Open" from "VMS" from the occurrences and finally got it down to 5 minutes and change.
> >
> > I was first in the session (right after a break so there were a few stragglers opening/closing the doors.
> >
> > I read from my script since I was tight on time. If I did it "off the cuff", I wasn't sure I'd finish on time. However, reading the script and advancing the slides was a little awkward. I'm not sure I'd do that again.
> >
> > My presentation was a little different than the rest. The others talked about USING a particular feature of LLVM or ADDING a new feature to LLVM. I was talking about retargeting frontends and mapping GEM to LLVM. I did discuss things we had (or will) change in LLVM and clang, but there is really nothing that is rocket science. Others have hooked their own frontends to LLVM in the past. The "interesting" part is that we're doing 6 frontends, RTL enhancements, and adapting clang for OpenVMS all at the same time.
> >
> > A friend in the audience said that people were listening and seemed impressed. I only had a few questions later on.
> >
> > Given the amount of information I presented, I think it might have been as a poster paper. The afternoon break on day #2 has about a dozen of posters and authors spread around the lobby area. It would have let interested folks read and absorb at their rate. However, I wouldn't have had a room full of captives.
>
>
> Thank you
>
> '...seemed impressed'?
>
> Come on!!! Pat yourself on the back a little more.
>
> While at their level they may not be falling off their chair in amazement, I happen to think it's pretty dam impressive, at least to someone at my low level
>
> It's probably going to take time to get known in those circles anyhow.
>
I hope to do a poster session at some point to let people absorb slower. Also, I didn't even get a chance to tell the crowd that we'll also be leveraging the LLVM JIT for the binary translator.
John Reagan
Oct 28, 2017, 12:13:20 PM10/28/17
to
Oh, while there were many sessions interesting to me, one was interesting for the community at large.
Android has been moving from gcc to clang to build. With the Oreo version on my new Pixel 2, the entire distribution (OS and user-mode applications) are now built with clang/llvm. They still link with the gnu linker but they are working to switching to lld as well. With clang/llvm, they were able to use many of the sanitizers and other clang-based tools to find and fix bugs in the Android code base as well as finding/fixing bugs in the compiler toolchain. When the talks show up on the LLVM YouTube channel, I'll send out some links to various talks that people might be interested in.
Android has been moving from gcc to clang to build. With the Oreo version on my new Pixel 2, the entire distribution (OS and user-mode applications) are now built with clang/llvm. They still link with the gnu linker but they are working to switching to lld as well. With clang/llvm, they were able to use many of the sanitizers and other clang-based tools to find and fix bugs in the Android code base as well as finding/fixing bugs in the compiler toolchain. When the talks show up on the LLVM YouTube channel, I'll send out some links to various talks that people might be interested in.
IanD
Nov 1, 2017, 10:37:25 AM11/1/17
to
On Sunday, October 29, 2017 at 3:13:20 AM UTC+11, John Reagan wrote:
> Oh, while there were many sessions interesting to me, one was interesting for the community at large.
>
> Android has been moving from gcc to clang to build. With the Oreo version on my new Pixel 2, the entire distribution (OS and user-mode applications) are now built with clang/llvm. They still link with the gnu linker but they are working to switching to lld as well. With clang/llvm, they were able to use many of the sanitizers and other clang-based tools to find and fix bugs in the Android code base as well as finding/fixing bugs in the compiler toolchain. When the talks show up on the LLVM YouTube channel,
I didn't realise Android was, err, had moved to LLVM
> Oh, while there were many sessions interesting to me, one was interesting for the community at large.
>
> Android has been moving from gcc to clang to build. With the Oreo version on my new Pixel 2, the entire distribution (OS and user-mode applications) are now built with clang/llvm. They still link with the gnu linker but they are working to switching to lld as well. With clang/llvm, they were able to use many of the sanitizers and other clang-based tools to find and fix bugs in the Android code base as well as finding/fixing bugs in the compiler toolchain. When the talks show up on the LLVM YouTube channel,
>>> I'll send out some links to various talks that people might be interested in.
I just found the below, which is your talk given at the LLVM convention :-)
Talk about having to condense a lot into a lightening talk! (Just over 5 mins)
https://www.youtube.com/watch?v=xTaBkCBYskA
Looks like your document got a little non-compliant at times :-)
IanD
Nov 1, 2017, 11:02:55 AM11/1/17
to
For those interested, there's a ton of talks listed (Including John's)
2017 LLVM Developers’ Meeting Talks:
https://www.youtube.com/channel/UCv2_41bSAa5Y_8BacJUZfjQ
I'm about to watch the Tensorflow one and get totally bamboozled I suspect
2017 LLVM Developers’ Meeting Talks:
https://www.youtube.com/channel/UCv2_41bSAa5Y_8BacJUZfjQ
I'm about to watch the Tensorflow one and get totally bamboozled I suspect
John Reagan
Nov 1, 2017, 12:18:56 PM11/1/17
to
I was difficult to flip through my script, keep on time, and reach over to press the button to advance the slides. Won't do that again.
John Reagan
Nov 1, 2017, 12:26:05 PM11/1/17
to
The "Falcon: An optimizing Java JIT" was interesting as well.
The "Writing Great Machine Schedulers" does give insight into the current machine schedulers inside of LLVM today.
The "Source-based Code Coverage BoF" was interesting and I had a lengthy talk with the speakers afterwards.
And the "lld: A Fast, Simple, and Portable Linker" was interesting in how they are using parallelization to improve linking speeds. Not sure if you'll ever see the VMS Linker use pthreads() however. :) :) :)
Talks from prior meetings are good also.
VAXman-
Nov 1, 2017, 12:53:12 PM11/1/17
to
In article <df58365d-83cd-4e2b...@googlegroups.com>, John Reagan <xyzz...@gmail.com> writes:
>On Wednesday, November 1, 2017 at 10:37:25 AM UTC-4, IanD wrote:
>On Wednesday, November 1, 2017 at 10:37:25 AM UTC-4, IanD wrote:
>> On Sunday, October 29, 2017 at 3:13:20 AM UTC+11, John Reagan wrote:
>> > Oh, while there were many sessions interesting to me, one was interesti=
>ng for the community at large.
>> >=20
>> > Android has been moving from gcc to clang to build. With the Oreo vers=
>ion on my new Pixel 2, the entire distribution (OS and user-mode applicatio=
>ns) are now built with clang/llvm. They still link with the gnu linker but=
> they are working to switching to lld as well. With clang/llvm, they were =
>able to use many of the sanitizers and other clang-based tools to find and =
>fix bugs in the Android code base as well as finding/fixing bugs in the com=
>piler toolchain. When the talks show up on the LLVM YouTube channel,=20
>>=20
>> I didn't realise Android was, err, had moved to LLVM
>>=20
>> >>> I'll send out some links to various talks that people might be intere=
>sted in.
>>=20
>> Please do :-)
>>=20
>> I just found the below, which is your talk given at the LLVM convention :=
>-)
>>=20
>> Talk about having to condense a lot into a lightening talk! (Just over 5 =
>mins)
>>=20
>> https://www.youtube.com/watch?v=3DxTaBkCBYskA
>>=20
>> Looks like your document got a little non-compliant at times :-)
>
>non-compliant? How so?
>
>
>I was difficult to flip through my script, keep on time, and reach over to =
>press the button to advance the slides. Won't do that again.
If you run into my buddy Iain S. in your LLVM dealings, let me know.
IanD
Nov 1, 2017, 12:53:13 PM11/1/17
to
I looked for various words trying to come up with a way of saying your document had a mind of it's own
I guess the humor part didn't come across like I intended
I watched the one on parallelism by the Cray guy, what a difference more time makes. He was able to back fill a lot of stuff. 5 mins is nowhere near long enough. 8-10 is more reasonable
John Reagan
Nov 1, 2017, 1:21:00 PM11/1/17
to
On Wednesday, November 1, 2017 at 12:53:12 PM UTC-4, VAXman- wrote:
>
> If you run into my buddy Iain S. in your LLVM dealings, let me know.
>
I saw him last year, but didn't bump into him this year. Perhaps I wasn't looking...
>
> If you run into my buddy Iain S. in your LLVM dealings, let me know.
>
0 new messages