From qihui.sun at gmail.com Fri Oct 19 19:52:58 2012 From: qihui.sun at gmail.com (Qihui Sun) Date: Sat, 20 Oct 2012 10:52:58 +0800 Subject: Hello, is there any progress or other messages ? Message-ID: Thank you! From john.r.rose at oracle.com Fri Oct 19 22:20:18 2012 From: john.r.rose at oracle.com (John Rose) Date: Fri, 19 Oct 2012 22:20:18 -0700 Subject: Hello, is there any progress or other messages ? In-Reply-To: References: Message-ID: On Oct 19, 2012, at 7:52 PM, Qihui Sun wrote: > Hello, is there any progress or other messages ? > Thank you! Yes, and congratulations; there are now two messages on this alias. Yours is the first and this is the second. There are a lot of point-to-point conversations going since shortly before JavaOne. These will upgrade to public discourse very soon, I promise. Meanwhile, please look at the links on the project page, if you haven't already. And check out the activity on mlvm-dev for related foundational work. Best wishes, ? John From mevans at alphatheory.com Fri Oct 19 22:26:20 2012 From: mevans at alphatheory.com (Matthew Evans) Date: Sat, 20 Oct 2012 05:26:20 +0000 Subject: JBoss Application Support? Message-ID: <5C34924866BCD34FB0F34DB9D7CFCB7C0AE891@exch01.alphatheory.local> Hello, As long as we?re running a Sumatra-capable JVM, will JBoss be GPU/APU accelerated? I would love to stick a bunch of APU/GPU devices into my servers to perform financial calculations under JBoss. Cheers, Matt From john.r.rose at oracle.com Fri Oct 26 17:22:58 2012 From: john.r.rose at oracle.com (John Rose) Date: Fri, 26 Oct 2012 17:22:58 -0700 Subject: infrastructure update: wiki location, etc. Message-ID: <73762FFA-B6AF-4691-B04A-760F96A06B3D@oracle.com> We have created a parent wiki page for Sumatra developers to host project documentation, here: https://wikis.oracle.com/display/OpenJDK/Project+Sumatra We will also create an initially empty, standalone repository. It will hold experimental copies of things like Aparapi. Serious JVM and JDK modifications will also require repositories, but we won't build those just yet. Some initial JVM development will be hosted by the mlvm project, as one or more patches, such as this one: http://hg.openjdk.java.net/mlvm/mlvm/hotspot/file/tip/value-obj-8001111.patch Have a good weekend! ? John Rose From Gary.Frost at amd.com Mon Oct 29 07:35:18 2012 From: Gary.Frost at amd.com (Frost, Gary) Date: Mon, 29 Oct 2012 14:35:18 +0000 Subject: infrastructure update: wiki location, etc. In-Reply-To: <73762FFA-B6AF-4691-B04A-760F96A06B3D@oracle.com> References: <73762FFA-B6AF-4691-B04A-760F96A06B3D@oracle.com> Message-ID: John, Thanks for kick-starting this. I plan on adding a page or two to track our initial experiments with Lambda and Aparapi. Hopefully I can get those out in the next few days. Generally these experiments will allow us to mockup up some potential Sumatra programming model ideas and use Aparapi as a proxy for Sumatra as we experiment with the models. Another suggestion might be to create a list of other experiments/research tasks that we might consider engaging in up-front. These may be centered on GC, Code Gen, allocation or wherever. But it might be good to collect such ideas in a central location. This way folk can contribute ideas, or volunteer to do some research/experimentation/prototyping. Gary -----Original Message----- From: sumatra-dev-bounces at openjdk.java.net [mailto:sumatra-dev-bounces at openjdk.java.net] On Behalf Of John Rose Sent: Friday, October 26, 2012 7:23 PM To: sumatra-dev at openjdk.java.net Subject: infrastructure update: wiki location, etc. We have created a parent wiki page for Sumatra developers to host project documentation, here: https://wikis.oracle.com/display/OpenJDK/Project+Sumatra We will also create an initially empty, standalone repository. It will hold experimental copies of things like Aparapi. Serious JVM and JDK modifications will also require repositories, but we won't build those just yet. Some initial JVM development will be hosted by the mlvm project, as one or more patches, such as this one: http://hg.openjdk.java.net/mlvm/mlvm/hotspot/file/tip/value-obj-8001111.patch Have a good weekend! - John Rose From Ryan.LaMothe at pnnl.gov Mon Oct 29 09:10:42 2012 From: Ryan.LaMothe at pnnl.gov (LaMothe, Ryan R) Date: Mon, 29 Oct 2012 09:10:42 -0700 Subject: infrastructure update: wiki location, etc. In-Reply-To: References: <73762FFA-B6AF-4691-B04A-760F96A06B3D@oracle.com> Message-ID: <0003FD2B9D9E364CBFD03658C2991E6301AC944A698C@EMAIL04.pnl.gov> Is there a plan to continue both Aparapi and Sumatra development in parallel? __________________________________________________ Ryan LaMothe -----Original Message----- From: sumatra-dev-bounces at openjdk.java.net [mailto:sumatra-dev-bounces at openjdk.java.net] On Behalf Of Frost, Gary Sent: Monday, October 29, 2012 7:35 AM To: John Rose; sumatra-dev at openjdk.java.net Subject: RE: infrastructure update: wiki location, etc. John, Thanks for kick-starting this. I plan on adding a page or two to track our initial experiments with Lambda and Aparapi. Hopefully I can get those out in the next few days. Generally these experiments will allow us to mockup up some potential Sumatra programming model ideas and use Aparapi as a proxy for Sumatra as we experiment with the models. Another suggestion might be to create a list of other experiments/research tasks that we might consider engaging in up-front. These may be centered on GC, Code Gen, allocation or wherever. But it might be good to collect such ideas in a central location. This way folk can contribute ideas, or volunteer to do some research/experimentation/prototyping. Gary -----Original Message----- From: sumatra-dev-bounces at openjdk.java.net [mailto:sumatra-dev-bounces at openjdk.java.net] On Behalf Of John Rose Sent: Friday, October 26, 2012 7:23 PM To: sumatra-dev at openjdk.java.net Subject: infrastructure update: wiki location, etc. We have created a parent wiki page for Sumatra developers to host project documentation, here: https://wikis.oracle.com/display/OpenJDK/Project+Sumatra We will also create an initially empty, standalone repository. It will hold experimental copies of things like Aparapi. Serious JVM and JDK modifications will also require repositories, but we won't build those just yet. Some initial JVM development will be hosted by the mlvm project, as one or more patches, such as this one: http://hg.openjdk.java.net/mlvm/mlvm/hotspot/file/tip/value-obj-8001111.patch Have a good weekend! - John Rose From Gary.Frost at amd.com Mon Oct 29 09:26:30 2012 From: Gary.Frost at amd.com (Frost, Gary) Date: Mon, 29 Oct 2012 16:26:30 +0000 Subject: infrastructure update: wiki location, etc. In-Reply-To: <0003FD2B9D9E364CBFD03658C2991E6301AC944A698C@EMAIL04.pnl.gov> References: <73762FFA-B6AF-4691-B04A-760F96A06B3D@oracle.com> <0003FD2B9D9E364CBFD03658C2991E6301AC944A698C@EMAIL04.pnl.gov> Message-ID: Ryan, My assumption is that Aparapi will ultimately be made redundant by a successful Sumatra implementation. However, in the meantime I do plan to maintained and augment Aparapi as an open source project for users wishing to tap into the compute potential of GPU devices today. I also expect to use whatever parts of Aparapi that are useful to help prototype Sumatra features/ideas and to track Sumatra suggestions/programing models and roll them back into Aparapi as we move forward. Basically when Sumatra is realized it should be far more convenient and performant than Aparapi for Java developers, and should be able to offer Aparapi's benefits without the current restrictions that Aparapi imposes. For Aparapi users we will also investigate moving the implementation of Aparapi API's on top of Sumatra as early as possible to establish an easy 'glide-path' to Sumatra features. Gary -----Original Message----- From: LaMothe, Ryan R [mailto:Ryan.LaMothe at pnnl.gov] Sent: Monday, October 29, 2012 11:11 AM To: Frost, Gary; John Rose; sumatra-dev at openjdk.java.net Subject: RE: infrastructure update: wiki location, etc. Is there a plan to continue both Aparapi and Sumatra development in parallel? __________________________________________________ Ryan LaMothe -----Original Message----- From: sumatra-dev-bounces at openjdk.java.net [mailto:sumatra-dev-bounces at openjdk.java.net] On Behalf Of Frost, Gary Sent: Monday, October 29, 2012 7:35 AM To: John Rose; sumatra-dev at openjdk.java.net Subject: RE: infrastructure update: wiki location, etc. John, Thanks for kick-starting this. I plan on adding a page or two to track our initial experiments with Lambda and Aparapi. Hopefully I can get those out in the next few days. Generally these experiments will allow us to mockup up some potential Sumatra programming model ideas and use Aparapi as a proxy for Sumatra as we experiment with the models. Another suggestion might be to create a list of other experiments/research tasks that we might consider engaging in up-front. These may be centered on GC, Code Gen, allocation or wherever. But it might be good to collect such ideas in a central location. This way folk can contribute ideas, or volunteer to do some research/experimentation/prototyping. Gary -----Original Message----- From: sumatra-dev-bounces at openjdk.java.net [mailto:sumatra-dev-bounces at openjdk.java.net] On Behalf Of John Rose Sent: Friday, October 26, 2012 7:23 PM To: sumatra-dev at openjdk.java.net Subject: infrastructure update: wiki location, etc. We have created a parent wiki page for Sumatra developers to host project documentation, here: https://wikis.oracle.com/display/OpenJDK/Project+Sumatra We will also create an initially empty, standalone repository. It will hold experimental copies of things like Aparapi. Serious JVM and JDK modifications will also require repositories, but we won't build those just yet. Some initial JVM development will be hosted by the mlvm project, as one or more patches, such as this one: http://hg.openjdk.java.net/mlvm/mlvm/hotspot/file/tip/value-obj-8001111.patch Have a good weekend! - John Rose From john.r.rose at oracle.com Mon Oct 29 09:34:54 2012 From: john.r.rose at oracle.com (John Rose) Date: Mon, 29 Oct 2012 09:34:54 -0700 Subject: infrastructure update: wiki location, etc. In-Reply-To: References: <73762FFA-B6AF-4691-B04A-760F96A06B3D@oracle.com> Message-ID: <25364D6A-D7C1-4EA0-B7A1-675141249C84@oracle.com> On Oct 29, 2012, at 7:35 AM, Frost, Gary wrote: > Thanks for kick-starting this. You are welcome. > I plan on adding a page or two to track our initial experiments with Lambda and Aparapi. Hopefully I can get those out in the next few days. Generally these experiments will allow us to mockup up some potential Sumatra programming model ideas and use Aparapi as a proxy for Sumatra as we experiment with the models. Good. Please do so. The wiki will have a limited number of "editors", mainly due to administration limitations; I have added you to the list. Anyone who registers with the Oracle wiki can add page comments, which should be enough to organize requests for more information. > Another suggestion might be to create a list of other experiments/research tasks that we might consider engaging in up-front. These may be centered on GC, Code Gen, allocation or wherever. But it might be good to collect such ideas in a central location. This way folk can contribute ideas, or volunteer to do some research/experimentation/prototyping. In general, email is best for information with a short shelf life. All email to sumatra-dev is archived here: http://mail.openjdk.java.net/pipermail/sumatra-dev/ The wiki should be used for information which is likely to be worth maintaining in the long term, or (in a few cases) for to-do lists that the community is committed to maintaining. ? John From jack at moxley.co.uk Mon Oct 29 09:57:22 2012 From: jack at moxley.co.uk (Jack Moxley) Date: Mon, 29 Oct 2012 16:57:22 +0000 Subject: infrastructure update: wiki location, etc. In-Reply-To: <25364D6A-D7C1-4EA0-B7A1-675141249C84@oracle.com> Message-ID: <20121029165722.26f9fdfe@mailgate.office.moxley.co.uk> If there is anything I can do to contribute just ask. By day I write fx/treasury systems for a bank, but, and despite its obvious use, I doubt I would be allowed to provide any sunlight inspired ideas in that capacity. :-) By night I am yet another java developer writing yet another language for the JVM, and having a common interface to every piece of hardware is core to its feature base. Project Sumatra fits that requirement, so if you need the half witted opinion of a DIY coder look no further :-). Jack. _____ From: John Rose [mailto:john.r.rose at oracle.com] To: Frost, Gary [mailto:Gary.Frost at amd.com] Cc: sumatra-dev at openjdk.java.net [mailto:sumatra-dev at openjdk.java.net] Sent: Mon, 29 Oct 2012 16:34:54 +0000 Subject: Re: infrastructure update: wiki location, etc. On Oct 29, 2012, at 7:35 AM, Frost, Gary wrote: > Thanks for kick-starting this. You are welcome. > I plan on adding a page or two to track our initial experiments with Lambda and Aparapi. Hopefully I can get those out in the next few days. Generally these experiments will allow us to mockup up some potential Sumatra programming model ideas and use Aparapi as a proxy for Sumatra as we experiment with the models. Good. Please do so. The wiki will have a limited number of "editors", mainly due to administration limitations; I have added you to the list. Anyone who registers with the Oracle wiki can add page comments, which should be enough to organize requests for more information. > Another suggestion might be to create a list of other experiments/research tasks that we might consider engaging in up-front. These may be centered on GC, Code Gen, allocation or wherever. But it might be good to collect such ideas in a central location. This way folk can contribute ideas, or volunteer to do some research/experimentation/prototyping. In general, email is best for information with a short shelf life. All email to sumatra-dev is archived here: http://mail.openjdk.java.net/pipermail/sumatra-dev/ The wiki should be used for information which is likely to be worth maintaining in the long term, or (in a few cases) for to-do lists that the community is committed to maintaining. ? John From Ryan.LaMothe at pnnl.gov Mon Oct 29 10:29:17 2012 From: Ryan.LaMothe at pnnl.gov (LaMothe, Ryan R) Date: Mon, 29 Oct 2012 10:29:17 -0700 Subject: infrastructure update: wiki location, etc. In-Reply-To: References: <73762FFA-B6AF-4691-B04A-760F96A06B3D@oracle.com> <0003FD2B9D9E364CBFD03658C2991E6301AC944A698C@EMAIL04.pnl.gov> Message-ID: <0003FD2B9D9E364CBFD03658C2991E6301AC944A69BE@EMAIL04.pnl.gov> Thanks Gary. I was also wondering if there will be some kind of planned regular sync with the Aparapi trunk, since our current and future work/research will primarily be in Aparapi. Also, since this was a Sumatra Wiki thread (sorry for the thread jacking) where is the ideal place for people to post Sumatra feature requests, Java language feature requests and ideally a list of existing frameworks and research. Would that location be the Wiki? JIRA? Elsewhere? If it would help, should we just use the Aparapi Google Code website for all feature requests, existing frameworks and research, etc. and just vet them from there? __________________________________________________ Ryan LaMothe -----Original Message----- From: Frost, Gary [mailto:Gary.Frost at amd.com] Sent: Monday, October 29, 2012 9:27 AM To: LaMothe, Ryan R; John Rose; sumatra-dev at openjdk.java.net Subject: RE: infrastructure update: wiki location, etc. Ryan, My assumption is that Aparapi will ultimately be made redundant by a successful Sumatra implementation. However, in the meantime I do plan to maintained and augment Aparapi as an open source project for users wishing to tap into the compute potential of GPU devices today. I also expect to use whatever parts of Aparapi that are useful to help prototype Sumatra features/ideas and to track Sumatra suggestions/programing models and roll them back into Aparapi as we move forward. Basically when Sumatra is realized it should be far more convenient and performant than Aparapi for Java developers, and should be able to offer Aparapi's benefits without the current restrictions that Aparapi imposes. For Aparapi users we will also investigate moving the implementation of Aparapi API's on top of Sumatra as early as possible to establish an easy 'glide-path' to Sumatra features. Gary -----Original Message----- From: LaMothe, Ryan R [mailto:Ryan.LaMothe at pnnl.gov] Sent: Monday, October 29, 2012 11:11 AM To: Frost, Gary; John Rose; sumatra-dev at openjdk.java.net Subject: RE: infrastructure update: wiki location, etc. Is there a plan to continue both Aparapi and Sumatra development in parallel? __________________________________________________ Ryan LaMothe -----Original Message----- From: sumatra-dev-bounces at openjdk.java.net [mailto:sumatra-dev-bounces at openjdk.java.net] On Behalf Of Frost, Gary Sent: Monday, October 29, 2012 7:35 AM To: John Rose; sumatra-dev at openjdk.java.net Subject: RE: infrastructure update: wiki location, etc. John, Thanks for kick-starting this. I plan on adding a page or two to track our initial experiments with Lambda and Aparapi. Hopefully I can get those out in the next few days. Generally these experiments will allow us to mockup up some potential Sumatra programming model ideas and use Aparapi as a proxy for Sumatra as we experiment with the models. Another suggestion might be to create a list of other experiments/research tasks that we might consider engaging in up-front. These may be centered on GC, Code Gen, allocation or wherever. But it might be good to collect such ideas in a central location. This way folk can contribute ideas, or volunteer to do some research/experimentation/prototyping. Gary -----Original Message----- From: sumatra-dev-bounces at openjdk.java.net [mailto:sumatra-dev-bounces at openjdk.java.net] On Behalf Of John Rose Sent: Friday, October 26, 2012 7:23 PM To: sumatra-dev at openjdk.java.net Subject: infrastructure update: wiki location, etc. We have created a parent wiki page for Sumatra developers to host project documentation, here: https://wikis.oracle.com/display/OpenJDK/Project+Sumatra We will also create an initially empty, standalone repository. It will hold experimental copies of things like Aparapi. Serious JVM and JDK modifications will also require repositories, but we won't build those just yet. Some initial JVM development will be hosted by the mlvm project, as one or more patches, such as this one: http://hg.openjdk.java.net/mlvm/mlvm/hotspot/file/tip/value-obj-8001111.patch Have a good weekend! - John Rose From Gary.Frost at amd.com Mon Oct 29 10:53:21 2012 From: Gary.Frost at amd.com (Frost, Gary) Date: Mon, 29 Oct 2012 17:53:21 +0000 Subject: infrastructure update: wiki location, etc. In-Reply-To: <0003FD2B9D9E364CBFD03658C2991E6301AC944A69BE@EMAIL04.pnl.gov> References: <73762FFA-B6AF-4691-B04A-760F96A06B3D@oracle.com> <0003FD2B9D9E364CBFD03658C2991E6301AC944A698C@EMAIL04.pnl.gov> <0003FD2B9D9E364CBFD03658C2991E6301AC944A69BE@EMAIL04.pnl.gov> Message-ID: Ryan, I don't want too imply that much of a link between Sumatra and Aparapi. It is natural for me (and perhaps you) to consider these as close cousins, but truthfully they are completely separate projects with similar goals and I would not want them any other way. We should borrow when we can, but truthfully I see Aparapi as just an 'influence' rather than anything else. I would definitely *not* recommend confusing folks by hosting any Aparapi specific stuff here, or for that matter any Sumatra resources over at the aparapi.googlecode site. The OpenJDK projects that I have seen/followed have great resources at their disposal (these email forums, mercurial repositories and the wiki page that John has set up etc) which will be more than appropriate for this OpenJDK project. My guess is we will be creating lots of links from aparapi site to here, to drive traffic and mindshare this way, and that is appropriate and hopefully beneficial. I would recommend everyone look around at some of the other OpenJDK projects to see the infrastructure that is being used. I am personally very impressed with the Lambda and the Graal projects, and I think we can learn a lot from both of these. Gary -----Original Message----- From: LaMothe, Ryan R [mailto:Ryan.LaMothe at pnnl.gov] Sent: Monday, October 29, 2012 12:29 PM To: Frost, Gary; John Rose; sumatra-dev at openjdk.java.net Subject: RE: infrastructure update: wiki location, etc. Thanks Gary. I was also wondering if there will be some kind of planned regular sync with the Aparapi trunk, since our current and future work/research will primarily be in Aparapi. Also, since this was a Sumatra Wiki thread (sorry for the thread jacking) where is the ideal place for people to post Sumatra feature requests, Java language feature requests and ideally a list of existing frameworks and research. Would that location be the Wiki? JIRA? Elsewhere? If it would help, should we just use the Aparapi Google Code website for all feature requests, existing frameworks and research, etc. and just vet them from there? __________________________________________________ Ryan LaMothe -----Original Message----- From: Frost, Gary [mailto:Gary.Frost at amd.com] Sent: Monday, October 29, 2012 9:27 AM To: LaMothe, Ryan R; John Rose; sumatra-dev at openjdk.java.net Subject: RE: infrastructure update: wiki location, etc. Ryan, My assumption is that Aparapi will ultimately be made redundant by a successful Sumatra implementation. However, in the meantime I do plan to maintained and augment Aparapi as an open source project for users wishing to tap into the compute potential of GPU devices today. I also expect to use whatever parts of Aparapi that are useful to help prototype Sumatra features/ideas and to track Sumatra suggestions/programing models and roll them back into Aparapi as we move forward. Basically when Sumatra is realized it should be far more convenient and performant than Aparapi for Java developers, and should be able to offer Aparapi's benefits without the current restrictions that Aparapi imposes. For Aparapi users we will also investigate moving the implementation of Aparapi API's on top of Sumatra as early as possible to establish an easy 'glide-path' to Sumatra features. Gary -----Original Message----- From: LaMothe, Ryan R [mailto:Ryan.LaMothe at pnnl.gov] Sent: Monday, October 29, 2012 11:11 AM To: Frost, Gary; John Rose; sumatra-dev at openjdk.java.net Subject: RE: infrastructure update: wiki location, etc. Is there a plan to continue both Aparapi and Sumatra development in parallel? __________________________________________________ Ryan LaMothe -----Original Message----- From: sumatra-dev-bounces at openjdk.java.net [mailto:sumatra-dev-bounces at openjdk.java.net] On Behalf Of Frost, Gary Sent: Monday, October 29, 2012 7:35 AM To: John Rose; sumatra-dev at openjdk.java.net Subject: RE: infrastructure update: wiki location, etc. John, Thanks for kick-starting this. I plan on adding a page or two to track our initial experiments with Lambda and Aparapi. Hopefully I can get those out in the next few days. Generally these experiments will allow us to mockup up some potential Sumatra programming model ideas and use Aparapi as a proxy for Sumatra as we experiment with the models. Another suggestion might be to create a list of other experiments/research tasks that we might consider engaging in up-front. These may be centered on GC, Code Gen, allocation or wherever. But it might be good to collect such ideas in a central location. This way folk can contribute ideas, or volunteer to do some research/experimentation/prototyping. Gary -----Original Message----- From: sumatra-dev-bounces at openjdk.java.net [mailto:sumatra-dev-bounces at openjdk.java.net] On Behalf Of John Rose Sent: Friday, October 26, 2012 7:23 PM To: sumatra-dev at openjdk.java.net Subject: infrastructure update: wiki location, etc. We have created a parent wiki page for Sumatra developers to host project documentation, here: https://wikis.oracle.com/display/OpenJDK/Project+Sumatra We will also create an initially empty, standalone repository. It will hold experimental copies of things like Aparapi. Serious JVM and JDK modifications will also require repositories, but we won't build those just yet. Some initial JVM development will be hosted by the mlvm project, as one or more patches, such as this one: http://hg.openjdk.java.net/mlvm/mlvm/hotspot/file/tip/value-obj-8001111.patch Have a good weekend! - John Rose From john.r.rose at oracle.com Mon Oct 29 11:39:51 2012 From: john.r.rose at oracle.com (John Rose) Date: Mon, 29 Oct 2012 11:39:51 -0700 Subject: infrastructure update: wiki location, etc. In-Reply-To: <0003FD2B9D9E364CBFD03658C2991E6301AC944A69BE@EMAIL04.pnl.gov> References: <73762FFA-B6AF-4691-B04A-760F96A06B3D@oracle.com> <0003FD2B9D9E364CBFD03658C2991E6301AC944A698C@EMAIL04.pnl.gov> <0003FD2B9D9E364CBFD03658C2991E6301AC944A69BE@EMAIL04.pnl.gov> Message-ID: <9FB9C764-391D-455A-AE88-044C3CADB89C@oracle.com> On Oct 29, 2012, at 10:29 AM, LaMothe, Ryan R wrote: > Java language feature requests Yes, Java language RFEs would be spam on sumatra-dev. We have plenty to do adjusting JVM architecture for GPUs, without touching the "third rail" of Java language design. It's amazing how much you can get done in unmodified Java. And even more amazing with lambdas in Java 8. Since the JVM is a multi-language VM, there are much easier options for experimenting with language design than modifying Java. For example, I would be delighted to see a project for adjusting Clojure libraries to execute on GPUs. Best wishes, ? John From Ryan.LaMothe at pnnl.gov Mon Oct 29 11:49:26 2012 From: Ryan.LaMothe at pnnl.gov (LaMothe, Ryan R) Date: Mon, 29 Oct 2012 11:49:26 -0700 Subject: infrastructure update: wiki location, etc. In-Reply-To: <9FB9C764-391D-455A-AE88-044C3CADB89C@oracle.com> References: <73762FFA-B6AF-4691-B04A-760F96A06B3D@oracle.com> <0003FD2B9D9E364CBFD03658C2991E6301AC944A698C@EMAIL04.pnl.gov> <0003FD2B9D9E364CBFD03658C2991E6301AC944A69BE@EMAIL04.pnl.gov> <9FB9C764-391D-455A-AE88-044C3CADB89C@oracle.com> Message-ID: <0003FD2B9D9E364CBFD03658C2991E6301AC944A69EB@EMAIL04.pnl.gov> Thanks John. The reason I asked is because I believe the two are intimately tied...no matter how much we can do in native Java to execute code on GPUs, there are a number of glaringly obvious things (at least to most people) in the language and JVM that simply need to be fixed at some point. __________________________________________________ Ryan LaMothe From: John Rose [mailto:john.r.rose at oracle.com] Sent: Monday, October 29, 2012 11:40 AM To: LaMothe, Ryan R Cc: Frost, Gary; sumatra-dev at openjdk.java.net Subject: Re: infrastructure update: wiki location, etc. On Oct 29, 2012, at 10:29 AM, LaMothe, Ryan R wrote: Java language feature requests Yes, Java language RFEs would be spam on sumatra-dev. We have plenty to do adjusting JVM architecture for GPUs, without touching the "third rail" of Java language design. It's amazing how much you can get done in unmodified Java. And even more amazing with lambdas in Java 8. Since the JVM is a multi-language VM, there are much easier options for experimenting with language design than modifying Java. For example, I would be delighted to see a project for adjusting Clojure libraries to execute on GPUs. Best wishes, - John From john.r.rose at oracle.com Mon Oct 29 12:17:05 2012 From: john.r.rose at oracle.com (John Rose) Date: Mon, 29 Oct 2012 12:17:05 -0700 Subject: infrastructure update: wiki location, etc. In-Reply-To: <0003FD2B9D9E364CBFD03658C2991E6301AC944A69EB@EMAIL04.pnl.gov> References: <73762FFA-B6AF-4691-B04A-760F96A06B3D@oracle.com> <0003FD2B9D9E364CBFD03658C2991E6301AC944A698C@EMAIL04.pnl.gov> <0003FD2B9D9E364CBFD03658C2991E6301AC944A69BE@EMAIL04.pnl.gov> <9FB9C764-391D-455A-AE88-044C3CADB89C@oracle.com> <0003FD2B9D9E364CBFD03658C2991E6301AC944A69EB@EMAIL04.pnl.gov> Message-ID: <942BD5E7-348D-4E7B-A485-7E35C535FCCE@oracle.com> On Oct 29, 2012, at 11:49 AM, LaMothe, Ryan R wrote: > The reason I asked is because I believe the two are intimately tied?no matter how much we can do in native Java to execute code on GPUs, there are a number of glaringly obvious things (at least to most people) in the language and JVM that simply need to be fixed at some point. I expect Sumatra to lead to a few well-chosen JVM changes. But I have to say that discomfort does not equal insight, and glaringly obvious is not the same as lucidly clear. Cue the story about the blind men and the elephant... Especially in the area of Java language design (but also with JVM design), although people can agree about the existence of pain points, the first suggested fix, especially if passionately urged as obvious even to a dunce, is wrong 99.99% of the time. The reason is simple: We are trying to optimize a complex artifact (JLS, JVM) used by many people (millions) in divergent, mutually inconsistent ways. If the problem is important, the hundredth suggested fix is usually wrong, too. Here's specific example. The way I see it, the 15-year-old JVM data model doesn't quite fit to GPU data types (e.g., vector values, complex numbers, fortran arrays). Java folks have been worrying about this problem for 15 years, which means that no "obvious" solution will be correct. ("Obviously we need a primitive type for complex numbers." "Obviously we need operator overloading." "Obviously we need a new array syntax for rectangular arrays." ? All wrong, in part because they are narrowly-conceived language changes.) Happily, I think there may be a correct-enough?though subtle?solution. See my value-types blog entry for some thinking, and a forthcoming JEP for a more concrete proposal. ? John From Ryan.LaMothe at pnnl.gov Mon Oct 29 12:39:57 2012 From: Ryan.LaMothe at pnnl.gov (LaMothe, Ryan R) Date: Mon, 29 Oct 2012 12:39:57 -0700 Subject: infrastructure update: wiki location, etc. In-Reply-To: <942BD5E7-348D-4E7B-A485-7E35C535FCCE@oracle.com> References: <73762FFA-B6AF-4691-B04A-760F96A06B3D@oracle.com> <0003FD2B9D9E364CBFD03658C2991E6301AC944A698C@EMAIL04.pnl.gov> <0003FD2B9D9E364CBFD03658C2991E6301AC944A69BE@EMAIL04.pnl.gov> <9FB9C764-391D-455A-AE88-044C3CADB89C@oracle.com> <0003FD2B9D9E364CBFD03658C2991E6301AC944A69EB@EMAIL04.pnl.gov> <942BD5E7-348D-4E7B-A485-7E35C535FCCE@oracle.com> Message-ID: <0003FD2B9D9E364CBFD03658C2991E6301AC944A69FC@EMAIL04.pnl.gov> Thanks again John. I can respect your opinions. The main point I was trying to highlight is that over the past 15 years most (all?) of the JLS and JVM has been dictated by individuals in an "ivory tower" down to the masses, even when the masses were complaining ad. nauseum on bug tickets that were getting closed without comment by Sun. There are many obvious items that do need to be addressed during Sumatra development, specifically because many of the decisions that may be wrong 99.99% of the time to some are the exact reasons Java is basically non-existent in high performance and scientific communities (e.g. non-contiguous multi-dimensional arrays, lack of unsigned types, etc.). Moving compute to GPUs is going to, at least initially, be most valuable to these exact people. So obvious doesn't mean wrong, but if we choose to ignore the obvious, we should at least take into account the reasons why millions of people choose not to use Java due to a lack of those exact features. __________________________________________________ Ryan LaMothe -----Original Message----- From: John Rose [mailto:john.r.rose at oracle.com] Sent: Monday, October 29, 2012 12:17 PM To: LaMothe, Ryan R Cc: Frost, Gary; sumatra-dev at openjdk.java.net Subject: Re: infrastructure update: wiki location, etc. On Oct 29, 2012, at 11:49 AM, LaMothe, Ryan R wrote: > The reason I asked is because I believe the two are intimately tied...no matter how much we can do in native Java to execute code on GPUs, there are a number of glaringly obvious things (at least to most people) in the language and JVM that simply need to be fixed at some point. I expect Sumatra to lead to a few well-chosen JVM changes. But I have to say that discomfort does not equal insight, and glaringly obvious is not the same as lucidly clear. Cue the story about the blind men and the elephant... Especially in the area of Java language design (but also with JVM design), although people can agree about the existence of pain points, the first suggested fix, especially if passionately urged as obvious even to a dunce, is wrong 99.99% of the time. The reason is simple: We are trying to optimize a complex artifact (JLS, JVM) used by many people (millions) in divergent, mutually inconsistent ways. If the problem is important, the hundredth suggested fix is usually wrong, too. Here's specific example. The way I see it, the 15-year-old JVM data model doesn't quite fit to GPU data types (e.g., vector values, complex numbers, fortran arrays). Java folks have been worrying about this problem for 15 years, which means that no "obvious" solution will be correct. ("Obviously we need a primitive type for complex numbers." "Obviously we need operator overloading." "Obviously we need a new array syntax for rectangular arrays." - All wrong, in part because they are narrowly-conceived language changes.) Happily, I think there may be a correct-enough-though subtle-solution. See my value-types blog entry for some thinking, and a forthcoming JEP for a more concrete proposal. - John From john.r.rose at oracle.com Mon Oct 29 13:01:19 2012 From: john.r.rose at oracle.com (John Rose) Date: Mon, 29 Oct 2012 13:01:19 -0700 Subject: infrastructure update: wiki location, etc. In-Reply-To: <0003FD2B9D9E364CBFD03658C2991E6301AC944A69FC@EMAIL04.pnl.gov> References: <73762FFA-B6AF-4691-B04A-760F96A06B3D@oracle.com> <0003FD2B9D9E364CBFD03658C2991E6301AC944A698C@EMAIL04.pnl.gov> <0003FD2B9D9E364CBFD03658C2991E6301AC944A69BE@EMAIL04.pnl.gov> <9FB9C764-391D-455A-AE88-044C3CADB89C@oracle.com> <0003FD2B9D9E364CBFD03658C2991E6301AC944A69EB@EMAIL04.pnl.gov> <942BD5E7-348D-4E7B-A485-7E35C535FCCE@oracle.com> <0003FD2B9D9E364CBFD03658C2991E6301AC944A69FC@EMAIL04.pnl.gov> Message-ID: <0A90784A-7746-465C-AD7C-8AF629808DD8@oracle.com> On Oct 29, 2012, at 12:39 PM, LaMothe, Ryan R wrote: > ...Java is basically non-existent in high performance and scientific communities (e.g. non-contiguous multi-dimensional arrays, lack of unsigned types, etc.). Moving compute to GPUs is going to, at least initially, be most valuable to these exact people. So obvious doesn't mean wrong, but if we choose to ignore the obvious, we should at least take into account the reasons why millions of people choose not to use Java due to a lack of those exact features. It's time, now, to fix some of that. I for one have been looking forward to this day for a long time. (I suppose there are reasons why the time is 2012 instead of 2002, but we'll let the historians work that out.) ? John From jocl at jocl.org Mon Oct 29 14:20:13 2012 From: jocl at jocl.org (Marco Hutter) Date: Mon, 29 Oct 2012 22:20:13 +0100 Subject: infrastructure update: wiki location, etc. In-Reply-To: <9FB9C764-391D-455A-AE88-044C3CADB89C@oracle.com> References: <73762FFA-B6AF-4691-B04A-760F96A06B3D@oracle.com> <0003FD2B9D9E364CBFD03658C2991E6301AC944A698C@EMAIL04.pnl.gov> <0003FD2B9D9E364CBFD03658C2991E6301AC944A69BE@EMAIL04.pnl.gov> <9FB9C764-391D-455A-AE88-044C3CADB89C@oracle.com> Message-ID: <508EF30D.108@jocl.org> Am 29.10.2012 19:39, schrieb John Rose: > Since the JVM is a multi-language VM, there are much easier options > for experimenting with language design than modifying Java. For > example, I would be delighted to see a project for adjusting Clojure > libraries to execute on GPUs. (Forwared from a mail, since it might be of general interest: ) There is an increasing awareness and demand for GPU support also in the 'JVM alien languages'. A while ago, I've been talking with the lead developer of the Groovy GPars (Parallel Systems) library about a support of low-level (GPU) data-parallelism, and we agreed that this would be a valuable feature that could fit nicely into the parallel collections infrastructure. This has also lead to the RFE at http://jira.codehaus.org/browse/GPARS-231 I also did some first experiments for translating Closures into OpenCL a few months ago ( http://jocl.org/GroovyGPU/ ) but these only have been a /very/ limited and naive proof of concept. The development has been adjourned due the recent projects (Rootbeer and Sumatra), assuming that the GPU support on JVM level will make such a specific approach redundant. Although I probably do not (yet) know enough about the OpenJDK to become a real contibutor, I'm looking forward to see how Sumatra evolves. Best regards, Marco Hutter