April 2020 - Ecp-aurora-sdk-discuss - mailman3.cels.anl.gov

Re: [Ecp-aurora-sdk-discuss] DIBlockByRefStruct on DICompositeType is no longer supported
by Applencourt, Thomas 30 Apr '20

30 Apr '20

Hi, If you want Argonne to track it and be motified when the bugs will be fixed, may I ask you to add an entry on the confluence page and a reproducer in jlse? (https://wiki.jlse.anl.gov/pages/viewpage.action?pageId=38797463) Thanks! Thomas Applencourt Leadership Computing Facility Argonne National Laboratory Building 240 Office - 1.D.18 9700 S Cass Av., Lemont Il, 60439 (630) 252-0036 ________________________________ From: Ecp-aurora-sdk-discuss <ecp-aurora-sdk-discuss-bounces(a)lists.jlse.anl.gov> on behalf of Vincent R. Pascuzzi via Ecp-aurora-sdk-discuss <ecp-aurora-sdk-discuss(a)lists.jlse.anl.gov> Sent: Thursday, April 30, 2020 07:49 To: Weiqun Zhang <weiqunzhang(a)lbl.gov> Cc: ecp-aurora-sdk-discuss(a)lists.jlse.anl.gov <ecp-aurora-sdk-discuss(a)lists.jlse.anl.gov> Subject: Re: [Ecp-aurora-sdk-discuss] DIBlockByRefStruct on DICompositeType is no longer supported Hi, Here’s a ticket: https://github.com/intel/llvm/issues/1617 I wanted to do a bit more testing before submitting it. Looks like things are bit hairier than I initially thought. How are you running the code that causes these issued warnings? What SYCL device are you using? I see them on an Iris node only when running with the Intel Neo. If you use the “CPU device”, does execution halt at the point you expect to see the warnings? Vince > On Apr 29, 2020, at 6:46 PM, Vincent R. Pascuzzi <vrpascuzzi(a)lbl.gov> wrote: > > Hi, > > Yes, I am experiencing the same thing: > > ``` > DIBlockByRefStruct on DICompositeType is no longer supported > !185 = !DICompositeType(tag: DW_TAG_structure_type, name: "CaloDetDescrElement", file: !186, line: 15, flags: DIFlagReservedBit4, elements: !170) > DIBlockByRefStruct on DICompositeType is no longer supported > !1032 = !DICompositeType(tag: DW_TAG_structure_type, name: "__private_class", scope: !892, file: !893, line: 34, flags: DIFlagReservedBit4, elements: !170) > ``` > > A friend of mine from Intel suggested I open a ticket on the intel/llvm. According to the compiler team, "This looks very strange... Most likely it's not from the DPC++ front-end compiler.” I will give you more details tomorrow (signing off for the evening). > > > Vince > >> On Apr 29, 2020, at 6:37 PM, Weiqun Zhang via Ecp-aurora-sdk-discuss <ecp-aurora-sdk-discuss(a)lists.jlse.anl.gov> wrote: >> >> Hello, >> >> With DPC++ beta6, I get a lot of messages like below at runtime in stderr. Does anyone know what that means? Is there a way to turn it off? >> >> DIBlockByRefStruct on DICompositeType is no longer supported >> !496 = !DICompositeType(tag: DW_TAG_structure_type, name: "initializer_list<int>", scope: !196, file: !497, line: 47, flags: DIFlagReservedBit4, elements: !138) >> DIBlockByRefStruct on DICompositeType is no longer supported >> !637 = !DICompositeType(tag: DW_TAG_structure_type, name: "reverse_iterator<__gnu_cxx::__normal_iterator<int *, std::vector<int, std::allocator<int> > > >", scope: !196, file: !519, line: 105, flags: DIF >> lagReservedBit4, elements: !138) >> DIBlockByRefStruct on DICompositeType is no longer supported >> !642 = !DICompositeType(tag: DW_TAG_structure_type, name: "reverse_iterator<__gnu_cxx::__normal_iterator<const int *, std::vector<int, std::allocator<int> > > >", scope: !196, file: !519, line: 105, flag >> s: DIFlagReservedBit4, elements: !138) >> DIBlockByRefStruct on DICompositeType is no longer supported >> !840 = !DICompositeType(tag: DW_TAG_structure_type, name: "reverse_iterator<int *>", scope: !196, file: !519, line: 105, flags: DIFlagReservedBit4, elements: !138) >> >> Weiqun >> >> PS. Sorry for sending this twice. The first time I sent to the wrong mailing list. >> >> >> -- >> Ecp-aurora-sdk-discuss mailing list >> Ecp-aurora-sdk-discuss(a)lists.jlse.anl.gov >> https://lists.jlse.anl.gov/mailman/listinfo/ecp-aurora-sdk-discuss > -- Ecp-aurora-sdk-discuss mailing list Ecp-aurora-sdk-discuss(a)lists.jlse.anl.gov https://lists.jlse.anl.gov/mailman/listinfo/ecp-aurora-sdk-discuss

1 0

Re: [Ecp-aurora-sdk-discuss] DIBlockByRefStruct on DICompositeType is no longer supported
by Vincent R. Pascuzzi 30 Apr '20

30 Apr '20

Hi, Here’s a ticket: https://github.com/intel/llvm/issues/1617 I wanted to do a bit more testing before submitting it. Looks like things are bit hairier than I initially thought. How are you running the code that causes these issued warnings? What SYCL device are you using? I see them on an Iris node only when running with the Intel Neo. If you use the “CPU device”, does execution halt at the point you expect to see the warnings? Vince > On Apr 29, 2020, at 6:46 PM, Vincent R. Pascuzzi <vrpascuzzi(a)lbl.gov> wrote: > > Hi, > > Yes, I am experiencing the same thing: > > ``` > DIBlockByRefStruct on DICompositeType is no longer supported > !185 = !DICompositeType(tag: DW_TAG_structure_type, name: "CaloDetDescrElement", file: !186, line: 15, flags: DIFlagReservedBit4, elements: !170) > DIBlockByRefStruct on DICompositeType is no longer supported > !1032 = !DICompositeType(tag: DW_TAG_structure_type, name: "__private_class", scope: !892, file: !893, line: 34, flags: DIFlagReservedBit4, elements: !170) > ``` > > A friend of mine from Intel suggested I open a ticket on the intel/llvm. According to the compiler team, "This looks very strange... Most likely it's not from the DPC++ front-end compiler.” I will give you more details tomorrow (signing off for the evening). > > > Vince > >> On Apr 29, 2020, at 6:37 PM, Weiqun Zhang via Ecp-aurora-sdk-discuss <ecp-aurora-sdk-discuss(a)lists.jlse.anl.gov> wrote: >> >> Hello, >> >> With DPC++ beta6, I get a lot of messages like below at runtime in stderr. Does anyone know what that means? Is there a way to turn it off? >> >> DIBlockByRefStruct on DICompositeType is no longer supported >> !496 = !DICompositeType(tag: DW_TAG_structure_type, name: "initializer_list<int>", scope: !196, file: !497, line: 47, flags: DIFlagReservedBit4, elements: !138) >> DIBlockByRefStruct on DICompositeType is no longer supported >> !637 = !DICompositeType(tag: DW_TAG_structure_type, name: "reverse_iterator<__gnu_cxx::__normal_iterator<int *, std::vector<int, std::allocator<int> > > >", scope: !196, file: !519, line: 105, flags: DIF >> lagReservedBit4, elements: !138) >> DIBlockByRefStruct on DICompositeType is no longer supported >> !642 = !DICompositeType(tag: DW_TAG_structure_type, name: "reverse_iterator<__gnu_cxx::__normal_iterator<const int *, std::vector<int, std::allocator<int> > > >", scope: !196, file: !519, line: 105, flag >> s: DIFlagReservedBit4, elements: !138) >> DIBlockByRefStruct on DICompositeType is no longer supported >> !840 = !DICompositeType(tag: DW_TAG_structure_type, name: "reverse_iterator<int *>", scope: !196, file: !519, line: 105, flags: DIFlagReservedBit4, elements: !138) >> >> Weiqun >> >> PS. Sorry for sending this twice. The first time I sent to the wrong mailing list. >> >> >> -- >> Ecp-aurora-sdk-discuss mailing list >> Ecp-aurora-sdk-discuss(a)lists.jlse.anl.gov >> https://lists.jlse.anl.gov/mailman/listinfo/ecp-aurora-sdk-discuss >

1 0

Re: [Ecp-aurora-sdk-discuss] DIBlockByRefStruct on DICompositeType is no longer supported
by Vincent R. Pascuzzi 29 Apr '20

29 Apr '20

Hi, Yes, I am experiencing the same thing: ``` DIBlockByRefStruct on DICompositeType is no longer supported !185 = !DICompositeType(tag: DW_TAG_structure_type, name: "CaloDetDescrElement", file: !186, line: 15, flags: DIFlagReservedBit4, elements: !170) DIBlockByRefStruct on DICompositeType is no longer supported !1032 = !DICompositeType(tag: DW_TAG_structure_type, name: "__private_class", scope: !892, file: !893, line: 34, flags: DIFlagReservedBit4, elements: !170) ``` A friend of mine from Intel suggested I open a ticket on the intel/llvm. According to the compiler team, "This looks very strange... Most likely it's not from the DPC++ front-end compiler.” I will give you more details tomorrow (signing off for the evening). Vince > On Apr 29, 2020, at 6:37 PM, Weiqun Zhang via Ecp-aurora-sdk-discuss <ecp-aurora-sdk-discuss(a)lists.jlse.anl.gov> wrote: > > Hello, > > With DPC++ beta6, I get a lot of messages like below at runtime in stderr. Does anyone know what that means? Is there a way to turn it off? > > DIBlockByRefStruct on DICompositeType is no longer supported > !496 = !DICompositeType(tag: DW_TAG_structure_type, name: "initializer_list<int>", scope: !196, file: !497, line: 47, flags: DIFlagReservedBit4, elements: !138) > DIBlockByRefStruct on DICompositeType is no longer supported > !637 = !DICompositeType(tag: DW_TAG_structure_type, name: "reverse_iterator<__gnu_cxx::__normal_iterator<int *, std::vector<int, std::allocator<int> > > >", scope: !196, file: !519, line: 105, flags: DIF > lagReservedBit4, elements: !138) > DIBlockByRefStruct on DICompositeType is no longer supported > !642 = !DICompositeType(tag: DW_TAG_structure_type, name: "reverse_iterator<__gnu_cxx::__normal_iterator<const int *, std::vector<int, std::allocator<int> > > >", scope: !196, file: !519, line: 105, flag > s: DIFlagReservedBit4, elements: !138) > DIBlockByRefStruct on DICompositeType is no longer supported > !840 = !DICompositeType(tag: DW_TAG_structure_type, name: "reverse_iterator<int *>", scope: !196, file: !519, line: 105, flags: DIFlagReservedBit4, elements: !138) > > Weiqun > > PS. Sorry for sending this twice. The first time I sent to the wrong mailing list. > > > -- > Ecp-aurora-sdk-discuss mailing list > Ecp-aurora-sdk-discuss(a)lists.jlse.anl.gov > https://lists.jlse.anl.gov/mailman/listinfo/ecp-aurora-sdk-discuss

1 0

DIBlockByRefStruct on DICompositeType is no longer supported
by Weiqun Zhang 29 Apr '20

29 Apr '20

Hello, With DPC++ beta6, I get a lot of messages like below at runtime in stderr. Does anyone know what that means? Is there a way to turn it off? DIBlockByRefStruct on DICompositeType is no longer supported !496 = !DICompositeType(tag: DW_TAG_structure_type, name: "initializer_list<int>", scope: !196, file: !497, line: 47, flags: DIFlagReservedBit4, elements: !138) DIBlockByRefStruct on DICompositeType is no longer supported !637 = !DICompositeType(tag: DW_TAG_structure_type, name: "reverse_iterator<__gnu_cxx::__normal_iterator<int *, std::vector<int, std::allocator<int> > > >", scope: !196, file: !519, line: 105, flags: DIF lagReservedBit4, elements: !138) DIBlockByRefStruct on DICompositeType is no longer supported !642 = !DICompositeType(tag: DW_TAG_structure_type, name: "reverse_iterator<__gnu_cxx::__normal_iterator<const int *, std::vector<int, std::allocator<int> > > >", scope: !196, file: !519, line: 105, flag s: DIFlagReservedBit4, elements: !138) DIBlockByRefStruct on DICompositeType is no longer supported !840 = !DICompositeType(tag: DW_TAG_structure_type, name: "reverse_iterator<int *>", scope: !196, file: !519, line: 105, flags: DIFlagReservedBit4, elements: !138) Weiqun PS. Sorry for sending this twice. The first time I sent to the wrong mailing list.

1 0

New A21 SDK available in JLSE
by Rosales-fernandez, Carlos 29 Apr '20

29 Apr '20

Hi all, We are pleased to announce that a new A21 SDK has been deployed in JLSE. This version 2020.04.30.001 includes multiple improvements and bug fixes. For additional details please refer to the latest release notes in the Wiki: https://wiki.jlse.anl.gov/display/inteldga/DGA+SDK+Release+Notes Modules will enable the newer SDK by default, but the older versions remain available for regression testing. Updated Performance Tools Interface (PTI) examples and GTPin files are also provided as part of this release, although no modules are associauted with them. Important notes =============== When using dynamic linking for threaded MKL calls, please include the tbb location in teh link line. An old tbb.so library is being picked up by the linker from /lib64 adn this is a workaround to avoid issues, so substitute your typicall "-ltbb" in the MKL build by (prepending location to LD_LIBRARY_PATH will not work): $ -L$TBBROOT/lib/intel64/gcc4.8 -ltbb VTune improvements in post-processing. Some workloads will see up to a 4x speedup in post-processing thanks to recent improvements. Please let us know of any issues with the release or the environment settings. Regards, Carlos

1 0

New A21 SDK coming to JLSE later today
by Rosales-fernandez, Carlos 29 Apr '20

29 Apr '20

Hi all, A new version of the A21 SDK will be deployed at 1pm Central Time. At that point default modulefiles will change to the 2020.04.30.001 version. Running jobs should not be affected. A confirmation email will be sent after the update is complete. Regards, Carlos

1 0

Re: [Ecp-aurora-sdk-discuss] DPC++ book
by Hammond, Jeff R 10 Apr '20

10 Apr '20

This question came up recently and the response from one of the authors was: “We are currently working on the rest of the book. The complete text should be out around Supercomputing 2020.” If there are specific topics you want covered, please post suggestions and questions here and we can forward to the right people. Jeff From: Ecp-aurora-sdk-discuss <ecp-aurora-sdk-discuss-bounces(a)lists.jlse.anl.gov> on behalf of "Habib, Salman via Ecp-aurora-sdk-discuss" <ecp-aurora-sdk-discuss(a)lists.jlse.anl.gov> Reply-To: "Habib, Salman" <habib(a)anl.gov> Date: Friday, April 10, 2020 at 9:04 AM To: "Kumaran, Kalyan" <kumaran(a)anl.gov>, "ecp-aurora-sdk-discuss(a)jlse.anl.gov" <ecp-aurora-sdk-discuss(a)jlse.anl.gov> Subject: Re: [Ecp-aurora-sdk-discuss] DPC++ book These chapters have been around for a while -- any more content expected? Salman On 4/10/20 10:53 AM, Kumaran, Kalyan via Ecp-aurora-sdk-discuss wrote: In case you are interested. We will add this on our confluence page under “Resources” too. -Kumar The first four chapters of Intel's forthcoming book, "Data Parallel C++: Mastering DPC++ for Programming of Heterogeneous Systems Using C++ and SYCL,” have been released in advance for free download. DPC++, an extension of C++ that incorporates SYCL and other new features, will be available on Aurora. To download the chapters, visit: https://www.apress.com/us/data-parallel-c-advanced-chapters-just-released/1…

1 0

Re: [Ecp-aurora-sdk-discuss] DPC++ book
by Habib, Salman 10 Apr '20

10 Apr '20

These chapters have been around for a while -- any more content expected? Salman On 4/10/20 10:53 AM, Kumaran, Kalyan via Ecp-aurora-sdk-discuss wrote: In case you are interested. We will add this on our confluence page under “Resources” too. -Kumar The first four chapters of Intel's forthcoming book, "Data Parallel C++: Mastering DPC++ for Programming of Heterogeneous Systems Using C++ and SYCL,” have been released in advance for free download. DPC++, an extension of C++ that incorporates SYCL and other new features, will be available on Aurora. To download the chapters, visit: https://www.apress.com/us/data-parallel-c-advanced-chapters-just-released/1…

1 0

DPC++ book
by Kumaran, Kalyan 10 Apr '20

10 Apr '20

In case you are interested. We will add this on our confluence page under “Resources” too. -Kumar The first four chapters of Intel's forthcoming book, "Data Parallel C++: Mastering DPC++ for Programming of Heterogeneous Systems Using C++ and SYCL,” have been released in advance for free download. DPC++, an extension of C++ that incorporates SYCL and other new features, will be available on Aurora. To download the chapters, visit: https://www.apress.com/us/data-parallel-c-advanced-chapters-just-released/1…

1 0

A21 offload compiler bug fix update in JLSE
by Rosales-fernandez, Carlos 08 Apr '20

08 Apr '20

Hi all, We have pushed a compiler bug fix to JLSE. This compiler drop just fixes the incorrect handling of the “-mcmodel” option, which was not being accepted by the 2020.03.30.001 compiler. The new version can be loaded as module “omp/2020.04.07.001” and has no other changes. As with the 03/30 drop this has not been made the default. Please let us know if you find any new issues. Regards, Carlos

1 0