pekka.jaaskelainen (Pekka Jääskeläinen)
User

Projects

User does not belong to any projects.

User Details

User Since
Nov 6 2013, 8:48 AM (262 w, 3 d)

Recent Activity

Oct 3 2018

pekka.jaaskelainen accepted D52117: Generate llvm.loop.parallel_accesses instead of llvm.mem.parallel_loop_access metadata..

I glimpsed over this without spotting anything crucial. My Clang code base knowledge is a bit lightweight though so you might want to wait for an another reviewer. On the other hand, the semantics seem to be retained so it might be safe to commit this in case the tests still pass.

Oct 3 2018, 11:20 PM
pekka.jaaskelainen accepted D52116: Introduce llvm.loop.parallel_accesses and llvm.access.group metadata..

Thanks for making this MD more robust! It is essential for the vectorization performance of pocl. Sorry for my slow response time.

Oct 3 2018, 12:47 AM

Jun 1 2017

pekka.jaaskelainen closed D33639: [OpenCL] Use SPIR_CC by default for kernels.

Committed in r304389.

Jun 1 2017, 12:19 AM

May 29 2017

pekka.jaaskelainen updated the diff for D33639: [OpenCL] Use SPIR_CC by default for kernels.

Added the requested AMDGCN test case.

May 29 2017, 9:55 PM
pekka.jaaskelainen added a comment to D33639: [OpenCL] Use SPIR_CC by default for kernels.

It seems the reviewer is blocked by accident.

May 29 2017, 9:45 PM
pekka.jaaskelainen removed 1 blocking reviewer(s) for D33639: [OpenCL] Use SPIR_CC by default for kernels: Anastasia.
May 29 2017, 9:44 PM
pekka.jaaskelainen created D33639: [OpenCL] Use SPIR_CC by default for kernels.
May 29 2017, 12:05 AM

Feb 13 2017

pekka.jaaskelainen accepted D29830: [OpenCL][Doc] Relase 4.0 notes for OpenCL.

LGTM.

Feb 13 2017, 6:29 AM

Feb 9 2017

pekka.jaaskelainen accepted D28080: [Docs][OpenCL] Added OpenCL feature description to user manual..
Feb 9 2017, 11:11 AM

Jan 23 2017

pekka.jaaskelainen added a comment to D28080: [Docs][OpenCL] Added OpenCL feature description to user manual..

Also, for me the command without -emit-llvm doesn't output anything.

What targets do you use?

Jan 23 2017, 10:11 PM

Jan 22 2017

pekka.jaaskelainen added a comment to D28080: [Docs][OpenCL] Added OpenCL feature description to user manual..

@pekka.jaaskelainen, I just compiled the release 4.0 branch and it all works for me as expected:
clang -cc1 -triple spir64-unknown-unknown test.cl
clang -target spir64-unknown-unknown test.cl

Jan 22 2017, 11:22 PM

Jan 18 2017

pekka.jaaskelainen added inline comments to D28080: [Docs][OpenCL] Added OpenCL feature description to user manual..
Jan 18 2017, 8:33 AM

Jan 17 2017

pekka.jaaskelainen requested changes to D28080: [Docs][OpenCL] Added OpenCL feature description to user manual..
Jan 17 2017, 12:00 AM

Jan 13 2017

pekka.jaaskelainen accepted D28080: [Docs][OpenCL] Added OpenCL feature description to user manual..
Jan 13 2017, 3:31 AM

Jan 11 2017

pekka.jaaskelainen added inline comments to D28080: [Docs][OpenCL] Added OpenCL feature description to user manual..
Jan 11 2017, 10:10 PM

Jan 10 2017

pekka.jaaskelainen requested changes to D28080: [Docs][OpenCL] Added OpenCL feature description to user manual..

Great work.

Jan 10 2017, 10:34 PM

Dec 22 2016

pekka.jaaskelainen accepted D28048: [OpenCL] Align fake address space map with the SPIR target maps..

@pekka.jaaskelainen, I think you are using x86 target in the POCL toolchain, do you envision any issues with this change?

Dec 22 2016, 1:44 PM

Nov 14 2016

pekka.jaaskelainen closed D26157: [OpenCL] always use SPIR address spaces for kernel_arg_addr_space MD.

Committed in r286819.

Nov 14 2016, 5:19 AM

Nov 7 2016

pekka.jaaskelainen added a comment to D26157: [OpenCL] always use SPIR address spaces for kernel_arg_addr_space MD.

target architecture). So if there are no memory segments, nothing can be done to optimize for this and therefore providing
the "fake" segment information doesn't seem to be useful? I am just trying to understand the use case.

Nov 7 2016, 10:58 PM

Nov 3 2016

pekka.jaaskelainen added a comment to D26157: [OpenCL] always use SPIR address spaces for kernel_arg_addr_space MD.

Indeed, it requires wider scale discussion to get it right, and e.g. to pass the info to AA. But to be honest, I think OpenCL and CUDA are still considered 'minority' languages in Clang/LLVM which makes me usually lean towards least intrusive implementation solutions whenever possible.

Nov 3 2016, 1:24 PM

Nov 2 2016

pekka.jaaskelainen added a comment to D26157: [OpenCL] always use SPIR address spaces for kernel_arg_addr_space MD.

@pekka.jaaskelainen, I have related question: what is the problem with retaining OpenCL address space information in LLVM IR?
My understanding is that target CodeGen can ignore this information for the machines with 'flat' address space model.
On the other hand I would expect this information be useful for the Optimizer to resolve pointer aliasing.
Does it make any sense to keep OpenCL address space information in LLVM IR generated for the machines with 'flat' address space (e.g. CPU)?

Nov 2 2016, 4:45 AM

Nov 1 2016

pekka.jaaskelainen added a comment to D26157: [OpenCL] always use SPIR address spaces for kernel_arg_addr_space MD.

Thanks. @tstellarAMD OK to commit for 3.9.1 as well?

Nov 1 2016, 11:10 AM

Oct 31 2016

pekka.jaaskelainen updated subscribers of D26157: [OpenCL] always use SPIR address spaces for kernel_arg_addr_space MD.
Oct 31 2016, 10:59 AM
pekka.jaaskelainen retitled D26157: [OpenCL] always use SPIR address spaces for kernel_arg_addr_space MD from to [OpenCL] always use SPIR address spaces for kernel_arg_addr_space MD.
Oct 31 2016, 10:18 AM

Sep 20 2016

pekka.jaaskelainen added a comment to D24735: Triple: Add opencl environment type.

Alright. LGTM!

Sep 20 2016, 8:35 AM
pekka.jaaskelainen added a comment to D24735: Triple: Add opencl environment type.

Makes sense to me. How to differentiate between OpenCL implementations? With the OS setting?

Sep 20 2016, 12:26 AM

May 17 2016

pekka.jaaskelainen added a comment to D20249: [OpenCL] Hierarchical/dynamic parallelism - enqueue kernel in OpenCL 2.0.

LGTM, FWIW.

May 17 2016, 6:13 AM

Mar 21 2016

pekka.jaaskelainen added inline comments to D17596: [OpenCL] Add ocl and spir version for spir target.
Mar 21 2016, 2:12 AM

Feb 23 2016

pekka.jaaskelainen accepted D16876: [OpenCL] Refine pipe builtin support.

@pekka, do you have any more comments?

Feb 23 2016, 8:54 AM

Feb 16 2016

pekka.jaaskelainen added inline comments to D16928: [OpenCL] Apply missing restrictions for Blocks in OpenCL v2.0.
Feb 16 2016, 1:37 AM
pekka.jaaskelainen accepted D16686: [OpenCL] Generate metadata for opencl_unroll_hint attribute.

LGTM.

Feb 16 2016, 1:19 AM

Feb 15 2016

pekka.jaaskelainen accepted D16865: [OpenCL] Support half precision floating point type literal.

LGTM!

Feb 15 2016, 10:56 PM

Feb 12 2016

pekka.jaaskelainen added a comment to D16865: [OpenCL] Support half precision floating point type literal.

Can you add a test that tries to use half precision constants in non OpenCL C mode? Otherwise LGTM.

Feb 12 2016, 2:56 AM
pekka.jaaskelainen accepted D16891: [OpenCL] Allow extern constant-address-space variable in non-kernel function.

LGTM.

Feb 12 2016, 2:51 AM
pekka.jaaskelainen accepted D16040: [OpenCL] Refine OpenCLImageAccessAttr to OpenCLAccessAttr.

Otherwise LGTM.

Feb 12 2016, 2:47 AM

Feb 11 2016

pekka.jaaskelainen accepted D16812: [OpenCL] Stop treating OpenCL builtin functions as standard C library functions.

LGTM

Feb 11 2016, 11:19 AM

Feb 1 2016

pekka.jaaskelainen accepted D16692: [OpenCL] Eliminate warning when declaring OpenCL builtin functions.
Feb 1 2016, 11:38 AM

Jan 25 2016

pekka.jaaskelainen added a comment to D15914: [OpenCL] Pipe builtin functions.

@pekka, do you think it would make sense to include a description of the new OpenCL features in the release note too? Or shall we wait for the completion of OpenCL 2.0 support (hopefully in next release)? Could do both as well I guess. :)

Jan 25 2016, 10:33 AM
pekka.jaaskelainen accepted D15914: [OpenCL] Pipe builtin functions.

LGTM too. Good job. You can try and ask the release manager if you can sneak it in. It's rather container patch so there might be a chance, but I don't know the policy at this stage of the release.

Jan 25 2016, 5:03 AM

Jan 14 2016

pekka.jaaskelainen added a comment to D15914: [OpenCL] Pipe builtin functions.

OK. Seems the LLVM 3.8 branching was done, it might be hard to get this in, but should we try and ask the release manager?

Jan 14 2016, 3:04 AM

Jan 11 2016

pekka.jaaskelainen added inline comments to D15914: [OpenCL] Pipe builtin functions.
Jan 11 2016, 10:54 AM
pekka.jaaskelainen added inline comments to D15914: [OpenCL] Pipe builtin functions.
Jan 11 2016, 10:16 AM
pekka.jaaskelainen added inline comments to D16047: [OpenCL] Add Sema checks for OpenCL 2.0.
Jan 11 2016, 10:05 AM

Jan 9 2016

pekka.jaaskelainen added a comment to D15914: [OpenCL] Pipe builtin functions.

By the way, I just got the access to the llvm svn, can I just commit the pipe type patch directly as I see you all accepted it. Or I should send it to the cfe-commit first?

Jan 9 2016, 1:36 AM
pekka.jaaskelainen added a comment to D15914: [OpenCL] Pipe builtin functions.

I wonder could we squeeze this in before the next week's LLVM 3.8 branching? It would be great.

Jan 9 2016, 1:04 AM

Dec 30 2015

pekka.jaaskelainen accepted D15691: [OpenCL] Improving OpenCL function pointer error checking to catch lone function designator.
Dec 30 2015, 10:39 AM

Dec 28 2015

pekka.jaaskelainen accepted D15603: [OpenCL] Pipe type support.
Dec 28 2015, 7:35 AM
pekka.jaaskelainen added inline comments to D15603: [OpenCL] Pipe type support.
Dec 28 2015, 6:51 AM
pekka.jaaskelainen added inline comments to D15603: [OpenCL] Pipe type support.
Dec 28 2015, 5:28 AM

Dec 15 2015

pekka.jaaskelainen added a comment to D14441: [OpenCL] Pipe types support..

Nice! I will keep on trying to help by reviewing patches whenever time allows.

Dec 15 2015, 7:25 AM

Dec 11 2015

pekka.jaaskelainen accepted D15293: [OpenCL] Correct NULL constant handling after applying default AS.
Dec 11 2015, 12:03 AM

Dec 4 2015

pekka.jaaskelainen added a comment to D14441: [OpenCL] Pipe types support..

Is this still waiting for some updates?

Dec 4 2015, 6:07 AM

Nov 12 2015

pekka.jaaskelainen accepted D14441: [OpenCL] Pipe types support..
Nov 12 2015, 11:49 AM

Nov 10 2015

pekka.jaaskelainen requested changes to D14441: [OpenCL] Pipe types support..

Only small issues found, plus it could use some more test cases.

Nov 10 2015, 9:07 AM

Oct 21 2015

pekka.jaaskelainen accepted D13168: [OpenCL] OpenCL2.0 - Apply default address space rules.
Oct 21 2015, 5:51 AM

Oct 2 2015

pekka.jaaskelainen added a comment to D13349: [OpenCL] Casting boolean to an integer vector in OpenCL should set all bits if boolean is true.

OK now.

Oct 2 2015, 6:30 AM
pekka.jaaskelainen accepted D13349: [OpenCL] Casting boolean to an integer vector in OpenCL should set all bits if boolean is true.
Oct 2 2015, 2:31 AM
pekka.jaaskelainen accepted D13280: [OpenCL] Adding reserved operator logical xor for OpenCL.
Oct 2 2015, 2:03 AM

Sep 29 2015

pekka.jaaskelainen requested changes to D13168: [OpenCL] OpenCL2.0 - Apply default address space rules.
Sep 29 2015, 7:38 AM

Sep 27 2015

pekka.jaaskelainen accepted D13105: [OpenCL] Enable program scope variables for OpenCL2.0.
Sep 27 2015, 11:18 AM

Sep 24 2015

pekka.jaaskelainen requested changes to D13105: [OpenCL] Enable program scope variables for OpenCL2.0.

Could only find style nitpicks in this one.

Sep 24 2015, 10:12 AM

Sep 15 2015

pekka.jaaskelainen added a comment to D12855: [OpenCL] Add new types for OpenCL 2.0..

We are interested in the pipe.

Sep 15 2015, 3:12 AM

Sep 14 2015

pekka.jaaskelainen accepted D12855: [OpenCL] Add new types for OpenCL 2.0..

The patch seems straightforward enough. BTW does someone know if anyone has worked on the 'pipe' qualifier?

Sep 14 2015, 10:47 PM

Sep 8 2015

pekka.jaaskelainen accepted D12591: Fix vector splitting for extract_vector_elt and vector elements of <8-bits..
Sep 8 2015, 1:23 PM

Sep 7 2015

pekka.jaaskelainen accepted D12678: [mips] Reserve address spaces 1-255 for software use..
Sep 7 2015, 11:32 AM

Sep 3 2015

pekka.jaaskelainen added a comment to D12591: Fix vector splitting for extract_vector_elt and vector elements of <8-bits..

The fix looks good, but maybe the test should be an .ir test as this probably affects other targets than MIPS too? Also could be a candidate for 3.7.1?

Sep 3 2015, 5:34 AM

Sep 2 2015

pekka.jaaskelainen added a comment to D10586: Add HSAIL support.

Is there a matching HLC-HSAIL branch for the latest version of the patch? It doesn't work with with hsail-stable-3.7 anymore.

Sep 2 2015, 11:31 AM

Aug 26 2015

pekka.jaaskelainen accepted D12378: [OpenCL] Improve diagnostics detecting implicit vector conversion..

Perhaps the assert() is unneccessary, but otherwise LGTM.

Aug 26 2015, 12:53 PM

Aug 3 2015

pekka.jaaskelainen added a comment to D10586: Add HSAIL support.

Can you please clarify me a bit where the actual device is needed/used (ATM only Kaveri) if the target is HSAIL, an "abstract machine"/IL? Just trying to get my head around on how the Clang/LLVM will work for HSAIL.

Aug 3 2015, 3:21 AM
pekka.jaaskelainen updated subscribers of D10586: Add HSAIL support.
Aug 3 2015, 3:20 AM

Feb 6 2015

pekka.jaaskelainen added inline comments to D7432: [RFC] Loop Interchane Pass.
Feb 6 2015, 5:32 AM
pekka.jaaskelainen added a comment to D7432: [RFC] Loop Interchane Pass.

Great to see this being worked on! It will help the implicit pocl work-group vectorization after the 2-level loop restricting is lifted. Related to this, be careful with the loop id metadata and especially the parallel loop metadata. The parallel loop metadata can be exploited to skip dependency analysis in some cases (At least when both of the loops are parallel? Needs to be thought through.). But if the metadata is moved to a wrong loop during the interchange it might result in a miscompilation.

Feb 6 2015, 12:58 AM

Jan 29 2015

pekka.jaaskelainen accepted D7245: Fix OpenCL 1.2 double as an optional core feature behaviour.

LGTM.

Jan 29 2015, 9:43 AM
pekka.jaaskelainen added a comment to D7245: Fix OpenCL 1.2 double as an optional core feature behaviour.

Are those if conditions that span to the next line indented correctly?

Jan 29 2015, 4:11 AM

Jan 18 2015

pekka.jaaskelainen added a comment to D7008: Proof-of-concept for a Loop fusion pass (WIP).

Is this missing the pass itself?

Jan 18 2015, 12:43 PM

Jun 10 2014

pekka.jaaskelainen added a comment to D4090: Use loop unrolling pragma metadata in the loop unroller .

Great, this feature will be really useful. How about naming these metadatas using the llvm.loop. as a prefix? That is, llvm.loop.unroll.enable etc? Maybe it's worth adding a test case with a negative unroll count to ensure nothing bad happens. Otherwise LGTM.

Jun 10 2014, 12:54 PM

Nov 12 2013

pekka.jaaskelainen awarded D2112: Add a Scalarize pass a Like token.
Nov 12 2013, 1:06 AM