♟ ABataev

[OPENMP] Fix the restrictions for in_reduction clause.

[OPENMP] Fix build problem, NFC.

[OPENMP] Add ast-print tests for in_reduction clause, NFC.

[OPENMP] Initial support for 'in_reduction' clause.

[OPENMP] Simplify analysis of data-sharing attributes.

[OPENMP] Fix DSA processing for member declaration.

[OPENMP] Initial support for 'task_reduction' clause.

[OPENMP] Generalization of sema analysis of reduction-based clauses,

[OPENMP] Pacify windows buildbots, NFC.

[OPENMP] Fix reduction combiner test

[OPENMP] Further fixes of the reduction codegen tests

[OPENMP] Further test fixes.

[OPENMP] Rework tests to pacify buildbots.

[OPENMP] Codegen for reduction clauses in 'taskloop' directives.

[OPENMP] Fix reduction tests, NFC.

[OPENMP] Further reduction test fix, NFC.

[OPENMP] Fix reduction tests, NFC.

[OPENMP] Generalization of codegen for reduction clauses.

[OPENMP] Emit implicit taskgroup block around taskloop directives.

[OPENMP] Skip BuildMemberExpr() in BuildFieldReferenceExpr(), NFC, by Kai Noda

[OPENMP] Add restriction for reduction clause in taskloop directives.

I'm working on the patch that stops vectorization if the parent basic block of the instruction is not BB or the instruction was processed already, but it's quite hard to add a test for this change. I can publish it as NFC, because we just limiting the number of analyzed instructions if this is acceptable.

[SLP] A test for limiting vectorization of instructions, NFC.

[OPENMP] Initial support for taskloop reductions.

In D33320#795834, @anna wrote:

Hi Alexey, Adam,

From what I can see in this algorithm, there is no limit on the actual size of the stack in the loop. The level variable controls just the recursion limit. So, in effect, IIUC, the max total number of operands being processed by the while loop is 2 ^ RecursionLimit (it's to the base 2 because we avoid phi nodes).

[OPENMP][DEBUG] Generate second function with correct arg types.

[OPENMP] Use MapVector instead of DenseMap for stable codegen, NFC.

[DebugInfo] Fix comment, NFC.

[DebugInfo] Add kind of ImplicitParamDecl for emission of FlagObjectPointer.

[SLP] More comments fix, NFC.

Removed FIXMEs and corrected comment

[SLP] Comment fix, NFC.

Update after review.

Improved reordering criteria.

[SLP] Change extension of the test, NFC.

[SLP] Add a test for fix of PR32164, NFC.

Address John comments.

Added DeclContext parameter to constructors of ImplicitParamDecl class.

[SLP] Improve comments and naming of functions/variables/members, NFC.

Revert "[SLP] Improve comments and naming of functions/variables/members, NFC."

[SLP] Improve comments and naming of functions/variables/members, NFC.

Added different kinds of ImplicitParamDecl.

Updates after comments.

Updates after review

In D33735#770333, @rjmccall wrote:

In D33735#770318, @aaron.ballman wrote:

In D33735#770296, @ABataev wrote:

In D33735#770288, @aaron.ballman wrote:

Can you help me to understand what problem is being solved with this new attribute? Under what circumstances would the first argument be an ImplicitParamDecl but not an implicit this or self?

For captured regions an outlined function is created, where all parameters are ImplicitParamDecls. And the very first parameter is wrongly treated as 'this' argument of the member function.

Ah, thank you! That makes sense to me, but it begs the question: why an attribute rather than a bit on ImplicitParamDecl?

I agree: it would make more sense for ImplicitParamDecl to store a Kind that would always be provided at construction.

In D33735#770288, @aaron.ballman wrote:

Can you help me to understand what problem is being solved with this new attribute? Under what circumstances would the first argument be an ImplicitParamDecl but not an implicit this or self?

In D29641#728047, @Gerolf wrote:

Hi,

I would like to understand this better. Could somebody explain what assumptions SCEV makes about its clients? Which assumption(s) is broken by SLP? It seems to me that this issue potentially touches fundamental design decisions/questions and I don't see any verifiers in place.

Thanks
Gerolf

[OpenMP][Driver] Put target binary for each offload target into a

[OPENMP] Allow 'use_device_ptr' clause in 'target data' alone.

Added a comment regarding assigning nullptr value to P

[OPENMP] Allow value of thread local variables in target regions.

Address comments from Adam.

Need cost analysis for this kind of transformation, moving it back to SLP vectorizer.

Michael, tried to implement it in InstCombiner, but we need cost analysis. Later passes are not suitable for this, because we need transformed code in InstCombiner. So, reopened it.

[OPENMP] Check DSA for variables captured by value.

[OPENMP] Add a check for iterator not reached the end of stack, NFC.

[OPENMP] Improve performance of the hasDSA() function, NFC.

[OPENMP] Fix failing test.

[OPENMP] Fix handling of OpenMP code during template instantiation.

ABataev (Alexey Bataev)
User

Projects

User Details

Recent Activity

Today

Fri, Jul 21

Thu, Jul 20

Tue, Jul 18

Mon, Jul 17

Thu, Jul 13

Wed, Jul 12

Tue, Jul 11

Fri, Jun 30

Thu, Jun 29

Tue, Jun 27

Jun 9 2017

Jun 8 2017

Jun 7 2017

Jun 6 2017

Jun 5 2017

Jun 2 2017

Jun 1 2017

May 31 2017

May 30 2017

May 26 2017

May 24 2017

May 18 2017

May 15 2017

May 11 2017

May 3 2017

Apr 27 2017

Apr 26 2017

ABataev (Alexey Bataev)User