The novice reader would associate shape with only two of the DerivedTypes: VectorType and ArrayType. How could a FunctionType or an IntegerType have a shape ?

Could you suggest a paper I could read to get to know this Zone Algorithm ?

Sorry about the delayed response.

This breaks my Julia build.

@bollu what is this patch about ?

Simpler check to determine the destination Module of the cloned instruction.

@grosser I was able to commit it.

I'm comitting on the release_50 branch only.

I got the following output on git svn dcommit -n,

@grosser could you please approve ? Also

restrict to 80 columns.

I'm using git-svn. Would doing a arc patch D36050 on the release_50 branch followed by a merge and then git svn dcommit push it to llvm.org/git/polly.git ?

remove the llvm.lifetime* declarations

How are invariant loads to pointers turned into scalar accesses ? Do we need -polly-invariant-load-hoisting for that ?

@grosser How does solving 1. relate to D35630 ?

Simpler, Unoptimized and readable testcase.

Added pseudo code to test case.

@PhilippSchaad Please go ahead.

Added test case.

@grosser Thanks for adding me as a reviewer ! It's helped me acquaint myself with the SPIR V codegen and understand GPUJIT better.

Sure. Will do in sometime.

Yes. I'll drop it.

1. Instructions and basic blocks of new test case have been named.
2. Removed format error with unittests/ScopPassManager/PassManagerTest.cpp

What do you mean by killing a scalar ?

@grosser I tried what you suggessted,

Drop Scop name from ptx_kernel's name.

1. Moved building the name into a routine.
2. ScopNameofIR is const now.
3. Rebased to trunk @ 306540.

@grosser ping

In D33985#797009, @grosser wrote:

Hi Sanjay,

adding the scop's name is a not good idea. Specifically, the scop's name may differ between release and debug builds, as names are only carried over in debug builds.

The name of the Scop is built from the boundary basic blocks of the region in Scop::getNameStr. How does that depend on the build type ? Are extra basic blocks inserted in debug builds ?

@grosser @lattner I was able to commit this patch.

Update diff to include changes till https://llvm.org/svn/llvm-project/polly/trunk@306540

Why wasn't it supported previously ?

This patch corrects the semantics of the accessor methods of the Scop class.

1. Clearly mention the source FUNCtion and SCOP and KERNEL number.
2. Changed name of SCOP prefixed to kernel to follow PTX identifier syntax acc. to http://docs.nvidia.com/cuda/parallel-thread-execution/index.html#identifiers (Raised by @philip.pfaffe )

@grosser ping.

Prefix SCoP name as well to uniquely identify a kernel, since a function can have more than one SCoP.

Yes, we can handle the metadata problem in a different patch.

@bollu Please refer this comment on my patch. I observed a pattern with functions that produced the buggy metadata, simple copy kernels (A[i][j][k] = B[i][j][k]) didn't produce buggy metadata but kernels that had computation (A[i][j][k] +=/-=/*= B[i][j][k]) resulted in Modules that didn't list the DICompileUnit in llvm.dbg.cu.

1. Rename TARGET_Hybrid to TARGET_HYBRID
2. Reduce code redundancy.

@bollu The patch did prevent passing the address of the debug intrinsic to the kernel, but couldn't guard against buggy metadata,

1. Punctuation consistency in comments.
2. Align code to LLVM's C/C++ format.
3. Initialized SkipScop at constructor like all other members.
4. Include code for GPU targeting only when GPU_CODEGEN is defined.

In D34054#777908, @bollu wrote:

@singam-sanjay: nitpick: polly-update-format changes the format of the file in a couple of places. For next time, could you please update the diff after running ninja/make polly-update-format? I've made a couple of stylistic comments that are upto personal choice :)

I'm still waiting for my commit access 😁.