Commit Graph

3495 Commits

Author SHA1 Message Date
Tobias Grosser b65ccc4302 [ScopInfo] Translate Scop::getParamSpace to isl++ [NFC]
llvm-svn: 310224
2017-08-06 20:11:59 +00:00
Tobias Grosser 8ea1fc19b3 [ScopInfo] Translate Scop::getContext to isl++ [NFC]
llvm-svn: 310221
2017-08-06 19:52:38 +00:00
Tobias Grosser 9a63570b13 [ScopInfo] Translate Scop::getIdForParam to isl++ [NFC]
llvm-svn: 310220
2017-08-06 19:31:27 +00:00
Tobias Grosser 5ab39ff224 [ScopInfo] Move get*Writes/getReads/getAccesses to isl++
llvm-svn: 310219
2017-08-06 19:22:27 +00:00
Tobias Grosser b2e6598a7f Remove functional changes that sneaked in by accident in r308892
llvm-svn: 310218
2017-08-06 18:59:19 +00:00
Tobias Grosser 132860afe5 [ScopInfo] Move ScopStmt::setAstBuild/getAstBuild to isl++
llvm-svn: 310216
2017-08-06 17:53:04 +00:00
Tobias Grosser 6ad1640a1d [ScopInfo] Move ScopStmt::getSchedule to isl++
llvm-svn: 310215
2017-08-06 17:45:28 +00:00
Tobias Grosser 2f3041fc6a [ScopInfo] Move getPredecessorDomainConstraints to isl++ [NFC]
llvm-svn: 310214
2017-08-06 17:31:38 +00:00
Tobias Grosser d16f927781 [ScopInfo] Move InvariantAccess to isl++ [NFC]
llvm-svn: 310213
2017-08-06 17:25:14 +00:00
Tobias Grosser dfd20b7949 [ScopInfo] Update comments to refer to isl++ [NFC]
llvm-svn: 310212
2017-08-06 17:25:09 +00:00
Tobias Grosser 27db02b247 [ScopInfo] Move ScopArrayInfo::ScopArrayInfo to isl++ [NFC]
llvm-svn: 310211
2017-08-06 17:25:05 +00:00
Tobias Grosser 85048eff1a [ScopInfo] Move ScopStmt::ScopStmt to isl++ [NFC]
llvm-svn: 310210
2017-08-06 17:24:59 +00:00
Tobias Grosser dcf8d696ff Move ScopInfo::getDomain(), getDomainSpace(), getDomainId() to isl++
llvm-svn: 310209
2017-08-06 16:39:52 +00:00
Tobias Grosser a9b5bbac78 Move ScopStmt::Domain to isl++
llvm-svn: 310207
2017-08-06 16:11:53 +00:00
Tobias Grosser cb0224ad59 Update to a newer version of isl++
llvm-svn: 310206
2017-08-06 15:56:45 +00:00
Tobias Grosser 8b40f8c6c7 Update to isl-0.18-812-g565da6e
This update is mostly a maintenance update, but also exposes a couple of new
functions that will be needed for the next version of the isl++ bindings.

llvm-svn: 310205
2017-08-06 15:51:16 +00:00
Tobias Grosser bfee458d0f [Scopinfo] Fix memory corruption issue that sneaked into the previous commit
llvm-svn: 310204
2017-08-06 15:47:04 +00:00
Tobias Grosser 2332fa3604 [ScopInfo] Move InvalidDomain to isl++ [NFC]
llvm-svn: 310203
2017-08-06 15:36:48 +00:00
Tobias Grosser 2b7479b1af [Polly] Fix for the JSON Exporter
Summary:
Small patch to fix the JSON exporter.

Currently, using "opt -polly-export-jscop" does not generate jscop files, but gives an error:
*** Error in `opt': corrupted double-linked list: 0x0000000000bc4bb0 ***

Updated the function getAccessRelationStr() to work with the current version of getAccessRelation(), fixing the JSON exporter

Reviewers: bollu, grosser

Reviewed By: grosser

Subscribers: grosser, llvm-commits, pollydev

Tags: #polly

Differential Revision: https://reviews.llvm.org/D36370

llvm-svn: 310199
2017-08-06 11:41:10 +00:00
Tobias Grosser aabfbfa5fc Add missing 'REQUIRES: pollyacc' line
llvm-svn: 310197
2017-08-06 11:21:09 +00:00
Tobias Grosser b99c11710c [GPGPU] Make sure managed arrays are prepared at the beginning of the scop
Summary:
This resolves some "instruction does not dominate use" errors, as we used to
prepare the arrays at the location of the first kernel, which not necessarily
dominated all other kernel calls.

Reviewers: Meinersbur, bollu, singam-sanjay

Subscribers: nemanjai, pollydev, llvm-commits, kbarton

Differential Revision: https://reviews.llvm.org/D36372

llvm-svn: 310196
2017-08-06 11:10:38 +00:00
Tobias Grosser 5b307cdb8a [GPGPU] Rename all, not only the first libdevice function
llvm-svn: 310194
2017-08-06 03:04:15 +00:00
Siddharth Bhat e53c924b0f [Polly] [PPCGCodeGeneration] Deal with loops outside the Scop correctly in PPCGCodeGeneration.
A Scop with a loop outside it is not handled currently by
PPCGCodeGeneration. The test case is such that the Scop has only one inner loop
that is detected. This currently breaks codegen.

The fix is to reuse the existing mechanism in `IslNodeBuilder` within
`GPUNodeBuilder.

Differential Revision: https://reviews.llvm.org/D36290

llvm-svn: 310193
2017-08-06 02:39:05 +00:00
Siddharth Bhat 0caed1fbe6 [IslNodeBuilder] [NFC] Refactor creation of loop induction variables of loops outside scops.
This logic is duplicated, so we refactor it into a separate function.
 This will be used in a later patch to teach PPCGCodeGen code generation
 for loops that are outside the scop.

 Differential Revision: https://reviews.llvm.org/D36310

llvm-svn: 310192
2017-08-06 02:07:11 +00:00
Tobias Grosser f2068ef7dd [Polly] Fix typo. NFC.
Reviewers: grosser, Meinersbur, bollu

Differential Revision: https://reviews.llvm.org/D36356

llvm-svn: 310187
2017-08-05 20:03:13 +00:00
Tobias Grosser f9308489eb Add forgotten CMakeLists.txt file in unit-test
llvm-svn: 310177
2017-08-05 09:44:11 +00:00
Tobias Grosser 00f25d0915 Fix spelling error in previous commit
llvm-svn: 310176
2017-08-05 09:39:00 +00:00
Tobias Grosser feae3dfe9f [unittests] Add unittest for getPartialTilePrefixes
In https://reviews.llvm.org/D36278 it was pointed out that the behavior of
getPartialTilePrefixes is not very well understood. To allow for a better
understanding, we first provide some basic unittests.

llvm-svn: 310175
2017-08-05 09:38:09 +00:00
Michael Kruse 138a3fbae1 [DeLICM] Refactor ZoneAlgorithm into ZoneAlgo.cpp. NFC.
Extract ZoneAlgorithm from DeLICM.cpp into its own file.
It will gain a second use by the load forwarding part of
-polly-optree.

llvm-svn: 310146
2017-08-04 22:51:23 +00:00
Siddharth Bhat 638316da5b [PPCGCodeGeneration] [NFC] Log every location from which PPCGCodegen bails.
This is useful when trying to understand why no GPU code was produced.

Differential Revision: https://reviews.llvm.org/D36318

llvm-svn: 310103
2017-08-04 19:36:40 +00:00
Michael Kruse a9a7086319 [ForwardOpTree] Refactor out forwardSpeculatable(). NFC.
The method forwardSpeculatable forwards speculatively executable
instructions and is currently the only way to forward an
instruction.

In the future we intend to add more methods.

llvm-svn: 310056
2017-08-04 12:28:42 +00:00
Philip Pfaffe 96d2143f20 [PM] Make the new-pm passes behave more like the legacy passes
Summary:
Testing the new-pm passes becomes much easier once they behave more like the
old passes in terms of the order in which Scops are processed and printed. This
requires three changes:
- ScopInfo: Use an ordered map to store scops
- ScopInfo: Iterate and print Scops in reverse order to match legacy PM behaviour
- ScopDetection: print function name in ScopAnalysisPrinter

Reviewers: grosser, Meinersbur, bollu

Reviewed By: grosser

Subscribers: pollydev, llvm-commits

Differential Revision: https://reviews.llvm.org/D36303

llvm-svn: 310052
2017-08-04 11:28:51 +00:00
Philip Pfaffe 6ea444e671 [NFC] Fix r310036: Appease clang-format
llvm-svn: 310039
2017-08-04 08:26:45 +00:00
Philip Pfaffe b24beb6f46 [NFC] ScopPass: Remove unused AnalysisKey from OwningInnerAnalysisManagerProxy
llvm-svn: 310036
2017-08-04 08:12:31 +00:00
Michael Kruse 1046aa3148 [VirtualInstruction] Handle MetadataAsValue as constant.
The complication of bspatch.cc of the AOSP buildbot currently fails
presumably because the occurance of a MetadataAsValue in an operand.
This kind of value can occur as operands of intrinsics, the typical
example being the debug intrinsics.

Polly currently ignores the debug intrinsics and it is not yet clear
which other intrinic might occur. For such cases, and to unbreak the
AOSP buildbot, treat a MetadataAsValue as a constant because it can be
referenced without modification in generated code.

llvm-svn: 309992
2017-08-03 22:00:01 +00:00
Michael Kruse 672c011460 [VirtualInstruction] Avoid use of getStmtFor(BB). NFC.
With this patch, we get rid of the last use of getStmtFor(BB). Here
this is done by getting the last statement of the incoming block in
case the user is a phi node; otherwise just fetching the statement
comprising the instruction for which the virtual use is being created.

Differential Revision: https://reviews.llvm.org/D36268

llvm-svn: 309947
2017-08-03 15:27:00 +00:00
Tobias Grosser c1cfe0a828 Add missing REQUIRES line
llvm-svn: 309943
2017-08-03 14:46:53 +00:00
Tobias Grosser b5563c6817 Make sure that all parameter dimensions are set in schedule
Summary:
In case the option -polly-ignore-parameter-bounds is set, not all parameters
will be added to context and domains. This is useful to keep the size of the
sets and maps we work with small. Unfortunately, for AST generation it is
necessary to ensure all parameters are part of the schedule tree. Hence,
we modify the GPGPU code generation to make sure this is the case.

To obtain the necessary information we expose a new function
Scop::getFullParamSpace(). We also make a couple of functions const to be
able to make SCoP::getFullParamSpace() const.

Reviewers: Meinersbur, bollu, gareevroman, efriedma, huihuiz, sebpop, simbuerg

Subscribers: nemanjai, kbarton, pollydev, llvm-commits

Tags: #polly

Differential Revision: https://reviews.llvm.org/D36243

llvm-svn: 309939
2017-08-03 13:51:15 +00:00
Michael Kruse 291fd8074e [test] Fix test case without Polly-ACC.
llvm-svn: 309938
2017-08-03 13:44:31 +00:00
Siddharth Bhat eadf76d34a [PPCGCodeGeneration] Construct `isl_multi_pw_aff` of PPCGArray.bounds even when polly-ignore-parameter-bounds is turned on.
When we have `-polly-ignore-parameter-bounds`, `Scop::Context` does not contain
all the paramters present in the program.

The construction of the `isl_multi_pw_aff` requires all the indivisual `pw_aff`
to have the same parameter dimensions. To achieve this, we used to realign
every `pw_aff` with `Scop::Context`. However, in conjunction with
`-polly-ignore-parameter-bounds`, this is now incorrect, since `Scop::Context`
does not contain all parameters.

We set this up correctly by creating a space that has all the parameters
used by all the `isl_pw_aff`. Then, we realign all `isl_pw_aff` to this space.

llvm-svn: 309934
2017-08-03 12:09:33 +00:00
Tobias Grosser a195576118 Enable simplify and forward-op-tree by default
These passes have been tested over the last month and should generally help
to remove scalar data dependences in Polly. We enable them to give them even
wider test coverage. Large performance regressions and any kind of correctness
regressions are not expected.

llvm-svn: 309878
2017-08-02 20:12:27 +00:00
Tobias Grosser 7b45af13ce Move setNewAccessRelation to isl++
llvm-svn: 309871
2017-08-02 19:27:25 +00:00
Tobias Grosser 6d58804cc2 Move ScopStmt::setAccessRelation to isl++
llvm-svn: 309870
2017-08-02 19:27:16 +00:00
Tobias Grosser 18ca9e5119 Replace asserts with llvm_unreachable to clarify intent
llvm-svn: 309856
2017-08-02 19:11:46 +00:00
Philip Pfaffe 33aef072c1 Fix r309826: Appease clang-format check.
llvm-svn: 309853
2017-08-02 18:26:48 +00:00
Singapuram Sanjay Srivallabh 1f9ab16c4e Fix code format on r309826
Summary:
Fix code format on r309826 / D35458

Reviewers: grosser, bollu

Reviewed By: grosser

Subscribers: pollydev

Tags: #polly

Differential Revision: https://reviews.llvm.org/D36232

llvm-svn: 309845
2017-08-02 17:56:39 +00:00
Philip Pfaffe 8f1872fb27 Fix r309826: Move intantiation and specialization of OwningScopAnalysisManagerFunctionProxy to the polly namespace.
When compiling with clang, explicit instantiation of the
OwningScopAnalysisManagerFunctionProxy needs to happen within the polly
namespace. Same goes with the specialization of its run method.

llvm-svn: 309835
2017-08-02 17:25:45 +00:00
Philip Pfaffe a70e2649ab [Polly][PM][WIP] Polly pass registration
Summary:
This patch is a first attempt at registering Polly passes with the LLVM tools. Tool plugins are still unsupported, but this registration is usable from the tools if Polly is linked into them (albeit requiring minimal patches to those tools). Registration requires a small amount of machinery (the owning analysis proxies), necessary for injecting ScopAnalysisManager objects into the calling tools.

This patch is marked WIP because the registration is incomplete. Parsing manual pipelines is fully supported, but default pass injection into the O3 pipeline is lacking, mostly because there is opportunity for some redesign here, I believe. The first point of order would be insertion points. I think it makes sense to run before the vectorizers. Running Polly Early, however, is weird. Mostly because it actually is the default (which to me is unexpected), and because Polly runs it's own O1 pipeline. Why not instead insert it at an appropriate place somewhere after simplification happend? Running after the loop optimizers seems intuitive, but it also seems wasteful, since multiple consecutive loops might well be a single scop, and we don't need to run for all of them.

My second request for comments would be regarding all those smallish helper passes we have,  like PollyViewer, PollyPrinter, PollyImportJScop. Right now these are controlled by command line options, deciding whether they should be part of the Polly pipeline. What is your opinion on treating them like real passes, and have the user write an appropriate pipeline if they want to use any of them?

Reviewers: grosser, Meinersbur, bollu

Reviewed By: grosser

Subscribers: llvm-commits, pollydev

Tags: #polly

Differential Revision: https://reviews.llvm.org/D35458

llvm-svn: 309826
2017-08-02 15:52:25 +00:00
Singapuram Sanjay Srivallabh 188053af5e Remove debug metadata from copied instruction to prevent GPUModule verification failure
Summary:
**Remove debug metadata from instruction to be copied to prevent the source file's debug metadata being copied into GPUModule and eventually failing Module verification and ASM string codegeneration.**

When copying the instruction onto the Module meant for the GPU, debug metadata attached to an instruction causes all related metadata to be pulled into the Module, including the DICompileUnit, which is not listed in llvm.dbg.cu of the Module. This fails the verification of the Module and generation of the ASM string.

The only debug metadata of the instruction, the DebugLoc, is unset by this patch.

This patch reattempts https://reviews.llvm.org/D35630 by targeting only those instructions that are to end up in a Module meant for the GPU.

Reviewers: grosser, bollu

Reviewed By: grosser

Subscribers: pollydev

Tags: #polly

Differential Revision: https://reviews.llvm.org/D36161

llvm-svn: 309822
2017-08-02 15:20:07 +00:00
Philip Pfaffe f081ec7609 [PM] Fix proxy invalidation
Summary: I made a mistake in handling transitive invalidation of analysis results. I've updated the list of preserved analyses as well as the correct result dependences.

The Invalidator passed through the invalidate() path can be used to
transitively invalidate analyses. It frequently happens that analysis
results depend on other analyses, and thus store references to their
results. When the dependee now gets invalidated, the depender needs to
be invalidated as well. This is the purpose of the Invalidator object,
which can be used to check whether some dependee analysis is in the
process of being invalidated. I originally was checking the wrong
dependee analyses, which is an actual error, you can only check analysis
results that are in the cache (which they are if you've captured their
reference). The invalidation I'm handling inside the proxy deals with
the standard analyses the proxy passes into the Scop pipeline, since I'm
capturing their reference.

This checking allows us to actually preserve a couple of results outside
of the proxy, since the Scop pipeline shouldn't break those, or
otherwise should update them accordingly.

Reviewers: grosser, Meinersbur, bollu

Reviewed By: grosser

Subscribers: pollydev, llvm-commits

Differential Revision: https://reviews.llvm.org/D36216

llvm-svn: 309811
2017-08-02 13:18:49 +00:00
Siddharth Bhat f23bb4a8ba [GPUJIT] Add GPUJIT APIs for allocating and freeing managed memory.
We introduce `polly_mallocManaged` and `polly_freeManaged` as
proxies for `cudaMallocManaged` / `cudaFree`. This is currently not
used by Polly. It is auxiliary code that is used in `COSMO`.

This is useful because `polly_mallocManaged` matches the signature of `malloc`,
while `cudaMallocManaged` does not. We introduce `polly_freeManaged` for
symmetry.

We use this in COSMO to use the unified memory feature of the newer
CUDA APIs (>= 6).

Differential Revision: https://reviews.llvm.org/D35991

llvm-svn: 309808
2017-08-02 12:23:22 +00:00
Philip Pfaffe ead67dbbd6 [SI][NewPM] Collect loop count statistics
llvm-svn: 309807
2017-08-02 11:14:41 +00:00
Philip Pfaffe f5a4394ad6 [SD] Set PollyUseRuntimeAliasChecks correctly
llvm-svn: 309805
2017-08-02 11:08:01 +00:00
Siddharth Bhat b1a52abd87 [GPUJIT] Teach GPUJIT to use a pre-existing CUDA context if available.
On mixing the driver and runtime APIs, it is quite possible that a
context already exists due to runtime API usage. In this case, Polly should
try to use the same context.

This patch teaches GPUJIT to detect that a context exists and how to
pick up this context.

Without this, calling `cudaMallocManaged`, for example, before a
polly-generated kernel launch causes P100 to *hang*.

This is a part of (https://reviews.llvm.org/D35991) that was extracted
out.

Differential Revision: https://reviews.llvm.org/D36162

llvm-svn: 309802
2017-08-02 09:19:42 +00:00
Michael Kruse fd35089689 [ForwardOpTree] Execute canForwardTree also in release builds.
Commit r309730 moved the call to canForwardTree into an assert(), even
though this function has side-effects if its DoIt parameter is true. To
avoid a warning in release builds, do an (void)Execution of its result
instead.

To avoid such confusion in the future, rename
canForwardTree() to forwardTree().

llvm-svn: 309753
2017-08-01 22:15:04 +00:00
Michael Kruse bc88a78cb4 [Simplify] Rewrite redundant write detection algorithm.
The previous algorithm was to search a writes and the sours of its value
operand, and see whether the write just stores the same read value back,
which includes a search whether there is another write access between
them. This is O(n^2) in the max number of accesses in a statement
(+ the complexity of isl comparing the access functions).

The new algorithm is more similar to the one used for searching for
overwrites and coalescable writes. It scans over all accesses in order
of execution while tracking which array elements still have the same
value since it was read. This is O(n), not counting the complexity
within isl. It should be more reliable than trying to catch all
non-conforming cases in the previous approach. It is also less code.

We now also support if the write is a partial write of the read's
domain, and to some extent non-affine subregions.

Differential Revision: https://reviews.llvm.org/D36137

llvm-svn: 309734
2017-08-01 20:01:34 +00:00
Reid Kleckner 859c1e606a Silence -Wunused-variable warning in NDEBUG builds
llvm-svn: 309730
2017-08-01 19:53:01 +00:00
Michael Kruse 693ef99935 [Simplify] Improve scalability.
With a lot of reads and writes to the same array in a statement,
some isl sets that capture the state between access can become
complex such that isl takes more considerable time and memory
for operations on them.

The problems identified were:

- is_subset() takes considerable time with many disjoints in the
  arguments. We limit the number of disjoints to 4, any additional
  information is thrown away.

- subtract() can lead to many disjoints. We instead assume that any
  array element is possibly accessed, which removes all disjoints.

- subtract_domain() may lead to considerable processing, even if all
  elements are are to be removed. Instead, we remove determine and
  remove the affected spaces manually. No behaviour is changed.

llvm-svn: 309728
2017-08-01 19:39:11 +00:00
Tobias Grosser e327eebccb Update to isl-0.18-809-gd5b4535
This fixes some undefined behavior in the isl schedule tree code.

llvm-svn: 309727
2017-08-01 19:37:50 +00:00
Siddharth Bhat 1ec9cba4e3 [NFC] Add 'REQUIRES: pollyacc' on 'test/GPGPU/invariant-load-hoisting-of-array.ll'
- Should fix broken build due to `r309681`.

llvm-svn: 309686
2017-08-01 14:52:18 +00:00
Siddharth Bhat 442e722c1e [GPUJIT] Call `cuProfilerStop` before destroying the context to flush profiler cache.
This is necessary to get accurate traces from `nvprof` / `nvcc`.
Otherwise, we lose some profiling information.

Differential Revision: https://reviews.llvm.org/D35940

llvm-svn: 309682
2017-08-01 14:36:24 +00:00
Siddharth Bhat edf9581e4c [PPCGCodeGeneration] Correct usage of llvm::Value with getLatestValue.
It is possible that the `HostPtr` that coresponds to an array could be
invariant load hoisted. Make sure we use the invariant load hoisted
value by using `IslNodeBuilder::getLatestValue`.

Differential Revision: https://reviews.llvm.org/D36001

llvm-svn: 309681
2017-08-01 14:26:39 +00:00
Siddharth Bhat f2cfd2a4db [NFC] [IslNodeBuilder, GPUNodeBuilder] Unify mechanism for looking up replacement Values.
We populate `IslNodeBuilder::ValueMap` which contains replacements for
`llvm::Value`s. There was no simple method to pick up a replacement if
it exists, otherwise fall back to the original.

Create a method `IslNodeBuilder::getLatestValue` which provides this
functionality.

This will be used in a later patch to fix bugs in `PPCGCodeGeneration`
where the latest value is not being used.

Differential Revision: https://reviews.llvm.org/D36000

llvm-svn: 309674
2017-08-01 12:15:51 +00:00
Siddharth Bhat 4d5820d171 [NFC] [PPCGCodeGeneration] Convert GPUNodeBuilder::getGridSizes to isl++.
llvm-svn: 309671
2017-08-01 10:45:41 +00:00
Siddharth Bhat ccbf4b509c [NFC] [PPCGCodeGeneration] Convert GPUNodeBuilder::getArrayOffset to isl++.
llvm-svn: 309669
2017-08-01 09:58:55 +00:00
Michael Kruse 9f6e41cdba [ForwardOpTree] Support synthesizable values.
This allows -polly-optree to move instructions that depend on
synthesizable values.

The difficulty for synthesizable values is that their value depends on
the location. When it is moved over a loop header, and the SCEV
expression depends on the loop induction variable (SCEVAddRecExpr), it
would use the current induction variable instead of the last one.

At the moment we cannot forward PHI nodes such that crossing the header
of loops referenced by SCEVAddRecExpr is not possible (assuming the loop
header has at least two incoming blocks: for entering the loop and the
backedge, such any instruction to be forwarded must have a phi between
use and definition).

A remaining issue is when the forwarded value is used after the loop,
but is only synthesizable inside the loop. This happens e.g. if
ScalarEvolution is unable to determine the number of loop iterations or
the initial loop value. We do not forward in this situation.

Differential Revision: https://reviews.llvm.org/D36102

llvm-svn: 309609
2017-07-31 19:46:21 +00:00
Michael Kruse 57cc92b790 [Simplify] Remove all kinds of redundant scalar writes.
In addition to array and PHI writes, also allow scalar value writes.
The only kind of write not allowed are writes by functions
(including memcpy/memmove/memset).

llvm-svn: 309582
2017-07-31 17:04:55 +00:00
Tobias Grosser 8fc6cdfb1c [GPGPU] Add support for NVIDIA libdevice
Summary:
This allows us to map functions such as exp, expf, expl, for which no
LLVM intrinsics exist. Instead, we link to NVIDIA's libdevice which provides
high-performance implementations of a wide range of (math) functions. We
currently link only a small subset, the exp, cos and copysign functions. Other
functions will be enabled as needed.

Reviewers: bollu, singam-sanjay

Reviewed By: bollu

Subscribers: tstellar, tra, nemanjai, pollydev, mgorny, llvm-commits, kbarton

Tags: #polly

Differential Revision: https://reviews.llvm.org/D35703

llvm-svn: 309560
2017-07-31 14:03:16 +00:00
Tobias Grosser 39977e4e76 Revert "Remove Debug metadata from copied instruction to prevent Module verification failure"
This reverts commit r309490 as it triggers on our AOSP buildbut error messages
of the form:

inlinable function call in a function with debug info must have a !dbg location

llvm-svn: 309556
2017-07-31 11:43:38 +00:00
Tobias Grosser 7639db8ed9 [IslNodeBuilder] Remove unused instruction
Suggested-by: Maximilian Falkenstein <falkensm@student.ethz.ch>
llvm-svn: 309533
2017-07-31 01:59:23 +00:00
Singapuram Sanjay Srivallabh cf9a813368 Remove Debug metadata from copied instruction to prevent Module verification failure
Summary:
**Remove debug metadata from instruction to be copied to prevent the source file's debug metadata being copied into GPUModule and eventually failing Module verification and ASM string codegeneration.**

When copying the instruction onto the Module meant for the GPU, debug metadata attached to an instruction causes all related metadata to be pulled into the Module, including the DICompileUnit, which is not listed in llvm.dbg.cu of the Module. This fails the verification of the Module and generation of the ASM string.

The only debug metadata of the instruction, the DebugLoc, is unset by this patch.

Reviewers: grosser, bollu, Meinersbur

Reviewed By: grosser, bollu

Subscribers: pollydev

Tags: #polly

Differential Revision: https://reviews.llvm.org/D35630

llvm-svn: 309490
2017-07-29 18:03:49 +00:00
Michael Kruse ce9617f4fe [Simplify] Implement write accesses coalescing.
Write coalescing combines write accesses that

- Write the same llvm::Value.
- Write to the same array.
- Unless they do not write anything in a statement instance (partial
  writes), write to the same element.
- There is no other access between them that accesses the same element.

This is particularly useful after DeLICM, which leaves partial writes to
disjoint domains.

Differential Revision: https://reviews.llvm.org/D36010

llvm-svn: 309489
2017-07-29 16:21:16 +00:00
Michael Kruse 4335c3992a [test] Add test case for -polly-simplify. NFC.
llvm-svn: 309458
2017-07-29 00:06:06 +00:00
Michael Kruse 8e41d2baab [Simplify] Do not remove dependencies of phis within region stmts.
These were wrongly assumed to be phi nodes that require
MemoryKind::PHI accesses.

llvm-svn: 309454
2017-07-28 23:22:32 +00:00
Michael Kruse fd7f40961b [VirtualInstruction] Do not iterate over a region statement's instruction list. NFC.
It should be empty anyways. In this case it would even be redundant
because we just all all instructions in region statements.

llvm-svn: 309453
2017-07-28 23:22:23 +00:00
Adrian Prantl 99c4a5fb8e Remove offset parameter from llvm.dbg.value intrinsics in testcase
llvm-svn: 309433
2017-07-28 21:08:53 +00:00
Michael Kruse 0137d80ad4 [VirtualInstruction] Remove assertion. NFC.
ScopStmt::contains is currently implemented on the basis of BasicBlock
and does not take the instruction list into account. Therefore any
instruction copied by -polly-optree into another statement currently
triggers that assertion.

Remove that assertion for now. We might re-enable it when the
implementation of ScopStmt::contains changes.

llvm-svn: 309421
2017-07-28 19:26:24 +00:00
Michael Kruse c99209b4b2 [test] Fix typo in filename. NFC.
llvm-svn: 309403
2017-07-28 16:57:56 +00:00
Michael Kruse 6c8f91b908 [Simplify] Fix typo in statistics output. NFC.
llvm-svn: 309402
2017-07-28 16:57:51 +00:00
Michael Kruse 34a77780c5 [Simplify] Remove empty partial accesses first. NFC.
So follow-up cleanup do not need special handling for such accesses.

llvm-svn: 309401
2017-07-28 16:57:45 +00:00
Siddharth Bhat 4ebeb3568a [PPCGCodeGeneration] Check that invariant load hoisting succeeded.
If we fail, throw an error for now. We can gracefully handle this later.

llvm-svn: 309387
2017-07-28 14:48:32 +00:00
Siddharth Bhat 0a1177b58e [ScopDetect] add `-polly-ignore-func` flag to ignore functions by name.
Ignore all functions whose name match a regex. Useful because creating a
regex that does *not* match a string is somewhat hard.

Example:
https://stackoverflow.com/questions/1240275/how-to-negate-specific-word-in-regex

llvm-svn: 309377
2017-07-28 11:47:24 +00:00
Tobias Grosser c0678c016f Add missing namespace comment
llvm-svn: 309373
2017-07-28 09:33:06 +00:00
Tobias Grosser 25271b91b2 [GPGPU] Do not require the Scop::Context to have information about all parameters
llvm-svn: 309368
2017-07-28 06:49:44 +00:00
Tobias Grosser 30caae6d23 [GPGPU] Fix compilation issue with latest CUDA upgrade to i128
llvm-svn: 309366
2017-07-28 06:38:49 +00:00
Hans Wennborg ce99589225 Tiny docs fix
llvm-svn: 309300
2017-07-27 18:14:00 +00:00
Tobias Grosser adcbee5433 Update isl to isl-0.18-800-g4018f45
This fixes a bug in isl_flow where triggering the compute out could result in
undefined or unexpected behavior. This fixes some recent regressions we saw
in the android buildbots. Thanks Eli Friedman for reducing the corresponding
test cases.

llvm-svn: 309274
2017-07-27 14:48:02 +00:00
Michael Kruse a508a4e619 [ScopBuilder/Simplify] Refactor isEscaping. NFC.
ScopBuilder and Simplify (through VirtualInstruction.cpp) previously
used this functionality in their own implementation. Refactor them
both into a common one into the Scop class.

BlockGenerator also makes use of a similiar functionality, but also
records outside users and takes place after region simplification.
Merging it as well would be more complicated.

llvm-svn: 309273
2017-07-27 14:39:52 +00:00
Michael Kruse 8a8aca4299 [Simplify] Count PHINodes in simplifiable exit nodes as escaping use.
After region exit simplification, the incoming block of a phi node in
the SCoP region's exit block lands outside of the region. Since we
treat SCoPs as if this already happened, we need to account for that
when looking for outside uses of scalars (i.e. escaping scalars).

llvm-svn: 309271
2017-07-27 14:09:31 +00:00
Michael Kruse eca86cee64 [ScopInfo] Never print instruction list of region stmts.
A region statement's instruction list is always empty and ignored by the code
generator. Don't give the impression that it means anything.

llvm-svn: 309197
2017-07-26 22:01:33 +00:00
Michael Kruse cedd7a74e1 [Simplify] Do not setInstructions() of region stmts. NFC.
The instruction list is ignored for region statements, there
is no reason to set it.

llvm-svn: 309196
2017-07-26 22:01:28 +00:00
Michael Kruse 95b39da8ae [Simplify] Fix invalid removal write for escaping values.
A PHI node's incoming block is the user of its operand, not the PHI's parent.

Assuming the PHINode's parent being the user lead to the removal of a
MemoryAccesses because its use was assumed to be inside of the SCoP.

llvm-svn: 309164
2017-07-26 19:58:15 +00:00
Roman Gareev 2e580538be [ScheduleOptimizer] Translate to C++ bindings
Translate the ScheduleOptimizer to use the new isl C++ bindings.

Reviewed-by: Michael Kruse <llvm@meinersbur.de>

Differential Revision: https://reviews.llvm.org/D35845

llvm-svn: 309119
2017-07-26 14:59:15 +00:00
Michael Kruse 1df1aac014 [ScopInfo] Avoid use of getStmtFor(BB). NFC.
Since there will be no more a 1:1 correspondence between statements and
basic blocks, we would like to get rid of the method getStmtFor(BB)
and its uses. Here we remove one of its uses in ScopInfo by fetching
the statement in which the call instruction lies.

Contributed-by: Nandini Singhal <cs15mtech01004@iith.ac.in>

Differential Revision: https://reviews.llvm.org/D35691

llvm-svn: 309110
2017-07-26 13:25:28 +00:00
Michael Kruse 11ed062258 [SCEVValidator] Loop exit values of loops before the SCoP are synthesizable.
In the following loop:

   int i;
   for (i = 0; i < func(); i+=1)
     ;
SCoP:
   for (int j = 0; j<n; j+=1)
     S(i, j)

The value i is synthesizable in the SCoP that includes only the j-loop.
This is because i is fixed within the SCoP, it is irrelevant whether
it originates from another loop.

This fixes a strange case where a PHI was synthesiable in a SCoP,
but not its incoming value, triggering an assertion.

This should fix MultiSource/Applications/sgefa/sgefa of the
perf-x86_64-penryn-O3-polly-before-vectorizer-unprofitable buildbot.

llvm-svn: 309109
2017-07-26 13:05:45 +00:00
Tobias Grosser 9ddcf8e6ac Revert accidental isl changes in 308923
It seems I still had some incomplete changes in the tree when committing.
In general, we only import changes from isl upstream. In this case, the
changes were especially unfortunate, as they broke the error management
in isl_flow.c and consequently caused regressions.

Thanks to Michael Kruse for spotting this mistake.

llvm-svn: 309039
2017-07-25 22:15:47 +00:00
Michael Kruse 8d89179e33 [ScopInfo] Rename ScopStmt::contains(BB) to represents(BB). NFC.
In future, there will be no more a 1:1 correspondence between statements
and basic blocks, the name `contains` does not correctly capture their
relationship. A BB may infact comprise of multiple statements; hence we
describe a statement 'representing' a basic block.

Differential Revision: https://reviews.llvm.org/D35838

llvm-svn: 308982
2017-07-25 16:25:37 +00:00
Philip Pfaffe 85cc5687df [IslAst] Untangle IslAst lit-testcases from specifics of the legacy-PM
Summary:
This consists instances of two changes:

- Accept any order of checks for a specific loop form, that appear in different order in the new vs legacy-PM.
- Remove checks for specific regions.

Reviewers: grosser

Reviewed By: grosser

Subscribers: pollydev, llvm-commits

Tags: #polly

Differential Revision: https://reviews.llvm.org/D35837

llvm-svn: 308976
2017-07-25 15:07:42 +00:00
Michael Kruse b6317007b4 [ScopInfo] Fix assertion for PHIs not in a region stmts entry.
A PHI node within a region statement is legal, but does not have
a MemoryKind::PHI access.

llvm-svn: 308973
2017-07-25 13:28:39 +00:00
Siddharth Bhat 43f178bbc9 [PPCGCodeGeneration] Skip arrays with empty extent.
Invariant load hoisted scalars, and arrays whose size we can statically compute
to be 0 do not need to be allocated as arrays.

Invariant load hoisted scalars are sent to the kernel directly as parameters.

Earlier, we used to allocate `0` bytes of memory for these because our
computation of size from `PPCGCodeGeneration::getArraySize` would result in `0`.

Now, since we don't invariant loads as arrays in PPCGCodeGeneration, this
problem does not occur anymore.

Differential Revision: https://reviews.llvm.org/D35795

llvm-svn: 308971
2017-07-25 12:35:36 +00:00