1793 Commits

Author SHA1 Message Date
Johannes Doerfert
45be64464b [NFC] Consistenly use commented and annotated ScopPass functions
The changes affect methods that are part of the Pass interface and
  include:
    - Comments that describe the methods purpose.
    - A consistent use of the keywords override and virtual.
  Additionally, the printScop method is now optional and removed from
  SCoP passes that do not implement it.

llvm-svn: 248685
2015-09-27 15:43:29 +00:00
Johannes Doerfert
0f37630849 [NFC] Use releaseMemory to release internal memory
llvm-svn: 248684
2015-09-27 15:42:28 +00:00
Tobias Grosser
0ff79e586d BlockGenerator: Simplify code generated for region statements
After having generated a new user statement a couple of inefficient or
trivially dead instructions may remain. This commit runs instruction
simplification over the newly generated blocks to ensure unneeded
instructions are removed right away.

This commit does adds simplification for non-affine subregions which was not
yet part of 248681.

llvm-svn: 248683
2015-09-27 11:35:00 +00:00
Tobias Grosser
412f9774f8 [CodeGen test] Replace undef values with some defined constants
Otherwise, part of the computation will be just simplified away when we add
instruction simplification support to the RegionGenerator.

llvm-svn: 248682
2015-09-27 11:34:53 +00:00
Tobias Grosser
1b9d25a42d BlockGenerator: Simplify code generated for scop statements
After having generated a new user statement a couple of inefficient or trivially
dead instructions may remain. This commit runs instruction simplification over
the newly generated blocks to ensure unneeded instructions are removed right
away.

This commit does not yet add simplification for non-affine subregions.

llvm-svn: 248681
2015-09-27 11:17:22 +00:00
Johannes Doerfert
284e67828f [NFC] Remove obsolete member
llvm-svn: 248675
2015-09-26 20:58:29 +00:00
Johannes Doerfert
fb19dd694c Create parallel code in a separate block
This commit basically reverts r246427 but still solves the issue
  tackled by that commit. Instead of emitting initialization code in the
  beginning of the start block we now generate parallel code in its own
  block and thereby guarantee separation. This is necessary as we cannot
  generate code for hoisted loads prior to the start block but it still
  needs to be placed prior to everything else.

llvm-svn: 248674
2015-09-26 20:57:59 +00:00
Michael Kruse
caac2b6930 Fix typo
llvm-svn: 248670
2015-09-26 15:51:44 +00:00
Tobias Grosser
06c495c2b0 Add test case from llvm.org/PR17187
The new domain construction algorithm now correctly models this test case (and
derives an empty run-time condition). Add this test case to ensure we do not
regress.

llvm-svn: 248669
2015-09-26 14:27:54 +00:00
Johannes Doerfert
12155a9ef4 Add test case from open bug
The bug (15771) was fixed already with the new domain generation
  but the test case was not added till now.

llvm-svn: 248668
2015-09-26 14:03:29 +00:00
Johannes Doerfert
c6987c18de [FIX] Use the surrounding loop for non-affine SCoP regions
When the whole SCoP is a non-affine region we need to use the
  surrounding loop in the construction of the schedule as that is
  the one that will be looked up after the schedule generation.

  This fixes bug 24947

llvm-svn: 248667
2015-09-26 13:41:43 +00:00
Tobias Grosser
bbda083c75 Add test case for delinearization through bitcasts
This was forgotten in r247928

llvm-svn: 248663
2015-09-26 08:55:59 +00:00
Tobias Grosser
99c70dd8d1 Ensure memory accesses to the same array have identical dimensionality
When recovering multi-dimensional memory accesses, it may happen that different
accesses to the same base array are recovered with different dimensionality.
This patch ensures that the dimensionalities are unified by adding zero valued
dimensions to acesses with lower dimensionality. When starting to model
fixed-size arrays as multi-dimensional in 247906, this has not been taken
care of.

llvm-svn: 248662
2015-09-26 08:55:54 +00:00
Michael Kruse
8d0b734e71 Let MemoryAccess remember its purpose
There are three possible reasons to add a memory memory access: For explicit load and stores, for llvm::Value defs/uses, and to emulate PHI nodes (the latter two called implicit accesses). Previously MemoryAccess only stored IsPHI. Register accesses could be identified through the isScalar() method if it was no IsPHI. isScalar() determined the number of dimensions of the underlaying array, scalars represented by zero dimensions.

For the work on de-LICM, implicit accesses can have more than zero dimensions, making the distinction of isScalars() useless, hence now stored explicitly in the MemoryAccess. Instead, we replace it by isImplicit() and avoid the term scalar for zero-dimensional arrays as it might be confused with llvm::Value which are also often referred to as scalars (or alternatively, as registers).

No behavioral change intended, under the condition that it was impossible to create explicit accesses to zero-dimensional "arrays".

llvm-svn: 248616
2015-09-25 21:21:00 +00:00
Michael Kruse
33d6c0bbc5 Use per-Purpose overloads for MemoryAccess creation
This makes the intent of each created object clearer and allows to add more specific asserts. The bug fixed in r248535 has been discovered this way.

No functional change intended; everything should behave as before.

llvm-svn: 248603
2015-09-25 18:53:27 +00:00
Tobias Grosser
c90ed44b31 [cmake] terminate isl/stdint.h with a newline
This avoids warnings of the form:
  stdint.h:1:20: warning: no newline at end of file [-Wnewline-eof]

llvm-svn: 248570
2015-09-25 11:45:26 +00:00
Tobias Grosser
c2bb0cbe00 Sort includes using Chandler's sort_includes.py script
llvm-svn: 248568
2015-09-25 09:49:19 +00:00
Tobias Grosser
8016f3a4f5 Add missing PHI to test case
llvm-svn: 248563
2015-09-25 05:41:30 +00:00
Tobias Grosser
da95a4a7c7 Handle read-only scalars used in PHI-nodes correctly
This change addresses three issues:

  - Read only scalars that enter a PHI node through an edge that comes from
    outside the scop are not modeled any more, as such PHI nodes will always
    be initialized to this initial value right before the SCoP is entered.
  - For PHI nodes that depend on a scalar value that is defined outside the
    scop, but where the scalar values is passed through an edge that itself
    comes from a BB that is part of the region, we introduce in this basic
    block a read of the out-of-scop value to ensure it's value is available
    to write it into the PHI alloc location.
  - Read only uses of scalars by PHI nodes are ignored in the general read only
    handling code, as they are taken care of by the general PHI node modeling
    code.

llvm-svn: 248535
2015-09-24 20:59:59 +00:00
Michael Kruse
26ed65e00d Fix comparison signed/unsigned mismatch warning; NFC
llvm-svn: 248520
2015-09-24 17:32:49 +00:00
Tobias Grosser
2553ebc79d Mark a couple of issues as done
llvm-svn: 248492
2015-09-24 14:30:14 +00:00
Michael Kruse
b280ded108 Rename Polly_isl to PollyISL
Library names in the LLVM framework usually do not contain underscores.

llvm-svn: 248487
2015-09-24 12:38:49 +00:00
Michael Kruse
2d0ece960f Remove Analysis Output of TempScopInfo
After the merge of TempScopInfo into ScopInfo the analysis output 
remained because of the existing unit tests. These remains are removed 
and the units tests converted to match the equivalent output of 
ScopInfo's analysis output. The unit tests are also moved into the
directory of ScopInfo tests.

Differential Revision: http://reviews.llvm.org/D13116

llvm-svn: 248485
2015-09-24 11:41:21 +00:00
Michael Kruse
519b3cfd27 Compile ISL into its own library
Refactor all ISL-related cmake build instructions into its own 
CMakeLists.txt and build as a separate library.

This is useful to apply ISL-related build flags to ISL only and not to 
Polly's files. Also, it the separation of both projects becomes clearer.

Proposed name of the library is Polly_isl. It is not "isl" to avoid 
mix-up with potentially installed libisl.{a|so}.

Tested configurations:
- Windows with cmake 3.2
- Ubuntu with cmake 3.0.2 
- Ubuntu with cmake 3.0.2  BUILD_SHARED_LIBS on
- Ubuntu with cmake 2.8.12.2 (LLVM minimum version)
- Ubuntu out-of-LLVM-tree

Differential Revision: http://reviews.llvm.org/D12810

llvm-svn: 248484
2015-09-24 11:30:22 +00:00
Johannes Doerfert
e526de5a47 Make MIN_LOOP_TRIP_COUNT a static constant
llvm-svn: 248192
2015-09-21 19:10:11 +00:00
Tobias Grosser
b1c39429d9 Do not model delinearized and linearized access relation for a single access
A missing return statement that previously did not have a visibly negative
effect caused after some data-structure changes in r248024 multi-dimensional
accesses to be modeled both multi-dimensional as well as linearized. This
commit adds the missing return to avoid the incorrect double modeling as
well as the compile time increases it caused.

llvm-svn: 248171
2015-09-21 16:19:25 +00:00
Johannes Doerfert
6a72a2af13 Use <nsw> AddRecs in the affinator to avoid bounded assumptions
If we encounter a <nsw> tagged AddRec for a loop we know the trip count of
  that loop has to be bounded or the semantics is undefined anyway. Hence, we
  only need to add unbounded assumptions if no such AddRec is known.

llvm-svn: 248128
2015-09-20 16:59:23 +00:00
Johannes Doerfert
707a406078 Add bounded loop assumption
So far we ignored the unbounded parts of the iteration domain, however
  we need to assume they do not occure at all to remain sound if they do.

llvm-svn: 248126
2015-09-20 16:38:19 +00:00
Johannes Doerfert
f2cc86edae Simplify domain generation
We now add loop carried information during the second traversal of the
  region instead of in a intermediate step in-between. This makes the
  generation simpler, removes code and should even be faster.

llvm-svn: 248125
2015-09-20 16:15:32 +00:00
Johannes Doerfert
0c1123a831 [FIX] Repair test case that was unprofitable
llvm-svn: 248124
2015-09-20 16:14:41 +00:00
Johannes Doerfert
06c57b594c Allow loops with multiple back edges
In order to allow multiple back edges we:
    - compute the conditions under which each back edge is taken
    - build the union over all these conditions, thus the condition that
      any back edge is taken
    - apply the same logic to the union we applied to a single back edge

llvm-svn: 248120
2015-09-20 15:00:20 +00:00
Johannes Doerfert
7175bdfbe4 Add loop trip count based heuristic for SCoP detection
As we currently do not perform any optimizations that targets (or is
  even aware) small trip counts we will skip them when we count the
  loops in a region.

llvm-svn: 248119
2015-09-20 14:56:54 +00:00
Johannes Doerfert
b276bde162 [NFC] Remove obsolete diagnostic for unstructured control flow
llvm-svn: 248118
2015-09-20 14:55:50 +00:00
Michael Kruse
84f70acd68 Remove unused variable Dimension [NFC]
llvm-svn: 248026
2015-09-18 20:03:32 +00:00
Michael Kruse
e2bccbbfb2 Merge IRAccess into MemoryAccess
All MemoryAccess objects will be owned by ScopInfo::AccFuncMap which 
previously stored the IRAccess objects. Instead of creating new 
MemoryAccess objects, the already created ones are reused, but their 
order might be different now. Some fields of IRAccess and MemoryAccess 
had the same meaning and are merged.

This is the last step of fusioning TempScopInfo.{h|cpp} and 
ScopInfo.{h.cpp}. Some refactoring might still make sense.

Differential Revision: http://reviews.llvm.org/D12843

llvm-svn: 248024
2015-09-18 19:59:43 +00:00
Tobias Grosser
b09455dee0 Store EscapeMap as Value* instead of AllocInst
This currently does not change the behavior in Polly, but it allows us to later
also overwrite the EscapeMap with our GlobalMap.

llvm-svn: 247970
2015-09-18 06:01:11 +00:00
Tobias Grosser
6f36d9ab01 Delinearize multi-dimensional arrays through bitcasts
In some cases instcombine introduces bitcasts that slightly obfuscate the
multi-dimensionality of an array. This patch teaches our fixed-size
delinearization how to look through bitcasts.

llvm-svn: 247928
2015-09-17 20:16:21 +00:00
Tobias Grosser
0537f41de5 Do not use the assumed context in the dependence analysis any more
This information is implicitly available through the multi-dimensionality of
memory accesses. This reduces compile time for 3mm from 430ms to 400ms and
should generally benefit compile time for cases where the assumed context
is complex.

llvm-svn: 247907
2015-09-17 17:28:19 +00:00
Tobias Grosser
5fd8c0961e Model fixed-size multi-dimensional arrays if possible multi-dimensional
If the GEP instructions give us enough insights, model scalar accesses as
multi-dimensional (and generate the relevant run-time checks to ensure
correctness). This will allow us to simplify the dependence computation in
a subsequent commit.

llvm-svn: 247906
2015-09-17 17:28:15 +00:00
Tobias Grosser
faf8f6f62e Extract function that derives the index expressions of a GEP instruction [NFC]
We currently use this functionality to add run-time assumptions that check its
in-bound property.

llvm-svn: 247893
2015-09-17 15:47:52 +00:00
Tobias Grosser
e375d8058a Add option to enable/disable reduction usage in dependence analysis
llvm-svn: 247781
2015-09-16 09:50:17 +00:00
Johannes Doerfert
883f8c1d2f Use modulo semantic to generate non-integer-overflow assumptions
This will allow to generate non-wrap assumptions for integer expressions
  that are part of the SCoP. We compare the common isl representation of
  the expression with one computed with modulo semantic. For all parameter
  combinations they are not equal we can have integer overflows.

  The nsw flags are respected when the modulo representation is computed,
  nuw and nw flags are ignored for now.

  In order to not increase compile time to much, the non-wrap assumptions
  are collected in a separate boundary context instead of the assumed
  context. This helps compile time as the boundary context can become
  complex and it is therefor not advised to use it in other operations
  except runtime check generation. However, the assumed context is e.g.,
  used to tighten dependences. While the boundary context might help to
  tighten the assumed context it is doubtful that it will help in practice
  (it does not effect lnt much) as the boundary (or no-wrap assumptions)
  only restrict the very end of the possible value range of parameters.

  PET uses a different approach to compute the no-wrap context, though lnt runs
  have shown that this version performs slightly better for us.

llvm-svn: 247732
2015-09-15 22:52:53 +00:00
Johannes Doerfert
cef616fe2d Use blocks instead of domains in SCEVAffinator
Due to the new domain generation, the SCoP keeps track of the domain
  for all blocks, thus the SCEVAffinator can now work with blocks to avoid
  duplication of the domains.

llvm-svn: 247731
2015-09-15 22:49:04 +00:00
Johannes Doerfert
b20f151d56 Coalesce the constructed domains early
llvm-svn: 247728
2015-09-15 22:11:49 +00:00
Michael Kruse
da8d6203f4 Fix after renamed CMake cache entry
LLVM_EXTERNAL_CLANG_BUILD was changed to LLVM_TOOL_CLANG_BUILD in
r242059.

llvm-svn: 247675
2015-09-15 10:51:15 +00:00
Michael Kruse
fa62b1763c Run polly-check-format with unit tests
Add polly-check-format as dependency of check-polly if clang-format is 
available in the same build.

Differential Revision: http://reviews.llvm.org/D12850

llvm-svn: 247600
2015-09-14 19:11:48 +00:00
Michael Kruse
69f3788c36 Revise polly-{update|check}-format targets
Summary:
Make clang-format run on each file independently using
add_custom_format (instead using a shell script in utils/). The targets
polly-{update|check}-format depend on these.

The primary motivation is to make them work on Windows, but also
improves them generally:
- Each file update/check can run in parallel (Although they do not take
  long to run anyway)
- Implicit dependency on clang-format, so it recompiles if necessary
- polly-check-format shows the formatting difference if failing

Differential Revision: http://reviews.llvm.org/D12837

llvm-svn: 247581
2015-09-14 16:59:50 +00:00
Michael Kruse
2846877d88 Replace some SmallVector-typed parameters by ArrayRef
ArrayRef avoids making implementation details such as the number of stack elements to be part of the function signature.

llvm-svn: 247572
2015-09-14 15:45:33 +00:00
Tobias Grosser
aaadc5302c [test] Load Polly before using the polly flags
llvm-svn: 247551
2015-09-14 11:49:05 +00:00
Johannes Doerfert
334f9e87c6 [FIX] XFAIL test that depends on pending LLVM commit
llvm-svn: 247550
2015-09-14 11:45:34 +00:00