Commit Graph

343 Commits

Author SHA1 Message Date
Johannes Doerfert 764b7e66f0 [FIX] Require base pointers of loads that might alias to be hoisted
Since the base pointer of a possibly aliasing pointer might not alias
  with any other pointer it (the base pointer) might not be tagged as
  "required invariant". However, we need it do be in order to compare
  the accessed addresses of the derived (possibly aliasing) pointer.

  This patch also tries to clean up the load hoisting a little bit.

llvm-svn: 270412
2016-05-23 09:26:46 +00:00
Johannes Doerfert a61eda7698 [FIX] Let ScalarEvolution forget hoisted values
We have to rethink the handling of escaping values in order to make
  this kind of "fixes" go away.

llvm-svn: 270409
2016-05-23 09:02:54 +00:00
Johannes Doerfert 1a4ad8f771 [FIX] Synthezise Sdiv/Srem/Udiv instructions correctly.
This patch simplifies the Sdiv/Srem/Udiv expansion and thereby
  prevents errors, e.g., regarding the insertion point.

llvm-svn: 270408
2016-05-23 08:55:43 +00:00
Johannes Doerfert 6f1bb7a9d9 Support truncate operations
Truncate operations are basically modulo operations, thus we can model
  them that way. However, for large types we assume the operand to fit
  in the new type size instead of introducing a modulo with a very large
  constant.

llvm-svn: 269300
2016-05-12 15:13:49 +00:00
Johannes Doerfert 404a0f81ea Check overflows in RTCs and bail accordingly
We utilize assumptions on the input to model IR in polyhedral world.
  To verify these assumptions we version the code and guard it with a
  runtime-check (RTC). However, since the RTCs are themselves generated
  from the polyhedral representation we generate them under the same
  assumptions that they should verify. In other words, the guarantees
  that we try to provide with the RTCs do not hold for the RTCs
  themselves. To this end it is necessary to employ a different check
  for the RTCs that will verify the assumptions did hold for them too.

Differential Revision: http://reviews.llvm.org/D20165

llvm-svn: 269299
2016-05-12 15:12:43 +00:00
Johannes Doerfert 2640454d1c Refactor simplifySCoP [NFC]
Remove obsolete code and decrease the indention in the
  Scop::simplifySCoP() function.

llvm-svn: 269049
2016-05-10 12:19:47 +00:00
Tobias Grosser 1022ca5646 Codegen: Enable the detection of min/max expressions
Min/max expressions are easier to read and can in some cases also result in
more concise IR that is generated as the min/max --- when lowered to a
cmp+select pattern -- commonly has a simpler condition then the ternary
condition isl would normally generate.

llvm-svn: 268855
2016-05-07 08:03:44 +00:00
Tobias Grosser 7ec06a86c1 test: Use CHECK-NEXT to not miss instructions in test output
llvm-svn: 268854
2016-05-07 08:03:32 +00:00
Johannes Doerfert 172dd8b923 Allow unsigned divisions
After zero-extend operations and unsigned comparisons we now allow
  unsigned divisions. The handling is basically the same as for signed
  division, except the interpretation of the operands. As the divisor
  has to be constant in both cases we can simply interpret it as an
  unsigned value without additional complexity in the representation.
  For the dividend we could choose from the different representation
  schemes introduced for zero-extend operations but for now we will
  simply use an assumption.

llvm-svn: 268032
2016-04-29 11:53:35 +00:00
Johannes Doerfert 64c69f79fb [FIX] Prevent division/modulo by zero in parameters -- test case
This commits a test case for r268023.

llvm-svn: 268026
2016-04-29 10:45:39 +00:00
Tobias Grosser 947dbe3aae test: Make test case independent of earlier instructions
Instead of matching for %6, we use a regexp to match for the result strings.
This test case caused unrelated noise in http://reviews.llvm.org/D15722.

llvm-svn: 267875
2016-04-28 12:36:39 +00:00
Johannes Doerfert 8ab2803b63 [FIX] Propagate execution domain of invariant loads
If the base pointer of an invariant load is is loaded conditionally, that
  condition needs to hold for the invariant load too. The structure of the
  program will imply this for domain constraints but not for imprecisions in
  the modeling. To this end we will propagate the execution context of base
  pointers during code generation and thus ensure the derived pointer does
  not access an invalid base pointer.

llvm-svn: 267707
2016-04-27 12:49:11 +00:00
Johannes Doerfert c3596284c3 Model zext-extend instructions
A zero-extended value can be interpreted as a piecewise defined signed
  value. If the value was non-negative it stays the same, otherwise it
  is the sum of the original value and 2^n where n is the bit-width of
  the original (or operand) type. Examples:
    zext i8 127 to i32 -> { [127] }
    zext i8  -1 to i32 -> { [256 + (-1)] } = { [255] }
    zext i8  %v to i32 -> [v] -> { [v] | v >= 0; [256 + v] | v < 0 }

  However, LLVM/Scalar Evolution uses zero-extend (potentially lead by a
  truncate) to represent some forms of modulo computation. The left-hand side
  of the condition in the code below would result in the SCEV
  "zext i1 <false, +, true>for.body" which is just another description
  of the C expression "i & 1 != 0" or, equivalently, "i % 2 != 0".

    for (i = 0; i < N; i++)
      if (i & 1 != 0 /* == i % 2 */)
        /* do something */

  If we do not make the modulo explicit but only use the mechanism described
  above we will get the very restrictive assumption "N < 3", because for all
  values of N >= 3 the SCEVAddRecExpr operand of the zero-extend would wrap.
  Alternatively, we can make the modulo in the operand explicit in the
  resulting piecewise function and thereby avoid the assumption on N. For the
  example this would result in the following piecewise affine function:
  { [i0] -> [(1)] : 2*floor((-1 + i0)/2) = -1 + i0;
    [i0] -> [(0)] : 2*floor((i0)/2) = i0 }
  To this end we can first determine if the (immediate) operand of the
  zero-extend can wrap and, in case it might, we will use explicit modulo
  semantic to compute the result instead of emitting non-wrapping assumptions.

  Note that operands with large bit-widths are less likely to be negative
  because it would result in a very large access offset or loop bound after the
  zero-extend. To this end one can optimistically assume the operand to be
  positive and avoid the piecewise definition if the bit-width is bigger than
  some threshold (here MaxZextSmallBitWidth).

  We choose to go with a hybrid solution of all modeling techniques described
  above. For small bit-widths (up to MaxZextSmallBitWidth) we will model the
  wrapping explicitly and use a piecewise defined function. However, if the
  bit-width is bigger than MaxZextSmallBitWidth we will employ overflow
  assumptions and assume the "former negative" piece will not exist.

llvm-svn: 267408
2016-04-25 14:01:36 +00:00
Johannes Doerfert 517d8d2f94 Check only loop control of loops that are part of the region
This also removes a duplicated line of code in the region generator
  that caused a SPEC benchmark to fail with the new SCoPs.

llvm-svn: 267404
2016-04-25 13:37:24 +00:00
Tobias Grosser b99b97420f Update two more test cases for r266445+r266446 II
llvm-svn: 266475
2016-04-15 21:02:35 +00:00
Tobias Grosser 8af5e2f7bb Update two more test cases for r266445+r266446
llvm-svn: 266474
2016-04-15 20:56:17 +00:00
Mandeep Singh Grang 0aed05ae4b [Polly] Remove unwanted --check-prefix=CHECK from unit tests. NFC.
Summary:
Removed unwanted --check-prefix=CHECK from the following unit tests:
      DeadCodeElimination/dead_iteration_elimination.ll
      Isl/CodeGen/simple_vec_cast.ll

Patch by: Mandeep Singh Grang (mgrang)

Reviewers: jdoerfert, zinob, spop, grosser

Projects: #polly

Differential Revision: http://reviews.llvm.org/D19143

llvm-svn: 266411
2016-04-15 06:12:29 +00:00
Johannes Doerfert 615e0b85f8 Record wrapping assumptions early
Utilizing the record option for assumptions we can simplify the wrapping
  assumption generation a lot. Additionally, we can now report locations
  together with wrapping assumptions, though they might not be accurate yet.

llvm-svn: 266069
2016-04-12 13:28:39 +00:00
Johannes Doerfert 561d36b320 Allow pointer expressions in SCEVs again.
In r247147 we disabled pointer expressions because the IslExprBuilder did not
  fully support them. This patch reintroduces them by simply treating them as
  integers. The only special handling for pointers that is left detects the
  comparison of two address_of operands and uses an unsigned compare.

llvm-svn: 265894
2016-04-10 09:50:10 +00:00
Johannes Doerfert fbb63b8028 [FIX] Do not allow select as a base pointer in the SCoP region
llvm-svn: 265884
2016-04-09 21:57:13 +00:00
Johannes Doerfert b3410db2b7 [FIX] Do not recompute SCEVs but pass them to subfunctions
This reverts commit 2879c53e80e05497f408f21ce470d122e9f90f94.
  Additionally, it adds SDiv and SRem instructions to the set of values
  discovered by the findValues function even if we add the operands to
  be able to recompute the SCEVs. In subfunctions we do not want to
  recompute SDiv and SRem instructions but pass them instead as they
  might have been created through the IslExprBuilder and are more
  complicated than simple SDiv/SRem instructions in the code.

llvm-svn: 265873
2016-04-09 14:30:11 +00:00
Johannes Doerfert 5155edc658 [FIX] Teach the ScopExpander about parallel subfunctions
llvm-svn: 265824
2016-04-08 18:16:58 +00:00
Johannes Doerfert 41cda15940 [FIX] Allow to lookup domains for non-affine subregion blocks
llvm-svn: 265779
2016-04-08 10:32:26 +00:00
Johannes Doerfert 7b81103589 [FIX] Look through div & srem instructions in SCEVs
The findValues() function did not look through div & srem instructions
  that were part of the argument SCEV. However, in different other
  places we already look through it. This mismatch caused us to preload
  values in the wrong order.

llvm-svn: 265775
2016-04-08 10:25:58 +00:00
Johannes Doerfert 57c5f0b1c4 [FIX] Ensure SAI objects for exit PHIs
If all exiting blocks of a SCoP are error blocks and therefor not
  represented we will not generate accesses and consequently no SAI
  objects for exit PHIs. However, they are needed in the code generation
  to generate the merge PHIs between the original and optimized region.
  With this patch we enusre that the SAI objects for exit PHIs exist
  even if all exiting blocks turn out to be eror blocks.

  This fixes the crash reported in PR27207.

llvm-svn: 265393
2016-04-05 13:44:21 +00:00
Johannes Doerfert 642594ae87 Exploit graph properties during domain generation
As a CFG is often structured we can simplify the steps performed during
  domain generation. When we push domain information we can utilize the
  information from a block A to build the domain of a block B, if A dominates B
  and there is no loop backede on a path from A to B. When we pull domain
  information we can use information from a block A to build the domain of a
  block B if B post-dominates A. This patch implements both ideas and thereby
  simplifies domains that were not simplified by isl. For the FINAL basic block
  in test/ScopInfo/complex-successor-structure-3.ll we used to build a universe
  set with 81 basic sets. Now it actually is represented as universe set.

  While the initial idea to utilize the graph structure depended on the
  dominator and post-dominator tree we can use the available region
  information as a coarse grained replacement. To this end we push the
  region entry domain to the region exit and pull it from the region
  entry for the region exit if applicable.

  With this patch we now successfully compile
    External/SPEC/CINT2006/400_perlbench/400_perlbench
  and
    SingleSource/Benchmarks/Adobe-C++/loop_unroll.

Differential Revision: http://reviews.llvm.org/D18450

llvm-svn: 265285
2016-04-04 07:57:39 +00:00
Johannes Doerfert 6ba927148d [FIX] Adjust the insert point for non-affine region PHIs
If a non-affine region PHI is generated we should not move the insert
  point prior to the synthezised value in the same block as we might
  split that block at the insert point later on. Only if the incoming
  value should be placed in a different block we should change the
  insertion point.

llvm-svn: 265132
2016-04-01 11:25:47 +00:00
Tobias Grosser 6deba4ea03 Revert 264782 and 264789
These caused LNT failures due to new assertions when running with
-polly-position=before-vectorizer -polly-process-unprofitable for:

FAIL: clamscan.compile_time
FAIL: cjpeg.compile_time
FAIL: consumer-jpeg.compile_time
FAIL: shapes.compile_time
FAIL: clamscan.execution_time
FAIL: cjpeg.execution_time
FAIL: consumer-jpeg.execution_time
FAIL: shapes.execution_time

The failures have been introduced by r264782, but r264789 had to be reverted
as it depended on the earlier patch.

llvm-svn: 264885
2016-03-30 18:18:31 +00:00
Johannes Doerfert a144fb148b Exploit graph properties during domain generation
As a CFG is often structured we can simplify the steps performed
  during domain generation. When we push domain information we can
  utilize the information from a block A to build the domain of a
  block B, if A dominates B. When we pull domain information we can
  use information from a block A to build the domain of a block B
  if B post-dominates A. This patch implements both ideas and thereby
  simplifies domains that were not simplified by isl. For the FINAL
  basic block in
    test/ScopInfo/complex-successor-structure-3.ll .
  we used to build a universe set with 81 basic sets. Now it actually is
  represented as universe set.

  While the initial idea to utilize the graph structure depended on the
  dominator and post-dominator tree we can use the available region
  information as a coarse grained replacement. To this end we push the
  region entry domain to the region exit and pull it from the region
  entry for the region exit.

Differential Revision: http://reviews.llvm.org/D18450

llvm-svn: 264789
2016-03-29 21:31:05 +00:00
Michael Kruse 88a2256a34 Revert "[ScopInfo] Fix domains after loops."
This reverts commit r264118. The approach is still under discussion.

llvm-svn: 264705
2016-03-29 07:50:52 +00:00
Tobias Grosser 37034db826 Update to isl-0.16.1-145-g243bf7c
Just an import to keep track with the latest version of isl. We are not looking
for specific features.

llvm-svn: 264452
2016-03-25 19:38:18 +00:00
Johannes Doerfert 549768c01a [FIX] Verify the alias group before returning it
Similar to r262612 we need to check not only the pointer SCEV and the
type of an alias group but also the actual access instruction. The
reason is again the same: The pointer SCEV is not flow sensitive but the
access function is. In r262612 we avoided consolidating alias groups
even though the pointer SCEV and the type were the same but the access
function was not. Here it is simpler as we can simply check all members
of an alias group against the given access instruction.

llvm-svn: 264274
2016-03-24 13:22:16 +00:00
Tobias Grosser 25e8ebe29d Drop explicit -polly-delinearize parameter
Delinearization is now enabled by default and does not need to explicitly need
to be enabled in our tests.

llvm-svn: 264154
2016-03-23 13:21:02 +00:00
Tobias Grosser bfb6a9683b Codegen:Do not invalidate dominator tree when bailing out during code generation
When codegenerating invariant loads in some rare cases we cannot generate code
and bail out. This change ensures that we maintain a valid dominator tree
in these situations. This fixes llvm.org/PR26736

Contributed-by: Matthias Reisinger <d412vv1n@gmail.com>
llvm-svn: 264142
2016-03-23 06:57:51 +00:00
Michael Kruse 49a59ca093 [ScopInfo] Fix domains after loops.
ISL can conclude additional conditions on parameters from restrictions
on loop variables. Such conditions persist when leaving the loop and the
loop variable is projected out. This results in a narrower domain for
exiting the loop than entering it and is logically impossible for
non-infinite loops.

We fix this by not adding a lower bound i>=0 when constructing BB
domains, but defer it to when also the upper bound it computed, which
was done redundantly even before this patch.

This reduces the number of LNT fails with -polly-process-unprofitable
-polly-position=before-vectorizer from 8 to 6.

llvm-svn: 264118
2016-03-22 23:27:42 +00:00
Michael Kruse faedfcbf6d [BlockGenerator] Fix PHI merges for MK_Arrays.
Value merging is only necessary for scalars when they are used outside
of the scop. While an array's base pointer can be used after the scop,
it gets an extra ScopArrayInfo of type MK_Value. We used to generate
phi's for both of them, where one was assuming the reault of the other
phi would be the original value, because it has already been replaced by
the previous phi. This resulted in IR that the current IR verifier
allows, but is probably illegal.

This reduces the number of LNT test-suite fails with
-polly-position=before-vectorizer -polly-process-unprofitable
from 16 to 10.

Also see llvm.org/PR26718.

llvm-svn: 262629
2016-03-03 17:20:43 +00:00
Michael Kruse c7e0d9c216 Fix non-synthesizable loop exit values.
Polly recognizes affine loops that ScalarEvolution does not, in
particular those with loop conditions that depend on hoisted invariant
loads. Check for SCEVAddRec dependencies on such loops and do not
consider their exit values as synthesizable because SCEVExpander would
generate them as expressions that depend on the original induction
variables. These are not available in generated code.

llvm-svn: 262404
2016-03-01 21:44:06 +00:00
Johannes Doerfert 066dbf3f8e Track assumptions and restrictions separatly
In order to speed up compile time and to avoid random timeouts we now
  separately track assumptions and restrictions. In this context
  assumptions describe parameter valuations we need and restrictions
  describe parameter valuations we do not allow. During AST generation
  we create a runtime check for both, whereas the one for the
  restrictions is negated before a conjunction is build.

  Except the In-Bounds assumptions we currently only track restrictions.

Differential Revision: http://reviews.llvm.org/D17247

llvm-svn: 262328
2016-03-01 13:06:28 +00:00
Johannes Doerfert abadd71da1 [FIX] Prevent compile time problems due to complex invariant loads
This cures the symptoms we see in h264 of SPEC2006 but not the cause.

llvm-svn: 262327
2016-03-01 13:05:14 +00:00
Michael Kruse f33c125dd2 Fix DomTree preservation for generated subregions.
The generated dedicated subregion exit block was assumed to have the same
dominance relation as the original exit block. This is incorrect if the exit
block receives other edges than only from the subregion, which results in that
e.g. the subregion's entry block does not dominate the exit block.

llvm-svn: 261865
2016-02-25 14:08:48 +00:00
Johannes Doerfert 9dd42ee7c1 Try to build alias checks even when non-affine accesses are allowed
From now on we bail only if a non-trivial alias group contains a non-affine
  access, not when we discover aliasing and non-affine accesses are allowed.

llvm-svn: 261863
2016-02-25 14:06:11 +00:00
Michael Kruse f8266fad8d Tidy test case. NFC.
The test style guide defines that opt should get its input from stdin.
(instead by file argument to avoid that the file name appears in its
output)

CHECK-FORCED is not recognized by FileCheck; remove it.

llvm-svn: 261786
2016-02-24 22:08:02 +00:00
Roman Gareev 11001e1534 Annotation of SIMD loops
Use 'mark' nodes annotate a SIMD loop during ScheduleTransformation and skip
parallelism checks.

The buildbot shows the following compile/execution time changes:

  Compile time:
    Improvements    Δ     Previous  Current  σ
    …/gesummv      -6.06% 0.2640    0.2480   0.0055
    …/gemver       -4.46% 0.4480    0.4280   0.0044
    …/covariance   -4.31% 0.8360    0.8000   0.0065
    …/adi          -3.23% 0.9920    0.9600   0.0065
    …/doitgen      -2.53% 0.9480    0.9240   0.0090
    …/3mm          -2.33% 1.0320    1.0080   0.0087

  Execution time:
    Regressions     Δ     Previous  Current  σ
    …/viterbi       1.70% 5.1840    5.2720   0.0074
    …/smallpt       1.06% 12.4920   12.6240  0.0040

Reviewed-by: Tobias Grosser <tobias@grosser.es>

Differential Revision: http://reviews.llvm.org/D14491

llvm-svn: 261620
2016-02-23 09:00:13 +00:00
Johannes Doerfert 85c06c80d1 Add test case for [FIX] commit r261474
llvm-svn: 261501
2016-02-21 21:53:39 +00:00
Johannes Doerfert 91bb5bc862 Use regular expressions instead of temporary names for IR test [NFC]
llvm-svn: 261488
2016-02-21 18:59:35 +00:00
Johannes Doerfert 4d9bb8d594 Allow all combinations of types and subscripts for memory accesses
To support non-aligned accesses we introduce a virtual element size
  for arrays that divides each access function used for this array. The
  adjustment of the access function based on the element size of the
  array was therefore moved after this virtual element size was
  determined, thus after all accesses have been created.

Differential Revision: http://reviews.llvm.org/D17246

llvm-svn: 261226
2016-02-18 16:50:12 +00:00
Johannes Doerfert 4cf1580f0c [FIX] Check the next base pointer for possible invariant loads
A load can only be invariant if its base pointer is invariant too. To
  this end, we check if the base pointer is defined inside the region or
  outside. In the former case we recursively check if we can (and
  therefore will) hoist the base pointer too. Only if that happends we
  can hoist the load.

llvm-svn: 260886
2016-02-15 12:42:05 +00:00
Johannes Doerfert f69162486b Revert "[FIX] Hoist accesses if AA stated they are invariant"
This reverts commit 98efa006c96ac981c00d2e386ec1102bce9f549a.

  The fix was broken since we do not use AA in the ScopDetection anymore to
  check for invariant accesses.

llvm-svn: 260884
2016-02-15 12:21:11 +00:00
Johannes Doerfert 2353e39e1f [FIX] Hoist accesses if AA stated they are invariant
Before this patch it could happen that we did not hoist a load that
  was a base pointer of another load even though AA already declared the
  first one as invariant (during ScopDetection). If this case arises we
  will now skipt the "can be overwriten" check because in this case the
  over-approximating nature causes us to generate broken code.

llvm-svn: 260862
2016-02-14 23:37:14 +00:00
Johannes Doerfert 965edde695 Separate more constant factors of parameters
So far we separated constant factors from multiplications, however,
  only when they are at the outermost level of a parameter SCEV. Now,
  we also separate constant factors from the parameter SCEV if the
  outermost expression is a SCEVAddRecExpr. With the changes to the
  SCEVAffinator we can now improve the extractConstantFactor(...)
  function at will without worrying about any other code part. Thus,
  if needed we can implement a more comprehensive
  extractConstantFactor(...) function that will traverse the SCEV
  instead of looking only at the outermost level.

  Four test cases were affected. One did not change much and the other
  three were simplified.

llvm-svn: 260859
2016-02-14 22:30:56 +00:00