1
0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-30 07:22:55 +01:00
Commit Graph

104795 Commits

Author SHA1 Message Date
Weiming Zhao
9975517643 Report error for non-zero data in .bss
User may initialize a var with non-zero value and specify .bss section.
E.g. : int a __attribute__((section(".bss"))) = 2;

This patch converts an assertion to error report for better user
experience.

Differential Revision: http://reviews.llvm.org/D4199

llvm-svn: 211455
2014-06-22 00:33:44 +00:00
Stepan Dyatkovskiy
13371efcb8 MergeFunctions Pass, FnSet has been replaced with FnTree.
Patch activates new implementation.
So from now, merging process should take time O(N*log(N)).
Where N size of module (we are free to measure it in
functions or in instructions). Internally FnTree represents
binary tree. So every lookup operation takes O(log(N)) time.

It is still not the last patch in series, we also have to
clean-up pass from old code, and update pass comments.

This patch belongs to patch series that improves MergeFunctions
performance time from O(N*N) to O(N*log(N)).

llvm-svn: 211445
2014-06-21 20:54:36 +00:00
Stepan Dyatkovskiy
e5ac34b416 MergeFunctions Pass, removed unused methods from old implementation.
Patch removed next old FunctionComparator methods:
    * enumerate
    * isEquivalentOperation
    * isEquivalentGEP
    * isEquivalentType 
    
This patch belongs to patch series that improves MergeFunctions
performance time from O(N*N) to O(N*log(N)).

llvm-svn: 211444
2014-06-21 20:13:24 +00:00
Stepan Dyatkovskiy
569e92615e MergeFunctions, doSanityCheck: fixed body comments.
llvm-svn: 211443
2014-06-21 19:07:51 +00:00
Stepan Dyatkovskiy
f2a0775dc2 MergeFunctions Pass, introduced sanity check, that checks order relation,
introduced among functions set.
    
This patch belongs to patch series that improves MergeFunctions
performance time from O(N*N) to O(N*log(N)).

llvm-svn: 211442
2014-06-21 18:58:11 +00:00
Stepan Dyatkovskiy
844ace0874 MergeFunctions Pass, introduced total ordering among top-level comparison
methods.
    
Patch changes return type of FunctionComparator::compare() and
FunctionComparator::compare(const BasicBlock*, const BasicBlock*)
methods from bool (equal or not) to {-1, 0, 1} (less, equal, great).
    
This patch belongs to patch series that improves MergeFunctions
performance time from O(N*N) to O(N*log(N)).

llvm-svn: 211437
2014-06-21 17:55:51 +00:00
Benjamin Kramer
7821fb4e74 LoopUnrollRuntime: Check for overflow in the trip count calculation.
Fixes PR19823.

llvm-svn: 211436
2014-06-21 13:46:25 +00:00
Benjamin Kramer
8edc3ff6a3 Legalizer: Add support for splitting insert_subvectors.
We handle this by spilling the whole thing to the stack and doing the
insertion as a store.

PR19492. This happens in real code because the vectorizer creates v2i128 when AVX is enabled.

llvm-svn: 211435
2014-06-21 12:56:42 +00:00
Benjamin Kramer
8d54e9ca1f SCEVExpander: Fold constant PHIs harder. The logic below only understands proper IVs.
PR20093.

llvm-svn: 211433
2014-06-21 11:47:18 +00:00
Richard Trieu
b7d5af56cb Add back functionality removed in r210497.
Instead of asserting, output a message stating that a null pointer was found.

llvm-svn: 211430
2014-06-21 02:43:02 +00:00
Andrea Di Biagio
53a4d382f7 [X86] Add ISel patterns to select SSE3/AVX ADDSUB instructions.
This patch adds ISel patterns to select SSE3/AVX ADDSUB instructions
from a sequence of "vadd + vsub + blend".

Example:

///
typedef float float4 __attribute__((ext_vector_type(4)));

float4 foo(float4 A, float4 B) {
  float4 X = A - B;
  float4 Y = A + B;
  return (float4){X[0], Y[1], X[2], Y[3]};
}
///

Before this patch, (with flag -mcpu=corei7) llc produced the following
assembly sequence:
  movaps  %xmm0, %xmm2
  addps   %xmm1, %xmm2
  subps   %xmm1, %xmm0
  blendps $10, %xmm2, %xmm0


With this patch, we now get a single
  addsubps  %xmm1, %xmm0

llvm-svn: 211427
2014-06-21 01:31:15 +00:00
Zachary Turner
a6805811e7 Fix the MinGW builder. Apparently std::call_once and
std::recursive_mutex are not available on MinGW and breaks the
builder.  Revert to using a function local static and sys::Mutex
just to get the tree green until we figure out a better solution.

llvm-svn: 211424
2014-06-21 00:24:51 +00:00
Rafael Espindola
001fc6d71f Always use a temp symbol for CIE.
Fixes pr19185.

llvm-svn: 211423
2014-06-20 23:54:32 +00:00
Rafael Espindola
c408251eb7 Use compact unwind for the iOS simulator.
Another step in fixing pr19185.

llvm-svn: 211416
2014-06-20 22:40:55 +00:00
Rafael Espindola
f82345a30b Use a helper function and clang-format.
No functionality change.

llvm-svn: 211415
2014-06-20 22:37:01 +00:00
Duncan P. N. Exon Smith
282a896d30 Support: ScaledNumber: Fix inconsistent test names
llvm-svn: 211414
2014-06-20 22:36:09 +00:00
Duncan P. N. Exon Smith
63810bb7e2 Support: Write ScaledNumbers::getLg{,Floor,Ceiling}()
llvm-svn: 211413
2014-06-20 22:33:40 +00:00
Rafael Espindola
28af33f6db Delete dead code.
The compact unwind info is only used by code that knows it is supported.

llvm-svn: 211412
2014-06-20 22:30:31 +00:00
Duncan P. N. Exon Smith
db0cbc8b8a Support: Write ScaledNumber::getQuotient() and getProduct()
llvm-svn: 211409
2014-06-20 21:47:47 +00:00
Duncan P. N. Exon Smith
6fde469206 Support: Cleanup ScaledNumber::getAdjusted() doc
llvm-svn: 211407
2014-06-20 21:44:36 +00:00
Duncan P. N. Exon Smith
276d22efba Support: Mark end of namespaces
This convinces clang-format to leave a newline.

llvm-svn: 211406
2014-06-20 21:43:20 +00:00
Kevin Enderby
1805a9f0de Fix some double printing of filenames for archives in llvm-nm when
the tool is given multiple files.  Also fix the same issue with Mach-O
universal files. And fix the newline spacing to separate the output
in these cases.

llvm-svn: 211405
2014-06-20 21:29:27 +00:00
Rafael Espindola
962bd7f649 Don't produce eh_frame relocations when targeting the IOS simulator.
First step for fixing pr19185.

llvm-svn: 211404
2014-06-20 21:15:27 +00:00
Zachary Turner
2e5e9cb0d7 Revert "Replace Execution Engine's mutex with std::recursive_mutex."
This reverts commit 1f502bd9d7d2c1f98ad93a09ffe435e11a95aedd, due to
GCC / MinGW's lack of support for C++11 threading.

It's possible this will go back in after we come up with a
reasonable solution.

llvm-svn: 211401
2014-06-20 21:07:14 +00:00
Reid Kleckner
40d9c6f936 Generate native unwind info on Win64
This patch enables LLVM to emit Win64-native unwind info rather than
DWARF CFI.  It handles all corner cases (I hope), including stack
realignment.

Because the unwind info is not flexible enough to describe stack frames
with a gap of unknown size in the middle, such as the one caused by
stack realignment, I modified register spilling code to place all spills
into the fixed frame slots, so that they can be accessed relative to the
frame pointer.

Patch by Vadim Chugunov!

Reviewed By: rnk

Differential Revision: http://reviews.llvm.org/D4081

llvm-svn: 211399
2014-06-20 20:35:47 +00:00
David Blaikie
f279d4f048 Fix some -Wsign-compare fallout from changing container count member functions to return unsigned instead of bool.
llvm-svn: 211393
2014-06-20 19:54:13 +00:00
Stepan Dyatkovskiy
9b141122e2 Commited patch from Björn Steinbrink:
Summary:
Different range metadata can lead to different optimizations in later
passes, possibly breaking the semantics of the merged function. So range
metadata must be taken into consideration when comparing Load
instructions.

Thanks!

llvm-svn: 211391
2014-06-20 19:11:56 +00:00
Adam Nemet
bfee3c0a59 [Make] Fix dependencies for td.expanded
Depend on all the .td files not just the main one.

llvm-svn: 211390
2014-06-20 19:00:41 +00:00
Ulrich Weigand
7ceb645df2 [RuntimeDyld] Fix ppc64 stub relocations on little-endian
When RuntimeDyldELF creates stub functions, it needs to install
relocations that will resolve to the final address of the target
routine. Since those are 16-bit relocs, they need to be applied to the
least-significant halfword of the instruction.  On big-endian ppc64,
this means that addresses have to be adjusted by 2, which is what the
code currently does.

However, on a little-endian system, the address must *not* be adjusted;
the least-significant halfword is the first one.  This patch updates the
RuntimeDyldELF code to take the target byte order into account.

llvm-svn: 211384
2014-06-20 18:17:56 +00:00
Kevin Enderby
a3633377f9 Fix a warning about the use of const being ignored with a cast.
llvm-svn: 211383
2014-06-20 18:07:34 +00:00
Ulrich Weigand
a5f93603cb [RuntimeDyld] Support more PPC64 relocations
This adds support for several missing PPC64 relocations in the
straight-forward manner to RuntimeDyldELF.cpp.

Note that this actually fixes a failure of a large-model test case on
PowerPC, allowing the XFAIL to be removed.

llvm-svn: 211382
2014-06-20 17:51:47 +00:00
Tom Stellard
ae7faa387d R600/SI: Add patterns for ctpop inside a branch
llvm-svn: 211378
2014-06-20 17:06:11 +00:00
Tom Stellard
2d56aec1cd R600/SI: Add a pattern for f32 ftrunc
llvm-svn: 211377
2014-06-20 17:06:09 +00:00
Tom Stellard
0c64d4a322 R600: Expand vector flog2
llvm-svn: 211376
2014-06-20 17:06:07 +00:00
Tom Stellard
f2ad4b6bb8 R600: Expand vector fexp2
llvm-svn: 211375
2014-06-20 17:06:05 +00:00
Tom Stellard
ce55c9c61a R600/SI: SI Control Flow Annotation bug fixed
Mixing of AddAvailableValue and GetValueAtEndOfBlock methods of SSAUpdater
leaded to the endless loop generation when the nested loops annotated.

This fixes a bug in the OCL_ML/KNN OpenCV test.  The test case is too
complex for FileCheck and would be very fragile.

Patch by: Elena Denisova

llvm-svn: 211374
2014-06-20 17:06:02 +00:00
Tom Stellard
d4aa49ad5e R600/SI: Add a VALU pattern for i64 xor
llvm-svn: 211373
2014-06-20 17:05:57 +00:00
Ulrich Weigand
3f880bd06e [PowerPC] Fix small argument stack slot offset for LE
When small arguments (structures < 8 bytes or "float") are passed in a
stack slot in the ppc64 SVR4 ABI, they must reside in the least
significant part of that slot.  On BE, this means that an offset needs
to be added to the stack address of the parameter, but on LE, the least
significant part of the slot has the same address as the slot itself.

This changes the PowerPC back-end ABI code to only add the small
argument stack slot offset for BE.  It also adds test cases to verify
the correct behavior on both BE and LE.

llvm-svn: 211368
2014-06-20 16:34:05 +00:00
Rafael Espindola
f4a9d636ea Move test so that it is skipped if the ARM target is not enabled.
llvm-svn: 211366
2014-06-20 15:30:38 +00:00
Rafael Espindola
cc59644b5c Allow a target to create a null streamer.
Targets can assume that a target streamer is present, so they have to be able
to construct a null streamer in order to set the target streamer in it to.

Fixes a crash when using the null streamer with arm.

llvm-svn: 211358
2014-06-20 13:11:28 +00:00
Yaron Keren
305b1c858b Code in LoopStrengthReduce.cpp depends on SmallBitVector::size() being size_t
and not unsigned.

llvm-svn: 211356
2014-06-20 12:57:44 +00:00
Yaron Keren
333ef0d2b6 Reverting size_type for the containers from size_type to unsigned.
Various places in LLVM assume that container size and count are unsigned
and do not use the container size_type. Therefore they break compilation
(or possibly executation) for LP64 systems where size_t is 64 bit while
unsigned is still 32 bit.

If we'll ever that many items in the container size_type could be made
size_t for a specific containers after reviweing its other uses.

llvm-svn: 211353
2014-06-20 12:20:56 +00:00
Yaron Keren
d3a024a162 Attempting to fix the 64 bit bots.
llvm-svn: 211351
2014-06-20 10:52:57 +00:00
Yaron Keren
e1172cbc66 The count() function for STL datatypes returns unsigned, even where it's
only 1/0 result like std::set. Some of the LLVM ADT already return unsigned
count(), while others still return bool count().

In continuation to r197879, this patch modifies DenseMap, DenseSet, 
ScopedHashTable, ValueMap:: count() to return size_type instead of bool,
1 instead of true and 0 instead of false.

size_type is typedef-ed locally within each class to size_t.

http://reviews.llvm.org/D4018

Reviewed by dblaikie.

llvm-svn: 211350
2014-06-20 10:26:56 +00:00
Oliver Stannard
8087ec46e0 Emit the ARM build attributes ABI_PCS_wchar_t and ABI_enum_size.
Emit the ARM build attributes ABI_PCS_wchar_t and ABI_enum_size based on
module flags metadata.

llvm-svn: 211349
2014-06-20 10:08:11 +00:00
Zoran Jovanovic
895809b0be ps][mips64r6] Added LSA/DLSA instructions
Differential Revision: http://reviews.llvm.org/D3897

llvm-svn: 211346
2014-06-20 09:28:09 +00:00
Matt Arsenault
f30877f744 R600: Trivial subtarget feature cleanups.
Remove an unused AMDIL leftover, correct extra periods
appearing in the help menu.

llvm-svn: 211341
2014-06-20 06:50:05 +00:00
Justin Bogner
770d258ab4 ArgList: use MakeArgList overloads in subclasses and clean up some calls.
llvm-svn: 211340
2014-06-20 04:36:29 +00:00
Karthik Bhat
fb8456f1af Add Support to Recognize and Vectorize NON SIMD instructions in SLPVectorizer.
This patch adds support to recognize patterns such as fadd,fsub,fadd,fsub.../add,sub,add,sub... and
vectorizes them as vector shuffles if they are profitable.
These patterns of vector shuffle can later be converted to instructions such as addsubpd etc on X86.
Thanks to Arnold and Hal for the reviews. http://reviews.llvm.org/D4015 

llvm-svn: 211339
2014-06-20 04:32:48 +00:00
Duncan P. N. Exon Smith
d63ac88ff6 Support: Clean up getRounded() tests
llvm-svn: 211337
2014-06-20 02:31:07 +00:00