| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
|
|
|
|
|
|
|
|
|
| |
Summary: Add basic support, i386 version will be added later.
Reviewers: emaste, joerg, clayborg, tfiala
Subscribers: lldb-commits
Differential Revision: http://reviews.llvm.org/D16508
llvm-svn: 259462
|
| |
|
|
|
|
|
|
|
|
|
|
| |
LVI has several separate sources of facts - edge local conditions, recursive queries, assumes, and control independent value facts - which all apply to the same value at the same location. The existing implementation was very conservative about exploiting all of these facts at once.
This change introduces an "intersect" function specifically to abstract the action of picking a good set of facts from all of the separate facts given. At the moment, this function is relatively simple (i.e. mostly just reuses the bits which were already there), but even the minor additions reveal the inherent power. For example, JumpThreading is now capable of doing an inductive proof that a particular value is always positive and removing a half range check.
I'm currently only using the new intersect function in one place. If folks are happy with the direction of the work, I plan on making a series of small changes without review to replace mergeIn with intersect at all the appropriate places.
Differential Revision: http://reviews.llvm.org/D14476
llvm-svn: 259461
|
| |
|
|
| |
llvm-svn: 259460
|
| |
|
|
| |
llvm-svn: 259459
|
| |
|
|
| |
llvm-svn: 259458
|
| |
|
|
| |
llvm-svn: 259457
|
| |
|
|
|
|
|
|
|
| |
Fix a crash in `getMemOpBaseRegImmOfs` that happens if the base of
`MemOp` is a frame index memory operand. The fix is to have
`getMemOpBaseRegImmOfs` bail out in such cases. We can possibly be more
clever here, if needed.
llvm-svn: 259456
|
| |
|
|
| |
llvm-svn: 259455
|
| |
|
|
| |
llvm-svn: 259454
|
| |
|
|
|
|
| |
Differential Revision: http://reviews.llvm.org/D15624
llvm-svn: 259453
|
| |
|
|
|
|
| |
The error message has been changed in code as of r258849 (http://reviews.llvm.org/D16546). Update the test as well.
llvm-svn: 259452
|
| |
|
|
| |
llvm-svn: 259451
|
| |
|
|
|
|
|
|
|
|
|
| |
When we do debug printing of atoms, they expect their parent file
to have an ordinal. Right now we trigger a bunch of asserts.
This doesn't actually manifest in differences in any real tests, but
even if the only thing it does is improve debuggability then its still
useful.
llvm-svn: 259450
|
| |
|
|
|
|
| |
FastISel counterpart to r259448.
llvm-svn: 259449
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Officially, we don't acknowledge non-default configurations of MXCSR,
as getting there would require usage of the FENV_ACCESS pragma (at
least insofar as rounding mode is concerned).
We don't support the pragma, so we can assume that the default
rounding mode - round to nearest, ties to even - is always used.
However, it's inconsistent with the rest of the instruction set,
where MXCSR is always effective (unless otherwise specified).
Also, it's an unnecessary obstacle to the few brave souls that use
fenv.h with LLVM.
Avoid the hard-coded rounding mode for fp_to_f16; use MXCSR instead.
llvm-svn: 259448
|
| |
|
|
|
|
|
|
|
|
| |
The unsafe stack pointer is only popped in moveStaticAllocasToUnsafeStack so it won't happen if there are no static allocas.
Fixes https://llvm.org/bugs/show_bug.cgi?id=26122
Differential Revision: http://reviews.llvm.org/D16339
llvm-svn: 259447
|
| |
|
|
|
|
|
|
| |
This routine was returning Undefined for most queries. This was utterly wrong. Amusingly, we do not appear to have any callers of this which are actually trying to exploit unreachable code or this would have broken the world.
A better approach would be to explicit describe the intersection of facts. That's blocked behind http://reviews.llvm.org/D14476 and I wanted to fix the current bug.
llvm-svn: 259446
|
| |
|
|
|
|
|
| |
When all the arguments of a template are elided, print "A<...>" instead of
"A<[2 * ...]>". Also remove comment fragment that means nothing.
llvm-svn: 259445
|
| |
|
|
|
|
|
|
| |
Instead of leave unused fields as is, set them to nullptr.
Currnetly this is NFC, but if you call writeResults more than
once, you should be able to see the difference.
llvm-svn: 259444
|
| |
|
|
|
|
|
|
| |
features
in the socket, so just returning freshly constructed objects is fine.
llvm-svn: 259443
|
| |
|
|
| |
llvm-svn: 259442
|
| |
|
|
|
|
|
|
| |
The ObjCPass is going to need to create pointer sized relocations in
the ObjC sections. This method will be used to give us a target independent
way of getting the correct kind for the refererence.
llvm-svn: 259441
|
| |
|
|
|
|
|
|
|
|
| |
An upcoming patch will use this to create lists of ObjC methods.
Adding it now to reduce the amount of code in that patch.
Test cases will follow in the other patch too.
llvm-svn: 259440
|
| |
|
|
|
|
|
|
|
|
|
| |
This adds the no_objc_category_merging cmdline option which will
be used in an upcoming commit to disable the category optimizer.
It is on by default in ld64 so we match that here.
Test case will come soon with the patch to make use of this option.
llvm-svn: 259439
|
| |
|
|
| |
llvm-svn: 259438
|
| |
|
|
| |
llvm-svn: 259437
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Most of the other methods to access Reference's were on DefinedAtom so
this just keeps them all together.
This will be used in a future patch in ObjCPass which needs to add
new references. The method is virtual because we may add references to
different data structures depending on whether we parsed a macho file or
yaml.
llvm-svn: 259436
|
| |
|
|
|
|
| |
http://reviews.llvm.org/D16648
llvm-svn: 259435
|
| |
|
|
|
|
|
| |
Enhance an existing test to also check that the ordering of the function
summary entries is stable.
llvm-svn: 259434
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Summary:
Checks using the result of getCompiler() will fail to identify the compiler
correctly if CC is a symlink path (ie /usr/bin/cc).
Reviewers: zturner, emaste
Subscribers: llvm-commits, sas
Differential Revision: http://reviews.llvm.org/D16488
Change by Francis Ricci <fjricci@fb.com>
llvm-svn: 259433
|
| |
|
|
|
|
| |
I'll submit a test case shortly which covers this, but it's causing clang self host problems in the builders so I wanted to get it removed.
llvm-svn: 259432
|
| |
|
|
|
|
|
|
|
| |
This will be used by a future commit which wanted to compare iterators
for equality.
Also changed operator!= to be in terms of the new operator==
llvm-svn: 259431
|
| |
|
|
| |
llvm-svn: 259430
|
| |
|
|
|
|
|
|
| |
Teach LVI to handle select instructions in the exact same way it handles PHI nodes. This is useful since various parts of the optimizer convert PHI nodes into selects and we don't want these transformations to cause inferior optimization.
Note that this patch does nothing to exploit the implied constraint on the inputs represented by the select condition itself. That will be a later patch and is blocked on http://reviews.llvm.org/D14476
llvm-svn: 259429
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
RecursiveASTVisitor::TraverseFunctionHelper() traverses a function's
ParmVarDecls by going to the function's getTypeSourceInfo if it exists, and
`DEF_TRAVERSE_TYPELOC(FunctionProtoType` then goes to the function's
ParmVarDecls.
For a function template that doesn't have parameters that explicitly depend on
the template parameter, we used to be clever and not build a new
TypeSourceInfo. That meant that when an instantiation of such a template is
visited, its TypeSourceInfo would point to the ParmVarDecls of the template,
not of the instantiation, which then confused clients of RecursiveASTVisitor.
So don't be clever for function templates that have parameters, even if none of
the parameters depend on the type.
Fixes PR26257.
http://reviews.llvm.org/D16478
llvm-svn: 259428
|
| |
|
|
| |
llvm-svn: 259427
|
| |
|
|
|
|
|
|
|
|
|
|
| |
When we parse a MachoFile, we set a number of members from the parsed
file, for example, subsectionsViaSymbols.
However, a number of passes, such as ObjCPass, create local copies of
MachoFile and don't get the benefit of setting flags and other fields in
the parser. Instead we can just give a more sensible default as the parser
will definitely get the correct value from the file anyway.
llvm-svn: 259426
|
| |
|
|
| |
llvm-svn: 259425
|
| |
|
|
| |
llvm-svn: 259424
|
| |
|
|
|
|
|
|
|
|
|
| |
A masked scatter with a zero mask means there's no store.
A masked gather with a zero mask means the passthru arg is returned.
This is a continuation of:
http://reviews.llvm.org/rL259369
http://reviews.llvm.org/rL259392
llvm-svn: 259421
|
| |
|
|
| |
llvm-svn: 259420
|
| |
|
|
|
|
|
|
|
|
| |
These sets do linear searching in small mode; It is not a good idea to
use huge numbers as the small value here, save people from themselves by
adding a static_assert.
Differential Revision: http://reviews.llvm.org/D16706
llvm-svn: 259419
|
| |
|
|
|
|
| |
> http://reviews.llvm.org/D16758
llvm-svn: 259418
|
| |
|
|
|
|
|
| |
Now that we remember the StringRefs when they are first added, we can
write the table in any order.
llvm-svn: 259417
|
| |
|
|
| |
llvm-svn: 259416
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
Now that MachoFile has classof(), we can use dyn_cast instead which
is actually the only safe way to handle this.
Turns out this actually manifests as a bug as we were incorrectly
casting instances which weren't MachoFile in to a MachoFile.
Unfortunately, there's no reliable way of checking for this as it
requires that the file we are looking for has a 0 at exactly the byte
we need for the load of subsectionsViaSymbols.
llvm-svn: 259413
|
| |
|
|
| |
llvm-svn: 259411
|
| |
|
|
|
|
|
|
| |
Add AVX512F/AVX512BW 512-bit tests.
Add AVX512F tests to existing 128/256-bit tests.
llvm-svn: 259410
|
| |
|
|
|
|
| |
Patch by Alexander Riccio.
llvm-svn: 259409
|
| |
|
|
|
|
|
| |
The former will be properly initialized in standalone CMake build of
compiler-rt.
llvm-svn: 259407
|