Commit Graph

125481 Commits

Author SHA1 Message Date
Teresa Johnson cc428573a2 Ensure MDNode used as key in metadata linking map cannot be RAUWed
As suggested in review for r255909, add a way to ensure that temporary
MD used as keys in the MetadataToID map during ThinLTO importing are not
RAUWed.

Add support for marking an MDNode as not replaceable. Clear the new
CanReplace flag when adding a temporary MD node to the MetadataToID map
and clear it when destroying the map.

llvm-svn: 256648
2015-12-30 19:32:24 +00:00
Teresa Johnson 26aa93586a [ThinLTO] Check MDNode values saved for metadata linking (NFC)
Add an assert suggested in review for r255909 to ensure that MDNodes
saved in the map used for metadata linking are either temporary or
resolved.

Also add a comment clarifying why we may need to save off non-MDNode
metadata.

llvm-svn: 256646
2015-12-30 19:13:57 +00:00
Sanjay Patel 16395dd709 fix formatting; NFC
llvm-svn: 256645
2015-12-30 18:31:30 +00:00
Tobias Grosser 666761afab Revert "[ptr-traits] Implement the base pointer traits using the actual"
The commit we revert is rather small, but it enables a larger piece of new
infrastructure that allows to detected misuses of pointer-traits at compile
time. Unfortunately, this change breaks with the use of incomplete types (e.g.
in Polly). As I am not aware of a simple fix on the Polly side, I temporarely
revert this commit to clean the bots and sync-up with Chandler how to best
adapt to these recent changes.

This reverts commit https://llvm.org/svn/llvm-project/llvm/trunk@256620.

llvm-svn: 256642
2015-12-30 14:02:58 +00:00
Craig Topper 13b2a4e35d [TableGen] Remove unnecessary conversion from StringRef to std::string when outputting to a raw_ostream. NFC
llvm-svn: 256628
2015-12-30 06:00:24 +00:00
Craig Topper 39311c7ed5 [TableGen] Remove raw_string_ostream by just emitting the header for the switch the first time we emit a case. If the header was never emitted just print the default at the end. NFC
llvm-svn: 256627
2015-12-30 06:00:22 +00:00
Craig Topper 77bd2b7581 [TableGen] Use range-based for loops. NFC
llvm-svn: 256626
2015-12-30 06:00:20 +00:00
Craig Topper c8b5b255f7 [TableGen] Move more things that come from variant into the AsmVariantInfo class so we can reduce some parameters. NFC
llvm-svn: 256625
2015-12-30 06:00:18 +00:00
Craig Topper ba61432a9f [TableGen] Use 'size_t' instead of 'unsigned' to better match the argument types of addAsmOperand. Simplify some code by using StringRef::find instead of std::find. These were previously done in r247527 and r247528, but another commit seems to have erased them. NFC
llvm-svn: 256624
2015-12-30 06:00:15 +00:00
Chandler Carruth f05e492cde [ptr-traits] Implement the base pointer traits using the actual
alignment of the pointee type!

This is the culmination of the ptr-traits work. Now the compiler will
catch me if I try to use a pointer to an empty struct as a key in
a dense map or inside a PointerIntPair or PointerUnion! This is much,
much better than sometimes corrupting data (and other times working
fine) due to insufficient alignment.

It also means that we will be much more diligent about rejecting other
uses of these constructs that aren't safe.

It also means that we can now be more aggressive with the constructs
when we actually have guaranteed higher alignment without specializing
stuff. I'll be going through and cleaning up all the current overrides
of these traits which are no longer necessary.

Many thanks to Richard, David, and others who helped me get all of this
together.

llvm-svn: 256620
2015-12-30 04:00:24 +00:00
Chandler Carruth c92a9645f2 [ptr-traits] Refactor how PointerIntPair does its pointer manipulation
to isolate it in a dependent helper class.

Without doing this, we end up requiring all of the pointer traits the
moment you even define a PointerIntPair. That makes them *incredibly*
hard to use, for example you can't use them at all inside a class for
pointers to that class!

This change sinks all the logic into a helper template class that only
needs to be fully instantiated when *using* the PointerIntPair. We still
get compile-time checking, but it is deferred long enough to make
tradition out-of-line method definitions (or just the normal deferred
method body parsing) sufficient to handle cycling references.

llvm-svn: 256618
2015-12-30 03:56:19 +00:00
Chandler Carruth a9030d5c16 [ptr-traits] Add one more #include necessary to do strict alignment
checking of pointers used in PointerIntPairs.

llvm-svn: 256617
2015-12-30 03:56:17 +00:00
Teresa Johnson 61b406ec75 Rename MDValue* to Metadata* (NFC)
Renamed MDValue* to Metadata*, and MDValueToValIDMap to MetadataToIDs,
as per review for r255909.

llvm-svn: 256593
2015-12-29 23:00:22 +00:00
Manuel Jacob 67f1d3ac63 [RS4GC] Use DenseMap::count() instead of DenseMap::find()/DenseMap::end(). NFC.
llvm-svn: 256586
2015-12-29 22:16:41 +00:00
Sanjay Patel ac6e910c42 don't repeat function names in comments; NFC
llvm-svn: 256584
2015-12-29 22:11:50 +00:00
Sanjay Patel d1d9db5889 use auto with dyn_casted values; NFC
llvm-svn: 256581
2015-12-29 22:00:37 +00:00
Manuel Jacob e3773d632e [PlaceSafepoints] Assert that the gc.safepoint_poll function is present in the module.
If running the PlaceSafepoints pass on a module which doesn't have the
gc.safepoint_poll function without disabling entry and backedge safepoints,
previously the pass crashed with an obscure error because of a null pointer.
Now it fails the assert instead.

llvm-svn: 256580
2015-12-29 21:57:55 +00:00
Sanjay Patel 7a7abc9a3b use auto with dyn_casted values; NFC
llvm-svn: 256579
2015-12-29 21:49:08 +00:00
Sanjay Patel b120ae96d3 fix formatting; NFC
llvm-svn: 256574
2015-12-29 19:34:53 +00:00
Sanjay Patel 4104f78640 use range-based for-loops; NFCI
llvm-svn: 256573
2015-12-29 19:14:23 +00:00
Sanjay Patel faeee6f44d use range-based for-loop; NFCI
llvm-svn: 256572
2015-12-29 18:30:09 +00:00
Chad Rosier 6b4326367a Add command line options to force function/loop alignments.
These are being added for testing purposes.
http://reviews.llvm.org/D15648

llvm-svn: 256571
2015-12-29 18:18:07 +00:00
Sanjay Patel 59309cc090 don't repeat function names in comments; NFC
llvm-svn: 256569
2015-12-29 18:14:06 +00:00
Geoff Berry 43dc285915 [JumpThreading] Fix opcode bonus in getJumpThreadDuplicationCost()
The code that was meant to adjust the duplication cost based on the
terminator opcode was not being executed in cases where the initial
threshold was hit inside the loop.

Subscribers: mcrosier, llvm-commits

Differential Revision: http://reviews.llvm.org/D15536

llvm-svn: 256568
2015-12-29 18:10:16 +00:00
Sanjay Patel 7dd45697ca use range-based for-loops; NFCI
llvm-svn: 256566
2015-12-29 17:15:22 +00:00
Philip Reames 87a8677e3d [MemoryBuiltins] Delete dead code [NFC]
llvm-svn: 256565
2015-12-29 17:04:43 +00:00
Michael Zuckerman c2e4bb421a [AVX512] add PSRLW Intrinsic
Fixing tab/space indentation.
Differential Revision: http://reviews.llvm.org/D15751

llvm-svn: 256561
2015-12-29 14:34:58 +00:00
Davide Italiano ed9d95b290 [llvm-objdump] Mark noreturn function as such.
Match attribute in the header to make MSVC happy.

llvm-svn: 256560
2015-12-29 13:41:02 +00:00
Michael Zuckerman 80821ee77c [AVX512] add PSRLW Intrinsic
Differential Revision: http://reviews.llvm.org/D15751

llvm-svn: 256558
2015-12-29 13:04:35 +00:00
Chandler Carruth 8db8e8b86f [ADT] Teach alignment helpers to work correctly for abstract classes.
This is necessary to use them as part of pointer traits and is generally
useful. I've added unit test coverage to isolate and ensure this works
correctly.

I'll watch the build bots to try to see if any compilers can't tolerate
this bit of magic (and much credit goes to Richard Smith for coming up
with this magical production!) but give a shout if you see issues.

llvm-svn: 256553
2015-12-29 09:52:41 +00:00
Chandler Carruth 5bd31b37ca [ptr-traits] Provide a real MCFragment address for the sentinel instead
of casting the integer '4' to such a pointer. There is no reason to
expect '4' to be a portable or reliable pointer of this form. The only
reason this ever worked is because the PointerIntPair that this actually
gets used with has an artificially *low* presumed alignment that allowed
it to work. When the alignment of PointerIntPair is derived from the
actual type's alignment, the asserts start firing on this pointer. I'm
amazed we never managed to do anything that triggered the alignment
sanitizer with it, as this is just flat out UB.

If folks dislike this approach to providing a sentinel fragment address,
there are a myriad of other alternatives, suggestions welcome. But this
one has the distinct advantage of not requiring the friend dance of
ilist's sentinel (which I'll point out is *also* in play for
MCFragment!) and seems to be using a nicely provided facility in
MCFragment to establish just such dummy nodes.

This is part of a series of patches to allow LLVM to check for complete
pointee types when computing its pointer traits. This is absolutely
necessary to get correct (or reproducible) results for things like how
many low bits are guaranteed to be zero.

llvm-svn: 256552
2015-12-29 09:32:18 +00:00
Chandler Carruth 6dd8f06db2 [ptr-traits] Sink several in-body method definitions to be out-of-line
inline definitions after the mutually recursive pair of types have been
defined. The two types mutually recurse specifically through
abstractions that require pointer traits which makes this kind of mutual
recursion especially tricky to get right in terms of ordering.

This is part of a series of patches to allow LLVM to check for complete
pointee types when computing its pointer traits. This is absolutely
necessary to get correct (or reproducible) results for things like how
many low bits are guaranteed to be zero.

llvm-svn: 256551
2015-12-29 09:24:42 +00:00
Chandler Carruth e0115344e6 [ptr-traits] Sink a constructor definition to the .cpp file and add
missing includes so that the pointee types for DenseMap pointer keys and
such are complete prior to us querying the pointer traits for them.

This is part of a series of patches to allow LLVM to check for complete
pointee types when computing its pointer traits. This is absolutely
necessary to get correct (or reproducible) results for things like how
many low bits are guaranteed to be zero.

llvm-svn: 256550
2015-12-29 09:24:39 +00:00
Chandler Carruth 388235fe85 [ptr-traits] Add a bunch of includes to provide complete types that are
used in pointer dense map key types or in other ways that require
pointer traits.

This is part of a series of patches to allow LLVM to check for complete
pointee types when computing its pointer traits. This is absolutely
necessary to get correct (or reproducible) results for things like how
many low bits are guaranteed to be zero.

llvm-svn: 256549
2015-12-29 09:06:21 +00:00
Chandler Carruth 8d736236d3 [ptr-traits] Split the MCFragment type hierarchy out of the MCAssembler
header to its own header, allowing users of fragments to have a narrower
header file, and avoid circular header dependencies when getting the
definition of MCSection prior to inspecting traits on MCSection
pointers.

This is part of a series of patches to allow LLVM to check for complete
pointee types when computing its pointer traits. This is absolutely
necessary to get correct (or reproducible) results for things like how
many low bits are guaranteed to be zero.

Note that this doesn't in any way change the design of MC, it is just
moving code around to allow the *header files* to be more fine grained.
Without this, it is impossible to get a complete type for MCSection
where it is needed.

If anyone would prefer a different slicing of the header files, I'm
happy to oblige of course. =]

llvm-svn: 256548
2015-12-29 09:06:16 +00:00
Chandler Carruth ebf579164f [MC] Fix the comment header for this header file. It was just
copy/pasted.

Happy for anyone to suggest a more precise or refined set of boilerplate
here, but the comments on the actual code seem descriptive and accurate.

llvm-svn: 256547
2015-12-29 09:06:10 +00:00
Craig Topper 3661c62e4b De-virtualize mnemonicIsValid and remove from the base class. It's not called by any common code.
llvm-svn: 256544
2015-12-29 07:43:03 +00:00
Craig Topper d270501a6e [TableGen] Remove MnemonicContainsDot from AsmParser. It isn't used. NFC
llvm-svn: 256542
2015-12-29 07:03:30 +00:00
Craig Topper 3294966ed7 [X86] Remove declaration of ATTAsmParser. Its equivalent to the DefaultAsmParser. NFC
llvm-svn: 256541
2015-12-29 07:03:27 +00:00
Craig Topper 884409a552 [TableGen] Add missing space to output.
llvm-svn: 256540
2015-12-29 07:03:25 +00:00
Craig Topper e4e7415705 [TableGen] Use range-based for loops. NFC
llvm-svn: 256539
2015-12-29 07:03:23 +00:00
James Y Knight d734aaa4ba [TrailingObjects] Dynamically realign under-aligned trailing objects.
Previously, the code enforced non-decreasing alignment of each trailing
type. However, it's easy enough to allow for realignment as needed, and
thus avoid the developer having to think about the possiblilities for
alignment requirements on all architectures.

(E.g. on Linux/x86, a struct with an int64 member is 4-byte aligned,
while on other 32-bit archs -- and even with other OSes on x86 -- it has
8-byte alignment. This sort of thing is irritating to have to manually
deal with.)

llvm-svn: 256533
2015-12-29 04:00:43 +00:00
James Y Knight 992904a0af Fix gold test after r256465.
That commit added a new pass, and this test is sensitive to what the
first pass after verify is called.

llvm-svn: 256532
2015-12-29 03:48:37 +00:00
Chandler Carruth a7dc087b36 [ptr-traits] Merge the MetadataTracking helpers into the Metadata
header.

This is part of a series of patches to allow LLVM to check for complete
pointee types when computing its pointer traits. This is absolutely
necessary to get correct (or reproducible) results for things like how
many low bits are guaranteed to be zero.

The MetadataTracking helpers aren't actually independent. They rely on
constructing a PointerUnion between Metadata and MetadataAsValue
pointers, which requires know the alignment of pointers to those types
which requires them to be complete.

The .cpp file even defined a method declared in Metadata.h! These really
don't seem like something that is separable, and there is no real
layering problem with just placing them together.

llvm-svn: 256531
2015-12-29 02:14:50 +00:00
Chandler Carruth dc82f13080 [ADT] Run clang-format over another file before I try to enhance it so
I don't commit the bug in r256518 again.

llvm-svn: 256530
2015-12-29 02:14:45 +00:00
Chandler Carruth b738855f8c [ADT] Use a nonce type with at least 4 byte alignment.
We didn't actually statically check this, and so it worked 25% of the
time for me. =/ Really sorry it took so long to fix, I shouldn't leave
the commit log editor window open without saving and landing the commit.
=[

llvm-svn: 256528
2015-12-29 00:03:24 +00:00
Eric Christopher 2ae7180b24 Accept dwarf version 5 for CIE versions.
llvm-svn: 256527
2015-12-28 23:02:42 +00:00
Artyom Skrobov 2aca0c622a [Thumb] Fix assembler error 'cannot honor width suffix pop {lr}'
Summary:
* avoid generating POP {LR} in Thumb1 epilogues
* combine MOV LR, Rx + BX LR -> BX Rx in a peephole optimization pass
* combine POP {LR} + B + BX LR -> POP {PC} on v5T+

Test cases by Ana Pazos

Differential Revision: http://reviews.llvm.org/D15707

llvm-svn: 256523
2015-12-28 21:40:45 +00:00
Sanjay Patel b3c53e512f [x86] lower calls to fmin and llvm.minnum.* using minss/minsd/minps/minpd (PR24475)
This is a follow-on to:
http://reviews.llvm.org/rL255700
http://reviews.llvm.org/rL256454
http://reviews.llvm.org/rL256510

llvm-svn: 256522
2015-12-28 21:16:55 +00:00
Easwaran Raman b9f7120e7a Refactor inline costs analysis by removing the InlineCostAnalysis class
InlineCostAnalysis is an analysis pass without any need for it to be one.
Once it stops being an analysis pass, it doesn't maintain any useful state
and the member functions inside can be made free functions. NFC.

Differential Revision: http://reviews.llvm.org/D15701

llvm-svn: 256521
2015-12-28 20:28:19 +00:00