We need to relax constraints on invariant loads so that they do not
create fake RAW dependences. So, we do not consider invariant loads as
scalar dependences in a region.
During these changes, it turned out that we do not consider `llvm::Value`
replacements correctly within `PPCGCodeGeneration` and `ISLNodeBuilder`.
The replacements dictated by `ValueMap` were not being followed in all
places. This was fixed in this commit. There is no clean way to decouple
this change because this bug only seems to arise when the relaxed
version of invariant load hoisting was enabled.
Differential Revision: https://reviews.llvm.org/D35120
llvm-svn: 307907
For multiprecision arithmetic on MIPS, rather than using ISD::ADDE / ISD::ADDC,
get SelectionDAG to break down the operation into ISD::ADDs and ISD::SETCCs.
For MIPS, only the DSP ASE has a carry flag, so in the general case it is not
useful to directly support ISD::{ADDE, ADDC, SUBE, SUBC} nodes.
Also improve the generation code in such cases for targets with
TargetLoweringBase::ZeroOrOneBooleanContent by directly using the result of the
comparison node rather than using it in selects. Similarly for ISD::SUBE /
ISD::SUBC.
Address optimization breakage by moving the generation of MIPS specific integer
multiply-accumulate nodes to before legalization.
This revolves PR32713 and PR33424.
Thanks to Simonas Kazlauskas and Pirama Arumuga Nainar for reporting the issue!
Reviewers: slthakur
Differential Revision: https://reviews.llvm.org/D33494
The previous version of this patch was too aggressive in producing fused
integer multiple-addition instructions.
llvm-svn: 307906
This boils down to not crashing in reg bank select due to the lack of
register operands on this instruction, and adding some tests. The
instruction selection is already covered by the TableGen'erated code.
llvm-svn: 307904
property and check for incompatible attributes
This commit changes the way ambiguous property synthesis (i.e. when synthesizing
a property that's declared in multiple protocols) is performed. Previously,
Clang synthesized the first property that was found. This lead to problems when
the property was synthesized in a class that conformed to two protocols that
declared that property and a second protocols had a 'readwrite' declaration -
the setter was not synthesized so the class didn't really conform to the second
protocol and user's code would crash at runtime when they would try to set the
property.
This commit ensures that a first readwrite property is selected. This is a
semantic change that changes users code in this manner:
```
@protocol P @property(readonly) int p; @end
@protocol P2 @property(readwrite) id p; @end
@interface I <P2> @end
@implementation I
@syntesize p; // Users previously got a warning here, and Clang synthesized
// readonly 'int p' here. Now Clang synthesizes readwrite 'id' p..
@end
```
To ensure that this change is safe, the warning about incompatible types is
promoted to an error when this kind of readonly/readwrite ambiguity is detected
in the @implementation. This will ensure that previous code that had this subtle
bug and ignored the warning now will fail to compile with an error, and users
should not get suprises at runtime once they resolve the error.
The commit also extends the ambiguity checker, and now it can detect conflicts
among the different property attributes. An error diagnostic is used for
conflicting attributes, to ensure that the user won't get "suprises" at runtime.
ProtocolPropertyMap is removed in favour of a a set + vector because the map's
order of iteration is non-deterministic, so it couldn't be used to select the
readwrite property.
rdar://31579994
Differential Revision: https://reviews.llvm.org/D35268
llvm-svn: 307903
Summary:
This patch replaces a bunch of iterator-based for loops with range-based
for loops. There are 2 iterator-based loops left in this file in
removeNotPreservedAnalysis, but I think those cannot be replaced by
range-based for loops as they modify the container they are iterating
over.
Unless I missed something, this schould be a NFC and I would appreciate
if someone could have a quick look to confirm that.
Reviewers: chandlerc, pcc, jhenderson
Reviewed By: jhenderson
Subscribers: llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D35310
llvm-svn: 307902
The first bit is actually the "untied" flag. That is why the condition was
wrong and has to be inverted to set the flag correctly.
Found and initial patch by Simon Convent!
llvm-svn: 307899
a declaration at location and for class that searches for all occurrences of
a specific declaration
This commit uses a single RecursiveSymbolVisitor class for both
USRLocFindingASTVisitor and NamedDeclOccurrenceFindingVisitor to avoid duplicate
traversal code. It also traverses nested name specifier locs in the new class
and remove the separate matching step.
Differential Revision: https://reviews.llvm.org/D34949
llvm-svn: 307898
This patch tidies up and organises ARM.td
so that it is easier to understandand
and extend in the future.
Reviewed by: @hahn, @rovka
Differential Revision: https://reviews.llvm.org/D35248
llvm-svn: 307897
Summary:
Similar to X86, it should be safe to inline callees if their
target-features are a subset of the caller. As some subtarget features
provide different instructions depending on whether they are set or
unset (e.g. ThumbMode and ModeSoftFloat), we use a whitelist of
target-features describing hardware capabilities only.
Reviewers: kristof.beyls, rengolin, t.p.northover, SjoerdMeijer, peter.smith, silviu.baranga, efriedma
Reviewed By: SjoerdMeijer, efriedma
Subscribers: dschuff, efriedma, aemerson, sdardis, javed.absar, arichardson, eraman, llvm-commits
Differential Revision: https://reviews.llvm.org/D34697
llvm-svn: 307889
As far as I can tell we can simply distinguish based on features rather than model number. Many of the strings we were previously using are treated the same by the backend.
llvm-svn: 307884
devirtualized.
The code to detect devirtualized calls is already in IRGen, so move the
code to lib/AST and make it a shared utility between Sema and IRGen.
This commit fixes a linkage error I was seeing when compiling the
following code:
$ cat test1.cpp
struct Base {
virtual void operator()() {}
};
template<class T>
struct Derived final : Base {
void operator()() override {}
};
Derived<int> *d;
int main() {
if (d)
(*d)();
return 0;
}
rdar://problem/33195657
Differential Revision: https://reviews.llvm.org/D34301
llvm-svn: 307883
All other code in MachODump.cpp uses the same comparison,
((r_length & 0x1) == 1), for distinguishing between the two,
while the code in llvm-objdump.cpp seemed to be incorrect.
Differential Revision: https://reviews.llvm.org/D35240
llvm-svn: 307882
Summary:
We were missing many feature flags that newer gcc supports and we had our own set of feature flags that gcc didnt' support that were overlapping. Clang's implementation assumes gcc's features list so a mismatch here is problematic.
I've also matched the cpu type/subtype lists with gcc and removed all the cpus that gcc doesn't support. I've also removed the fallback autodetection logic that was taken from Host.cpp. It was the main reason we had extra feature flags relative to gcc. I don't think gcc does this in libgcc.
Once this support is in place we can consider implementing __builtin_cpu_is in clang. This could also be needed for function dispatching that Erich Keane is working on.
Reviewers: echristo, asbirlea, RKSimon, erichkeane, zvi
Reviewed By: asbirlea
Subscribers: dberris, llvm-commits
Differential Revision: https://reviews.llvm.org/D35214
llvm-svn: 307878
Summary: Add target hooks for printing and parsing target MMO flags.
Targets may override getSerializableMachineMemOperandTargetFlags() to
return a mapping from string to flag value for target MMO values that
should be serialized/parsed in MIR output.
Add implementation of this hook for AArch64 SuppressPair MMO flag.
Reviewers: bogner, hfinkel, qcolombet, MatzeB
Subscribers: mcrosier, javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D34962
llvm-svn: 307877
Tests would go flaky if;
1. Using %T (not %t)
2. Put a file with common name like header.h into %T
3. Other tests (eg. misc-unused-parameters.cpp) are doing as well
We should avoid using %T unless it really makes sense.
llvm-svn: 307876
I don't know a reliable way of crafting a test for this case,
but I'll try a little harder. In the meanwhile, let's get the
bots green again. Please note this will be tested by `check-cfi`
once r307215 relands.
llvm-svn: 307874
Code to convert MachO - specific section debug section names to standard DWARF v5
section names was in the wrong place.
Differential Revision: https://reviews.llvm.org/D35321
llvm-svn: 307872
The instrumentation tracks the return address and not that of the
call so we remove one to compensate. Thanks for Peter Collingbourne
for confirming the analysis of the problem.
llvm-svn: 307871
I have a patch to let DwarfContext defer to lld for getting section
contents and relocations.
That is a pretty big performance improvement.
This is just a refactoring to make that easier to do.
This change makes the *creation* of gdb index a dedicated step and
makes that templated. That is so that we can uses Elf_Rel in the code.
llvm-svn: 307867
When we fail to sink an instruction, we must make sure not to modify
the function; otherwise, we end up in an infinite loop because
CodeGenPrepare iterates until it doesn't make any changes.
Fixes https://bugs.llvm.org/show_bug.cgi?id=33608 .
llvm-svn: 307866
On iOS/AArch64, the address space is very limited and has a dynamic maximum address based on the configuration of the device. We're already using a dynamic shadow, and we find a large-enough "gap" in the VM where we place the shadow memory. In some cases and some device configuration, we might not be able to find a large-enough gap: E.g. if the main executable is linked against a large number of libraries that are not part of the system, these libraries can fragment the address space, and this happens before ASan starts initializing.
This patch has a solution, where we have a "backup plan" when we cannot find a large-enough gap: We will restrict the address space (via MmapFixedNoAccess) to a limit, for which the shadow limit will fit.
Differential Revision: https://reviews.llvm.org/D35098
llvm-svn: 307865
This is an incremental change to the promotion feature.
There are two problems with the current behavior:
1) loops with multiple exiting blocks are totally disabled
2) a counter update can only be promoted one level up in
the loop nest -- which does help much for short trip
count inner loops inside a high trip-count outer loops.
Due to this limitation, we still saw very large profile
count fluctuations from run to run for the affected loops
which are usually very hot.
This patch adds the support for promotion counters iteratively
across the loop nest. It also turns on the promotion for
loops with multiple exiting blocks (with a limit).
For single-threaded applications, the performance impact is flat
on average. For instance, dealII improves, but povray regresses.
llvm-svn: 307863
Summary:
Some programs run into a stack overflow issue. This change avoids this
problem by replacing the recursive algorithm with the iterative version.
Reviewers: MatzeB, t.p.northover, dblaikie
Reviewed By: MatzeB
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D35105
llvm-svn: 307860
Alex Lorenz