Previously, only signed comparisons were being handled.
Decrements could also be handled, but 'sub nuw %x, 1' is currently canonicalized to
'add %x, -1' in InstCombineAddSub, losing the nuw flag. Removing that canonicalization
seems like it might have far-reaching ramifications so I kept this simple for now.
Tim Northover [Fri, 13 Jan 2017 23:11:37 +0000 (23:11 +0000)]
[GlobalISel] track predecessor mapping during switch lowering.
Correctly populating Machine PHIs relies on knowing exactly how the IR level
CFG was lowered to MachineIR. This needs to be tracked by any translation
phases that meddle (currently only SwitchInst handling).
Daniel Berlin [Fri, 13 Jan 2017 22:40:01 +0000 (22:40 +0000)]
NewGVN: Move leaders around properly to ensure we have a canonical dominating leader. Fixes PR 31613.
Summary:
This is a testcase where phi node cycling happens, and because we do
not order the leaders by domination or anything similar, the leader
keeps changing.
Using std::set for the members is too expensive, and we actually don't
need them sorted all the time, only at leader changes.
We could keep both a set and a vector, and keep them mostly sorted and
resort as necessary, or use a set and a fibheap, but all of this seems
premature.
After running some statistics, we are able to avoid the vast majority
of sorting by keeping a "next leader" field. Most congruence classes only have
leader changes once or twice during GVN.
Greg Clayton [Fri, 13 Jan 2017 21:08:18 +0000 (21:08 +0000)]
Cleanup how DWARFDie attributes are accessed and decoded.
Removed all DWARFDie::getAttributeValueAs*() calls.
Renamed:
Optional<DWARFFormValue> DWARFDie::getAttributeValue(dwarf::Attribute);
To:
Optional<DWARFFormValue> DWARFDie::find(dwarf::Attribute);
Added:
Optional<DWARFFormValue> DWARFDie::findRecursively(dwarf::Attribute);
All decoding of Optional<DWARFFormValue> values are now done using the dwarf::to*() functions from DWARFFormValue.h:
Old code:
auto DeclLine = DWARFDie.getAttributeValueAsSignedConstant(DW_AT_decl_line).getValueOr(0);
New code:
auto DeclLine = toUnsigned(DWARFDie.find(DW_AT_decl_line), 0);
This composition helps us since we can now easily do:
auto DeclLine = toUnsigned(DWARFDie.findRecursively(DW_AT_decl_line), 0);
This allows us to easily find attribute values in the current DIE only (the first new code above) or in any DW_AT_abstract_origin or DW_AT_specification Dies using the line above. Note that the code line length is shorter and more concise.
David L. Jones [Fri, 13 Jan 2017 21:02:41 +0000 (21:02 +0000)]
"Use" lambda captures which are otherwise only used in asserts. NFC
Summary:
The LLVM coding standards recommend "using" values that are only
needed by asserts:
http://llvm.org/docs/CodingStandards.html#assert-liberally
Without this change, LLVM cannot bootstrap with -Werror as the second
stage fails with this new warning:
https://reviews.llvm.org/rL291905
See also the previous fixes:
https://reviews.llvm.org/rL291916
https://reviews.llvm.org/rL291939
https://reviews.llvm.org/rL291940
https://reviews.llvm.org/rL291941
Artem Belevich [Fri, 13 Jan 2017 20:56:17 +0000 (20:56 +0000)]
[NVPTX] Added support for half-precision floating point.
Only scalar half-precision operations are supported at the moment.
- Adds general support for 'half' type in NVPTX.
- fp16 math operations are supported on sm_53+ GPUs only
(can be disabled with --nvptx-no-f16-math).
- Type conversions to/from fp16 are supported on all GPU variants.
- On GPU variants that do not have full fp16 support (or if it's disabled),
fp16 operations are promoted to fp32 and results are converted back
to fp16 for storage.
Michael Liao [Fri, 13 Jan 2017 18:28:30 +0000 (18:28 +0000)]
[SCEV] Limit recursion depth of constant evolving.
- For a loop body with VERY complicated exit condition evaluation, constant
evolving may run out of stack on platforms such as Windows. Need to limit the
recursion depth.
Ivan Krasin [Fri, 13 Jan 2017 17:30:10 +0000 (17:30 +0000)]
Fix UBSan bots by blacklisting bits/stl_tree.h.
Summary:
libstdc++ has some undefined behavior in bits/stl_tree.h that
has recently became excercised by some of the LLVM code.
Given that fixing libstdc++ will take years, adding the file
into a blacklist to fix bots seems like a necessity.
Ivan Krasin [Fri, 13 Jan 2017 16:45:15 +0000 (16:45 +0000)]
Revert r291903 and r291898. Reason: they break check-lld on the bots.
Summary:
Revert [ARM] Fix ubig32_t read in ARMAttributeParser
Now using support functions to read data instead of trying to
perform casts.
===========================================================
Revert [ARM] Enable objdump to construct triple for ARM
Now that The ARMAttributeParser has been moved into the library,
it has been modified so that it can parse the attributes without
printing them and stores them in a map. ELFObjectFile now queries
the attributes to fill out the architecture details of a provided
triple for 'arm' and 'thumb' targets. llvm-objdump uses this new
functionality.
GCC changes the CC between the user-code and the builtins based on the
value of `-target` rather than `-mfloat-abi`. When a HF target is used,
the VFP variant of the AAPCS CC is used. Otherwise, the AAPCS variant
is used. In all cases, the AEABI functions use the AAPCS CC. Adjust
the calling convention based on the target.
Simon Pilgrim [Fri, 13 Jan 2017 13:16:19 +0000 (13:16 +0000)]
[X86][AVX512] Add support for variable ASHR v2i64/v4i64 support without VLX
Use v8i64 variable ASHR instructions if we don't have VLX.
This is a reduced version of D28537 that just adds support for variable shifts - I'll continue with that patch (for just constant/uniform shifts) once I've fixed the type legalization issue in avx512-cvt.ll.
Daniel Sanders [Fri, 13 Jan 2017 11:50:34 +0000 (11:50 +0000)]
[aarch64][globalisel] Move getValueMapping/getCopyMapping to AArch64GenRegisterBankInfo. NFC.
Summary:
We did lose a little specificity in the assertion messages for the
PartialMappingIdx enumerators in this change but this was necessary to
avoid unnecessary use of 'public:' and we haven't lost anything that
can't be discovered easily in lldb. Once this is tablegen-erated we could
also safely remove the assertions.
Sam Parker [Fri, 13 Jan 2017 11:04:21 +0000 (11:04 +0000)]
[ARM] Enable objdump to construct triple for ARM
Now that The ARMAttributeParser has been moved into the library,
it has been modified so that it can parse the attributes without
printing them and stores them in a map. ELFObjectFile now queries
the attributes to fill out the architecture details of a provided
triple for 'arm' and 'thumb' targets. llvm-objdump uses this new
functionality.
Diana Picus [Fri, 13 Jan 2017 10:37:37 +0000 (10:37 +0000)]
[ARM] CodeGen: Replace AddDefaultT1CC and AddNoT1CC. NFC
For AddDefaultT1CC, we add a new helper t1CondCodeOp, which creates the
appropriate register operand. For AddNoT1CC, we use the existing condCodeOp
helper - we only had two uses of AddNoT1CC, so at this point it's probably not
worth having yet another helper just for them.
Diana Picus [Fri, 13 Jan 2017 10:18:01 +0000 (10:18 +0000)]
[ARM] CodeGen: Remove AddDefaultCC. NFC.
Replace all uses of AddDefaultCC with add(condCodeOp()).
The transformation has been done automatically with a custom tool based on Clang
AST Matchers + RefactoringTool.
Diana Picus [Fri, 13 Jan 2017 09:37:56 +0000 (09:37 +0000)]
[ARM] CodeGen: Remove AddDefaultPred. NFC.
Replace all uses of AddDefaultPred with MachineInstrBuilder::add(predOps()).
This makes the code building MachineInstrs more readable, because it allows us
to write code like:
This commit also adds the predOps helper in the ARM backend, as well as the add
method taking a variable number of operands to the MachineInstrBuilder.
The transformation has been done mostly automatically with a custom tool based
on Clang AST Matchers + RefactoringTool.
[X86][AVX512] Adding missing shuffle lowering to blend mask instructions
Some shuffles can be lowered to blend mask instruction (VPBLENDMB/VPBLENDMW/VPBLENDMD/VPBLENDMQ) .
In this patch, I added new pattern match for this case.
Craig Topper [Fri, 13 Jan 2017 07:28:56 +0000 (07:28 +0000)]
[AVX-512] Remove unmasked BLENDM instructions from the wrong load folding table. The unmasked versions read memory from operand 2, but were in the operand 3 table.
These aren't the most interesting set of blendm instructions as the unmasked version isn't useful. We were also missing the B and W forms. I'll add the masked versions of all sizes in a future patch.
Craig Topper [Fri, 13 Jan 2017 06:26:18 +0000 (06:26 +0000)]
[IR] Don't call assertModuleIsMaterialized in release builds
Summary:
To fix a release vs debug build linking error, r259695 made the body of assertModuleIsMaterialized empty if Value.cpp gets compiled in a release build. This way any code compiled as a debug build can still link against a release version of the function.
This patch takes this a step farther and removes all calls to it from Value.h in any code that includes it in a relase build.
This shrinks the opt binary on my macbook build by 17240 bytes.
Serge Pavlov [Fri, 13 Jan 2017 06:09:54 +0000 (06:09 +0000)]
Track validity of pass results
Running tests with expensive checks enabled exhibits some problems with
verification of pass results.
First, the pass verification may require results of analysis that are not
available. For instance, verification of loop info requires results of dominator
tree analysis. A pass may be marked as conserving loop info but does not need to
be dependent on DominatorTreePass. When a pass manager tries to verify that loop
info is valid, it needs dominator tree, but corresponding analysis may be
already destroyed as no user of it remained.
Another case is a pass that is skipped. For instance, entities with linkage
available_externally do not need code generation and such passes are skipped for
them. In this case result verification must also be skipped.
To solve these problems this change introduces a special flag to the Pass
structure to mark passes that have valid results. If this flag is reset,
verifications dependent on the pass result are skipped.
Easwaran Raman [Fri, 13 Jan 2017 01:34:00 +0000 (01:34 +0000)]
ProfileSummaryInfo improvements.
* Add is{Hot|Cold}CallSite methods
* Fix a bug in isHotBB where it was looking for MD_prof on a return instruction
* Use MD_prof data only if sample profiling was used to collect profiles.
* Add an unit test to ProfileSummaryInfo
NAKAMURA Takumi [Fri, 13 Jan 2017 01:13:10 +0000 (01:13 +0000)]
Revert r291503, "Lift the 10-type limit for AlignedCharArrayUnion", and followings.
r291503, "Lift the 10-type limit for AlignedCharArrayUnion"
r291514, "Fix MSVC build of AlignedCharArrayUnion"
r291515, "Revert the attempt to optimize the constexpr functions. MSVC does not handle this yet"
r291519, "Try once again to fix the MSVC build of AlignedCharArrayUnion"
Teresa Johnson [Thu, 12 Jan 2017 22:04:45 +0000 (22:04 +0000)]
[ThinLTO] Import static functions from the same module as caller
Summary:
We can sometimes end up with multiple copies of a local function that
have the same GUID in the index. This happens when there are local
functions with the same name that are in different source files with the
same name (but in different directories), and they were compiled in
their own directory so had the same path at compile time.
In this case make sure we import the copy in the caller's module. While
it isn't a correctness problem (the renamed reference which is based on the
module IR hash will be unique since the module must have had an
externally visible function that was imported), importing the wrong copy
will result in lost performance opportunity since it won't be referenced
and inlined.
Chris Bieneman [Thu, 12 Jan 2017 21:35:21 +0000 (21:35 +0000)]
[ObjectYAML] Pull yaml2dwarf out of yaml2obj for reuse
This patch pulls the yaml2dwarf code out of yaml2obj into a new set of DWARF emitter functions in the DWARFYAML namespace. This will enable the YAML->DWARF code to be used inside DWARF tests by populating the DWARFYAML structs and calling the Emitter functions.
Robert Lougher [Thu, 12 Jan 2017 20:34:35 +0000 (20:34 +0000)]
[DebugInfo] Handle same locations in DILocation::getMergedLocation
Revision 289661 introduced the function DILocation::getMergedLocation for
merging of debug locations. At the time is was simply a stub which always
returned no location. This patch modifies getMergedLocation to handle the
case where the two locations are the same or can't be discriminated.
Nikolai Bozhenov [Thu, 12 Jan 2017 19:54:27 +0000 (19:54 +0000)]
[X86] Replace AND+IMM64 with SRL/SHL
Emit SHRQ/SHLQ instead of ANDQ with a 64 bit constant mask if the result
is unused and the mask has only higher/lower bits set. For example, with
this patch LLVM emits
shrq $41, %rdi
je
instead of
movabsq $0xFFFFFE0000000000, %rcx
testq %rcx, %rdi
je
This reduces number of instructions, code size and register pressure.
The transformation is applied only for cases where the mask cannot be
encoded as an immediate value within TESTQ instruction.
[llvm-config] Fix obviously wrong code in parsing DyLib components.
The code parsing the string was using the offset returned from
StringRef::find() wrong, assuming it was relative to the staring
offset that is passed to the function, but the returned offset
is always relative to the beginning of the line.
This causes odd behaviour while parsing the component string.
Spotted thanks to the newly added test:
Nikolai Bozhenov [Thu, 12 Jan 2017 19:41:27 +0000 (19:41 +0000)]
[X86] Rename tests for bypassing slow division (NFC)
For tests on bypassing slow division there's no need to be
Atom-specific. The patch renames all tests on division bypassing
and makes their names more consistent:
atom-bypass-slow-division.ll -> bypass-slow-division-32.ll
(tests verifying correctness of divl-to-divb bypassing)
atom-bypass-slow-division-64.ll -> bypass-slow-division-64.ll
(tests verifying correctness of divq-to-divl bypassing)
slow-div.ll -> bypass-slow-division-tune.ll
(tests verifying that bypassing is enabled only when appropriate)
Nikolai Bozhenov [Thu, 12 Jan 2017 19:34:15 +0000 (19:34 +0000)]
[X86] Tune bypassing of slow division for Intel CPUs
64-bit integer division in Intel CPUs is extremely slow, much slower
than 32-bit division. On the other hand, 8-bit and 16-bit divisions
aren't any faster. The only important exception is Atom where DIV8
is fastest. Because of that, the patch
1) Enables bypassing of 64-bit division for Atom, Silvermont and
all big cores.
2) Modifies 64-bit bypassing to use 32-bit division instead of
16-bit one. This doesn't make the shorter division slower but
increases chances of taking it. Moreover, it's much more likely
to prove at compile-time that a value fits 32 bits and doesn't
require a run-time check (e.g. zext i32 to i64).
Switch some additional library call setup to be table driven. This
makes it more immediately obvious what the library call looks like.
This is important for ARM since the calling conventions for the builtins
change based on the target/libcall name. NFC
Daniel Sanders [Thu, 12 Jan 2017 16:11:23 +0000 (16:11 +0000)]
[globalisel] Move as much RegisterBank initialization to the constructor as possible
Summary:
The register bank is now entirely initialized in the constructor. However,
we still have the hardcoded number of register classes which will be
dealt with in the TableGen patch (D27338) since we do not have access
to this information to resolve this at this stage. The number of register
classes is known to the TRI and to TableGen but the RegisterBank
constructor is too early for the former and too late for the latter.
This will be fixed when the data is tablegen-erated.
Daniel Sanders [Thu, 12 Jan 2017 15:32:10 +0000 (15:32 +0000)]
[globalisel] Initialize RegisterBanks with static data.
Summary:
Refactor the RegisterBank initialization to use static data. This requires
GlobalISel implementations to rewrite calls to createRegisterBank() and
addRegBankCoverage() into a call to setRegBankData().
Out of tree targets can use diff 4 of D27807
(https://reviews.llvm.org/D27807?id=84117) to have addRegBankCoverage() dump
the register classes and other data that needs to be provided to
setRegBankData(). This is the method that was used to generate the static data
in this patch.
Tablegen-eration of this static data will follow after some refactoring.
Summary:
Memory Dependence Analysis was limited to return only local dependencies
for invariant.group handling. Now it returns NonLocal when it finds it
and then by asking getNonLocalPointerDependency we get found dep.
Thanks to this we are able to devirtualize loops!
void indirect(A &a, int n) {
for (int i = 0 ; i < n; i++)
a.foo();
}
void test(int n) {
A a;
indirect(a);
}
After inlining a.foo() will be changed to direct call, even if foo and A::A()
is external (but only if vtable definition is be available).
[XRay] Implement the `llvm-xray account` subcommand
Summary:
This is the third of a multi-part change to implement subcommands for
the `llvm-xray` tool.
Here we define the `account` subcommand which does simple function call
accounting, generating basic statistics on function calls we find in an
XRay log/trace. We support text output and csv output for this
subcommand.
This change also supports sorting, summing, and filtering the top N
results.
Part of this tool will later be turned into a library that could be used
for basic function call accounting.
Elad Cohen [Thu, 12 Jan 2017 06:49:03 +0000 (06:49 +0000)]
[X86][AVX512] Fix PR31515 - Do not flip vselect condition if it's not a vXi1 mask
r289653 added a case where `vselect <cond> <vector1> <all-zeros>`
is transformed to:
`vselect xor(cond, DAG.getConstant(1, DL, CondVT) <all-zeros> <vector1>`
This was not aimed to catch cases where Cond is not a vXi1
mask but it does. Moreover, when Cond type is VxiN (N > 1)
then xor(cond, DAG.getConstant(1, DL, CondVT) != NOT(cond).
This patch changes the above to xor with allones, and avoids
entering the case for non-mask Conds.
Chandler Carruth [Thu, 12 Jan 2017 03:49:07 +0000 (03:49 +0000)]
Make a test actually test what it set out to test.
This test seems to have largely been relying on asserts being tripped.
It had a very specific and somewhat uninteresting grep of the output,
but it never really did anything to cause SCEV to be preserved across
loop simplify, certainly not explicitly. And a later addition to it
actually added CHECK lines despite the test never running FileCheck.
Now we actually print SCEV before and after loop simplify to make sure
it is *changing* and being *updated*. Which seems to be much more likely
the point of the test.
projects / llvm / log