Bill Schmidt [Tue, 3 Feb 2015 16:16:01 +0000 (16:16 +0000)]
[PowerPC] Yet another approach to __tls_get_addr
This patch is a third attempt to properly handle the local-dynamic and
global-dynamic TLS models.
In my original implementation, calls to __tls_get_addr were hidden
from view until the asm-printer phase, at which point the underlying
branch-and-link instruction was created with proper relocations. This
mostly worked well, but I used some repellent techniques to ensure
that the TLS_GET_ADDR nodes at the SD and MI levels correctly received
input from GPR3 and produced output into GPR3. This proved to work
badly in the presence of multiple TLS variable accesses, with the
copies to and from GPR3 being scheduled incorrectly and generally
creating havoc.
In r221703, I addressed that problem by representing the calls to
__tls_get_addr as true calls during instruction lowering. This had
the advantage of removing all of the bad hacks and relying on the
existing call machinery to properly glue the copies in place. It
looked like this was going to be the right way to go.
However, as a side effect of the recent discovery of problems with
linker optimizations for TLS, we discovered cases of suboptimal code
generation with this strategy. The problem comes when tls_get_addr is
called for the same address, and there is a resulting CSE
opportunity. It turns out that in such cases MachineCSE will common
the addis/addi instructions that set up the input value to
tls_get_addr, but will not common the calls themselves. MachineCSE
does not have any machinery to common idempotent calls. This is
perfectly sensible, since presumably this would be done at the IR
level, and introducing calls in the back end isn't commonplace. In
any case, we end up with two calls to __tls_get_addr when one would
suffice, and that isn't good.
I presumed that the original design would have allowed commoning of
the machine-specific nodes that hid the __tls_get_addr calls, so as
suggested by Ulrich Weigand, I went back to that design and cleaned it
up so that the copies were properly held together by glue
nodes. However, it turned out that this didn't work either...the
presence of copies to physical registers kept the machine-specific
nodes from being commoned also.
All of which leads to the design presented here. This is a return to
the original design, except that no attempt is made to introduce
copies to and from GPR3 during instruction lowering. Virtual registers
are used until prior to register allocation. At that point, a special
pass is run that identifies the machine-specific nodes that hide the
tls_get_addr calls and introduces the copies to and from GPR3 around
them. The register allocator then coalesces these copies away. With
this design, MachineCSE succeeds in commoning tls_get_addr calls where
possible, and we get nice optimal code generation (better than GCC at
the moment, which does not common these calls).
One additional problem must be dealt with: After introducing the
mentions of the physical register GPR3, the aggressive anti-dependence
breaker sees opportunities to improve scheduling by selecting a
different register instead. Flags must be used on the instruction
descriptions to tell the anti-dependence breaker to keep its hands in
its pockets.
One thing missing from the original design was recording a definition
of the link register on the GET_TLS_ADDR nodes. Doing this was found
to be insufficient to force a stack frame to be created, which led to
looping behavior because two different LR values were stored at the
same address. This appears to have been an oversight in
PPCFrameLowering::determineFrameLayout(), which is repaired here.
Because MustSaveLR() returns true for calls to builtin_return_address,
this changed the expected behavior of
test/CodeGen/PowerPC/retaddr2.ll, which now stacks a frame but
formerly did not. I've fixed the test case to reflect this.
There are existing TLS tests to catch regressions; the checks in
test/CodeGen/PowerPC/tls-store2.ll proved to be too restrictive in the
face of instruction scheduling with these changes, so I fixed that
up.
I've added a new test case based on the PrettyStackTrace module that
demonstrated the original problem. This checks that we get correct
code generation and that CSE of the calls to __get_tls_addr has taken
place.
Sanjay Patel [Tue, 3 Feb 2015 15:37:18 +0000 (15:37 +0000)]
Improve test to actually check for a folded load.
This test was checking for lack of a "movaps" (an aligned load)
rather than a "movups" (an unaligned load). It also included
a store which complicated the checking.
Add specific CPU runs to prevent subtarget feature flag overrides
from inhibiting this optimization.
Renato Golin [Tue, 3 Feb 2015 11:20:45 +0000 (11:20 +0000)]
Adding AArch64 support to ASan instrumentation
For the time being, it is still hardcoded to support only the 39 VA bits
variant, I plan to work on supporting 42 and 48 VA bits variants, but I
don't have access to such hardware at the moment.
Craig Topper [Tue, 3 Feb 2015 11:03:43 +0000 (11:03 +0000)]
[X86] Add Requires[In64BitMode] around MOVSX64rr32/MOVSX64rm32. This makes it more strictly mutexed with the ARPL instruction 32-bit mode. Helps with some disassembler changes I'm experimenting with. Should be NFC.
Lang Hames [Tue, 3 Feb 2015 06:14:06 +0000 (06:14 +0000)]
[PBQP Regalloc] Pre-spill vregs that have no legal physregs.
The PBQP::RegAlloc::MatrixMetadata class assumes that matrices have at least two
rows/columns (for the spill option plus at least one physreg). This patch
ensures that that invariant is met by pre-spilling vregs that have no physreg
options so that no node (and no corresponding edges) need be added to the PBQP
graph.
This fixes a bug in an out-of-tree target that was identified by Jonas Paulsson.
Thanks for tracking this down Jonas!
Justin Bogner [Tue, 3 Feb 2015 00:20:11 +0000 (00:20 +0000)]
InstrProf: Simplify RawCoverageMappingReader's API slightly
This is still kind of a weird API, but dropping the (partial) update
of the passed in CoverageMappingRecord makes it a little easier to
understand and use.
Remove unused class variables and update all callers/uses from
the HexagonSplitTFRCondSet pass. Use the subtarget off the machine
function at the same time.
Migrate the HexagonSplitConst32AndConst64 pass from TargetMachine
based getSubtarget to the one cached on the MachineFunction.
Remove unused class variables and update all callers/uses.
LLVM ToT produces poor MMX code compared to 3.5. However, part of the previous
functionality can be achieved by using -x86-experimental-vector-widening-legalization.
Add tests to be sure we don't regress again.
IR: Allow GenericDebugNode construction from MDString
Allow `GenericDebugNode` construction directly from `MDString`, rather
than requiring `StringRef`s. I've refactored the `StringRef`
constructors to use these. There's no real functionality change here,
except for exposing the lower-level API.
The purpose of this is to simplify construction of string operands when
reading bitcode. It's unnecessarily indirect to parse an `MDString` ID,
lookup the `MDString` in the bitcode reader list, get the `StringRef`
out of that, and then have `GenericDebugNode::getImpl()` use
`MDString::get()` to acquire the original `MDString`. Instead, this
allows the bitcode reader to directly pass in the `MDString`.
Lang Hames [Mon, 2 Feb 2015 19:51:18 +0000 (19:51 +0000)]
[Orc] Make OrcMCJITReplacement::addObject calls transfer buffer ownership to the
ObjectLinkingLayer.
There are a two of overloads for addObject, one of which transfers ownership of
the underlying buffer to OrcMCJITReplacement. This commit makes the ownership
transfering version pass ownership down to the ObjectLinkingLayer in order to
prevent the issue described in r227778.
I think this commit will fix the sanitizer bot failures that necessitated the
removal of the load-object-a.ll regression test in r227785, so I'm reinstating
that test.
Move debug-info-centred `Metadata` subclasses into their own
header/source file. A couple of private template functions are needed
from both `Metadata.cpp` and `DebugInfoMetadata.cpp`, so I've moved them
to `lib/IR/MetadataImpl.h`.
Adrian Prantl [Mon, 2 Feb 2015 18:31:58 +0000 (18:31 +0000)]
Debug Info: Relax assertion in isUnsignedDIType() to allow floats to be
described by integer constants. This is a bit ugly, but if the source
language allows arbitrary type casting, the debug info must follow suit.
For example:
void foo() {
float a;
*(int *)&a = 0;
}
For the curious: SROA replaces the float alloca with an i32 alloca, which
is then optimized away and described via dbg.value(i32 0, ...).
Tom Stellard [Mon, 2 Feb 2015 18:02:28 +0000 (18:02 +0000)]
R600/SI: 64-bit and larger memory access must be at least 4-byte aligned
This is true for SI only. CI+ supports unaligned memory accesses,
but this requires driver support, so for now we disallow unaligned
accesses for all GCN targets.
Jan Wen Voung [Mon, 2 Feb 2015 16:56:50 +0000 (16:56 +0000)]
Fix ARM peephole optimizeCompare to avoid optimizing unsigned cmp to 0.
Summary:
Previously it only avoided optimizing signed comparisons to 0.
Sometimes the DAGCombiner will optimize the unsigned comparisons
to 0 before it gets to the peephole pass, but sometimes it doesn't.
Erik Eckstein [Mon, 2 Feb 2015 12:45:34 +0000 (12:45 +0000)]
Fix: SLPVectorizer crashes with assertion when vectorizing a cmp instruction.
The commit r225977 uncovered this bug. The problem was that the vectorizer tried to
read the second operand of an already deleted instruction.
The bug didn't show up before r225977 because the freed memory still contained a non-null pointer.
With r225977 deletion of instructions is delayed and the read operand pointer is always null.
Lang Hames [Mon, 2 Feb 2015 04:34:02 +0000 (04:34 +0000)]
[Orc] Remove the OwnedModules list from OrcMCJITReplacement and use
ExecutionEngine's Modules list instead.
This makes the owned modules visibile to ExecutionEngine. In particular,
it is required for ExecutionEngine::runStaticConstructorsAndDestructors to
work.
Regression tests for Orc (which test this issue) will be committed shortly.
Lang Hames [Mon, 2 Feb 2015 04:32:17 +0000 (04:32 +0000)]
[Orc] Make the ObjectLinkingLayer take ownership of object files until
finalization time.
As currently implemented, RuntimeDyldELF requires the original object
file to be avaible when relocations are being resolved. This patch
ensures that the ObjectLinkingLayer preserves it until then. In the
future RuntimeDyldELF should be rewritten to remove this requirement, at
which point this patch can be reverted.
Regression test cases for Orc (which include coverage of this bug) will
be committed shortly.
Hal Finkel [Sun, 1 Feb 2015 22:58:46 +0000 (22:58 +0000)]
[PowerPC] Put PPCEarlyReturn into its own source file
PPCInstrInfo.cpp has ended up containing several small MI-level passes, and
this is making the file harder to read than necessary. Split out
PPCEarlyReturn into its own source file. NFC.
Now that PPCInstrInfo.cpp does not also contain pass implementations, I hope
that it will be slightly less unwieldy.
Hal Finkel [Sun, 1 Feb 2015 22:01:29 +0000 (22:01 +0000)]
[PowerPC] Put PPCVSXCopy into its own source file
PPCInstrInfo.cpp has ended up containing several small MI-level passes, and
this is making the file harder to read than necessary. Split out
PPCVSXCopy into its own source file. NFC.
Hal Finkel [Sun, 1 Feb 2015 21:51:22 +0000 (21:51 +0000)]
[PowerPC] Put PPCVSXFMAMutate into its own source file
PPCInstrInfo.cpp has ended up containing several small MI-level passes, and
this is making the file harder to read than necessary. Split out
PPCVSXFMAMutate into its own source file. NFC.
Hal Finkel [Sun, 1 Feb 2015 21:20:58 +0000 (21:20 +0000)]
[PowerPC] Remove the PPCVSXCopyCleanup pass
This MI-level pass was necessary when VSX support was first being developed,
specifically, before the ABI code had been updated to use VSX registers for
arguments (the register assignments did not change, in a physical sense, but
the VSX super-registers are now used). Unfortunately, I never went back and
removed this pass after that was done. I believe this code is now effectively
dead.
Hal Finkel [Sun, 1 Feb 2015 20:16:10 +0000 (20:16 +0000)]
[PowerPC] Add implicit ops to conditional returns in PPCEarlyReturn
When PPCEarlyReturn, it should really copy implicit ops from the old return
instruction to the new one. This currently does not matter much, because we run
PPCEarlyReturn very late in the pipeline (there is nothing to do DCE on
definitions of those registers). However, for completeness, we should do it
anyway.
Noticed by inspection (and there should be no functional change); thus, no
test case.
Hal Finkel [Sun, 1 Feb 2015 19:07:41 +0000 (19:07 +0000)]
[PowerPC] VSX stores don't also read
The VSX store instructions were also picking up an implicit "may read" from the
default pattern, which was an intrinsic (and we don't currently have a way of
specifying write-only intrinsics).
This was causing MI verification to fail for VSX spill restores.
Hal Finkel [Sun, 1 Feb 2015 17:52:16 +0000 (17:52 +0000)]
[PowerPC] Better scheduling for isel on P7/P8
isel is actually a cracked instruction on the P7/P8, and must start a dispatch
group. The scheduling model should reflect this so that we don't bunch too many
of them together when possible.
Thanks to Bill Schmidt and Pat Haugen for helping to sort this out.
Adam Nemet [Sun, 1 Feb 2015 16:56:15 +0000 (16:56 +0000)]
[LoopVectorize] Move LoopAccessAnalysis to its own module
Other than moving code and adding the boilerplate for the new files, the code
being moved is unchanged.
There are a few global functions that are shared with the rest of the
LoopVectorizer. I moved these to the new module as well (emitLoopAnalysis,
stripIntegerCast, replaceSymbolicStrideSCEV) along with the Report class used
by emitLoopAnalysis. There is probably room for further improvement in this
area.
I kept DEBUG_TYPE "loop-vectorize" because it's used as the PassName with
emitOptimizationRemarkAnalysis. This will obviously have to change.
NFC. This is part of the patchset that splits out the memory dependence logic
from LoopVectorizationLegality into a new class LoopAccessAnalysis.
LoopAccessAnalysis will be used by the new Loop Distribution pass.
Adam Nemet [Sun, 1 Feb 2015 16:56:11 +0000 (16:56 +0000)]
[LoopVectorize] Move RuntimePointerCheck under LoopAccessAnalysis
This class needs to remain public because it's used by
LoopVectorizationLegality::addRuntimeCheck.
NFC. This is part of the patchset that splits out the memory dependence logic
from LoopVectorizationLegality into a new class LoopAccessAnalysis.
LoopAccessAnalysis will be used by the new Loop Distribution pass.
Adam Nemet [Sun, 1 Feb 2015 16:56:09 +0000 (16:56 +0000)]
[LoopVectorize] Pass parameters explicitly to MemoryDepChecker
Rather than using globals use a structure to pass parameters from the
vectorizer. This prepares the class to be moved outside the LoopVectorizer.
It's not great how all this is passed through in LoopAccessAnalysis but this
is all expected to change once the class start servicing the Loop Distribution
pass as well where some of these parameters make no sense.
NFC. This is part of the patchset that splits out the memory dependence logic
from LoopVectorizationLegality into a new class LoopAccessAnalysis.
LoopAccessAnalysis will be used by the new Loop Distribution pass.
Adam Nemet [Sun, 1 Feb 2015 16:56:05 +0000 (16:56 +0000)]
[LoopVectorize] Make hasVectorInstrinsicScalarOpd inline
VectorUtils.h needs to be included in LoopAccessAnalysis.cpp for
getIntrinsicIDForCall but hasVectorInstrinsicScalarOpd is not used by this
module.
NFC. This is part of the patchset that splits out the memory dependence logic
from LoopVectorizationLegality into a new class LoopAccessAnalysis.
LoopAccessAnalysis will be used by the new Loop Distribution pass.
[X86] Convert esp-relative movs of function arguments to pushes, step 2
This moves the transformation introduced in r223757 into a separate MI pass.
This allows it to cover many more cases (not only cases where there must be a
reserved call frame), and perform rudimentary call folding. It still doesn't
have a heuristic, so it is enabled only for optsize/minsize, with stack
alignment <= 8, where it ought to be a fairly clear win.
Adam Nemet [Sun, 1 Feb 2015 16:56:04 +0000 (16:56 +0000)]
[LoopVectorize] Split out LoopAccessAnalysis from LoopVectorizationLegality
Move the canVectorizeMemory functionality from LoopVectorizationLegality to a
new class LoopAccessAnalysis and forward users.
Currently the collection of the symbolic stride information is kept with
LoopVectorizationLegality and it becomes an input to LoopAccessAnalysis.
NFC. This is part of the patchset that splits out the memory dependence logic
from LoopVectorizationLegality into a new class LoopAccessAnalysis.
LoopAccessAnalysis will be used by the new Loop Distribution pass.
Adam Nemet [Sun, 1 Feb 2015 16:56:02 +0000 (16:56 +0000)]
[LoopVectorize] Add accessors for Num{Stores,Loads,PredStores} in AccessAnalysis
These members are moving to LoopAccessAnalysis. The accessors help to hide
this.
NFC. This is part of the patchset that splits out the memory dependence logic
from LoopVectorizationLegality into a new class LoopAccessAnalysis.
LoopAccessAnalysis will be used by the new Loop Distribution pass.
Adam Nemet [Sun, 1 Feb 2015 16:56:00 +0000 (16:56 +0000)]
[LoopVectorize] Rename the Report class to VectorizationReport
This class will become public in the new LoopAccessAnalysis header so the name
needs to be more global.
NFC. This is part of the patchset that splits out the memory dependence logic
from LoopVectorizationLegality into a new class LoopAccessAnalysis.
LoopAccessAnalysis will be used by the new Loop Distribution pass.
Adam Nemet [Sun, 1 Feb 2015 16:55:58 +0000 (16:55 +0000)]
[LoopVectorize] Factor out duplicated code into Report::emitAnalysis
The logic in emitAnalysis is duplicated across multiple functions. This
splits it into a function. Another use will be added by the patchset.
NFC. This is part of the patchset that splits out the memory dependence logic
from LoopVectorizationLegality into a new class LoopAccessAnalysis.
LoopAccessAnalysis will be used by the new Loop Distribution pass.
Adam Nemet [Sun, 1 Feb 2015 16:55:56 +0000 (16:55 +0000)]
[LoopVectorize] Split out RuntimePointerCheck from LoopVectorizationLegality
RuntimePointerCheck will be used through LoopAccessAnalysis in
LoopVectorizationLegality. Later in the patchset it will become a local class
of LoopAccessAnalysis.
NFC. This is part of the patchset that splits out the memory dependence logic
from LoopVectorizationLegality into a new class LoopAccessAnalysis.
LoopAccessAnalysis will be used by the new Loop Distribution pass.
Hal Finkel [Sun, 1 Feb 2015 15:03:28 +0000 (15:03 +0000)]
[PowerPC] Make r2 allocatable on PPC64/ELF for some leaf functions
The TOC base pointer is passed in r2, and we normally reserve this register so
that we can depend on it being there. However, for leaf functions, and
specifically those leaf functions that don't do any TOC access of their own
(which is generally due to accessing the constant pool, using TLS, etc.),
we can treat r2 as an ordinary callee-saved register (it must be callee-saved
because, for local direct calls, the linker will not insert any save/restore
code).
The allocation order has been changed slightly for PPC64/ELF systems to put r2
at the end of the list (while leaving it near the beginning for Darwin systems
to prevent unnecessary output changes). While r2 is allocatable, using it still
requires spill/restore traffic, and thus comes at the end of the list.
Sylvestre Ledru [Sun, 1 Feb 2015 14:55:43 +0000 (14:55 +0000)]
Fix some bashims. More information on https://wiki.ubuntu.com/DashAsBinSh. Reported initially on https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=772302 & https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=772301
projects / llvm / log