Pete Cooper [Wed, 10 Jun 2015 22:38:41 +0000 (22:38 +0000)]
Add User::growHungoffUses and use it to grow the hung off uses. NFC.
PhiNode, SwitchInst, LandingPad and IndirectBr all had virtually identical
logic for growing the hung off uses.
Move it to User so that they can all call a single shared implementation.
Their destructors were all empty after this change and were deleted. They all
have virtual clone_impl methods which can be used as vtable anchors.
Pete Cooper [Wed, 10 Jun 2015 22:38:38 +0000 (22:38 +0000)]
Delete User::dropHungOffUses and move it in to ~User which is the only caller. NFC.
Now that the subclasses which care about hung off uses let ~User clean it up,
there's no need for a separate method. Just inline it to ~User and delete it.
Pete Cooper [Wed, 10 Jun 2015 22:38:34 +0000 (22:38 +0000)]
Make User track whether a class has 'hung off uses' and delete them in its destructor.
Currently all of the logic for deleting hung off uses, which PHI/switch/etc use,
is in their classes.
This adds a bit to Value which tracks whether that user had hung off uses,
then User can be responsible for clearing them instead of the sub classes.
Note, the bit used here was taken from NumOperands which was 30-bits.
Given the reduction to 29 bits, and the average User being just over 100 bytes,
a single User with 29-bits of num operands would need 50GB of RAM for itself
so its reasonable to assume that 29-bits is enough for now.
This is a step towards hiding all the hung off uses logic in the User.
Pete Cooper [Wed, 10 Jun 2015 22:38:30 +0000 (22:38 +0000)]
Move the special Phi logic for hung off uses in to User::allocHungOffUses. NFC.
PhiNode's need to allocate space for an array of Use[N] and then BasicBlock*[N].
They had their own allocHungOffUses to handle all of this. This moves the logic
in to User::allocHungOffUses and PhiNode passes in a bool to say to allocate
the BB* space too.
ArgumentPromotion: Drop sret attribute on functions that are only called directly.
If the first argument to a function is a 'this' argument and the second
has the sret attribute, the ArgumentPromotion pass may promote the 'this'
argument to more than one argument, violating the IR constraint that 'sret'
may only be applied to the first or second argument.
Although this IR constraint is arguably unnecessary, it highlighted the fact
that ArgPromotion does not need to preserve this attribute. Dropping the
attribute reduces register pressure in the backend by avoiding the register
copy required by sret. Because sret implies noalias, we also replace the
former with the latter.
Sanjay Patel [Wed, 10 Jun 2015 20:32:21 +0000 (20:32 +0000)]
[x86] Add a reassociation optimization to increase ILP via the MachineCombiner pass
This is a reimplementation of D9780 at the machine instruction level rather than the DAG.
Use the MachineCombiner pass to reassociate scalar single-precision AVX additions (just a
starting point; see the TODO comments) to increase ILP when it's safe to do so.
The code is closely based on the existing MachineCombiner optimization that is implemented
for AArch64.
This patch should not cause the kind of spilling tragedy that led to the reversion of r236031.
Teresa Johnson [Wed, 10 Jun 2015 17:49:28 +0000 (17:49 +0000)]
Add new EliminateAvailableExternally module pass, which is performed in
O2 compiles just before GlobalDCE, unless we are preparing for LTO.
This pass eliminates available externally globals (turning them into
declarations), regardless of whether they are dead/unreferenced, since
we are guaranteed to have a copy available elsewhere at link time.
This enables additional opportunities for GlobalDCE.
If we are preparing for LTO (e.g. a -flto -c compile), the pass is not
included as we want to preserve available externally functions for possible
link time inlining. The FE indicates whether we are doing an -flto compile
via the new PrepareForLTO flag on the PassManagerBuilder.
Alexey Samsonov [Wed, 10 Jun 2015 17:37:38 +0000 (17:37 +0000)]
[GVN] Set proper debug locations for some instructions created by GVN.
Determining proper debug locations for instructions created in
PHITransAddr is tricky. We use a simple approach here and simply copy
debug locations from instructions computing load address to
"corresponding" instructions re-creating the address computation
in predecessor basic blocks.
This may not always be correct, given all the rearrangement and
simplification going on, and debug locations may jump around a lot,
as the basic blocks we copy locations between may be very far from
each other.
Still, this would work good in most simple cases (e.g. when chain
of address computing instruction is short, or our mapping turns out
to be 1-to-1), and we desire to have *some* reasonable debug locations
associated with newly inserted instructions.
See http://reviews.llvm.org/D10351 review thread for more details.
Igor Laevsky [Wed, 10 Jun 2015 12:31:53 +0000 (12:31 +0000)]
[StatepointLowering] Reuse stack slots across basic blocks
During statepoint lowering we can sometimes avoid spilling of the value if we know that it was already spilled for previous statepoint.
We were doing this by checking if incoming statepoint value was lowered into load from stack slot. This was working only in boundaries of one basic block.
But instead of looking at the lowered node we can look directly at the llvm-ir value and if it was gc.relocate (or some simple modification of it) look up stack slot for it's derived pointer and reuse stack slot from it. This allows us to look across basic block boundaries.
Reid Kleckner [Wed, 10 Jun 2015 01:02:30 +0000 (01:02 +0000)]
[WinEH] Emit .safeseh directives for all 32-bit exception handlers
Use a "safeseh" string attribute to do this. You would think we chould
just accumulate the set of personalities like we do on dwarf, but this
fails to account for the LSDA-loading thunks we use for
__CxxFrameHandler3. Each of those needs to make it into .sxdata as well.
The string attribute seemed like the most straightforward approach.
[RegisterScavenger] Fix handling of predicated instructions
Summary:
The RegisterScavenger explicitly ignores <kill> flags on operands of
predicated instructions and therefore assumes that such registers remain
live. When it then scavenges such a register, it inserts a spill of this
(killed) register. This is invalid code and gets flagged up by the
verifier.
Nowadays kill flags are set correctly on predicated instructions. This
patch makes the Scavenger respect them.
The bug has so far only been triggered by an internal pass, so I don't
have a test case unfortunately.
Jingyue Wu [Tue, 9 Jun 2015 21:50:32 +0000 (21:50 +0000)]
[NVPTX] fix a crash bug in NVPTXFavorNonGenericAddrSpaces
Summary:
We used to assume V->RAUW only modifies the operand list of V's user.
However, if V and V's user are Constants, RAUW may replace and invalidate V's
user entirely.
This patch fixes the above issue by letting the caller replace the
operand instead of calling RAUW on Constants.
Test Plan: @nested_const_expr and @rauw in access-non-generic.ll
llvm-lib is intended to be a lib.exe compatible utility that also
understands bitcode. The implementation lives in a library so that
lld can use it to implement /lib.
Reid Kleckner [Tue, 9 Jun 2015 21:42:19 +0000 (21:42 +0000)]
[WinEH] Add 32-bit SEH state table emission prototype
This gets all the handler info through to the asm printer and we can
look at the .xdata tables now. I've convinced one small catch-all test
case to work, but other than that, it would be a stretch to say this is
functional.
The state numbering algorithm avoids doing any scope reconstruction as
we do for C++ to simplify the implementation.
Chad Rosier [Tue, 9 Jun 2015 20:59:41 +0000 (20:59 +0000)]
[AArch64] Remove an overly conservative check when generating store pairs.
Store instructions do not modify register values and therefore it's safe
to form a store pair even if the source register has been read in between
the two store instructions.
Previously, the read of w1 (see below) prevented the formation of a stp.
Akira Hatanaka [Tue, 9 Jun 2015 19:07:19 +0000 (19:07 +0000)]
Remove DisableTailCalls from TargetOptions and the code in resetTargetOptions
that was resetting it.
Remove the uses of DisableTailCalls in subclasses of TargetLowering and use
the value of function attribute "disable-tail-calls" instead. Also,
unconditionally add pass TailCallElim to the pipeline and check the function
attribute at the start of runOnFunction to disable the pass on a per-function
basis.
This is part of the work to remove TargetMachine::resetTargetOptions, and since
DisableTailCalls was the last non-fast-math option that was being reset in that
function, we should be able to remove the function entirely after the work to
propagate IR-level fast-math flags to DAG nodes is completed.
Out-of-tree users should remove the uses of DisableTailCalls and make changes
to attach attribute "disable-tail-calls"="true" or "false" to the functions in
the IR.
Pete Cooper [Tue, 9 Jun 2015 18:36:13 +0000 (18:36 +0000)]
Allocate space for MCSymbol::Name only if required.
Similarly to User which allocates a number of Use's prior to the this pointer,
allocate space for the Name* for MCSymbol only when we need a name.
Given that an MCSymbol is 48-bytes on 64-bit systems, this saves a decent % of space.
Given the verify_uselistorder test case with debug info and llc, 50k symbols have names
out of 700k so this optimises for the common case of temporary unnamed symbols.
Samuel Antao [Tue, 9 Jun 2015 16:29:34 +0000 (16:29 +0000)]
The constant initialization for globals in NVPTX is generated as an
array of bytes. The generation of this byte arrays was expecting
the host to be little endian, which prevents big endian hosts to be
used in the generation of the PTX code. This patch fixes the
problem by changing the way the bytes are extracted so that it
works for either little and big endian.
Rui Ueyama [Tue, 9 Jun 2015 15:20:42 +0000 (15:20 +0000)]
Remove object_error::success and use std::error_code() instead
make_error_code(object_error) is slow because object::object_category()
uses a ManagedStatic variable. But the real problem is that the function is
called too frequently. This patch uses std::error_code() instead of
object_error::success. In most cases, we return "success", so this patch
reduces number of function calls to that function.
Toma Tabacu [Tue, 9 Jun 2015 10:34:31 +0000 (10:34 +0000)]
[mips] [IAS] Add support for BNE and BEQ with an immediate operand.
Summary:
For some branches, GAS accepts an immediate instead of the 2nd register operand.
We only implement this for BNE and BEQ for now. Other branch instructions can be added later, if needed.
Daniel Sanders [Tue, 9 Jun 2015 09:47:46 +0000 (09:47 +0000)]
[ADT] Assert that SmallVectorBase::grow_pod() successfully reallocates memory.
Summary:
If malloc/realloc fails then the SmallVector becomes unusable since begin() and
end() will return NULL. This is unlikely to occur but was the cause of recent
bugpoint test failures on my machine.
It is not clear whether not checking for malloc/realloc failure is a deliberate
decision and adding checks has the potential to impact compiler performance.
Therefore, this patch only adds the check to builds with assertions enabled for
the moment.
Keno Fischer [Tue, 9 Jun 2015 01:53:59 +0000 (01:53 +0000)]
[DWARF] Fix a few corner cases in expression emission
Summary: I noticed an object file with `DW_OP_reg4 DW_OP_breg4 0` as a DWARF expression,
which I traced to a missing break (and `++I`) in this code snippet.
While I was at it, I also added support for a few other corner cases
along the same lines that I could think of.
Test Plan: Hand-crafted test case to exercises these cases is included.
Anna Zaks [Tue, 9 Jun 2015 00:58:08 +0000 (00:58 +0000)]
[asan] Prevent __attribute__((annotate)) triggering errors on Darwin
The following code triggers a fatal error in the compiler instrumentation
of ASan on Darwin because we place the attribute into llvm.metadata section,
which does not have the proper MachO section name.
This commit reorders the checks so that we skip everything in llvm.metadata
first. It also removes the hard failure in case the section name does not
parse. That check will be done lower in the compilation pipeline anyway.
MergeFunctions: Impose a total order on the replacement of functions
We don't want to replace function A by Function B in one module and Function B
by Function A in another module.
If these functions are marked with linkonce_odr we would end up with a function
stub calling B in one module and a function stub calling A in another module. If
the linker decides to pick these two we will have two stubs calling each other.
Reid Kleckner [Mon, 8 Jun 2015 22:12:44 +0000 (22:12 +0000)]
[MC] Use unsigned for the Kind bitfield in MCSymbol
Fixes most of the test suite on Windows with clang-cl.
I'm not sure why the test suite was passing with MSVC 2013. Maybe they
changed their behavior and we are emulating their old sign extension
behavior. I think this deserves more investigation, but I want to green
the bot first.
Keno Fischer [Mon, 8 Jun 2015 20:09:58 +0000 (20:09 +0000)]
[InstrInfo] Refactor foldOperandImpl to thread through InsertPt. NFC
Summary:
This was a longstanding FIXME and is a necessary precursor to cases
where foldOperandImpl may have to create more than one instruction
(e.g. to constrain a register class). This is the split out NFC changes from
D6262.
Akira Hatanaka [Mon, 8 Jun 2015 18:50:43 +0000 (18:50 +0000)]
[ARM] Pass a callback to FunctionPass constructors to enable skipping execution
on a per-function basis.
Previously some of the passes were conditionally added to ARM's pass pipeline
based on the target machine's subtarget. This patch makes changes to add those
passes unconditionally and execute them conditonally based on the predicate
functor passed to the pass constructors. This enables running different sets of
passes for different functions in the module.
Pete Cooper [Mon, 8 Jun 2015 17:17:16 +0000 (17:17 +0000)]
Move COFF Type in to the MCSymbolCOFF class.
The flags field in MCSymbol only needs to be 16-bits on ELF and MachO.
This moves the 16-bit Type out of there so that it can be reduced in size in a future commit.
Matthias Braun [Mon, 8 Jun 2015 16:56:23 +0000 (16:56 +0000)]
X86: Reject register operands with obvious type mismatches.
While we have some code to transform specification like {ax} into
{eax}/{rax} if the operand type isn't 16bit, we should reject cases
where there is no sane way to do this, like the i128 type in the
example.
Oliver Stannard [Mon, 8 Jun 2015 16:55:31 +0000 (16:55 +0000)]
Fix assertion failure in global-merge with unused ConstantExpr
The global-merge pass was crashing because it assumes that all ConstantExprs
(reached via the global variables that they use) have at least one user.
I haven't worked out a way to test this, as an unused ConstantExpr cannot be
represented by serialised IR, and global-merge can only be run in llc, which
does not run any passes which can make a ConstantExpr dead.
This (reduced to the point of silliness) C code triggers this bug when compiled
for arm-none-eabi at -O1:
static a = 7;
static volatile b[10] = {&a};
c;
main() {
c = 0;
for (; c < 10;)
printf(b[c]);
}
Colin LeMahieu [Mon, 8 Jun 2015 16:34:47 +0000 (16:34 +0000)]
[Hexagon] Adding functionality for searching for compound instruction pairs. Compound instructions reduce slot resource requirements freeing those packet slots up for more instructions.
Javed Absar [Mon, 8 Jun 2015 15:01:11 +0000 (15:01 +0000)]
ARM]: Add support for MMFR4_EL1 in assembler
This patch adds support for system register MMFR4_EL1 (memory model feature register) in the assembler.
This register provides information about the implemented memory model and memory management support.
Igor Breger [Mon, 8 Jun 2015 14:03:17 +0000 (14:03 +0000)]
AVX-512: Implemented 256/128bit VALIGND/Q instructions for SKX and KNL
Implemented DAG lowering for all these forms.
Added tests for DAG lowering and encoding.
Artur Pilipenko [Mon, 8 Jun 2015 11:58:13 +0000 (11:58 +0000)]
Minor refactoring of GEP handling in isDereferenceablePointer
For GEP instructions isDereferenceablePointer checks that all indices are constant and within bounds. Replace this index calculation logic to a call to accumulateConstantOffset. Separated from the http://reviews.llvm.org/D9791
projects / llvm / log