As indicated by the tests, it is possible to feed the AsmParser an
invalid datalayout string. We should verify the result of parsing this
string regardless of whether or not we have assertions enabled.
llvm-svn: 223898
We used to only combine intrinsics, and turn them into VLD1_UPD/VST1_UPD
when the base pointer is incremented after the load/store.
We can do the same thing for generic load/stores.
Note that we can only combine the first load/store+adds pair in
a sequence (as might be generated for a v16f32 load for instance),
because other combines turn the base pointer addition chain (each
computing the address of the next load, from the address of the last
load) into independent additions (common base pointer + this load's
offset).
Differential Revision: http://reviews.llvm.org/D6585
llvm-svn: 223862
It was missing from the VLD1/VST1 handling logic, even though the
corresponding instructions exist (same form as v2i64).
In preparation for a future patch.
llvm-svn: 223832
The load/store value type is currently not available when lowering the memcpy
intrinsic. Add the missing nullptr check to support this in 'computeAddress'.
Fixes rdar://problem/19178947.
llvm-svn: 223818
patterns.
This is causing Clang to miscompile itself for 32-bit x86 somehow, and likely
also on ARM and PPC. I really don't know how, but reverting now that I've
confirmed this is actually the culprit. I have a reproduction as well and so
should be able to restore this shortly.
This reverts commit r223764.
Original commit log follows:
Teach instcombine to canonicalize "element extraction" from a load of an
integer and "element insertion" into a store of an integer into actual
element extraction, element insertion, and vector loads and stores.
Previously various parts of LLVM (including instcombine itself) would
introduce integer loads and stores into the code as a way of opaquely
loading and storing "bits". In some cases (such as a memcpy of
std::complex<float> object) we will eventually end up using those bits
in non-integer types. In order for SROA to effectively promote the
allocas involved, it splits these "store a bag of bits" integer loads
and stores up into the constituent parts. However, for non-alloca loads
and tsores which remain, it uses integer math to recombine the values
into a large integer to load or store.
All of this would be "fine", except that it forces LLVM to go through
integer math to combine and split up values. While this makes perfect
sense for integers (and in fact is critical for bitfields to end up
lowering efficiently) it is *terrible* for non-integer types, especially
floating point types. We have a much more canonical way of representing
the act of concatenating the bits of two SSA values in LLVM: a vector
and insertelement. This patch teaching InstCombine to use this
representation.
With this patch applied, LLVM will no longer introduce integer math into
the critical path of every loop over std::complex<float> operations such
as those that make up the hot path of ... oh, most HPC code, Eigen, and
any other heavy linear algebra library.
For the record, I looked *extensively* at fixing this in other parts of
the compiler, but it just doesn't work:
- We really do want to canonicalize memcpy and other bit-motion to
integer loads and stores. SSA values are tremendously more powerful
than "copy" intrinsics. Not doing this regresses massive amounts of
LLVM's scalar optimizer.
- We really do need to split up integer loads and stores of this form in
SROA or every memcpy of a trivially copyable struct will prevent SSA
formation of the members of that struct. It essentially turns off
SROA.
- The closest alternative is to actually split the loads and stores when
partitioning with SROA, but this has all of the downsides historically
discussed of splitting up loads and stores -- the wide-store
information is fundamentally lost. We would also see performance
regressions for bitfield-heavy code and other places where the
integers aren't really intended to be split without seemingly
arbitrary logic to treat integers totally differently.
- We *can* effectively fix this in instcombine, so it isn't that hard of
a choice to make IMO.
llvm-svn: 223813
Lowering patterns were written through avx512_broadcast_pat multiclass as pattern generates VBROADCAST and COPY_TO_REGCLASS nodes.
Added lowering tests.
llvm-svn: 223804
Split `Metadata` away from the `Value` class hierarchy, as part of
PR21532. Assembly and bitcode changes are in the wings, but this is the
bulk of the change for the IR C++ API.
I have a follow-up patch prepared for `clang`. If this breaks other
sub-projects, I apologize in advance :(. Help me compile it on Darwin
I'll try to fix it. FWIW, the errors should be easy to fix, so it may
be simpler to just fix it yourself.
This breaks the build for all metadata-related code that's out-of-tree.
Rest assured the transition is mechanical and the compiler should catch
almost all of the problems.
Here's a quick guide for updating your code:
- `Metadata` is the root of a class hierarchy with three main classes:
`MDNode`, `MDString`, and `ValueAsMetadata`. It is distinct from
the `Value` class hierarchy. It is typeless -- i.e., instances do
*not* have a `Type`.
- `MDNode`'s operands are all `Metadata *` (instead of `Value *`).
- `TrackingVH<MDNode>` and `WeakVH` referring to metadata can be
replaced with `TrackingMDNodeRef` and `TrackingMDRef`, respectively.
If you're referring solely to resolved `MDNode`s -- post graph
construction -- just use `MDNode*`.
- `MDNode` (and the rest of `Metadata`) have only limited support for
`replaceAllUsesWith()`.
As long as an `MDNode` is pointing at a forward declaration -- the
result of `MDNode::getTemporary()` -- it maintains a side map of its
uses and can RAUW itself. Once the forward declarations are fully
resolved RAUW support is dropped on the ground. This means that
uniquing collisions on changing operands cause nodes to become
"distinct". (This already happened fairly commonly, whenever an
operand went to null.)
If you're constructing complex (non self-reference) `MDNode` cycles,
you need to call `MDNode::resolveCycles()` on each node (or on a
top-level node that somehow references all of the nodes). Also,
don't do that. Metadata cycles (and the RAUW machinery needed to
construct them) are expensive.
- An `MDNode` can only refer to a `Constant` through a bridge called
`ConstantAsMetadata` (one of the subclasses of `ValueAsMetadata`).
As a side effect, accessing an operand of an `MDNode` that is known
to be, e.g., `ConstantInt`, takes three steps: first, cast from
`Metadata` to `ConstantAsMetadata`; second, extract the `Constant`;
third, cast down to `ConstantInt`.
The eventual goal is to introduce `MDInt`/`MDFloat`/etc. and have
metadata schema owners transition away from using `Constant`s when
the type isn't important (and they don't care about referring to
`GlobalValue`s).
In the meantime, I've added transitional API to the `mdconst`
namespace that matches semantics with the old code, in order to
avoid adding the error-prone three-step equivalent to every call
site. If your old code was:
MDNode *N = foo();
bar(isa <ConstantInt>(N->getOperand(0)));
baz(cast <ConstantInt>(N->getOperand(1)));
bak(cast_or_null <ConstantInt>(N->getOperand(2)));
bat(dyn_cast <ConstantInt>(N->getOperand(3)));
bay(dyn_cast_or_null<ConstantInt>(N->getOperand(4)));
you can trivially match its semantics with:
MDNode *N = foo();
bar(mdconst::hasa <ConstantInt>(N->getOperand(0)));
baz(mdconst::extract <ConstantInt>(N->getOperand(1)));
bak(mdconst::extract_or_null <ConstantInt>(N->getOperand(2)));
bat(mdconst::dyn_extract <ConstantInt>(N->getOperand(3)));
bay(mdconst::dyn_extract_or_null<ConstantInt>(N->getOperand(4)));
and when you transition your metadata schema to `MDInt`:
MDNode *N = foo();
bar(isa <MDInt>(N->getOperand(0)));
baz(cast <MDInt>(N->getOperand(1)));
bak(cast_or_null <MDInt>(N->getOperand(2)));
bat(dyn_cast <MDInt>(N->getOperand(3)));
bay(dyn_cast_or_null<MDInt>(N->getOperand(4)));
- A `CallInst` -- specifically, intrinsic instructions -- can refer to
metadata through a bridge called `MetadataAsValue`. This is a
subclass of `Value` where `getType()->isMetadataTy()`.
`MetadataAsValue` is the *only* class that can legally refer to a
`LocalAsMetadata`, which is a bridged form of non-`Constant` values
like `Argument` and `Instruction`. It can also refer to any other
`Metadata` subclass.
(I'll break all your testcases in a follow-up commit, when I propagate
this change to assembly.)
llvm-svn: 223802
replaceDbgDeclareForAlloca() replaces an alloca by a value storing the
address of what was the alloca. If there is a dbg.declare corresponding
to that alloca, we need to lower it to a dbg.value describing the additional
dereference operation to be performed to get to the underlying variable.
This is done by adding a DW_OP_deref to the complex location part of the
location description. This deref was added to the end of the operation list,
which is wrong. The expression applies to what is described by the
dbg.{declare,value}, and as we are changing this, we need to apply the
DW_OP_deref as the first operation in the list.
Part of the fix for rdar://19162268.
llvm-svn: 223799
The goal of this tool is to replicate Darwin's dsymutil functionality
based on LLVM. dsymutil is a DWARF linker. Darwin's linker (ld64) does
not link the debug information, it leaves it in the object files in
relocatable form, but embbeds a `debug map` into the executable that
describes where to find the debug information and how to relocate it.
When releasing/archiving a binary, dsymutil is called to link all the DWARF
information into a `dsym bundle` that can distributed/stored along with
the binary.
With this commit, the LLVM based dsymutil is just able to parse the STABS
debug maps embedded by ld64 in linked binaries (and not all of them, for
example archives aren't supported yet).
Note that the tool directory is called dsymutil, but the executable is
currently called llvm-dsymutil. This discrepancy will disappear once the
tool will be feature complete. At this point the executable will be renamed
to dsymutil, but until then you do not want it to override the system one.
Differential Revision: http://reviews.llvm.org/D6242
llvm-svn: 223793
With the foregoing three patches, VSX instructions can be used for
little endian. This patch removes the restriction that prevented
this, and re-enables the test cases from the first three patches.
llvm-svn: 223792
When performing instruction selection for ISD::VECTOR_SHUFFLE, there
is special code for handling v2f64 and v2i64 using VSX instructions.
This code must be adjusted for little-endian. Because the two inputs
are treated as a double-wide register, we must swap their order for
little endian. To get the appropriate mask elements to use with the
big-endian biased XXPERMDI instruction, we must reverse their order
and invert the bits.
A new test is added to test the 16 possible values of the shuffle
mask. It is initially disabled for reasons specified in the test. It
is re-enabled by patch 4/4.
llvm-svn: 223791
This allows it to work with non trivial manglings like the one in COFF.
Amusingly, this can be tested with gold, as emit-llvm causes the plugin to
exit before any COFF is generated.
llvm-svn: 223790
This optimization transforms code like:
bb1:
%0 = icmp ne i32 %a, 0
%1 = icmp ne i32 %b, 0
%or.cond = or i1 %0, %1
br i1 %or.cond, label %TrueBB, label %FalseBB
into a multiple branch instructions like:
bb1:
%0 = icmp ne i32 %a, 0
br i1 %0, label %TrueBB, label %bb2
bb2:
%1 = icmp ne i32 %b, 0
br i1 %1, label %TrueBB, label %FalseBB
This optimization is already performed by SelectionDAG, but not by FastISel.
FastISel cannot perform this optimization, because it cannot generate new
MachineBasicBlocks.
Performing this optimization at CodeGenPrepare time makes it available to both -
SelectionDAG and FastISel - and the implementation in SelectiuonDAG could be
removed. There are currenty a few differences in codegen for X86 and PPC, so
this commmit only enables it for FastISel.
Reviewed by Jim Grosbach
This fixes rdar://problem/19034919.
llvm-svn: 223786
This patch addresses the inherent big-endian bias in the lxvd2x,
lxvw4x, stxvd2x, and stxvw4x instructions. These instructions load
vector elements into registers left-to-right (with the first element
loaded into the high-order bits of the register), regardless of the
endian setting of the processor. However, these are the only
vector memory instructions that permit unaligned storage accesses, so
we want to use them for little-endian.
To make this work, a lxvd2x or lxvw4x is replaced with an lxvd2x
followed by an xxswapd, which swaps the doublewords. This works for
lxvw4x as well as lxvd2x, because for lxvw4x on an LE system the
vector elements are in LE order (right-to-left) within each
doubleword. (Thus after lxvw2x of a <4 x float> the elements will
appear as 1, 0, 3, 2. Following the swap, they will appear as 3, 2,
0, 1, as desired.) For stores, an stxvd2x or stxvw4x is replaced
with an stxvd2x preceded by an xxswapd.
Introduction of extra swap instructions provides correctness, but
obviously is not ideal from a performance perspective. Future patches
will address this with optimizations to remove most of the introduced
swaps, which have proven effective in other implementations.
The introduction of the swaps is performed during lowering of LOAD,
STORE, INTRINSIC_W_CHAIN, and INTRINSIC_VOID operations. The latter
are used to translate intrinsics that specify the VSX loads and stores
directly into equivalent sequences for little endian. Thus code that
uses vec_vsx_ld and vec_vsx_st does not have to be modified to be
ported from BE to LE.
We introduce new PPCISD opcodes for LXVD2X, STXVD2X, and XXSWAPD for
use during this lowering step. In PPCInstrVSX.td, we add new SDType
and SDNode definitions for these (PPClxvd2x, PPCstxvd2x, PPCxxswapd).
These are recognized during instruction selection and mapped to the
correct instructions.
Several tests that were written to use -mcpu=pwr7 or pwr8 are modified
to disable VSX on LE variants because code generation changes with
this and subsequent patches in this set. I chose to include all of
these in the first patch than try to rigorously sort out which tests
were broken by one or another of the patches. Sorry about that.
The new test vsx-ldst-builtin-le.ll, and the changes to vsx-ldst.ll,
are disabled until LE support is enabled because of breakages that
occur as noted in those tests. They are re-enabled in patch 4/4.
llvm-svn: 223783
missing barcelona CPU which that test uncovered, and remove the 32-bit
x86 CPUs which I really wasn't prepared to audit and test thoroughly.
If anyone wants to clean up the 32-bit only x86 CPUs, go for it.
Also, if anyone else wants to try to de-duplicate the AMD CPUs, that'd
be cool, but from the looks of it wouldn't save as much as it did for
the Intel CPUs.
llvm-svn: 223774
Instructions of the form [ADD Rd, pc, #imm] are manually aliased
in processInstruction() to use ADR. To accomodate this, mod_imm handling
had to be tweaked a bit. Turns out it was the manual aliasing that must
be tweaked to accommodate mod_imms instead. More information about the
parsed instruction is available at the point where processInstruction()
is invoked, which makes it easier to detect a mod_imm at that point rather
than trying to detect a potential alias when a mod_imm is being prepped.
Added a test case and fixed some white spaces as well.
llvm-svn: 223772
Removed some duplicate test cases from the file /test/Transforms/InstCombine/shift.ll.
test54 and test57 were duplicates of each other.
test55 and test58 were duplicates of each other.
(Removed test57 and test58)
llvm-svn: 223767
integer and "element insertion" into a store of an integer into actual
element extraction, element insertion, and vector loads and stores.
Previously various parts of LLVM (including instcombine itself) would
introduce integer loads and stores into the code as a way of opaquely
loading and storing "bits". In some cases (such as a memcpy of
std::complex<float> object) we will eventually end up using those bits
in non-integer types. In order for SROA to effectively promote the
allocas involved, it splits these "store a bag of bits" integer loads
and stores up into the constituent parts. However, for non-alloca loads
and tsores which remain, it uses integer math to recombine the values
into a large integer to load or store.
All of this would be "fine", except that it forces LLVM to go through
integer math to combine and split up values. While this makes perfect
sense for integers (and in fact is critical for bitfields to end up
lowering efficiently) it is *terrible* for non-integer types, especially
floating point types. We have a much more canonical way of representing
the act of concatenating the bits of two SSA values in LLVM: a vector
and insertelement. This patch teaching InstCombine to use this
representation.
With this patch applied, LLVM will no longer introduce integer math into
the critical path of every loop over std::complex<float> operations such
as those that make up the hot path of ... oh, most HPC code, Eigen, and
any other heavy linear algebra library.
For the record, I looked *extensively* at fixing this in other parts of
the compiler, but it just doesn't work:
- We really do want to canonicalize memcpy and other bit-motion to
integer loads and stores. SSA values are tremendously more powerful
than "copy" intrinsics. Not doing this regresses massive amounts of
LLVM's scalar optimizer.
- We really do need to split up integer loads and stores of this form in
SROA or every memcpy of a trivially copyable struct will prevent SSA
formation of the members of that struct. It essentially turns off
SROA.
- The closest alternative is to actually split the loads and stores when
partitioning with SROA, but this has all of the downsides historically
discussed of splitting up loads and stores -- the wide-store
information is fundamentally lost. We would also see performance
regressions for bitfield-heavy code and other places where the
integers aren't really intended to be split without seemingly
arbitrary logic to treat integers totally differently.
- We *can* effectively fix this in instcombine, so it isn't that hard of
a choice to make IMO.
Differential Revision: http://reviews.llvm.org/D6548
llvm-svn: 223764
This handles the simplest case for mov -> push conversion:
1. x86-32 calling convention, everything is passed through the stack.
2. There is no reserved call frame.
3. Only registers or immediates are pushed, no attempt to combine a mem-reg-mem sequence into a single PUSHmm.
Differential Revision: http://reviews.llvm.org/D6503
llvm-svn: 223757
The aggressive anti-dep breaker, used by the PowerPC backend during post-RA
scheduling (but is available to all targets), did not handle early-clobber MI
operands (at all). When constructing the list of available registers for the
replacement of some def operand, check the using instructions, and remove
registers assigned to early-clobbered defs from the set.
Fixes PR21452.
llvm-svn: 223727
This fixes an issue with ScheduleDAGInstrs::buildSchedGraph
where stores without an underlying object would not be added
as a predecessor to the current BarrierChain.
llvm-svn: 223717
GCC accepts 'cc' as an alias for 'cr0', and we need to do the same when
processing inline asm constraints. This had previously been implemented using a
non-allocatable register, named 'cc', that was listed as an alias of 'cr0', but
the infrastructure does not seem to support this properly (neither the register
allocator nor the scheduler properly accounts for the alias). Instead, we can
just process this as a naming alias inside of the inline asm
constraint-processing code, so we'll do that instead.
There are two regression tests, one where the post-RA scheduler did the wrong
thing with the non-allocatable alias, and one where the register allocator did
the wrong thing. Fixes PR21742.
llvm-svn: 223708
We were already lazily linking functions, but all GlobalValues can be treated
uniformly for this.
The test updates are to ensure that a given GlobalValue is still linked in.
This fixes pr21494.
llvm-svn: 223681
Fix a compact unwind encoding logic bug which would try to encode
more callee saved registers than it should, leading to early bail out
in the encoding logic and abusive use of DWARF frame mode unnecessarily.
Also remove no-compact-unwind.ll which was testing the wrong thing
based on this bug and move it to valid 'compact unwind' tests. Added
other few more tests too.
llvm-svn: 223676
Introduce the ``llvm.instrprof_increment`` intrinsic and the
``-instrprof`` pass. These provide the infrastructure for writing
counters for profiling, as in clang's ``-fprofile-instr-generate``.
The implementation of the instrprof pass is ported directly out of the
CodeGenPGO classes in clang, and with the followup in clang that rips
that code out to use these new intrinsics this ends up being NFC.
Doing the instrumentation this way opens some doors in terms of
improving the counter performance. For example, this will make it
simple to experiment with alternate lowering strategies, and allows us
to try handling profiling specially in some optimizations if we want
to.
Finally, this drastically simplifies the frontend and puts all of the
lowering logic in one place.
llvm-svn: 223672
Teach ISel how to match a TZCNT/LZCNT from a conditional move if the
condition code is X86_COND_NE.
Existing tablegen patterns only allowed to match TZCNT/LZCNT from a
X86cond with condition code equal to X86_COND_E. To avoid introducing
extra rules, I added an 'ImmLeaf' definition that checks if the
condition code is COND_E or COND_NE.
llvm-svn: 223668
Before this patch, the backend sub-optimally expanded the non-constant shift
count of a v8i16 shift into a sequence of two 'movd' plus 'movzwl'.
With this patch the backend checks if the target features sse4.1. If so, then
it lets the shuffle legalizer deal with the expansion of the shift amount.
Example:
;;
define <8 x i16> @test(<8 x i16> %A, <8 x i16> %B) {
%shamt = shufflevector <8 x i16> %B, <8 x i16> undef, <8 x i32> zeroinitializer
%shl = shl <8 x i16> %A, %shamt
ret <8 x i16> %shl
}
;;
Before (with -mattr=+avx):
vmovd %xmm1, %eax
movzwl %ax, %eax
vmovd %eax, %xmm1
vpsllw %xmm1, %xmm0, %xmm0
retq
Now:
vpxor %xmm2, %xmm2, %xmm2
vpblendw $1, %xmm1, %xmm2, %xmm1
vpsllw %xmm1, %xmm0, %xmm0
retq
llvm-svn: 223660
As a fixup to r223616, follow the convention of naming the files after
the LLVM release whose bitcode they're maintaining compatability with.
llvm-svn: 223623
Add assembly and bitcode tests that I neglected to add in r223564 (IR:
Disallow complicated function-local metadata) and r223574 (IR: Disallow
function-local metadata attachments).
Found a couple of bugs:
- The error message for function-local attachments gave the wrong line
number -- it indicated the next token (typically on the next line)
instead of the token that started the attachment. Fixed.
- Metadata arguments of the form `!{i32 0, i32 %v}` (or with the
arguments reversed) fired an assertion in `ValueEnumerator` in LLVM
v3.5, so I suppose this never really worked. I suppose this was
"fixed" by r223564.
(Thanks to dblaikie for pointing out my omission.)
Part of PR21532.
llvm-svn: 223616
matching offsets. I don't expect this to really matter, but its what the
latest incarnation of my script for maintaining these tests happens to
produce, and so its simpler for me if everything matches.
llvm-svn: 223613
Consider:
void f() {}
void __attribute__((weak)) g() {}
bool b = &f != &g;
It's possble for g to resolve to f if --defsym=g=f is passed on to the
linker.
llvm-svn: 223585
This can significantly reduce the size of the switch, allowing for more
efficient lowering.
I also worked with the idea of exploiting unreachable defaults by
omitting the range check for jump tables, but always ended up with a
non-neglible binary size increase. It might be worth looking into some more.
SimplifyCFG currently does this transformation, but I'm working towards changing
that so we can optimize harder based on unreachable defaults.
Differential Revision: http://reviews.llvm.org/D6510
llvm-svn: 223566
Disallow complex types of function-local metadata. The only valid
function-local metadata is an `MDNode` whose sole argument is a
non-metadata function-local value.
Part of PR21532.
llvm-svn: 223564
Fix the poor codegen seen in PR21710 ( http://llvm.org/bugs/show_bug.cgi?id=21710 ).
Before we crack 32-byte build vectors into smaller chunks (and then subsequently
glue them back together), we should look for the easy case where we can just load
all elements in a single op.
An example of the codegen change is:
From:
vmovss 16(%rdi), %xmm1
vmovups (%rdi), %xmm0
vinsertps $16, 20(%rdi), %xmm1, %xmm1
vinsertps $32, 24(%rdi), %xmm1, %xmm1
vinsertps $48, 28(%rdi), %xmm1, %xmm1
vinsertf128 $1, %xmm1, %ymm0, %ymm0
retq
To:
vmovups (%rdi), %ymm0
retq
Differential Revision: http://reviews.llvm.org/D6536
llvm-svn: 223518
Summary:
Follow up to [x32] "Use ebp/esp as frame and stack pointer":
http://reviews.llvm.org/D4617
In that earlier patch, NaCl64 was made to always use rbp.
That's needed for most cases because rbp should hold a full
64-bit address within the NaCl sandbox so that load/stores
off of rbp don't require sandbox adjustment (zeroing the top
32-bits, then filling those by adding r15).
However, llvm.frameaddress returns a pointer and pointers
are 32-bit for NaCl64. In this case, use ebp instead, which
will make the register copy type check. A similar mechanism
may be needed for llvm.eh.return, but is not added in this change.
Test Plan: test/CodeGen/X86/frameaddr.ll
Reviewers: dschuff, nadav
Subscribers: jfb, llvm-commits
Differential Revision: http://reviews.llvm.org/D6514
llvm-svn: 223510
Update of some of the VSX test cases for Power to check fast-isel codegen as well as the regular codegen.
http://reviews.llvm.org/D6357
llvm-svn: 223509
SSE2/AVX non-constant packed shift instructions only use the lower 64-bit of
the shift count.
This patch teaches function 'getTargetVShiftNode' how to deal with shifts
where the shift count node is of type MVT::i64.
Before this patch, function 'getTargetVShiftNode' only knew how to deal with
shift count nodes of type MVT::i32. This forced the backend to wrongly
truncate the shift count to MVT::i32, and then zero-extend it back to MVT::i64.
llvm-svn: 223505
When a loop gets bundled up, its outgoing edges are quite large, and can
just barely overflow 64-bits. If one successor has multiple incoming
edges -- and that successor is getting all the incoming mass --
combining just its edges can overflow. Handle that by saturating rather
than asserting.
This fixes PR21622.
llvm-svn: 223500
No functional changes. Got myself bitten in r223113 when adding support for
modified immediate syntax (regressions reported by joerg@britannica.bec.de,
fixes in r223366 and r223381). Our assembler tests did not cover serveral
different syntax variants. This patch expands the test coverage to check for
the following cases:
1. Modified immediate operands may be expressed with expressions, as in #(4 * 2)
instead of #8.
2. Modified immediate operands may be _optionally_ prefixed by a '#' symbol or a
'$' symbol.
3. Certain instructions (e.g. ADD) support single input register variants;
[ADD r0, #mod_imm] is same as [ADD r0, r0, #mod_imm].
4. Certain instructions have aliases which convert plain immediates to modified
immediates. For an example, [ADD r0, -10] is not valid because -10 (in two's
complement) cannot be encoded as a modified immediate, but ARMInstrInfo.td
defines an alias which can transform this into a [SUB r0, 10].
llvm-svn: 223475
Do not realign origin address if the corresponding application
address is at least 4-byte-aligned.
Saves 2.5% code size in track-origins mode.
llvm-svn: 223464
When lowering a vector shift node, the backend checks if the shift count is a
shuffle with a splat mask. If so, then it introduces an extra dag node to
extract the splat value from the shuffle. The splat value is then used
to generate a shift count of a target specific shift.
However, if we know that the shift count is a splat shuffle, we can use the
splat index 'I' to extract the I-th element from the first shuffle operand.
The advantage is that the splat shuffle may become dead since we no longer
use it.
Example:
;;
define <4 x i32> @example(<4 x i32> %a, <4 x i32> %b) {
%c = shufflevector <4 x i32> %b, <4 x i32> undef, <4 x i32> zeroinitializer
%shl = shl <4 x i32> %a, %c
ret <4 x i32> %shl
}
;;
Before this patch, llc generated the following code (-mattr=+avx):
vpshufd $0, %xmm1, %xmm1 # xmm1 = xmm1[0,0,0,0]
vpxor %xmm2, %xmm2
vpblendw $3, %xmm1, %xmm2, %xmm1 # xmm1 = xmm1[0,1],xmm2[2,3,4,5,6,7]
vpslld %xmm1, %xmm0, %xmm0
retq
With this patch, the redundant splat operation is removed from the code.
vpxor %xmm2, %xmm2
vpblendw $3, %xmm1, %xmm2, %xmm1 # xmm1 = xmm1[0,1],xmm2[2,3,4,5,6,7]
vpslld %xmm1, %xmm0, %xmm0
retq
llvm-svn: 223461
The test file test/CodeGen/ARM/build-attributes.ll was missing several
floating-point build attribute tests. The intention of this commit is that for
each CPU / architecture currently tested, there are now tests that make sure
the following attributes are sufficiently checked,
* Tag_ABI_FP_rounding
* Tag_ABI_FP_denormal
* Tag_ABI_FP_exceptions
* Tag_ABI_FP_user_exceptions
* Tag_ABI_FP_number_model
Also in this commit, the -unsafe-fp-math flag has been augmented with the full
suite of flags Clang sends to LLVM when you pass -ffast-math to Clang. That is,
`-unsafe-fp-math' has been changed to `-enable-unsafe-fp-math -disable-fp-elim
-enable-no-infs-fp-math -enable-no-nans-fp-math -fp-contract=fast'
Change-Id: I35d766076bcbbf09021021c0a534bf8bf9a32dfc
llvm-svn: 223454
Reverting this because, while it fixes the problem in the reduced test case, it
does not fix the problem in the full test case from the bug report.
llvm-svn: 223442
The scheduling dependency graph is built bottom-up within each scheduling
region, and ScheduleDAGInstrs::addPhysRegDeps is called to add output/anti
dependencies, based on physical registers, to the SUs for instructions
based on those that come before them.
In the test case, we start before post-RA scheduling with a block that looks
like this:
...
INLINEASM <...
andc $0,$0,$2
stdcx. $0,0,$3
bne- 1b
> [sideeffect] [mayload] [maystore] [attdialect], $0:[regdef-ec:G8RC], %X6<earlyclobber,def,dead>, $1:[mem], %X3<kill>, $2:[reguse:G8RC], %X5<kill>, $3:[reguse:G8RC], %X3, $4:[mem], %X3, $5:[clobber], %CC<earlyclobber,imp-def,dead>, <<badref>>
...
%X4<def,dead> = ANDIo8 %X4<kill>, 1, %CR0<imp-def,dead>, %CR0GT<imp-def>
...
%R29<def> = ISEL %R3<undef>, %R4<kill>, %CR0GT<kill>
where it is relevant that %CC is an alias to %CR0, and that %CR0GT is a
subregister of %CR0. However, for post-RA scheduling, no dependency was added
to prevent the INLINEASM from being scheduled in between the ANDIo8 and the
ISEL (which communicate via the %CR0GT register).
In ScheduleDAGInstrs::addPhysRegDeps, when called for the %CC operand, we'd
iterate over all of its aliases (which include %CC itself and also %CR0), and
look for previously-encountered defs of those registers. We'd find the ANDIo8,
but decide not to add a dependency between the INLINEASM and the ANDIo8 because
both the INLINEASM's def of %CC is dead, and also the ANDIo8 def of %CR0 is
dead. This ignores, however, that ANDIo8 has a non-dead def of %CR0GT, a
subregister of %CR0, and thus a dependency still must exist.
To fix this problem, when calling registerDefIsDead on the SU with the def, we
also check all subregisters for possible non-dead defs, and add the dependency
if any are found.
Fixes PR21742.
llvm-svn: 223440
with fixes. Includes the move of tests for llvm-objdump for universal files to an X86
directory. And the fix where it was failing on linux Rafael tracked down with asan.
I had both Jim Grosbach and Adam Hemet look over the second fix since I could not
set up asan to reproduce with the old version but not with the fix.
llvm-svn: 223416
r223113 added support for ARM modified immediate assembly syntax. Which
assumes all immediate operands are prefixed with a '#'. This assumption
is wrong as per the ARMARM - which recommends that all '#' characters be
treated optional. The current patch fixes this regression and adds a test
case. A follow-up patch will expand the test coverage to other instructions.
llvm-svn: 223381
So there are a couple of issues with indirect calls on thumbv4t. First, the most
'obvious' instruction, 'blx' isn't available until v5t. And secondly, the
next-most-obvious sequence: 'mov lr, pc; bx rN' doesn't DTRT in thumb code
because the saved off pc has its thumb bit cleared, so when the callee returns
we end up in ARM mode.... yuck.
The solution is to 'bl' to a nearby landing pad with a 'bx rN' in it.
We could cut down on code size by sharing the landing pads between call sites
that are close enough, but for the moment let's do correctness first and look at
performance later.
Patch by: Iain Sandoe
http://reviews.llvm.org/D6519
llvm-svn: 223380
Reapply r223347, with a fix to not crash on uninserted instructions (or more
precisely, instructions in uninserted blocks). bugpoint was able to reduce the
test case somewhat, but it is still somewhat large (and relies on setting
things up to be simplified during inlining), so I've not included it here.
Nevertheless, it is clear what is going on and why.
Original commit message:
Restrict somewhat the memory-allocation pointer cmp opt from r223093
Based on review comments from Richard Smith, restrict this optimization from
applying to globals that might resolve lazily to other dynamically-loaded
modules, and also from dynamic allocas (which might be transformed into malloc
calls). In short, take extra care that the compared-to pointer is really
simultaneously live with the memory allocation.
llvm-svn: 223371
r223113 added support for ARM modified immediate assembly syntax. That patch
has broken support for immediate expressions, as in:
add r0, #(4 * 4)
It wasn't caught because we don't have any tests for this feature. This patch
fixes this regression and adds test cases.
llvm-svn: 223366
The current DAG combine turns a sequence of extracts from <4 x i32> followed by zexts into a store followed by scalar loads.
According to measurements by Martin Krastev (see PR 21269) for x86-64, a sequence of an extract, movs and shifts gives better performance. However, for 32-bit x86, the previous sequence still seems better.
Differential Revision: http://reviews.llvm.org/D6501
llvm-svn: 223360
According to a previous FIXME comment we now not only look at MBB
successors, but also handle code sinking past them:
x = computation
if () {} else {}
use x
The instruction could be sunk over the whole diamond for the
if/then/else (or loop, etc), allowing it to be sunk into other blocks
after that.
Modified test added in r204522, due to one spill less present.
Minor fixes in comments.
Patch provided by Jonas Paulsson. Reviewed by Hal Finkel.
llvm-svn: 223350
Added instcombine optimizations for BSWAP with AND/OR/XOR ops:
OP( BSWAP(x), BSWAP(y) ) -> BSWAP( OP(x, y) )
OP( BSWAP(x), CONSTANT ) -> BSWAP( OP(x, BSWAP(CONSTANT) ) )
Since its just a one liner, I've also added BSWAP to the DAGCombiner equivalent as well:
fold (OP (bswap x), (bswap y)) -> (bswap (OP x, y))
Refactored bswap-fold tests to use FileCheck instead of just checking that the bswaps had gone.
Differential Revision: http://reviews.llvm.org/D6407
llvm-svn: 223349
I'm recommiting the codegen part of the patch.
The vectorizer part will be send to review again.
Masked Vector Load and Store Intrinsics.
Introduced new target-independent intrinsics in order to support masked vector loads and stores. The loop vectorizer optimizes loops containing conditional memory accesses by generating these intrinsics for existing targets AVX2 and AVX-512. The vectorizer asks the target about availability of masked vector loads and stores.
Added SDNodes for masked operations and lowering patterns for X86 code generator.
Examples:
<16 x i32> @llvm.masked.load.v16i32(i8* %addr, <16 x i32> %passthru, i32 4 /* align */, <16 x i1> %mask)
declare void @llvm.masked.store.v8f64(i8* %addr, <8 x double> %value, i32 4, <8 x i1> %mask)
Scalarizer for other targets (not AVX2/AVX-512) will be done in a separate patch.
http://reviews.llvm.org/D6191
llvm-svn: 223348
Based on review comments from Richard Smith, restrict this optimization from
applying to globals that might resolve lazily to other dynamically-loaded
modules, and also from dynamic allocas (which might be transformed into malloc
calls). In short, take extra care that the compared-to pointer is really
simultaneously live with the memory allocation.
llvm-svn: 223347
Summary: Add rpath load command support in Mach-O object and update llvm-objdump to use it.
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6512
llvm-svn: 223343
Commit on
- This patch fixes the bug described in
http://lists.cs.uiuc.edu/pipermail/llvmdev/2013-May/062343.html
The fix allocates an extra slot just below the GPRs and stores the base pointer
there. This is done only for functions containing llvm.eh.sjlj.setjmp that also
need a base pointer. Because code containing llvm.eh.sjlj.setjmp saves all of
the callee-save GPRs in the prologue, the offset to the extra slot can be
computed before prologue generation runs.
Impact at run-time on affected functions is::
- One extra store in the prologue, The store saves the base pointer.
- One extra load after a llvm.eh.sjlj.setjmp. The load restores the base pointer.
Because the extra slot is just above a gap between frame-pointer-relative and
base-pointer-relative chunks of memory, there is no impact on other offset
calculations other than ensuring there is room for the extra slot.
http://reviews.llvm.org/D6388
Patch by Arch Robison <arch.robison@intel.com>
llvm-svn: 223329
We had mistakenly believed that GCC's 'cc' referred to the entire
condition-code register (cr0 through cr7) -- and implemented this in r205630 to
fix PR19326, but 'cc' is actually an alias only to 'cr0'. This is causing LLVM
to clobber too much with legacy code with inline asm using the 'cc' clobber.
Fixes PR21451.
llvm-svn: 223328
On PowerPC, inline asm memory operands might be expanded as 0($r), where $r is
a register containing the address. As a result, this register cannot be r0, and
we need to enforce this register subclass constraint to prevent miscompiling
the code (we'd get this constraint for free with the usual instruction
definitions, but that scheme has no knowledge of how we end up printing inline
asm memory operands, and so here we need to do it 'by hand'). We can accomplish
this within the current address-mode selection framework by introducing an
explicit COPY_TO_REGCLASS node.
Fixes PR21443.
llvm-svn: 223318
Prior to this commit, physical registers defined implicitly were considered free
right after their definition, i.e.. like dead definitions. Therefore, their uses
had to immediately follow their definitions, otherwise the related register may
be reused to allocate a virtual register.
This commit fixes this assumption by keeping implicit definitions alive until
they are actually used. The downside is that if the implicit definition was dead
(and not marked at such), we block an otherwise available register. This is
however conservatively correct and makes the fast register allocator much more
robust in particular regarding the scheduling of the instructions.
Fixes PR21700.
llvm-svn: 223317
The non-opaque part can be structurally uniqued. To keep this to just
a hash lookup, we don't try to unique cyclic types.
Also change the type mapping algorithm to be optimistic about a type
not being recursive and only create a new type when proven to be wrong.
This is not as strong as trying to speculate that we can keep the source
type, but is simpler (no speculation to revert) and more powerfull
than what we had before (we don't copy non-recursive types at least).
I initially wrote this to try to replace the name based type merging.
It is not strong enough to replace it, but is is a useful addition.
With this patch the number of named struct types is a clang lto bootstrap goes
from 49674 to 15986.
llvm-svn: 223278
This allows cases like float x; fmin(1.0, x); to be optimized to fminf(1.0f, x);
rdar://19049359
Differential Revision: http://reviews.llvm.org/D6496
llvm-svn: 223270
Try to convert two compares of a signed range check into a single unsigned compare.
Examples:
(icmp sge x, 0) & (icmp slt x, n) --> icmp ult x, n
(icmp slt x, 0) | (icmp sgt x, n) --> icmp ugt x, n
llvm-svn: 223224
Almost all immediates in PowerPC assembly (both 32-bit and 64-bit) are signed
numbers, and it is important that we print them as such. To make sure that
happens, we change PPCTargetLowering::LowerAsmOperandForConstraint so that it
does all intermediate checks on a signed-extended int64_t value, and then
creates the resulting target constant using MVT::i64. This will ensure that all
negative values are printed as negative values (mirroring what is done in other
backends to achieve the same sign-extension effect).
This came up in the context of inline assembly like this:
"add%I2 %0,%0,%2", ..., "Ir"(-1ll)
where we used to print:
addi 3,3,4294967295
and gcc would print:
addi 3,3,-1
and gas accepts both forms, but our builtin assembler (correctly) does not. Now
we print -1 like gcc does.
While here, I replaced a bunch of custom integer checks with isInt<16> and
friends from MathExtras.h.
Thanks to Paul Hargrove for the bug report.
llvm-svn: 223220
LLVM understands a -enable-sign-dependent-rounding-fp-math codegen option. When
the user has specified this option, the Tag_ABI_FP_rounding attribute should be
emitted with value 1. This option currently does not appear to disable
transformations and optimizations that assume default floating point rounding
behavior, AFAICT, but the intention should be recorded in the build attributes,
regardless of what the compiler actually does with the intention.
Change-Id: If838578df3dc652b6f2796b8d152545674bcb30e
llvm-svn: 223218
When lazy reading a module, the types used in a function will not be visible to
a TypeFinder until the body is read.
This patch fixes that by asking the module for its identified struct types.
If a materializer is present, the module asks it. If not, it uses a TypeFinder.
This fixes pr21374.
I will be the first to say that this is ugly, but it was the best I could find.
Some of the options I looked at:
* Asking the LLVMContext. This could be made to work for gold, but not currently
for ld64. ld64 will load multiple modules into a single context before merging
them. This causes us to see types from future merges. Unfortunately,
MappedTypes is not just a cache when it comes to opaque types. Once the
mapping has been made, we have to remember it for as long as the key may
be used. This would mean moving MappedTypes to the Linker class and having
to drop the Linker::LinkModules static methods, which are visible from C.
* Adding an option to ignore function bodies in the TypeFinder. This would
fix the PR by picking the worst result. It would work, but unfortunately
we are currently quite dependent on the upfront type merging. I will
try to reduce our dependency, but it is not clear that we will be able
to get rid of it for now.
The only clean solution I could think of is making the Module own the types.
This would have other advantages, but it is a much bigger change. I will
propose it, but it is nice to have this fixed while that is discussed.
With the gold plugin, this patch takes the number of types in the LTO clang
binary from 52817 to 49669.
llvm-svn: 223215
Select i1 logical ops directly to 64-bit SALU instructions.
Vector i1 values are always really in SGPRs, with each
bit for each item in the wave. This saves about 4 instructions
when and/or/xoring any condition, and also helps write conditions
that need to be passed in vcc.
This should work correctly now that the SGPR live range
fixing pass works. More work is needed to eliminate the VReg_1
pseudo regclass and possibly the entire SILowerI1Copies pass.
llvm-svn: 223206
We were assuming that each back-edge in a region represented a unique
loop, which is not always the case. We need to use LoopInfo to
correctly determine which back-edges are loops.
llvm-svn: 223199
We just needed to remove the assertion in
AMDGPURegisterInfo::getFrameRegister(), which is called when
initializing the parser for inline assembly.
llvm-svn: 223197
Patch by Ben Gamari!
This redefines the `prefix` attribute introduced previously and
introduces a `prologue` attribute. There are a two primary usecases
that these attributes aim to serve,
1. Function prologue sigils
2. Function hot-patching: Enable the user to insert `nop` operations
at the beginning of the function which can later be safely replaced
with a call to some instrumentation facility
3. Runtime metadata: Allow a compiler to insert data for use by the
runtime during execution. GHC is one example of a compiler that
needs this functionality for its tables-next-to-code functionality.
Previously `prefix` served cases (1) and (2) quite well by allowing the user
to introduce arbitrary data at the entrypoint but before the function
body. Case (3), however, was poorly handled by this approach as it
required that prefix data was valid executable code.
Here we redefine the notion of prefix data to instead be data which
occurs immediately before the function entrypoint (i.e. the symbol
address). Since prefix data now occurs before the function entrypoint,
there is no need for the data to be valid code.
The previous notion of prefix data now goes under the name "prologue
data" to emphasize its duality with the function epilogue.
The intention here is to handle cases (1) and (2) with prologue data and
case (3) with prefix data.
References
----------
This idea arose out of discussions[1] with Reid Kleckner in response to a
proposal to introduce the notion of symbol offsets to enable handling of
case (3).
[1] http://lists.cs.uiuc.edu/pipermail/llvmdev/2014-May/073235.html
Test Plan: testsuite
Differential Revision: http://reviews.llvm.org/D6454
llvm-svn: 223189
The X86AsmParser intel handling was refactored in r216481, making it
try each different memory operand size to see which one matches.
Operand sizes larger than 80 ("[xyz]mmword ptr") were forgotten, which
led to an "invalid operand" error for code such as:
movdqa [rax], xmm0
llvm-svn: 223187
We need to use the custom expansion of readcyclecounter on all 32-bit targets
(even those with 64-bit registers). This should fix the ppc64 buildbot.
llvm-svn: 223182
A global variable without an explicit alignment specified should be assumed to
be ABI-aligned according to its type, like on other platforms. This allows us
to use better memory operations when accessing it.
rdar://18533701
llvm-svn: 223180
This frequently leads to cases like:
ldr xD, [xN, :lo12:var]
add xA, xN, :lo12:var
ldr xD, [xA, #8]
where the ADD would have been needed anyway, and the two distinct addressing
modes can prevent the formation of an ldp. Because of how we handle ADRP
(aggressively forming an ADRP/ADD pseudo-inst at ISel time), this pattern also
results in duplicated ADRP instructions (one on its own to cover the ldr, and
one combined with the add).
llvm-svn: 223172
Such loops shouldn't be vectorized due to the loops form.
After applying loop-rotate (+simplifycfg) the tests again start to check
what they are intended to check.
llvm-svn: 223170
4i32 shuffles for single insertions into zero vectors lowers to X86vzmovl which was using (v)blendps - causing domain switch stalls. This patch fixes this by using (v)pblendw instead.
The updated tests on test/CodeGen/X86/sse41.ll still contain a domain stall due to the use of insertps - I'm looking at fixing this in a future patch.
Differential Revision: http://reviews.llvm.org/D6458
llvm-svn: 223165
We've long supported readcyclecounter on PPC64, but it is easier there (the
read of the 64-bit time-base register can be accomplished via a single
instruction). This now provides an implementation for PPC32 as well. On PPC32,
the time-base register is still 64 bits, but can only be read 32 bits at a time
via two separate SPRs. The ISA manual explains how to do this properly (it
involves re-reading the upper bits and looping if the counter has wrapped while
being read).
This requires PPC to implement a custom integer splitting legalization for the
READCYCLECOUNTER node, turning it into a target-specific SDAG node, which then
gets turned into a pseudo-instruction, which is then expanded to the necessary
sequence (which has three SPR reads, the comparison and the branch).
Thanks to Paul Hargrove for pointing out to me that this was still unimplemented.
llvm-svn: 223161
Reduce the number of nops emitted for stackmap shadows on AArch64 by counting
non-stackmap instructions up to the next branch target towards the requested
shadow.
<rdar://problem/14959522>
llvm-svn: 223156
Summary:
Like N32/N64, they must be passed in the upper bits of the register.
The new code could be merged with the existing if-statements but I've
refrained from doing this since it will make porting the O32 implementation
to tablegen harder later.
Reviewers: vmedic
Reviewed By: vmedic
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6463
llvm-svn: 223148
Previously .cpu directive in ARM assembler didnt switch to the new CPU and
therefore acted as a nop. This implemented real action for .cpu and eg.
allows to assembler FreeBSD kernel with -integrated-as.
llvm-svn: 223147
This is the third patch in a small series. It contains the CodeGen support for lowering the gc.statepoint intrinsic sequences (223078) to the STATEPOINT pseudo machine instruction (223085). The change also includes the set of helper routines and classes for working with gc.statepoints, gc.relocates, and gc.results since the lowering code uses them.
With this change, gc.statepoints should be functionally complete. The documentation will follow in the fourth change, and there will likely be some cleanup changes, but interested parties can start experimenting now.
I'm not particularly happy with the amount of code or complexity involved with the lowering step, but at least it's fairly well isolated. The statepoint lowering code is split into it's own files and anyone not working on the statepoint support itself should be able to ignore it.
During the lowering process, we currently spill aggressively to stack. This is not entirely ideal (and we have plans to do better), but it's functional, relatively straight forward, and matches closely the implementations of the patchpoint intrinsics. Most of the complexity comes from trying to keep relocated copies of values in the same stack slots across statepoints. Doing so avoids the insertion of pointless load and store instructions to reshuffle the stack. The current implementation isn't as effective as I'd like, but it is functional and 'good enough' for many common use cases.
In the long term, I'd like to figure out how to integrate the statepoint lowering with the register allocator. In principal, we shouldn't need to eagerly spill at all. The register allocator should do any spilling required and the statepoint should simply record that fact. Depending on how challenging that turns out to be, we may invest in a smarter global stack slot assignment mechanism as a stop gap measure.
Reviewed by: atrick, ributzka
llvm-svn: 223137
Follow up from r222926. Also handle multiple destinations from merged
cases on multiple and subsequent phi instructions.
rdar://problem/19106978
llvm-svn: 223135
Go through implicit defs of CSMI and MI, and clear the kill flags on
their uses in all the instructions between CSMI and MI.
We might have made some of the kill flags redundant, consider:
subs ... %NZCV<imp-def> <- CSMI
csinc ... %NZCV<imp-use,kill> <- this kill flag isn't valid anymore
subs ... %NZCV<imp-def> <- MI, to be eliminated
csinc ... %NZCV<imp-use,kill>
Since we eliminated MI, and reused a register imp-def'd by CSMI
(here %NZCV), that register, if it was killed before MI, should have
that kill flag removed, because it's lifetime was extended.
Also, add an exhaustive testcase for the motivating example.
Reviewed by: Juergen Ributzka <juergen@apple.com>
llvm-svn: 223133
The blocking code originated in ARM, which is more aggressive about casting
types to a canonical representative before doing anything else, so I missed out
most vector HFAs and broke the ABI. This should fix it.
llvm-svn: 223126
Load instructions are inserted into loop preheaders when sinking stores
and later removed if not used by the SSA updater. Avoid sinking if the
loop has no preheader and avoid crashes. This fixes one more side effect
of not handling indirectbr instructions properly on LoopSimplify.
llvm-svn: 223119
Certain ARM instructions accept 32-bit immediate operands encoded as a 8-bit
integer value (0-255) and a 4-bit rotation (0-30, even). Current ARM assembly
syntax support in LLVM allows the decoded (32-bit) immediate to be specified
as a single immediate operand for such instructions:
mov r0, #4278190080
The ARMARM defines an extended assembly syntax allowing the encoding to be made
more explicit, as in:
mov r0, #255, #8 ; (same 32-bit value as above)
The behaviour of the two instructions can be different w.r.t flags, which is
documented under "Modified immediate constants" in ARMARM. This patch enables
support for this extended syntax at the MC layer.
llvm-svn: 223113
The default ARM floating-point mode does not support IEEE 754 mode exactly. Of
relevance to this patch is that input denormals are flushed to zero. The way in
which they're flushed to zero depends on the architecture,
* For VFPv2, it is implementation defined as to whether the sign of zero is
preserved.
* For VFPv3 and above, the sign of zero is always preserved when a denormal
is flushed to zero.
When FP support has been disabled, the strategy taken by this patch is to
assume the software support will mirror the behaviour of the hardware support
for the target *if it existed*. That is, for architectures which can only have
VFPv2, it is assumed the software will flush to positive zero. For later
architectures it is assumed the software will flush to zero preserving sign.
Change-Id: Icc5928633ba222a4ba3ca8c0df44a440445865fd
llvm-svn: 223110
System memory allocation functions, which are identified at the IR level by the
noalias attribute on the return value, must return a pointer into a memory region
disjoint from any other memory accessible to the caller. We can use this
property to simplify pointer comparisons between allocated memory and local
stack addresses and the addresses of global variables. Neither the stack nor
global variables can overlap with the region used by the memory allocator.
Fixes PR21556.
llvm-svn: 223093
The statepoint intrinsics are intended to enable precise root tracking through the compiler as to support garbage collectors of all types. The addition of the statepoint intrinsics to LLVM should have no impact on the compilation of any program which does not contain them. There are no side tables created, no extra metadata, and no inhibited optimizations.
A statepoint works by transforming a call site (or safepoint poll site) into an explicit relocation operation. It is the frontend's responsibility (or eventually the safepoint insertion pass we've developed, but that's not part of this patch series) to ensure that any live pointer to a GC object is correctly added to the statepoint and explicitly relocated. The relocated value is just a normal SSA value (as seen by the optimizer), so merges of relocated and unrelocated values are just normal phis. The explicit relocation operation, the fact the statepoint is assumed to clobber all memory, and the optimizers standard semantics ensure that the relocations flow through IR optimizations correctly.
This is the first patch in a small series. This patch contains only the IR parts; the documentation and backend support will be following separately. The entire series can be seen as one combined whole in http://reviews.llvm.org/D5683.
Reviewed by: atrick, ributzka
llvm-svn: 223078
Summary:
".weak" symbols cannot be consumed by ptxas (PR21685). This patch makes the
weak directive in MCAsmPrinter customizable, and disables emitting ".weak"
symbols for NVPTX.
Test Plan: weak-linkage.ll
Reviewers: jholewinski
Reviewed By: jholewinski
Subscribers: majnemer, jholewinski, llvm-commits
Differential Revision: http://reviews.llvm.org/D6455
llvm-svn: 223077
r208210 introduced an optimization that improves the vector select
codegen by doing the setcc on vectors directly.
This is a problem they the setcc operands are i1s, because the
optimization would create vectors of i1, which aren't legal.
Part of PR21549.
Differential Revision: http://reviews.llvm.org/D6308
llvm-svn: 223075
r213378 improved f16 bitcasts, so that they go directly through subregs,
instead of through the stack. That code now causes an assertion failure
for bitcasts from other 16-bits types (most importantly v2i8).
Correct that by doing the custom lowering for i16 bitcasts only when the
input is an f16.
Part of PR21549.
Differential Revision: http://reviews.llvm.org/D6307
llvm-svn: 223074
The MachineVerifier used to check that there was always exactly one
unconditional branch to a non-landingpad (normal) successor.
If that normal successor to an invoke BB is unreachable, it seems
reasonable to only have one successor, the landing pad.
On targets other than AArch64 (and on AArch64 with a different testcase),
the branch folder turns the branch to the landing pad into a fallthrough.
The MachineVerifier, which relies on AnalyzeBranch, is unable to check
the condition, and doesn't complain. However, it does in this specific
testcase, where the branch to the landing pad remained.
Make the MachineVerifier accept it.
llvm-svn: 223059
An unreachable default destination can be exploited by other optimizations, and
SDag lowering is now prepared to handle them efficiently.
For example, branches to the unreachable destination will be optimized away,
such as in the case of range checks for switch lookup tables.
On 64-bit Linux, this reduces the size of a clang bootstrap by 80 kB (and
Chromium by 30 kB).
llvm-svn: 223050
This can significantly reduce the size of the switch, allowing for more
efficient lowering.
I also worked with the idea of exploiting unreachable defaults by
omitting the range check for jump tables, but always ended up with a
non-neglible binary size increase. It might be worth looking into some more.
llvm-svn: 223049
The explicit set of destination types is not fully redundant when lazy loading
since the TypeFinder will not find types used only in function bodies.
This keeps the logic to drop the name of mapped types since it still helps
with avoiding further renaming.
llvm-svn: 223043
Summary:
PowerPC DWARF unwind info defined CFA as SP + offset even in a function
where the stack had been dynamically realigned. This clearly doesn't
work because the offset from SP to CFA is not a constant. Fix it by
defining CFA as BP instead.
This was causing the AddressSanitizer null_deref test to fail 50% of
the time, depending on whether SP happened to be 32-byte aligned on
entry to a particular function or not.
Reviewers: willschm, uweigand, hfinkel
Reviewed By: hfinkel
Subscribers: llvm-commits
Differential Revision: http://reviews.llvm.org/D6410
llvm-svn: 222996
Add checkDecodedInstruction for post-decode checking of instructions, to catch
the corner cases like HVC that don't fit into the general pattern. Needed to
check for an invalid condition field in instruction encoding despite HVC not
taking a predicate.
Patch by Matthew Wahab.
Change-Id: I48e28de981d7a9e43569594da3c45fb478b4f795
llvm-svn: 222992
This commit fixes a bug in stack protector pass where edge weights were not set
when new basic blocks were added to lists of successor basic blocks.
Differential Revision: http://reviews.llvm.org/D5766
llvm-svn: 222987
Instead of keeping an explicit set, just drop the names of types we choose
to map to some other type.
This has the advantage that the name of the unused will not cause the context
to rename types on module read.
llvm-svn: 222986
Add assembler support for the fixed-point cache-inhibited load/store
instructions. These are hypervisor-level only, so don't get too excited ;)
Fixes PR21650.
llvm-svn: 222976
This reverts commit r222632 (and follow-up r222636), which caused a host
of LNT failures on an internal bot. I'll respond to the commit on the
list with a reproduction of one of the failures.
Conflicts:
lib/Target/X86/X86TargetTransformInfo.cpp
llvm-svn: 222936
We may be in a situation where the icmps might not be near each other in
a tree of or instructions. Try to dig out related compare instructions
and see if they combine.
N.B. This won't fire on deep trees of compares because rewritting the
tree might end up creating a net increase of IR. We may have to resort
to something more sophisticated if this is a real problem.
llvm-svn: 222928
Loop simplify skips exit-block insertion when exits contain indirectbr
instructions. This leads to an assertion in LICM when trying to sink
stores out of non-dedicated loop exits containing indirectbr
instructions. This patch fix this issue by re-checking for dedicated
exits in LICM prior to store sink attempts.
Differential Revision: http://reviews.llvm.org/D6414
rdar://problem/18943047
llvm-svn: 222927
Switch cases statements with sequential values that branch to the same
destination BB may often be handled together in a single new source BB.
In this scenario we need to remove remaining incoming values from PHI
instructions in the destination BB, as to match the number of source
branches.
Differential Revision: http://reviews.llvm.org/D6415
rdar://problem/19040894
llvm-svn: 222926
Allow unaligned 16-byte memop codegen for btver2. No functional changes for any other subtargets.
Replace the existing supposed small memcpy test with an actual test of a small memcpy.
The previous test wasn't using FileCheck either.
This patch should allow us to close PR21541 ( http://llvm.org/bugs/show_bug.cgi?id=21541 ).
Differential Revision: http://reviews.llvm.org/D6360
llvm-svn: 222925
The original patch would fail when:
* A dst opaque type (%A) is matched with a src type (%A).
* A src opaque (%E) type is then speculatively matched with %A and the
speculation fails afterward.
* When rolling back the speculation we would cancel the source %A to dest
%A mapping.
The fix is to keep an explicit list of which resolutions are speculative.
Original message:
Fix overly aggressive type merging.
If we find out that two types are *not* isomorphic, we learn nothing about
opaque sub types in both the source and destination.
llvm-svn: 222923
MSan does not assign origin for instrumentation temps (i.e. the ones that do
not come from the application code), but "select" instrumentation erroneously
tried to use one of those.
https://code.google.com/p/memory-sanitizer/issues/detail?id=78
llvm-svn: 222918
Add more tests to make sure the encoding/decoding of build attributes works
correctly for all permissible values of build attributes. For cases where there
are an infinite number of such values, a representative subset has been settled
for.
Change-Id: I2643c9624c211b2d56405306e16eec2d487bc5d6
llvm-svn: 222917
The AAPCS treats small structs and homogeneous floating (or vector) aggregates
specially, and guarantees they either get passed as a contiguous block of
registers, or prevent any future use of those registers and get passed on the
stack.
This concept can fit quite neatly into LLVM's own type system, mapping an HFA
to [N x float] and so on, and small structs to [N x i64]. Doing so allows
front-ends to emit AAPCS compliant code without having to duplicate the
register counting logic.
llvm-svn: 222903
I also added a test.
Original message:
Allow FDE references outside the +/-2GB range supported by PC relative
offsets for code models other than small/medium. For JIT application,
memory layout is less controlled and can result in truncations
otherwise.
Patch from Akos Kiss.
Differential Revision: http://reviews.llvm.org/D6079
llvm-svn: 222897
This reverts commit r222760.
It changed our behaviour on PIC so we don't match gas anymore. It also
included lots of unnecessary changes to tests.
If those changes are desirable, there should be an independent discussion
as they are out of scope for that patch.
I will recommit the other bits.
llvm-svn: 222896
This reverts commit r222727, which causes LTO bootstrap failures.
Last passing @ r222698:
http://lab.llvm.org:8080/green/job/clang-Rlto_master_build/532/
First failing @ r222843:
http://lab.llvm.org:8080/green/job/clang-Rlto_master_build/533/
Internal bootstraps pointed at a much narrower range: r222725 is
passing, and r222731 is failing.
LTO crashes while handling libclang.dylib:
http://lab.llvm.org:8080/green/job/clang-Rlto_master_build/533/consoleFull#-158682280549ba4694-19c4-4d7e-bec5-911270d8a58c
GEP is not of right type for indices!
%InfoObj.i.i = getelementptr inbounds %"class.llvm::OnDiskIterableChainedHashTable"* %.lcssa, i64 0, i32 0, i32 4, !dbg !123627
%"class.clang::serialization::reader::ASTIdentifierLookupTrait" = type { %"class.clang::ASTReader.31859"*, %"class.clang::serialization::ModuleFile.31870"*, %"class.clang::IdentifierInfo"* }LLVM ERROR: Broken function found, compilation aborted!
clang: error: linker command failed with exit code 1 (use -v to see invocation)
Looks like the new algorithm doesn't merge types aggressively enough.
llvm-svn: 222895
Fixed missing dominance check.
Original commit message:
This optimization tries to reuse the generated compare instruction, if there is a comparison against the default value after the switch.
Example:
if (idx < tablesize)
r = table[idx]; // table does not contain default_value
else
r = default_value;
if (r != default_value)
...
Is optimized to:
cond = idx < tablesize;
if (cond)
r = table[idx];
else
r = default_value;
if (cond)
...
Jump threading will then eliminate the second if(cond).
llvm-svn: 222891
The string data for string-valued build attributes were being unconditionally
uppercased. There is no mention in the ARM ABI addenda about case conventions,
so it's technically implementation defined as to whether the data are
capitialised in some way or not. However, there are good reasons not to
captialise the data.
* It's less work.
* Some vendors may legitimately have case-sensitive checks for these
attributes which would fail on LLVM generated object files.
* There could be locale issues with uppercasing.
The original reasons for uppercasing appear to have stemmed from an
old codesourcery toolchain behaviour, see
http://comments.gmane.org/gmane.comp.compilers.llvm.cvs/87133
This patch makes the object file emitted no longer captialise string
data, it encodes as seen in the assembly source.
Change-Id: Ibe20dd6e60d2773d57ff72a78470839033aa5538
llvm-svn: 222882
This optimization tries to reuse the generated compare instruction, if there is a comparison against the default value after the switch.
Example:
if (idx < tablesize)
r = table[idx]; // table does not contain default_value
else
r = default_value;
if (r != default_value)
...
Is optimized to:
cond = idx < tablesize;
if (cond)
r = table[idx];
else
r = default_value;
if (cond)
...
\endcode
Jump threading will then eliminate the second if(cond).
llvm-svn: 222872
This restores our ability to optimize:
(X & C) ? X & ~C : X into X & ~C
(X & C) ? X : X & ~C into X
(X & C) ? X | C : X into X
(X & C) ? X : X | C into X | C
llvm-svn: 222868
This reverts commit r210006, it miscompiled libapr which is used in who
knows how many projects.
A test has been added to ensure that we don't regress again.
I'll work on a rewrite of what the optimization was trying to do later.
llvm-svn: 222856
This mostly entails adding relocations, however there are a couple of
changes to existing relocations:
1. R_AARCH64_NONE is defined to be zero rather than 256
R_AARCH64_NONE has been defined to be zero for a long time elsewhere
e.g. binutils and glibc since the submission of the AArch64 port in
2012 so this is required for compatibility.
2. R_AARCH64_TLSDESC_ADR_PAGE renamed to R_AARCH64_TLSDESC_ADR_PAGE21
I don't think there is any way for relocation names to leak out of LLVM
so this should not break anything.
Tested with check-all with no regressions.
llvm-svn: 222821
including SAE mode and memory operand.
Added AVX512_maskable_scalar template, that should cover all scalar instructions in the future.
The main difference between AVX512_maskable_scalar<> and AVX512_maskable<> is using X86select instead of vselect.
I need it, because I can't create vselect node for MVT::i1 mask for scalar instruction.
http://reviews.llvm.org/D6378
llvm-svn: 222820
Since (v)pslldq / (v)psrldq instructions resolve to a single input argument it is useful to match it much earlier than we currently do - this prevents more complicated shuffles (notably insertion into a zero vector) matching before it.
Differential Revision: http://reviews.llvm.org/D6409
llvm-svn: 222796
If solveBlockValue() needs results from predecessors that are not already
computed, it returns false with the intention of resuming when the dependencies
have been resolved. However, the computation would never be resumed since an
'overdefined' result had been placed in the cache, preventing any further
computation.
The point of placing the 'overdefined' result in the cache seems to have been
to break cycles, but we can check for that when inserting work items in the
BlockValue stack instead. This makes the "stop and resume" mechanism of
solveBlockValue() work as intended, unlocking more analysis.
Using this patch shaves 120 KB off a 64-bit Chromium build on Linux.
I benchmarked compiling bzip2.c at -O2 but couldn't measure any difference in
compile time.
Tests by Jiangning Liu from r215343 / PR21238, Pete Cooper, and me.
Differential Revision: http://reviews.llvm.org/D6397
llvm-svn: 222768
On LP64 platforms, it will work or not depending on the choosen memory
layout, so neither PASS nor XFAIL is appropiate.
As UNSUPPORTED as per-test target doesn't exist (yet), remove the test
instead to unbreak the builds.
llvm-svn: 222767
This changes the order in which different types are passed to get, but
one order is not inherently better than the other.
The main motivation is that this simplifies linkDefinedTypeBodies now that
it is only linking "real" opaque types. It is also means that we only have to
call it once and that we don't need getImpl.
A small change in behavior is that we don't copy type names when resolving
opaque types. This is an improvement IMHO, but it can be added back if
desired. A test is included with the new behavior.
llvm-svn: 222764
and PIC:
Allow FDE references outside the +/-2GB range supported by PC relative
offsets for code models other than small/medium. For JIT application,
memory layout is less controlled and can result in truncations
otherwise.
Patch from Akos Kiss.
Differential Revision: http://reviews.llvm.org/D6079
llvm-svn: 222760
stored rather than the pointer type.
This change is analogous to r220138 which changed the canonicalization
for loads. The rationale is the same: memory does not have a type,
operations (and thus the values they produce) have a type. We should
match that type as closely as possible rather than reading some form of
semantics into the pointer type.
With this change, loads and stores should no longer be made with
nonsensical types for the values that tehy load and store. This is
particularly important when trying to match specific loaded and stored
types in the process of doing other instcombines, which is what led me
down this twisty maze of miscanonicalization.
I've put quite some effort into looking through IR to find places where
LLVM's optimizer was being unreasonably conservative in the face of
mismatched load and store types, however it is possible (let's say,
likely!) I have missed some. If you see regressions here, or from
r220138, the likely cause is some part of LLVM failing to cope with load
and store types differing. Test cases appreciated, it is important that
we root all of these out of LLVM.
llvm-svn: 222748
clearly only exactly equal width ptrtoint and inttoptr casts are no-op
casts, it says so right there in the langref. Make the code agree.
Original log from r220277:
Teach the load analysis to allow finding available values which require
inttoptr or ptrtoint cast provided there is datalayout available.
Eventually, the datalayout can just be required but in practice it will
always be there today.
To go with the ability to expose available values requiring a ptrtoint
or inttoptr cast, helpers are added to perform one of these three casts.
These smarts are necessary to finish canonicalizing loads and stores to
the operational type requirements without regressing fundamental
combines.
I've added some test cases. These should actually improve as the load
combining and store combining improves, but they may fundamentally be
highlighting some missing combines for select in addition to exercising
the specific added logic to load analysis.
llvm-svn: 222739
The pattern matching failed to recognize all instances of "-1", because when
comparing against "-1" we didn't use an APInt of the same bitwidth.
This commit fixes this and also adds inverse versions of the conditon to catch
more cases.
llvm-svn: 222722
This handles cases where we are comparing a masked value against itself.
The analysis could be further improved by making it recursive but such
expense is not currently justified.
llvm-svn: 222716
The attn instruction is not part of the Power ISA, but is documented in the A2
user manual, and is accepted by the GNU assembler for the A2 and the POWER4+.
Reported as part of PR21650.
llvm-svn: 222712
This does not matter on newer cores (where we can use reciprocal estimates in
fast-math mode anyway), but for older cores this allows us to generate better
fast-math code where we have multiple FDIVs with a common divisor.
llvm-svn: 222710
When processing an assignment in the integrated assembler that sets
a symbol to the value of another symbol, we need to copy the st_other
bits that encode the local entry point offset.
Modeled after MipsTargetELFStreamer::emitAssignment handling of the
ELF::STO_MIPS_MICROMIPS flag.
llvm-svn: 222672
We would create an instruction but not inserting it.
Not inserting the unused instruction would lead us to verification
failure.
This fixes PR21653.
llvm-svn: 222659
Fix JRADDIUSP instruction, remove delay slot flag because this instruction
doesn't have delay slot.
Differential Revision: http://reviews.llvm.org/D6365
llvm-svn: 222658
With the help of new method readInstruction16() two bytes are read and
decodeInstruction() is called with DecoderTableMicroMips16, if this fails
four bytes are read and decodeInstruction() is called with
DecoderTableMicroMips32.
Differential Revision: http://reviews.llvm.org/D6149
llvm-svn: 222648
This patch teaches function 'transformVSELECTtoBlendVECTOR_SHUFFLE' how to
convert VSELECT dag nodes to shuffles on targets that do not have SSE4.1.
On pre-SSE4.1 targets, we can still perform blend operations using movss/movsd.
Also, removed a target specific combine that performed a premature lowering of
VSELECT nodes to target specific MOVSS/MOVSD nodes.
llvm-svn: 222647
We tried to get the result of DataLayout::getLargestLegalIntTypeSize but
we didn't have a DataLayout. This resulted in opt crashing.
This fixes PR21651.
llvm-svn: 222645
r222375 made some improvements to build_vector lowering of v4x32 and v4xf32 into an insertps, but it missed a case where:
1. A single extracted element is used twice.
2. The lower of the two non-zero indexes should be preserved, and the higher should be used for the dest mask.
This caused a crash, since the source value for the insertps ends-up uninitialized.
Differential Revision: http://reviews.llvm.org/D6377
llvm-svn: 222635
Introduced new target-independent intrinsics in order to support masked vector loads and stores. The loop vectorizer optimizes loops containing conditional memory accesses by generating these intrinsics for existing targets AVX2 and AVX-512. The vectorizer asks the target about availability of masked vector loads and stores.
Added SDNodes for masked operations and lowering patterns for X86 code generator.
Examples:
<16 x i32> @llvm.masked.load.v16i32(i8* %addr, <16 x i32> %passthru, i32 4 /* align */, <16 x i1> %mask)
declare void @llvm.masked.store.v8f64(i8* %addr, <8 x double> %value, i32 4, <8 x i1> %mask)
Scalarizer for other targets (not AVX2/AVX-512) will be done in a separate patch.
http://reviews.llvm.org/D6191
llvm-svn: 222632
has a remarkably unique and efficient lowering.
While we get this some of the time already, we miss a few cases and
there wasn't a principled reason we got it. We should at least test
this. v8 already has tests for this pattern.
llvm-svn: 222607
Fixes the self-host fail. Note that this commit activates dominator
analysis in the combiner by default (like the original commit did).
llvm-svn: 222590
This s_mov_b32 will write to a virtual register from the M0Reg
class and all the ds instructions now take an extra M0Reg explicit
argument.
This change is necessary to prevent issues with the scheduler
mixing together instructions that expect different values in the m0
registers.
llvm-svn: 222583
filler such as if delay slot filler have to put NOP instruction into the
delay slot of microMIPS BEQ or BNE instruction which uses the register $0,
then instead of emitting NOP this instruction is replaced by the corresponding
microMIPS compact branch instruction, i.e. BEQZC or BNEZC.
Differential Revision: http://reviews.llvm.org/D3566
llvm-svn: 222580
This patch adds a feature flag to avoid unaligned 32-byte load/store AVX codegen
for Sandy Bridge and Ivy Bridge. There is no functionality change intended for
those chips. Previously, the absence of AVX2 was being used as a proxy to detect
this feature. But that hindered codegen for AVX-enabled AMD chips such as btver2
that do not have the 32-byte unaligned access slowdown.
Performance measurements are included in PR21541 ( http://llvm.org/bugs/show_bug.cgi?id=21541 ).
Differential Revision: http://reviews.llvm.org/D6355
llvm-svn: 222544
shuffle lowering to allow much better blend matching.
Specifically, with the new structure the code seems clearer to me and we
correctly can hit the cases where merging two 128-bit lanes is a clear
win and can be shuffled cheaply afterward.
llvm-svn: 222539
a bunch more improvements.
Non-lane-crossing is fine, the key is that lane merging only makes sense
for single-input shuffles. Not sure why I got so turned around here. The
code all works, I was just using the wrong model for it.
This only updates v4 and v8 lowering. The v16 and v32 lowering requires
restructuring the entire check sequence.
llvm-svn: 222537
Before this patch, the DAGCombiner only tried to convert build_vector dag nodes
into shuffles if all operands were either extract_vector_elt or undef.
This patch improves that logic and teaches the DAGCombiner how to deal with
build_vector dag nodes where one or more operands are zero. A build_vector
dag node with some zero operands is turned into a shuffle only if the resulting
shuffle mask is legal for the target.
llvm-svn: 222536
lanes.
By special casing these we can often either reduce the total number of
shuffles significantly or reduce the number of (high latency on Haswell)
AVX2 shuffles that potentially cross 128-bit lanes. Even when these
don't actually cross lanes, they have much higher latency to support
that. Doing two of them and a blend is worse than doing a single insert
across the 128-bit lanes to blend and then doing a single interleaved
shuffle.
While this seems like a narrow case, it kept cropping up on me and the
difference is *huge* as you can see in many of the test cases. I first
hit this trying to perfectly fix the interleaving shuffle patterns used
by Halide for AVX2.
llvm-svn: 222533
merging 128-bit subvectors and also shuffling all the elements of those
subvectors. Currently we generate pretty bad code for many of these, but
I'm testing a patch that should dramatically improve this in addition to
making the shuffle lowering robust to other changes.
llvm-svn: 222525
E.g., ( a / D; b / D ) -> ( recip = 1.0 / D; a * recip; b * recip)
A hook is added to allow the target to control whether it needs to do such combine.
Reviewed in http://reviews.llvm.org/D6334
llvm-svn: 222510
This mirrors r222331, which enabled SeparateConstOffsetFromGEP on AArch64, in
the PowerPC backend. Yields, on a POWER7 machine, a 30% speedup on
SingleSource/Benchmarks/Shootout/nestedloop (this might just be from LICM,
there is a store moved out of the inner loop) and a potential speedup on
MultiSource/Benchmarks/mediabench/mpeg2/mpeg2dec/mpeg2decode. Regardless, it
makes some code look cleaner, and synchronizing the backends in this regard
seems like a generally good thing.
llvm-svn: 222504
The alloca's type is irrelevant, only those types which are used in a
load or store of the exact size of the slice should be considered.
This manifested as an assertion failure when we compared the various
types: we had a size mismatch.
This fixes PR21480.
llvm-svn: 222499
Currently LoopUnroll generates a prologue loop before the main loop
body to execute first N%UnrollFactor iterations. Also, this loop is
used if trip-count can overflow - it's determined by a runtime check.
However, we've been mistakenly optimizing this loop to a linear code for
UnrollFactor = 2, not taking into account that it also serves as a safe
version of the loop if its trip-count overflows.
llvm-svn: 222451
Windows itanium targets the MSVCRT, and the stack probe symbol is provided by
MSVCRT. This corrects the emission of stack probes on i686-windows-itanium.
llvm-svn: 222439
This reverts commit r222142. This is causing/exposing an execution-time regression
in spec2006/gcc and coremark on AArch64/A57/Ofast.
Conflicts:
test/Transforms/Reassociate/optional-flags.ll
llvm-svn: 222398
This patch improves the lowering of v4f32 and v4i32 build_vector dag nodes
that are known to have at least two non-zero elements.
With this patch, a build_vector that performs a blend with zero is
converted into a shuffle. This is done to let the shuffle legalizer expand
the dag node in a optimal way. For example, if we know that a build_vector
performs a blend with zero, we can try to lower it as a movq/blend instead of
always selecting an insertps.
This patch also improves the logic that lowers a build_vector into a insertps
with zero masking. See for example the extra test cases added to test sse41.ll.
Differential Revision: http://reviews.llvm.org/D6311
llvm-svn: 222375
A register operand that has a common sub-class with its instruction's
defined register class is not always legal. For example,
SReg_32 and M0Reg both have a common sub-class, but we can't
use an SReg_32 in instructions that expect a M0Reg.
This prevents the llvm.SI.sendmsg.ll test from failing when the fold
operand pass is added.
llvm-svn: 222368
When the BasicBlock containing the return instrution has a PHI with 2
incoming values, FoldReturnIntoUncondBranch will remove the no longer
used incoming value and remove the no longer needed phi as well. This
leaves us with a BB that no longer has a PHI, but the subsequent call
to FoldReturnIntoUncondBranch from FoldReturnAndProcessPred will not
remove the return instruction (which still uses the result of the call
instruction). This prevents EliminateRecursiveTailCall to remove
the value, as it is still being used in a basicblock which has no
predecessors.
The basicblock can not be erased on the spot, because its iterator is
still being used in runTRE.
This issue was exposed when removing the threshold on size for lifetime
marker insertion for named temporaries in clang. The testcase is a much
reduced version of peelOffOuterExpr(const Expr*, const ExplodedNode *)
from clang/lib/StaticAnalyzer/Core/BugReporterVisitors.cpp.
llvm-svn: 222354
This patch builds on http://reviews.llvm.org/D5598 to perform byte rotation shuffles (lowerVectorShuffleAsByteRotate) on pre-SSSE3 (palignr) targets - pre-SSSE3 is only enabled on i8 and i16 vector targets where it is a more definite performance gain.
I've also added a separate byte shift shuffle (lowerVectorShuffleAsByteShift) that makes use of the ability of the SLLDQ/SRLDQ instructions to implicitly shift in zero bytes to avoid the need to create a zero register if we had used palignr.
Differential Revision: http://reviews.llvm.org/D5699
llvm-svn: 222340
AliasSetTracker::addUnknown may create an AliasSet devoid of pointers
just to contain an instruction if no suitable AliasSet already exists.
It will then AliasSet::addUnknownInst and we will be done.
However, it's possible for addUnknown to choose an existing AliasSet to
addUnknownInst.
If this were to occur, we are in a bit of a pickle: removing pointers
from the AliasSet can cause the entire AliasSet to become destroyed,
taking our unknown instructions out with them.
Instead, keep track whether or not our AliasSet has any unknown
instructions.
This fixes PR21582.
llvm-svn: 222338
SeparateConstOffsetFromGEP can gives more optimizaiton opportunities related to GEPs, which benefits EarlyCSE
and LICM. By enabling these passes we can have better address calculations and generate a better addressing
mode. Some SPEC 2006 benchmarks (astar, gobmk, namd) have obvious improvements on Cortex-A57.
Reviewed in http://reviews.llvm.org/D5864.
llvm-svn: 222331
It printed out base relocation table header as table entry.
This patch also makes llvm-readobj to not skip ABSOLUTE entries
becuase it was confusing.
llvm-svn: 222299
This partially makes up for not having address spaces
used for alias analysis in some simple cases.
This is not yet enabled by default so shouldn't change anything yet.
llvm-svn: 222286
Under many circumstances the stack is not 32-byte aligned, resulting in the use of the vmovups/vmovupd/vmovdqu instructions when inserting ymm reloads/spills.
This minor patch adds these instructions to the isFrameLoadOpcode/isFrameStoreOpcode helpers so that they can be correctly identified and not be treated as folded reloads/spills.
This has also been noticed by http://llvm.org/bugs/show_bug.cgi?id=18846 where it was causing redundant spills - I've added a reduced test case at test/CodeGen/X86/pr18846.ll
Differential Revision: http://reviews.llvm.org/D6252
llvm-svn: 222281
shift-right for booleans (i1).
Arithmetic shift-right immediate with sign-/zero-extensions also works for
boolean values. Update the assert and the test cases to reflect that fact.
llvm-svn: 222272
shift-right for booleans (i1).
Logical shift-right immediate with sign-/zero-extensions also works for boolean
values. Update the assert and the test cases to reflect that fact.
llvm-svn: 222270
We would attempt to replace an frem's operand with the same operand.
This would cause InstCombine to think real work was done, causing
InstCombine to enter an infinite loop.
This fixes the second part of PR21576.
llvm-svn: 222265
Shifts also perform sign-/zero-extends to larger types, which requires us to emit
an integer extend instead of a simple COPY.
Related to PR21594.
llvm-svn: 222257
This should expose more of the actually used VALU
instructions to the machine optimization passes.
This also should help getting i1 handling into a better state.
For not entirly understood reasons, this fixes the split-scalar-i64-add.ll
test where a 64-bit add would only partially be moved to the VALU
resulting in use of undefined VCC.
llvm-svn: 222256
This change emits a COPY for a shift-immediate with a "zero" shift value.
This fixes PR21594 where we emitted a shift instruction with an incorrect
immediate operand.
llvm-svn: 222247
EarlyCSE is giving up on the current instruction immediately when it recognizes that the current instruction makes a previous store trivially dead. There's no reason to do this. Once the previous store has been deleted, it's perfectly legal to remember the value of the current store (for value forwarding) and the fact the store occurred (it could be dead too!).
Reviewed by: Hal
Differential Revision: http://reviews.llvm.org/D6301
llvm-svn: 222241
It is impossible for (x & INT_MAX) == 0 && x == INT_MAX to ever be true.
While this sort of reasoning should normally live in InstSimplify,
the machinery that derives this result is not trivial to split out.
llvm-svn: 222230
Usually global variables are in a retain list and instanciated before
any call to constructImportedEntityDIE is made. This isn't true for
forward declarations though.
The testcase for this change is generated by a clang patched to emit
such forward declarations (patch at http://reviews.llvm.org/D6173
which will land soon). The updated testcase tests more than just
global variables, it now tests every type of 'using' clause we
support.
llvm-svn: 222217
I added a pessimization in r217102 to prevent miscompiles when the
incremented induction variable was used in a comparison; it would be
poison.
Try to use the incremented induction variable more often when we can be
sure that the increment won't end in poison.
Differential Revision: http://reviews.llvm.org/D6222
llvm-svn: 222213
use DIScopeRef.
A paired commit at clang will follow to show cases where we will use an
identifer for the context of a global variable.
rdar://18958417
llvm-svn: 222195
Change uniquing from a `FoldingSet` to a `DenseSet` with custom
`DenseMapInfo`. Unfortunately, this doesn't save any memory, since
`DenseSet<T>` is a simple wrapper for `DenseMap<T, char>`, but I'll come
back to fix that later.
I used the name `GenericDenseMapInfo` to the custom `DenseMapInfo` since
I'll be splitting `MDNode` into two classes soon: `MDNodeFwdDecl` for
temporaries, and `GenericMDNode` for everything else.
I also added a non-debug-info reduced version of a type-uniquing test
that started failing on an earlier draft of this patch.
Part of PR21532.
llvm-svn: 222191
When converting a switch to a lookup table we might have to generate a bitmaks
to encode and check for holes in the original switch statement.
The type of this mask depends on the number of switch statements, which can
result in illegal types for pretty much all architectures.
To avoid unnecessary type legalization and help FastISel this commit increases
the size of the bitmask to next power-of-2 value when necessary.
This fixes rdar://problem/18984639.
llvm-svn: 222168
The triple parser should only accept existing architecture names
when the triple starts with armv, armebv, thumbv or thumbebv.
Patch by Gabor Ballabas.
llvm-svn: 222129
This was motivated by a bug which caused code like this to be
miscompiled:
declare void @take_ptr(i8*)
define void @test() {
%addr1.32 = alloca i8
%addr2.32 = alloca i32, i32 1028
call void @take_ptr(i8* %addr1)
ret void
}
This was emitting the following assembly to get the value of %addr1:
add r0, sp, #1020
add r0, r0, #8
However, "add r0, r0, #8" is not a valid Thumb1 instruction, and this
could not be assembled. The generated object file contained this,
resulting in r0 holding SP+8 rather tha SP+1028:
add r0, sp, #1020
add r0, sp, #8
This function looked like it could have caused miscompilations for
other combinations of registers and offsets (though I don't think it is
currently called with these), and the heuristic it used did not match
the emitted code in all cases.
llvm-svn: 222125
We were a little lax in a few areas:
- We pretended that import libraries were like any old COFF file, they
are not. In fact, they aren't really COFF files at all, we should
probably grow some specialized functionality to handle them smarter.
- Our symbol iterators were more than happy to attempt to go past the
end of the symbol table if you had a symbol with a bad list of
auxiliary symbols.
llvm-svn: 222124
Some optimisations in DAGCombiner cause miscompilations for targets that use
TargetLowering::UndefinedBooleanContent, because they assume that the results
of a SELECT_CC node are boolean values, and can be safely ANDed, ORed and
XORed. These optimisations are only valid for targets that use
ZeroOrOneBooleanContent or ZeroOrNegativeOneBooleanContent.
This is a follow-up to D6210/r221693.
llvm-svn: 222123
This is a simple optimization for switch table lookup:
It computes the output value directly with an (optional) mul and add if there is a linear mapping between index and output.
Example:
int f1(int x) {
switch (x) {
case 0: return 10;
case 1: return 11;
case 2: return 12;
case 3: return 13;
}
return 0;
}
generates:
define i32 @f1(i32 %x) #0 {
entry:
%0 = icmp ult i32 %x, 4
br i1 %0, label %switch.lookup, label %return
switch.lookup:
%switch.offset = add i32 %x, 10
ret i32 %switch.offset
return:
ret i32 0
}
llvm-svn: 222121
This adds back r222061, but now calls initializePAEvalPass from the correct
library to avoid link problems.
Original message:
Don't make assumptions about the name of private global variables.
Private variables are can be renamed, so it is not reliable to make
decisions on the name.
The name is also dropped by the assembler before getting to the
linker, so using the name causes a disconnect between how llvm makes a
decision (var name) and how the linker makes a decision (section it is
in).
This patch changes one case where we were looking at the variable name to use
the section instead.
Test tuning by Michael Gottesman.
llvm-svn: 222117
Summary:
Several places in DependenceAnalysis assumes both SCEVs in a subscript pair
share the same integer type. For instance, isKnownPredicate calls
SE->getMinusSCEV(X, Y) which asserts X and Y share the same type. However,
DependenceAnalysis fails to ensure this assumption when producing a subscript
pair, causing tests such as NonCanonicalizedSubscript to crash. With this
patch, DependenceAnalysis runs unifySubscriptType before producing any
subscript pair, ensuring the assumption.
Test Plan:
Added NonCanonicalizedSubscript.ll on which DependenceAnalysis before the fix
crashed because subscripts have different types.
Reviewers: spop, sebpop, jingyue
Reviewed By: jingyue
Subscribers: eliben, meheff, llvm-commits
Differential Revision: http://reviews.llvm.org/D6289
llvm-svn: 222100
HowFarToZero was supposed to use unsigned division in order to calculate
the backedge taken count. However, SCEVDivision::divide performs signed
division. Unless I am mistaken, no users of SCEVDivision actually want
signed arithmetic: switch to udiv and urem.
This fixes PR21578.
llvm-svn: 222093
This patch teaches the DAGCombiner how to combine shuffles according to rules:
shuffle(shuffle(A, Undef, M0), B, M1) -> shuffle(B, A, M2)
shuffle(shuffle(A, B, M0), B, M1) -> shuffle(B, A, M2)
shuffle(shuffle(A, B, M0), A, M1) -> shuffle(B, A, M2)
llvm-svn: 222090
Updated X86TargetLowering::isShuffleMaskLegal to match SHUFP masks with commuted inputs and PSHUFD masks that reference the second input.
As part of this I've refactored isPSHUFDMask to work in a more general manner and allow it to match against either the first or second input vector.
Differential Revision: http://reviews.llvm.org/D6287
llvm-svn: 222087
This gets the correct NaN behavior based on the compare type
the hardware uses. This now passes the new piglit test I have
for this on SI.
Add stricter tests for the operand order.
llvm-svn: 222079
Private variables are can be renamed, so it is not reliable to make
decisions on the name.
The name is also dropped by the assembler before getting to the
linker, so using the name causes a disconnect between how llvm makes a
decision (var name) and how the linker makes a decision (section it is
in).
This patch changes one case where we were looking at the variable name to use
the section instead.
Test tuning by Michael Gottesman.
llvm-svn: 222061
We use to track quite a few "adjusted" offsets through the FrameLowering code
to account for changes in the prologue instructions as we went and allow the
emission of correct CFA annotations. However, we were missing a couple of cases
and the code was almost impenetrable.
It's easier to just add any stack-adjusting instruction to a list and emit them
together.
llvm-svn: 222057
When we folded the DPR alignment gap into a push, we weren't noting the extra
distance from the beginning of the push to the FP, and so FP ended up pointing
at an incorrect offset.
The .cfi_def_cfa_offset directives are still wrong in this case, but I think
that can be improved by refactoring.
llvm-svn: 222056
The test's DWARF stubs were there just to trigger the emission of .cfi
directives. Fortunately, the NetBSD ABI already demands proper DWARF unwind
info, so it's easier to just use that triple.
llvm-svn: 222055
FYI, removed the unused MCInstrAnalysis as it does not exist for 64-bit ARM and
was causing a “couldn't initialize disassembler for target” error.
llvm-svn: 222045
We would attempt to replace a fptrunc of an frem with an identical
fptrunc. This would cause the new fptrunc to be added to the worklist.
Of course, this results in an infinite loop because we will keep
visiting the newly created fptruncs.
This fixes PR21576.
llvm-svn: 222040
doing Load PRE"
This commit updates the failing test in
Analysis/TypeBasedAliasAnalysis/gvn-nonlocal-type-mismatch.ll
The failing test is sensitive to the order in which we process loads. This
version turns on the RPO traversal instead of the while DT traversal in GVN.
The new test code is functionally same just the order of loads that are
eliminated is swapped.
This new version also fixes an issue where GVN splits a critical edge and
potentially invalidate the RPO/DT iterator.
llvm-svn: 222039
If we have spilled the value of the m0 register, then we need to restore
it with v_readlane_b32 to a regular sgpr, because v_readlane_b32 can't
write to m0.
v_readlane_b32 can't write to m0, so
llvm-svn: 222036
ELF targets (and maybe COFF) use relocations when referring
to strings in the .debug_str section. Handle that in the
accelerator table dumper. This commit restores the
test/DebugInfo/cross-cu-inlining.ll test to its expected
platform independant form, validating that the fix works
(this test failed on linux boxes).
llvm-svn: 222029
If this workaround gets the bots green, then we have to find out
why the -dwarf-accel-tables=Enable option doesn't work as
expected on non-darwin platforms.
llvm-svn: 222007
Prior to this commit fmul and fadd binary operators were being canonicalized for
both scalar and vector versions. We now canonicalize add, mul, and, or, and xor
vector instructions.
llvm-svn: 222006
This reverts commit r221842 which was a revert of r221836 and of the
test parts of r221837.
This new version fixes an UB bug pointed out by David (along with
addressing some other review comments), makes some dumping more
resilient to broken input data and forces the accelerator tables
to be dumped in the tests where we use them (this decision is
platform specific otherwise).
llvm-svn: 222003
This patch adds builtin support for xvdivdp and xvdivsp, along with a
test case. Straightforward stuff.
There's a companion patch for Clang.
llvm-svn: 221983
getTargetConstant should only be used when you can guarantee the instruction
selected will be able to cope with the raw value. BUILD_VECTOR is rather too
generic for this so we should use getConstant instead. In that case, an
instruction can still consume the constant, but if it doesn't it'll be
materialised through its own round of ISel.
Should fix PR21352.
llvm-svn: 221961
Summary:
This has most of what is needed for mips fast-isel call lowering for O32.
What is missing I will add on the next patch because this patch is already too large.
It should not be doing anything wrong but it will punt on some cases that it is basically
capable of doing.
The mechanism is there for parameters to be passed on the stack but I have not enabled it because it serves as a way for now to prevent some of the strange cases of O32 register passing that I have not fully checked yet and have some issues.
The Mips O32 abi rules are very complicated as far how data is passed in floating and integer registers.
However there is a way to think about this all very simply and this implementation reflects that.
Basically, the ABI rules are written as if everything is passed on the stack and aligned as such.
Once that is conceptually done, it is nearly trivial to reassign those locations to registers and
then all the complexity disappears.
So I have told tablegen that all the data is passed on the stack and during the lowering I fix
this by assigning to registers as per the ABI doc.
This has been my approach and you can line up what I did with the ABI document and see 1 to 1 what
is going on.
Test Plan: callabi.ll
Reviewers: dsanders
Reviewed By: dsanders
Subscribers: jholewinski, echristo, ahatanak, llvm-commits, rfuhler
Differential Revision: http://reviews.llvm.org/D5714
llvm-svn: 221948
Fix for LLI failure on Windows\X86: http://llvm.org/PR5053
LLI.exe crashes on Windows\X86 when single precession floating point
intrinsics like the following are used: acos, asin, atan, atan2, ceil,
copysign, cos, cosh, exp, floor, fmin, fmax, fmod, log, pow, sin, sinh,
sqrt, tan, tanh
The above intrinsics are defined as inline-expansions in math.h, and are
not exported by msvcr120.dll (Win32 API GetProcAddress returns null).
For an FREM instruction, the JIT compiler generates a call to a stub for
the fmodf() intrinsic, and adds a relocation to fixup at load time. The
loader searches the libraries for the function, but fails because the
symbol is not exported. So, the call target remains NULL and the
execution crashes.
Since the math functions are loaded at JIT/runtime, the JIT can patch
CALL instruction directly instead of the searching the libraries'
exported symbols. However, this fix caused build failures due to
unresolved symbols like _fmodf at link time.
Therefore, the current fix defines helper functions in the Runtime
link/load library to perform the above operations. The address of these
helper functions are used to patch up the CALL instruction at load time.
Reviewers: lhames, rnk
Reviewed By: rnk
Differential Revision: http://reviews.llvm.org/D5387
Patch by Swaroop Sridhar!
llvm-svn: 221947
in-lane shuffles that aren't always handled well by the current vector
shuffle lowering.
No functionality change yet, that will follow in a subsequent commit.
llvm-svn: 221938
The generic FastISel code would bail, because it can't emit a sign-extend for
AArch64. This copies the code over and uses AArch64 specific emit functions.
This is not ideal and 'computeAddress' should handles this, so it can fold the
address computation into the memory operation.
I plan to clean up 'computeAddress' anyways, so I will add that in a future
commit.
Related to rdar://problem/18962471.
llvm-svn: 221923
If a function is just an unreachable, this would hit a
"this is not a MachO target" assertion because of setting
HasSubsectionViaSymbols.
llvm-svn: 221920
e.g. v_mad_f32 a, b, c -> v_mad_f32 b, a, c
This simplifies matching v_madmk_f32.
This looks somewhat surprising, but it appears to be
OK to do this. We can commute src0 and src1 in all
of these instructions, and that's all that appears
to matter.
llvm-svn: 221910
Normally entries can only move to a lower address, but when that wasn't viable,
the user's block was considered anyway. Unfortunately, it went via
createNewWater which wasn't designed to handle the case where there's already
an island after the block.
Unfortunately, the test we have is slow and fragile, and I couldn't reduce it
to anything sane even with the @llvm.arm.space intrinsic. The test change here
is recreating the previous one after the change.
rdar://problem/18545506
llvm-svn: 221905
We were using a naive heuristic to determine whether a basic block already had
an unconditional branch at the end. This mostly corresponded to reality
(assuming branches got optimised) because there's not much point in a branch to
the next block, but could go wrong.
llvm-svn: 221904
Creating tests for the ConstantIslands pass is very difficult, since it depends
on precise layout details. Having the ability to precisely inject a number of
bytes into the stream helps greatly.
llvm-svn: 221903
Let's try this again...
This reverts r219432, plus a bug fix.
Description of the bug in r219432 (by Nick):
The bug was using AllPositive to break out of the loop; if the loop break
condition i != e is changed to i != e && AllPositive then the
test_modulo_analysis_with_global test I've added will fail as the Modulo will
be calculated incorrectly (as the last loop iteration is skipped, so Modulo
isn't updated with its Scale).
Nick also adds this comment:
ComputeSignBit is safe to use in loops as it takes into account phi nodes, and
the == EK_ZeroEx check is safe in loops as, no matter how the variable changes
between iterations, zero-extensions will always guarantee a zero sign bit. The
isValueEqualInPotentialCycles check is therefore definitely not needed as all
the variable analysis holds no matter how the variables change between loop
iterations.
And this patch also adds another enhancement to GetLinearExpression - basically
to convert ConstantInts to Offsets (see test_const_eval and
test_const_eval_scaled for the situations this improves).
Original commit message:
This reverts r218944, which reverted r218714, plus a bug fix.
Description of the bug in r218714 (by Nick):
The original patch forgot to check if the Scale in VariableGEPIndex flipped the
sign of the variable. The BasicAA pass iterates over the instructions in the
order they appear in the function, and so BasicAliasAnalysis::aliasGEP is
called with the variable it first comes across as parameter GEP1. Adding a
%reorder label puts the definition of %a after %b so aliasGEP is called with %b
as the first parameter and %a as the second. aliasGEP later calculates that %a
== %b + 1 - %idxprom where %idxprom >= 0 (if %a was passed as the first
parameter it would calculate %b == %a - 1 + %idxprom where %idxprom >= 0) -
ignoring that %idxprom is scaled by -1 here lead the patch to incorrectly
conclude that %a > %b.
Revised patch by Nick White, thanks! Thanks to Lang to isolating the bug.
Slightly modified by me to add an early exit from the loop and avoid
unnecessary, but expensive, function calls.
Original commit message:
Two related things:
1. Fixes a bug when calculating the offset in GetLinearExpression. The code
previously used zext to extend the offset, so negative offsets were converted
to large positive ones.
2. Enhance aliasGEP to deduce that, if the difference between two GEP
allocations is positive and all the variables that govern the offset are also
positive (i.e. the offset is strictly after the higher base pointer), then
locations that fit in the gap between the two base pointers are NoAlias.
Patch by Nick White!
llvm-svn: 221876
between splitting a vector into 128-bit lanes and recombining them vs.
decomposing things into single-input shuffles and a final blend.
This handles a large number of cases in AVX1 where the cross-lane
shuffles would be much more expensive to represent even though we end up
with a fast blend at the root. Instead, we can do a better job of
shuffling in a single lane and then inserting it into the other lanes.
This fixes the remaining bits of Halide's regression captured in PR21281
for AVX1. However, the bug persists in AVX2 because I've made this
change reasonably conservative. The cases where it makes sense in AVX2
to split into 128-bit lanes are much more rare because we can often do
full permutations across all elements of the 256-bit vector. However,
the particular test case in PR21281 is an example of one of the rare
cases where it is *always* better to work in a single 128-bit lane. I'm
going to try to teach the logic to detect and form the good code even in
AVX2 next, but it will need to use a separate heuristic.
Finally, there is one pesky regression here where we previously would
craftily use vpermilps in AVX1 to shuffle both high and low halves at
the same time. We no longer pull that off, and not for any really good
reason. Ultimately, I think this is just another missing nuance to the
selection heuristic that I'll try to add in afterward, but this change
already seems strictly worth doing considering the magnitude of the
improvements in common matrix math shuffle patterns.
As always, please let me know if this causes a surprising regression for
you.
llvm-svn: 221861
David Majnemer