Ordinarily (shl (add x, c1), c2) -> (add (shl x, c2), c1 << c2)
is only done if the add has one use. If the resulting constant
add can be folded into an addressing mode, force this to happen
for the pointer operand.
This ends up happening a lot because of how LDS objects are allocated.
Since the globals are allocated next to each other, acessing the first
element of the second object is directly indexed by a shifted pointer.
llvm-svn: 215739
The default assumes that a 16-bit signed offset is used.
LDS instruction use a 16-bit unsigned offset, so it wasn't
being used in some cases where it was assumed a negative offset
could be used.
More should be done here, but first isLegalAddressingMode needs
to gain an addressing mode argument. For now, copy most of the rest
of the default implementation with the immediate offset change.
llvm-svn: 215732
In a previous iteration of the pass, we would try to compensate for
writeback by updating later instructions and/or inserting a SUBS to
reset the base register if necessary.
Since such a SUBS sets the condition flags it's not generally safe to do
this. For now, only merge LDR/STRs if there is no writeback to the base
register (LDM that loads into the base register) or the base register is
killed by one of the merged instructions. These cases are clear wins
both in terms of instruction count and performance.
Also add three new test cases, and update the existing ones accordingly.
llvm-svn: 215729
the new shuffle lowering and an implementation for v4 shuffles.
This allows us to handle non-half-crossing shuffles directly for v4
shuffles, both integer and floating point. This currently misses places
where we could perform the blend via UNPCK instructions, but otherwise
generates equally good or better code for the test cases included to the
existing vector shuffle lowering. There are a few cases that are
entertainingly better. ;]
llvm-svn: 215702
target-specific shuffl DAG combines.
We were recognizing the paired shuffles backwards. This code needs to be
replaced anyways as we have the same functionality elsewhere, but I'll
do the refactoring in a follow-up, this is the minimal fix to the
behavior.
In addition to fixing miscompiles with the new vector shuffle lowering,
it also causes the canonicalization to kick in much better, selecting
the smaller encoding variants in lots of places in the new AVX path.
This still isn't quite ideal as we don't need both the shufpd and the
punpck instructions, but that'll get fixed in a follow-up patch.
llvm-svn: 215690
broken logic for merging shuffle masks in the face of SM_SentinelZero
mask operands.
While these are '-1' they don't mean 'undef' the way '-1' means in the
pre-legalized shuffle masks. Instead, they mean that the shuffle
operation is forcibly zeroing that lane. Reflect this and explicitly
handle it in a bunch of places. In one place the effect is equivalent
but much more clear. In the rest it was really weirdly broken.
Also, rewrite the entire merging thing to be a more directy operation
with a single loop and just doing math to map the indices through the
various masks.
Also add a bunch of asserts to try to make in extremely clear what the
different masks can possibly look like.
Finally, add some comments to clarify that we're merging shuffle masks
*up* here rather than *down* as we do everywhere else, and thus the
logic is quite confusing.
Thanks to several different people for sending test cases, and for
Robert Khasanov for an initial attempt at fixing.
llvm-svn: 215687
FastEmit_i won't always succeed to materialize an i32 constant and just fail.
This would trigger a fall-back to SelectionDAG, which is really not necessary.
This fix will first fall-back to a constant pool load to materialize the constant
before giving up for good.
This fixes <rdar://problem/18022633>.
llvm-svn: 215682
This reverts:
r215595 "[FastISel][X86] Add large code model support for materializing floating-point constants."
r215594 "[FastISel][X86] Use XOR to materialize the "0" value."
r215593 "[FastISel][X86] Emit more efficient instructions for integer constant materialization."
r215591 "[FastISel][AArch64] Make use of the zero register when possible."
r215588 "[FastISel] Let the target decide first if it wants to materialize a constant."
r215582 "[FastISel][AArch64] Cleanup constant materialization code. NFCI."
llvm-svn: 215673
This patch allows a vector fneg of a bitcasted integer value to be optimized in the same way that we already optimize a scalar fneg. If the integer variable is a constant, we can precompute the result and not require any logic ops.
This patch is very similar to a fabs patch committed at r214892.
Differential Revision: http://reviews.llvm.org/D4852
llvm-svn: 215646
Summary:
This is done by removing some hardcoded registers like $at or expecting a single digit register to be selected.
Contains work done by Matheus Almeida.
Reviewers: matheusalmeida, dsanders
Reviewed By: dsanders
Subscribers: tomatabacu
Differential Revision: http://reviews.llvm.org/D4227
llvm-svn: 215640
lowering scheme.
Currently, this just directly bails to the fallback path of splitting
the 256-bit vector into two 128-bit vectors, operating there, and then
joining the results back together. While the results are far from
perfect, they are *shockingly* good for what we're doing here. I'll be
layering the rest of the functionality on top of this piece by piece and
updating tests as I go.
Note that 256-bit vectors in this mode are still somewhat WIP. While
I think the code paths that I'm adding here are clean and good-to-go,
there are still a lot of 128-bit assumptions that I'll need to stomp out
as I march through the functional spread here.
llvm-svn: 215637
input node after manually adding it to the worklist and using CombineTo.
Once we use CombineTo the input node may have been deleted. Despite this
being *completely confusing* and somewhat broken, the only way to
"correctly" return from a DAG combine after potentially deleting the
input node is to return *that exact node*....
But really, this code should just never have used CombineTo. It won't do
what it wants (returning the node as mentioned above just causes the
combine to infloop). The correct way to combine away a casted load to
a load of the correct type is to RAUW the chain directly and then return
the loaded value to replace the actual value node.
I managed to find this with the vector shuffle fuzzer even though it
clearly has nothing at all to do with vector shuffles and rather those
happen to trigger a load of a constant pool that hits this combine *just
right*. I've included the test as it is small and a nice stress test
that the infrastructure isn't asserting.
llvm-svn: 215622
combining by replacing it with something else but not re-process the
node afterward to remove it.
In a truly remarkable stroke of bad luck, this would (in the test case
attached) end up getting some other node combined into it without ever
getting re-processed. By adding it back on to the worklist, in addition
to deleting the dead nodes more quickly we also ensure that if it
*stops* being dead for any reason it makes it back through the
legalizer. Without this, the test case will end up failing during
instruction selection due to an and node with a type we don't have an
instruction pattern for.
It took many million runs of the shuffle fuzz tester to find this.
llvm-svn: 215611
Certain functions such as objc_autoreleaseReturnValue have to be called as
tail-calls even at -O0. Since normal fast-isel doesn't emit calls as tail calls,
we have to fall back to SelectionDAG to select calls that are marked as tail.
<rdar://problem/17991614>
llvm-svn: 215600
FastISel didn't take much advantage of the different addressing modes available
to it on AArch64. This commit allows the ComputeAddress method to recognize more
addressing modes that allows shifts and sign-/zero-extensions to be folded into
the memory operation itself.
For Example:
lsl x1, x1, #3 --> ldr x0, [x0, x1, lsl #3]
ldr x0, [x0, x1]
sxtw x1, w1
lsl x1, x1, #3 --> ldr x0, [x0, x1, sxtw #3]
ldr x0, [x0, x1]
llvm-svn: 215597
In the large code model for X86 floating-point constants are placed in the
constant pool and materialized by loading from it. Since the constant pool
could be far away, a PC relative load might not work. Therefore we first
materialize the address of the constant pool with a movabsq and then load
from there the floating-point value.
Fixes <rdar://problem/17674628>.
llvm-svn: 215595
This mostly affects the i64 value type, which always resulted in an 15byte
mobavsq instruction to materialize any constant. The custom code checks the
value of the immediate and tries to use a different and smaller mov
instruction when possible.
This fixes <rdar://problem/17420988>.
llvm-svn: 215593
This change materializes now the value "0" from the zero register.
The zero register can be folded by several instruction, so no
materialization is need at all.
Fixes <rdar://problem/17924413>.
llvm-svn: 215591
This changes the order in which FastISel tries to materialize a constant.
Originally it would try to use a simple target-independent approach, which
can lead to the generation of inefficient code.
On X86 this would result in the use of movabsq to materialize any 64bit
integer constant - even for simple and small values such as 0 and 1. Also
some very funny floating-point materialization could be observed too.
On AArch64 it would materialize the constant 0 in a register even the
architecture has an actual "zero" register.
On ARM it would generate unnecessary mov instructions or not use mvn.
This change simply changes the order and always asks the target first if it
likes to materialize the constant. This doesn't fix all the issues
mentioned above, but it enables the targets to implement such
optimizations.
Related to <rdar://problem/17420988>.
llvm-svn: 215588
This change is also in preparation for a future change to make sure that
the constant materialization uses MOVT/MOVW when available and not a load
from the constant pool.
llvm-svn: 215584
This for some reason fixes v1i64 kernel arguments on pre-SI. This
currently breaks some other cases in the kernel-args.ll test for R600,
but I'm not particularly confident in the new output. VTX_READ_* are not
used for some of the scalarized cases, and the code reading from the
constant buffer doesn't make much sense to me.
llvm-svn: 215564
This patch improves the existing algorithm in DAGCombiner that
attempts to fold shuffles according to rule:
shuffle(shuffle(x, y, M1), undef, M2) -> shuffle(y, undef, M3)
Before this change, there were cases where the DAGCombiner conservatively
avoided folding shuffles even if the resulting mask would have been legal.
That is because the algorithm wrongly assumed that commuting
an illegal shuffle mask would always produce an illegal mask.
With this change, we now correctly compute the commuted shuffle mask before
calling method 'isShuffleMaskLegal' on it.
On X86, this improves for example the codegen for the following function:
define <4 x i32> @test(<4 x i32> %A, <4 x i32> %B) {
%1 = shufflevector <4 x i32> %B, <4 x i32> %A, <4 x i32> <i32 1, i32 2, i32 6, i32 7>
%2 = shufflevector <4 x i32> %1, <4 x i32> undef, <4 x i32> <i32 2, i32 3, i32 2, i32 3>
ret <4 x i32> %2
}
Before this change the X86 backend (-mcpu=corei7) generated
the following assembly code for function @test:
shufps $-23, %xmm0, %xmm1 # xmm1 = xmm1[1,2],xmm0[2,3]
movhlps %xmm1, %xmm1 # xmm1 = xmm1[1,1]
movaps %xmm1, %xmm0
Now we produce:
movhlps %xmm0, %xmm0 # xmm0 = xmm0[1,1]
Added extra test cases in combine-vec-shuffle-2.ll to verify that we correctly
fold according to the above-mentioned rule.
llvm-svn: 215555
Added avx512_movnt_vl multiclass for handling 256/128-bit forms of instruction.
Added encoding and lowering tests.
Reviewed by Elena Demikhovsky <elena.demikhovsky@intel.com>
llvm-svn: 215536
one pesky test case correctly.
This test case caused the old code to infloop occilating between solving
the low-half and the high-half. The 'side balancing' part of
single-input v8 shuffle lowering didn't handle the one pattern which can
cause it to occilate. Fortunately the fuzz testing found this case.
Unfortuately it was *terrible* to handle. I'm really sorry for the
amount and density of the code here, I'd love suggestions on how to
simplify it. I feel like there *must* be a simpler form here, but after
a lot of days I've not found it. This is the only one I've found that
even works. I've added the one pesky test case along with some nice
comments explaining the core problem that we have to solve here.
So far this has survived approximately 32k test cases. More strenuous
fuzzing commencing.
llvm-svn: 215519
This implements PPCTargetLowering::getTgtMemIntrinsic for Altivec load/store
intrinsics. As with the construction of the MachineMemOperands for the
intrinsic calls used for unaligned load/store lowering, the only slight
complication is that we need to represent a larger memory range than the
loaded/stored value-type size (because the address is rounded down to an
aligned address, and we need to conservatively represent the entire possible
range of the actual access). This required adding an extra size field to
TargetLowering::IntrinsicInfo, and this was done in a way that required no
modifications to other targets (the size defaults to the store size of the
provided memory data type).
This fixes test/CodeGen/PowerPC/unal-altivec-wint.ll (so it can be un-XFAILed).
llvm-svn: 215512
Unfortunately, our use of the SDNode class hierarchy for INTRINSIC_W_CHAIN and
INTRINSIC_VOID nodes is somewhat broken right now. These nodes sometimes are
used for memory intrinsics (those with MachineMemOperands), and sometimes not.
When not, the nodes are not created as instances of MemIntrinsicSDNode, but
rather created as some other subclass of SDNode using DAG::getNode. When they
are memory intrinsics, they are created using DAG::getMemIntrinsicNode as
instances of MemIntrinsicSDNode. MemIntrinsicSDNode is a subclass of
MemSDNode, but prior to r214452, we had a non-self-consistent setup whereby
MemIntrinsicSDNode::classof on INTRINSIC_W_CHAIN and INTRINSIC_VOID would
return true but MemSDNode::classof on INTRINSIC_W_CHAIN and INTRINSIC_VOID
would return false. In r214452, MemSDNode::classof was changed to return true
for INTRINSIC_W_CHAIN and INTRINSIC_VOID, which is now self-consistent. The
problem is that neither the pre-r214452 logic and the post-r214452 logic are
really right. The truth is that not all INTRINSIC_W_CHAIN and INTRINSIC_VOID
nodes are instances of MemIntrinsicSDNode (or MemSDNode for that matter), and
the return value from classof needs to reflect that. This was broken before
r214452 (because MemIntrinsicSDNode::classof always returned true), and was
broken afterward (because MemSDNode::classof also always returned true), and
will now be correct.
The minimal solution is to grab one of the SubclassData bits (there is one left
for MemIntrinsicSDNode nodes) and use it to store whether or not a particular
INTRINSIC_W_CHAIN or INTRINSIC_VOID is really an instance of
MemIntrinsicSDNode or not. Doing this allows both MemIntrinsicSDNode::classof
and MemSDNode::classof to return the correct answer for the underlying object
for both the memory-intrinsic and non-memory-intrinsic cases.
This fixes the problem that r214452 created in the SelectionDAGDumper (thanks
to Matt Arsenault for pointing it out).
Because PowerPC does not implement getTgtMemIntrinsic, this change breaks
test/CodeGen/PowerPC/unal-altivec-wint.ll. I've XFAILed it for now, and will
fix it in a follow-up commit.
llvm-svn: 215511
I think that this will scale better in most cases than adding a Pat<> for each
mapping from the intrinsic DAG to the intruction (i.e. rri, rrik, rrikz). We
can just lower to the SDNode and have the resulting DAG be matches by the DAG
patterns.
Alternatively (long term), we could keep the Pat<>s but generate them via the
new AVX512_masking multiclass. The difficulty is that in order to formulate
that we would have to concatenate DAGs. Currently this is only supported if
the operators of the input DAGs are identical.
llvm-svn: 215473
v2: drop enum keyword
use correct extension mode
don't bother computing the sign in unsinged case
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
llvm-svn: 215462
v2: add tests
rename LowerSDIV24 to LowerSDIVREM24
handle the rem part in this function
Signed-off-by: Jan Vesely <jan.vesely@rutgers.edu>
llvm-svn: 215460
The combiner ignored DBG nodes when checking
the uses of a virtual register.
It combined a sequence like
%vreg1 = madd %vreg2, %vreg3,...
DBG_VALUE (%vreg1 ...)
%vreg4 = add %vreg1,...
to
%vreg4 = madd %vreg2, %vreg3
leaving behind a dangling DBG_VALUE with
a definition. This triggered an assertion
in the MachineTraceMetrics.cpp module.
llvm-svn: 215431
There are no variable values like registers encoded in the low 32 bits of MUBUF
instructions, so it is relatively easy to check these bits, and it will
help prevent us from introducing encoding bugs.
llvm-svn: 215397
This bit was left uninitialized, which was causing some random failures
of piglit tests.
NOTE: This is a candidate for the 3.5 branch.
llvm-svn: 215396
For many Thumb-1 register register instructions, setting the CPSR is not
permitted inside an IT block. We would not correctly flag those instructions.
The previous change to identify this scenario was insufficient as it did not
actually catch all the instances. The current list is formed by manual
inspection of the ARMv6M ARM.
The change to the Thumb2 IT block test is due to the fact that the new more
stringent checking of the MIs results in the If Conversion pass being prevented
from executing (since not all the instructions in the BB are predicable). This
results in code gen changes.
Thanks to Tim Northover for pointing out that the previous patch was
insufficient and hinting that the use of the v6M ARM would be much easier to use
than the v7 or v8!
llvm-svn: 215382
By default, LLVM uses the "C" calling convention for all runtime
library functions. The half-precision FP conversion functions use the
soft-float calling convention, and are needed for some targets which
use the hard-float convention by default, so must have their calling
convention explicitly set.
llvm-svn: 215348
be propagated to all its users, and this propagation could increase the
probability of finding common subexpressions. If the COPY has only one user,
the COPY itself can be removed.
llvm-svn: 215344
Follow up to r214266. Add missing case in ScalarizeVectorResult() for
cttz_zero_undef.
Differential Revision: http://reviews.llvm.org/D4813
llvm-svn: 215330
The ARM ARM states that CPSR may not be updated by a MUL in thumb mode. Due to
an ordering of Thumb 2 Size Reduction and If Conversion, we would end up
generating a THUMB MULS inside an IT block.
The If Conversion pass uses the TTI isPredicable method to ensure that it can
transform a Basic Block. However, because we only check for IT handling on
Thumb2 functions, we may miss some cases. Even then, it only validates that the
CPSR is not *live* rather than it is not accessed. This corrects the handling
for that particular case since the same restriction does not hold on the vast
majority of the instructions.
This does prevent the IfConversion optimization from kicking in in certain
cases, but generating correct code is more valuable. Addresses PR20555.
llvm-svn: 215328
These tests were using SI-NOT: MOVREL to make sure concat vectors
weren't being lowered to stack loads and stores, but we are using
scratch buffers for the stack now instead of registers, so we need
to add an additional SI-NOT check for scratch buffers.
With this change I was able to uncover one broken test which will
be fixed in a future commit.
llvm-svn: 215269
I accidentally also used INC/DEC for unsigned arithmetic which doesn't work,
because INC/DEC don't set the required flag which is used for the overflow
check.
llvm-svn: 215237
Also added the testcase that should have been in r215194.
This behaviour has surprised me a few times now. The problem is that the
generated MipsSubtarget::ParseSubtargetFeatures() contains code like this:
if ((Bits & Mips::FeatureABICalls) != 0) IsABICalls = true;
so '-abicalls' means 'leave it at the default' and '+abicalls' means 'set it to
true'. In this case, (and the similar -modd-spreg case) I'd like the code to be
IsABICalls = (Bits & Mips::FeatureABICalls) != 0;
or possibly:
if ((Bits & Mips::FeatureABICalls) != 0)
IsABICalls = true;
else
IsABICalls = false;
and preferably arrange for 'Bits & Mips::FeatureABICalls' to be true by default
(on some triples).
llvm-svn: 215211
For best-case performance on Cortex-A57, we should try to use a balanced mix of odd and even D-registers when performing a critical sequence of independent, non-quadword FP/ASIMD floating-point multiply or multiply-accumulate operations.
This pass attempts to detect situations where the register allocation may adversely affect this load balancing and to change the registers used so as to better utilize the CPU.
Ideally we'd just take each multiply or multiply-accumulate in turn and allocate it alternating even or odd registers. However, multiply-accumulates are most efficiently performed in the same functional unit as their accumulation operand. Therefore this pass tries to find maximal sequences ("Chains") of multiply-accumulates linked via their accumulation operand, and assign them all the same "color" (oddness/evenness).
This optimization affects S-register and D-register floating point multiplies and FMADD/FMAs, as well as vector (floating point only) muls and FMADD/FMA. Q register instructions (and 128-bit vector instructions) are not affected.
llvm-svn: 215199
This short-circuited our error reporting for incorrectly specified
target triples (you'd get AArch64 code instead).
Should fix PR20567.
llvm-svn: 215191
This completes one item from the todo-list of r215125 "Generate masking
instruction variants with tablegen".
The AddedComplexity is needed just like for the k variant.
Added a codegen test based on valignq.
llvm-svn: 215173
__stack_chk_guard.
Handle the case where the pointer operand of the load instruction that loads the
stack guard is not a global variable but instead a bitcast.
%StackGuard = load i8** bitcast (i64** @__stack_chk_guard to i8**)
call void @llvm.stackprotector(i8* %StackGuard, i8** %StackGuardSlot)
Original test case provided by Ana Pazos.
This fixes PR20558.
llvm-svn: 215167
a base GOT entry.
Summary:
get tip of tree mips fast-isel to pass test-suite
Two bugs were fixed:
1) one bit booleans were treated as 1 bit signed integers and so the literal '1' could become sign extended.
2) mips uses got for pic but in certain cases, as with string constants for example, many items can be referenced from the same got entry and this case was not handled properly.
Test Plan: test-suite
Reviewers: dsanders
Reviewed By: dsanders
Subscribers: mcrosier
Differential Revision: http://reviews.llvm.org/D4801
llvm-svn: 215155
Re-commit of r214832,r21469 with a work-around that
avoids the previous problem with gcc build compilers
The work-around is to use SmallVector instead of ArrayRef
of basic blocks in preservesResourceLen()/MachineCombiner.cpp
llvm-svn: 215151
BranchFolderPass was not correctly setting the basic block branch weights when
tail-merging created or merged blocks. This patch recomutes the weights of
tail-merged blocks using the following formula:
branch_weight(merged block to successor j) =
sum(block_frequency(bb) * branch_probability(bb -> j))
bb is a block that is in the set of merged blocks.
<rdar://problem/16256423>
llvm-svn: 215135
shuffle lowering.
This is closely related to the previous one. Here we failed to use the
source offset when swapping in the other case -- where we end up
swapping the *final* shuffle. The cause of this bug is a bit different:
I simply wasn't thinking about the fact that this mask is actually
a slice of a wide mask and thus has numbers that need SourceOffset
applied. Simple fix. Would be even more simple with an algorithm-y thing
to use here, but correctness first. =]
llvm-svn: 215095
via the fuzz tester.
Here I missed an offset when round-tripping a value through a shuffle
mask. I got it right 2 lines below. See a problem? I do. ;] I'll
probably be adding a little "swap" algorithm which accepts a range and
two values and swaps those values where they occur in the range. Don't
really have a name for it, let me know if you do.
llvm-svn: 215094
through the new fuzzer.
This one is great: bad operator precedence led the modulus to happen at
the wrong point. All the asserts didn't fire because there were usually
the right values past the end of the 4 element region we were looking
at. Probably could have gotten a crash here with ASan + fuzzing, but the
correctness tests pinpointed this really nicely.
llvm-svn: 215092
Summary:
Since pointers are 32-bit on x32 we can use ebp and esp as frame and stack
pointer. Some operations like PUSH/POP and CFI_INSTRUCTION still
require 64-bit register, so using 64-bit MachineFramePtr where required.
X86_64 NaCl uses 64-bit frame/stack pointers, however it's been found that
both isTarget64BitLP64 and isTarget64BitILP32 are true for NaCl. Addressing
this issue here as well by making isTarget64BitLP64 false.
Also mark hasReservedSpillSlot unreachable on X86. See inlined comments.
Test Plan: Add one new simple test and upgrade 2 existing with x32 target case.
Reviewers: nadav, dschuff
Subscribers: llvm-commits, zinovy.nis
Differential Revision: http://reviews.llvm.org/D4617
llvm-svn: 215091
fuzz testing.
The function which tested for adjacency did what it said on the tin, but
when I called it, I wanted it to do something more thorough: I wanted to
know if the *pairs* of shuffle elements were adjacent and started at
0 mod 2. In one place I had the decency to try to test for this, but in
the other it was completely skipped, miscompiling this test case. Fix
this by making the helper actually do what I wanted it to do everywhere
I called it (and removing the now redundant code in one place).
I *really* dislike the name "canWidenShuffleElements" for this
predicate. If anyone can come up with a better name, please let me know.
The other name I thought about was "canWidenShuffleMask" but is it
really widening the mask to reduce the number of lanes shuffled? I don't
know. Naming things is hard.
llvm-svn: 215089
Otherwise we can end up with an argument frame size that is not a
multiple of stack slot size, which is very awkward.
This fixes PR20547, which was a bug in x86_64 Sys V vararg handling.
However, it's much easier to test this with x86 callee-cleanup
functions, which previously ended in "retl $6" instead of "retl $8".
This does affect behavior of all backends, but it presumably fixes the
same bug in all of them.
llvm-svn: 214980
Particularly on MachO, we were generating "blx _dest" instructions on M-class
CPUs, which don't actually exist. They happen to get fixed up by the linker
into valid "bl _dest" instructions (which is why such a massive issue has
remained largely undetected), but we shouldn't rely on that.
llvm-svn: 214959
test case to actually generate correct code.
The primary miscompile fixed here is that we weren't correctly handling
in-place elements in one half of a single-input v8i16 shuffle when
moving a dword of elements from that half to the other half. Some times,
we would clobber the in-place elements in forming the dword to move
across halves.
The fix to this involves forcibly marking the in-place inputs even when
there is no need to gather them into a dword, and to much more carefully
re-arrange the elements when grouping them into a dword to move across
halves. With these two changes we would generate correct shuffles for
the test case, but found another miscompile. There are also some random
perturbations of the generated shuffle pattern in SSE2. It looks like
a wash; more instructions in some cases fewer in others.
The second miscompile would corrupt the results into nonsense. This is
a buggy pattern in one of the added DAG combines. Mapping elements
through a PSHUFD when pairing redundant half-shuffles is *much* harder
than this code makes it out to be -- it requires reasoning about *all*
of where the input is used in the PSHUFD, not just one part of where it
is used. Plus, we can't combine a half shuffle *into* a PSHUFD but the
code didn't guard against it. I think this was just a bad idea and I've
just removed that aspect of the combine. No tests regress as
a consequence so seems OK.
llvm-svn: 214954
This partially fixes weird looking load scheduling
in memcpy test. The load clustering doesn't seem
particularly smart, but this method seems to be partially
deprecated so it might not be worth trying to fix.
llvm-svn: 214943
This currently has a noticable effect on the kernel argument loads.
LDS and global loads are more problematic, I think because of how copies
are currently inserted to ensure that the address is a VGPR.
llvm-svn: 214942
This was coming in weird debug info that had variables (and hence
debug_locs) but was in GMLT mode (because it was missing the 13th field
of the compile_unit metadata) so no ranges were constructed. We should
always have at least one range for any CU with a debug_loc in it -
because the range should cover the debug_loc.
The assertion just ensures that the "!= 1" range case inside the
subsequent loop doesn't get entered for the case where there are no
ranges at all, which should never reach here in the first place.
llvm-svn: 214939
Without the 13th field, the "emission kind" field defaults to 0 (which
is not equal to either of the values of the emission kind enum (1 ==
full debug info, 2 == line tables only)).
In this particular instance, the comparison with "FullDebugInfo" was
done when adding elements to the ranges list - so for these test cases
no values were added to the ranges list.
This got weirder when emitting debug_loc entries as the addresses should
be relative to the range of the CU if the CU has only one range (the
reasonable assumption is that if we're emitting debug_loc lists for a CU
that CU has at least one range - but due to the above situation, it has
zero) so the ranges were emitted relative to the start of the section
rather than relative to the start of the CU's singular range.
Fix these tests by accounting for the difference in the description of
debug_loc entries (in some cases making the test ignorant to these
differences, in others adding the extra label difference expression,
etc) or the presence/absence of high/low_pc on the CU, and add the 13th
field to their CUs to enable proper "full debug info" emission here.
In a future commit I'll fix up a bunch of other test cases that are not
so rigorously depending on this behavior, but still doing similarly
weird things due to the missing 13th field.
llvm-svn: 214937
This reverts r214893, re-applying r214881 with the test case relaxed a bit to
satiate the build bots.
POP on armv4t cannot be used to change thumb state (unilke later non-m-class
architectures), therefore we need a different return sequence that uses 'bx'
instead:
POP {r3}
ADD sp, #offset
BX r3
This patch also fixes an issue where the return value in r3 would get clobbered
for functions that return 128 bits of data. In that case, we generate this
sequence instead:
MOV ip, r3
POP {r3}
ADD sp, #offset
MOV lr, r3
MOV r3, ip
BX lr
http://reviews.llvm.org/D4748
llvm-svn: 214928
Commits r213915 and r214718 fix recognition of shuffle masks for vmrg*
and vpku*um instructions for a little-endian target, by swapping the
input arguments. The vsldoi instruction requires similar treatment,
and also needs its shift count adjusted for little endian.
Reviewed by Ulrich Weigand.
This is a bug fix candidate for release 3.5 (and hopefully the last of
those for PowerPC).
llvm-svn: 214923
a test case.
We also miscompile this test case which is showing a serious flaw in the
single-input v8i16 shuffle code. I've left the specific instruction
checks FIXME-ed out until I can address the bug in the single-input
code, but I wanted to separate out a significant functionality change to
produce correct code from a very simple and targeted crasher fix.
The miscompile problem stems from keeping track of inputs by value
rather than by index. As a consequence of doing this, we can't reliably
update those inputs because they might swap and we can't detect this
without copying the mask.
The blend code now uses indices for the input lists and this seems
strictly better. It also should make it easier to sort things and do
other cleanups. I think the time has come to simplify The Great Lambda
here.
llvm-svn: 214914
Allow vector fabs operations on bitcasted constant integer values to be optimized
in the same way that we already optimize scalar fabs.
So for code like this:
%bitcast = bitcast i64 18446744069414584320 to <2 x float> ; 0xFFFF_FFFF_0000_0000
%fabs = call <2 x float> @llvm.fabs.v2f32(<2 x float> %bitcast)
%ret = bitcast <2 x float> %fabs to i64
Instead of generating something like this:
movabsq (constant pool loadi of mask for sign bits)
vmovq (move from integer register to vector/fp register)
vandps (mask off sign bits)
vmovq (move vector/fp register back to integer return register)
We should generate:
mov (put constant value in return register)
I have also removed a redundant clause in the first 'if' statement:
N0.getOperand(0).getValueType().isInteger()
is the same thing as:
IntVT.isInteger()
Testcases for x86 and ARM added to existing files that deal with vector fabs.
One existing testcase for x86 removed because it is no longer ideal.
For more background, please see:
http://reviews.llvm.org/D4770
And:
http://llvm.org/bugs/show_bug.cgi?id=20354
Differential Revision: http://reviews.llvm.org/D4785
llvm-svn: 214892
This is similar to what I did with the two-source permutation recently. (It's
almost too similar so that we should consider generating the masking variants
with some tablegen help.)
Both encoding and intrinsic tests are added as well. For the latter, this is
what the IR that the intrinsic test on the clang side generates.
Part of <rdar://problem/17688758>
llvm-svn: 214890
POP on armv4t cannot be used to change thumb state (unilke later non-m-class
architectures), therefore we need a different return sequence that uses 'bx'
instead:
POP {r3}
ADD sp, #offset
BX r3
This patch also fixes an issue where the return value in r3 would get clobbered
for functions that return 128 bits of data. In that case, we generate this
sequence instead:
MOV ip, r3
POP {r3}
ADD sp, #offset
MOV lr, r3
MOV r3, ip
BX lr
http://reviews.llvm.org/D4748
llvm-svn: 214881
Matt Arsenault