This requires a knowledge of the stack size which is not known until
the frame is complete, hence the need for the XCoreFTAOElim pass
which lowers the XCoreISD::FRAME_TO_ARGS_OFFSET instrution into its
final form.
llvm-svn: 198614
Longer term, we want to move users to "*-*-*-macho" for embedded work, but for
now people are relying on the last thing we told them, which is unfortunately
"*-*-darwin-eabi".
rdar://problem/15703934
llvm-svn: 198602
The 0x66 prefix toggles between 16-bit and 32-bit addressing mode.
So in 32-bit mode it is used to switch to 16-bit addressing mode for the
following instruction, while in 16-bit mode it's the other way round — it's
used to switch to 32-bit mode instead.
Thus, emit the 0x66 prefix byte for OpSize only in 32-bit (and 64-bit) mode,
and introduce a new OpSize16 bit which is used in 16-bit mode instead.
This is just the basic infrastructure for that change; a subsequent patch
will add the new OpSize16 bit to the 32-bit instructions that need it.
Patch from David Woodhouse.
llvm-svn: 198586
This is not really expected to work right yet. Mostly because we will
still emit the OpSize (0x66) prefix in all the wrong places, along with
a number of other corner cases. Those will all be fixed in the subsequent
commits.
Patch from David Woodhouse.
llvm-svn: 198584
There is a wrong assumption that the vector element type and the
type of each ConstantSDNode in the build_vector were the same.
However, when promoting the integer operand of a legally typed
build_vector, the operand type and the vector element type do not
need to be the same
(See method 'DAGTypeLegalizer::PromoteIntOp_BUILD_VECTOR' in
LegalizeIntegerTypes.cpp).
in AArch64 backend, the following dag sequence:
C0: i1 = Constant<0>
C1: i1 = Constant<-1>
V: v8i1 = BUILD_VECTOR C1, C1, C0, C0, C0, C0, C0, C0
is type-legalized into:
NewC0: i32 = Constant<0>
NewC1: i32 = Constant<1>
V: v8i8 = BUILD_VECTOR NewC1, NewC1, NewC0, NewC0, NewC0, NewC0, NewC0, NewC0
Forcing a getZeroExtend to VTBits to ensure that the new constant
is correctly.
llvm-svn: 198582
Add some tests to validate correct register selection, including a fix
to an existing test which was requiring the *wrong* output.
Patch from David Woodhouse.
llvm-svn: 198566
__builtin_returnaddress requires that the value passed into is be a constant.
However, at -O0 even a constant expression may not be converted to a constant.
Emit an error message intead of crashing.
llvm-svn: 198531
This commit was the source of crasher PR18384:
While deleting: label %for.cond127
An asserting value handle still pointed to this value!
UNREACHABLE executed at llvm/lib/IR/Value.cpp:671!
Reverting to get the builders green, feel free to re-land after fixing up.
(Renato has a handy isolated repro if you need it.)
This reverts commit r198478.
llvm-svn: 198503
getSCEV for an ashr instruction creates an intermediate zext
expression when it truncates its operand.
The operand is initially inside the loop, so the narrow zext
expression has a non-loop-invariant loop disposition.
LoopSimplify then runs on an outer loop, hoists the ashr operand, and
properly invalidate the SCEVs that are mapped to value.
The SCEV expression for the ashr is now an AddRec with the hoisted
value as the now loop-invariant start value.
The LoopDisposition of this wide value was properly invalidated during
LoopSimplify.
However, if we later get the ashr SCEV again, we again try to create
the intermediate zext expression. We get the same SCEV that we did
earlier, and it is still cached because it was never mapped to a
Value. When we try to create a new AddRec we abort because we're using
the old non-loop-invariant LoopDisposition.
I don't have a solution for this other than to clear LoopDisposition
when LoopSimplify hoists things.
I think the long-term strategy should be to perform LoopSimplify on
all loops before computing SCEV and before running any loop opts on
individual loops. It's possible we may want to rerun LoopSimplify on
individual loops, but it should rarely do anything, so rarely require
invalidating SCEV.
llvm-svn: 198478
This reverts commit r198398, thus reapplying r198397.
I had accidentally introduced an endianness issue when applying the hash
to the type unit. Using support::ulittle64_t in the reinterpret_cast in
addDwarfTypeUnitType fixes this issue.
Original commit message:
Debug Info: Type Units: Simplify type hashing using IR-provided unique
names.
What's good for LTO metadata size problems ought to be good for non-LTO
debug info size too, so let's rely on the same uniqueness in both cases.
If it's insufficient for non-LTO for whatever reason (since we now won't
be uniquing CU-local types or any C types - but these are likely to not
be the most significant contributors to type bloat) we should consider a
frontend solution that'll help both LTO and non-LTO alike, rather than
using DWARF-level DIE-hashing that only helps non-LTO debug info size.
It's also much simpler this way and benefits C++ even more since we can
deduplicate lexically separate definitions of the same C++ type since
they have the same mangled name.
llvm-svn: 198436
The loop rerolling pass was failing with an assertion failure from a
failed cast on loops like this:
void foo(int *A, int *B, int m, int n) {
for (int i = m; i < n; i+=4) {
A[i+0] = B[i+0] * 4;
A[i+1] = B[i+1] * 4;
A[i+2] = B[i+2] * 4;
A[i+3] = B[i+3] * 4;
}
}
The code was casting the SCEV-expanded code for the new
induction variable to a phi-node. When the loop had a non-constant
lower bound, the SCEV expander would end the code expansion with an
add insted of a phi node and the cast would fail.
It looks like the cast to a phi node was only needed to get the
induction variable value coming from the backedge to compute the end
of loop condition. This patch changes the loop reroller to compare
the induction variable to the number of times the backedge is taken
instead of the iteration count of the loop. In other words, we stop
the loop when the current value of the induction variable ==
IterationCount-1. Previously, the comparison was comparing the
induction variable value from the next iteration == IterationCount.
This problem only seems to occur on 32-bit targets. For some reason,
the loop is not rerolled on 64-bit targets.
PR18290
llvm-svn: 198425
cycles
This allows the value equality check to work even if we don't have a dominator
tree. Also add some more comments.
I was worried about compile time impacts and did not implement reachability but
used the dominance check in the initial patch. The trade-off was that the
dominator tree was required.
The llvm utility function isPotentiallyReachable cuts off the recursive search
after 32 visits. Testing did not show any compile time regressions showing my
worries unjustfied.
No compile time or performance regressions at O3 -flto -mavx on test-suite +
externals.
Addresses review comments from r198290.
llvm-svn: 198400
What's good for LTO metadata size problems ought to be good for non-LTO
debug info size too, so let's rely on the same uniqueness in both cases.
If it's insufficient for non-LTO for whatever reason (since we now won't
be uniquing CU-local types or any C types - but these are likely to not
be the most significant contributors to type bloat) we should consider a
frontend solution that'll help both LTO and non-LTO alike, rather than
using DWARF-level DIE-hashing that only helps non-LTO debug info size.
It's also much simpler this way and benefits C++ even more since we can
deduplicate lexically separate definitions of the same C++ type since
they have the same mangled name.
llvm-svn: 198397
The cgo problem was that it wants dwarf2 which doesn't support direct
constant encoding of the location. So let's add support for dwarf2
encoding (using a location expression) of data member locations.
This reverts commit r198385.
llvm-svn: 198389
The greedy register allocator tries to split a live-range around each
instruction where it is used or defined to relax the constraints on the entire
live-range (this is a last chance split before falling back to spill).
The goal is to have a big live-range that is unconstrained (i.e., that can use
the largest legal register class) and several small local live-range that carry
the constraints implied by each instruction.
E.g.,
Let csti be the constraints on operation i.
V1=
op1 V1(cst1)
op2 V1(cst2)
V1 live-range is constrained on the intersection of cst1 and cst2.
tryInstructionSplit relaxes those constraints by aggressively splitting each
def/use point:
V1=
V2 = V1
V3 = V2
op1 V3(cst1)
V4 = V2
op2 V4(cst2)
Because of how the coalescer infrastructure works, each new variable (V3, V4)
that is alive at the same time as V1 (or its copy, here V2) interfere with V1.
Thus, we end up with an uncoalescable copy for each split point.
To make tryInstructionSplit less aggressive, we check if the split point
actually relaxes the constraints on the whole live-range. If it does not, we do
not insert it.
Indeed, it will not help the global allocation problem:
- V1 will have the same constraints.
- V1 will have the same interference + possibly the newly added split variable
VS.
- VS will produce an uncoalesceable copy if alive at the same time as V1.
<rdar://problem/15570057>
llvm-svn: 198369
I originally had these using opt -verify, and I never removed the
-verify when converting them to use llvm-as instead, so these were
failing because of using the -verify argument which llvm-as doesn't have
instead of what it's actually supposed to be testing.
llvm-svn: 198352
Even within a multiclass, we had been generating concrete implicit anonymous
defs when parsing values (generally in value lists). This behavior was
incorrect, and led to errors when multiclass parameters were used in the
parameter list of the implicit anonymous def.
If we had some multiclass:
multiclass mc<string n> {
... : SomeClass<SomeOtherClass<n> >
The capture of the multiclass parameter 'n' would not work correctly, and
depending on how the implicit SomeOtherClass was used, either TableGen would
ignore something it shouldn't, or would crash.
To fix this problem, when inside a multiclass, we generate prototype anonymous
defs for implicit anonymous defs (just as we do for explicit anonymous defs).
Within the multiclass, the current record prototype is populated with a node
that is essentially: !cast<SomeOtherClass>(!strconcat(NAME, anon_value_name)).
This is then resolved to the correct concrete anonymous def, in the usual way,
when NAME is resolved during multiclass instantiation.
llvm-svn: 198348
Plugins need to go in build/Debug/lib as well (rather than build/lib/Debug).
Also, fix the SHLIBDIR path for Xcode, which by default includes Xcode build
settings rather than a simple %(build_mode)s parameter.
llvm-svn: 198344
TableGen had been generating a different name for an anonymous multiclass's
NAME for every def in the multiclass. This had an unfortunate side effect: it
was impossible to reference one def within the multiclass from another (in the
parameter list, for example). By making sure we only generate an anonymous name
once per multiclass (which, as it turns out, requires only changing the name
parameter to reference type), we can now concatenate NAME within the multiclass
with a def name in order to generate a reference to that def.
This does not matter so much, in and of itself, but is necessary for a
follow-up commit that will fix variable capturing in implicit anonymous
multiclass defs (and that is important).
llvm-svn: 198340
When widening an IV to remove s/zext, we generally try to eliminate
the original narrow IV. However, LCSSA phi nodes outside the loop were
still using the original IV. Clean this up more aggressively to avoid
redundancy in generated code.
llvm-svn: 198338
When there are cycles in the value graph we have to be careful interpreting
"Value*" identity as "value" equivalence. We interpret the value of a phi node
as the value of its operands.
When we check for value equivalence now we make sure that the "Value*" dominates
all cycles (phis).
%0 = phi [%noaliasval, %addr2]
%l = load %ptr
%addr1 = gep @a, 0, %l
%addr2 = gep @a, 0, (%l + 1)
store %ptr ...
Before this patch we would return NoAlias for (%0, %addr1) which is wrong
because the value of the load is from different iterations of the loop.
Tested on x86_64 -mavx at O3 and O3 -flto with no performance or compile time
regressions.
PR18068
radar://15653794
llvm-svn: 198290
Robert Lytton