The code that builds the dependence graph assumes that two PseudoSourceValues
don't alias. In a tail calling function two FixedStackObjects might refer to the
same location. Worse 'immutable' fixed stack objects like function arguments are
not immutable and will be clobbered.
Change this so that a load from a FixedStackObject is not invariant in a tail
calling function and don't return a PseudoSourceValue for an instruction in tail
calling functions when building the dependence graph so that we handle function
arguments conservatively.
Fix for PR23459.
rdar://20740035
llvm-svn: 236916
The test here was sinking the AND here to a lower BB:
%vreg7<def> = ANDWri %vreg8, 0; GPR32common:%vreg7,%vreg8
TBNZW %vreg8<kill>, 0, <BB#1>; GPR32common:%vreg8
which meant that vreg8 was read after it was killed.
This commit changes the code from clearing kill flags on the AND to clearing flags on all registers used by the AND.
llvm-svn: 236886
We were accidentally folding a sign/zero extend in to address arithmetic in a different BB when the extend wasn't available there.
Cross BB fast-isel isn't safe, so restrict this to only when the extend is in the same BB as the use.
llvm-svn: 236764
It's quite possible to encounter an insertvalue instruction that's more deeply
nested than the value we're looking for, but when that happens we really
mustn't compare beyond the end of the index array.
Since I couldn't see any guarantees about what comparisons std::equal makes, we
probably need to directly check the size beforehand. In practice, I suspect
most std::equal implementations would probably bail early, which would be OK.
But just in case...
rdar://20834485
llvm-svn: 236635
This patch introduces a new pass that computes the safe point to insert the
prologue and epilogue of the function.
The interest is to find safe points that are cheaper than the entry and exits
blocks.
As an example and to avoid regressions to be introduce, this patch also
implements the required bits to enable the shrink-wrapping pass for AArch64.
** Context **
Currently we insert the prologue and epilogue of the method/function in the
entry and exits blocks. Although this is correct, we can do a better job when
those are not immediately required and insert them at less frequently executed
places.
The job of the shrink-wrapping pass is to identify such places.
** Motivating example **
Let us consider the following function that perform a call only in one branch of
a if:
define i32 @f(i32 %a, i32 %b) {
%tmp = alloca i32, align 4
%tmp2 = icmp slt i32 %a, %b
br i1 %tmp2, label %true, label %false
true:
store i32 %a, i32* %tmp, align 4
%tmp4 = call i32 @doSomething(i32 0, i32* %tmp)
br label %false
false:
%tmp.0 = phi i32 [ %tmp4, %true ], [ %a, %0 ]
ret i32 %tmp.0
}
On AArch64 this code generates (removing the cfi directives to ease
readabilities):
_f: ; @f
; BB#0:
stp x29, x30, [sp, #-16]!
mov x29, sp
sub sp, sp, #16 ; =16
cmp w0, w1
b.ge LBB0_2
; BB#1: ; %true
stur w0, [x29, #-4]
sub x1, x29, #4 ; =4
mov w0, wzr
bl _doSomething
LBB0_2: ; %false
mov sp, x29
ldp x29, x30, [sp], #16
ret
With shrink-wrapping we could generate:
_f: ; @f
; BB#0:
cmp w0, w1
b.ge LBB0_2
; BB#1: ; %true
stp x29, x30, [sp, #-16]!
mov x29, sp
sub sp, sp, #16 ; =16
stur w0, [x29, #-4]
sub x1, x29, #4 ; =4
mov w0, wzr
bl _doSomething
add sp, x29, #16 ; =16
ldp x29, x30, [sp], #16
LBB0_2: ; %false
ret
Therefore, we would pay the overhead of setting up/destroying the frame only if
we actually do the call.
** Proposed Solution **
This patch introduces a new machine pass that perform the shrink-wrapping
analysis (See the comments at the beginning of ShrinkWrap.cpp for more details).
It then stores the safe save and restore point into the MachineFrameInfo
attached to the MachineFunction.
This information is then used by the PrologEpilogInserter (PEI) to place the
related code at the right place. This pass runs right before the PEI.
Unlike the original paper of Chow from PLDI’88, this implementation of
shrink-wrapping does not use expensive data-flow analysis and does not need hack
to properly avoid frequently executed point. Instead, it relies on dominance and
loop properties.
The pass is off by default and each target can opt-in by setting the
EnableShrinkWrap boolean to true in their derived class of TargetPassConfig.
This setting can also be overwritten on the command line by using
-enable-shrink-wrap.
Before you try out the pass for your target, make sure you properly fix your
emitProlog/emitEpilog/adjustForXXX method to cope with basic blocks that are not
necessarily the entry block.
** Design Decisions **
1. ShrinkWrap is its own pass right now. It could frankly be merged into PEI but
for debugging and clarity I thought it was best to have its own file.
2. Right now, we only support one save point and one restore point. At some
point we can expand this to several save point and restore point, the impacted
component would then be:
- The pass itself: New algorithm needed.
- MachineFrameInfo: Hold a list or set of Save/Restore point instead of one
pointer.
- PEI: Should loop over the save point and restore point.
Anyhow, at least for this first iteration, I do not believe this is interesting
to support the complex cases. We should revisit that when we motivating
examples.
Differential Revision: http://reviews.llvm.org/D9210
<rdar://problem/3201744>
llvm-svn: 236507
When deciding whether a value comes from the aggregate or inserted value of an
insertvalue instruction, we compare the indices against those of the location
we're interested in. One of the lists needs reversing because the input data is
backwards (so that modifications take place at the end of the SmallVector), but
we were reversing both before leading to incorrect results.
Should fix PR23408
llvm-svn: 236457
Finish off PR23080 by renaming the debug info IR constructs from `MD*`
to `DI*`. The last of the `DIDescriptor` classes were deleted in
r235356, and the last of the related typedefs removed in r235413, so
this has all baked for about a week.
Note: If you have out-of-tree code (like a frontend), I recommend that
you get everything compiling and tests passing with the *previous*
commit before updating to this one. It'll be easier to keep track of
what code is using the `DIDescriptor` hierarchy and what you've already
updated, and I think you're extremely unlikely to insert bugs. YMMV of
course.
Back to *this* commit: I did this using the rename-md-di-nodes.sh
upgrade script I've attached to PR23080 (both code and testcases) and
filtered through clang-format-diff.py. I edited the tests for
test/Assembler/invalid-generic-debug-node-*.ll by hand since the columns
were off-by-three. It should work on your out-of-tree testcases (and
code, if you've followed the advice in the previous paragraph).
Some of the tests are in badly named files now (e.g.,
test/Assembler/invalid-mdcompositetype-missing-tag.ll should be
'dicompositetype'); I'll come back and move the files in a follow-up
commit.
llvm-svn: 236120
After legalization, scalar SETCC has an i32 result type on AArch64.
The i1 requirement seems too conservative, replace it with an assert.
This also means that we now can run after legalization. That should also
be fine, since the ops legalizer runs again after each combine, and
all types created all have the same sizes as the (legal) inputs.
Exposed by r235917; while there, robustize its tests (bsl also uses the
register it defines).
llvm-svn: 235922
When the setcc has f64 operands, we can't build a vector setcc mask
to feed a vselect, because f64 doesn't divide v3f32 evenly.
Just bail out when that happens.
llvm-svn: 235917
Summary:
Constant stores of f16 vectors can create NvCast nodes from various
operand types to v4f16 or v8f16 depending on patterns in the stored
constants. This patch adds nvcast rules with v4f16 and v8f16 values.
AArchISelLowering::LowerBUILD_VECTOR has the details on which constant
patterns generate the nvcast nodes.
Reviewers: jmolloy, srhines, ab
Subscribers: rengolin, aemerson, llvm-commits
Differential Revision: http://reviews.llvm.org/D9201
llvm-svn: 235610
Summary:
Set operation action for SINT_TO_FP and UINT_TO_FP nodes with v4i32,
v8i8, v8i16 inputs to allow promotion of v4f16 results.
Add tests for sitofp and uitofp for vec4, vec8, vec16, and i8, i16, i32,
and i64 vectors. Only missing tests are for v16i8 and v16i16 as the
shift operations are too complicated to write a proper check sequence.
The conversions from v4i64 to v4f16 do not depend on this patch - v4i64
is split and the conversion gets handled while lowering v2i64. I am
adding a test here for completeness.
Reviewers: aemerson, rengolin, ab, jmolloy, srhines
Subscribers: rengolin, aemerson, llvm-commits
Differential Revision: http://reviews.llvm.org/D9166
llvm-svn: 235609
Enough concerns were raised that this optimization is pessimising some code patterns.
The obvious fix, to add a Reassociate run afterwards, causes even more pessimisation in some cases due to fewer complex addressing modes being matched. As there isn't a trivial fix for this, backing this out by default until someone gets a chance to fix the addressing mode matcher.
llvm-svn: 235491
Instead of merging everything together, look at the users of
GlobalVariables, and try to group them by function, to create
sets of globals used "together".
Using that information, a less-aggressive alternative is to keep merging
everything together *except* globals that are only ever used alone, that
is, those for which it's clearly non-profitable to merge with others.
In my testing, grouping by Function is too aggressive, but grouping by
BasicBlock is too conservative. Anything in-between isn't trivially
available, so stick with Function grouping for now.
cl::opts are added for testing; both enabled by default.
A few of the testcases aren't testing the merging proper, but just
various edge cases when merging does occur. Update them to use the
previous grouping behavior. Also, one of the tests is unrelated to
GlobalMerge; change it accordingly.
While there, switch to r234666' flags rather than the brutal -O3.
Differential Revision: http://reviews.llvm.org/D8070
llvm-svn: 235249
The result is either an Untyped reg sequence, on ldN with N > 1, or
just the type of the input vector, on ld1. Don't force Untyped.
Instead, just use the type of the reg sequence.
This mirrors the behavior of createTuple, which feeds the LD1*_POST.
The narrow code path wasn't actually covered by tests, because V64
insert_vector_elt are widened to V128 before the LD1LANEpost combine
has the chance to run, usually.
The only case where it does run on V64 vectors is if the vector ops
legalizer ran. So, tickle the code with a ctpop.
Fixes PR23265.
llvm-svn: 235243
This is a followon to r233681 - I'd misunderstood the semantics of FTRUNC,
and had confused it with (FP_ROUND ..., 0).
Thanks for Ahmed Bougacha for his post-commit review!
llvm-svn: 235191
Found by code inspection, but breaking i16 at least breaks other tests.
They aren't checking this in particular though, so also add some
explicit tests for the already working types.
llvm-svn: 235148
See r230786 and r230794 for similar changes to gep and load
respectively.
Call is a bit different because it often doesn't have a single explicit
type - usually the type is deduced from the arguments, and just the
return type is explicit. In those cases there's no need to change the
IR.
When that's not the case, the IR usually contains the pointer type of
the first operand - but since typed pointers are going away, that
representation is insufficient so I'm just stripping the "pointerness"
of the explicit type away.
This does make the IR a bit weird - it /sort of/ reads like the type of
the first operand: "call void () %x(" but %x is actually of type "void
()*" and will eventually be just of type "ptr". But this seems not too
bad and I don't think it would benefit from repeating the type
("void (), void () * %x(" and then eventually "void (), ptr %x(") as has
been done with gep and load.
This also has a side benefit: since the explicit type is no longer a
pointer, there's no ambiguity between an explicit type and a function
that returns a function pointer. Previously this case needed an explicit
type (eg: a function returning a void() function was written as
"call void () () * @x(" rather than "call void () * @x(" because of the
ambiguity between a function returning a pointer to a void() function
and a function returning void).
No ambiguity means even function pointer return types can just be
written alone, without writing the whole function's type.
This leaves /only/ the varargs case where the explicit type is required.
Given the special type syntax in call instructions, the regex-fu used
for migration was a bit more involved in its own unique way (as every
one of these is) so here it is. Use it in conjunction with the apply.sh
script and associated find/xargs commands I've provided in rr230786 to
migrate your out of tree tests. Do let me know if any of this doesn't
cover your cases & we can iterate on a more general script/regexes to
help others with out of tree tests.
About 9 test cases couldn't be automatically migrated - half of those
were functions returning function pointers, where I just had to manually
delete the function argument types now that we didn't need an explicit
function type there. The other half were typedefs of function types used
in calls - just had to manually drop the * from those.
import fileinput
import sys
import re
pat = re.compile(r'((?:=|:|^|\s)call\s(?:[^@]*?))(\s*$|\s*(?:(?:\[\[[a-zA-Z0-9_]+\]\]|[@%](?:(")?[\\\?@a-zA-Z0-9_.]*?(?(3)"|)|{{.*}}))(?:\(|$)|undef|inttoptr|bitcast|null|asm).*$)')
addrspace_end = re.compile(r"addrspace\(\d+\)\s*\*$")
func_end = re.compile("(?:void.*|\)\s*)\*$")
def conv(match, line):
if not match or re.search(addrspace_end, match.group(1)) or not re.search(func_end, match.group(1)):
return line
return line[:match.start()] + match.group(1)[:match.group(1).rfind('*')].rstrip() + match.group(2) + line[match.end():]
for line in sys.stdin:
sys.stdout.write(conv(re.search(pat, line), line))
llvm-svn: 235145
A big-endian vector return needs a byte-swap which we aren't doing right now.
For now just bail on these cases to get correctness back.
llvm-svn: 235133
Fix for test case found by James Molloy - TRUNCATE of constant build vectors can be more simply achieved by simply replacing with a new build vector node with the truncated value type - no need to touch the scalar operands at all.
llvm-svn: 235079
The only type that isn't an integer, isn't floating point, and isn't
a vector; ladies and gentlemen, the gift that keeps on giving: x86_mmx!
Fixes PR23246.
Original message (reverted in r235062):
[CodeGen] Combine concat_vectors of scalars into build_vector.
Combine something like:
(v8i8 concat_vectors (v2i8 bitcast (i16)) x4)
into:
(v8i8 (bitcast (v4i16 BUILD_VECTOR (i16) x4)))
If any of the scalars are floating point, use that throughout.
Differential Revision: http://reviews.llvm.org/D8948
llvm-svn: 235072
This commit makes LLVM not estimate branch probabilities when doing a
single bit bitmask tests.
The code that originally made me discover this is:
if ((a & 0x1) == 0x1) {
..
}
In this case we don't actually have any branch probability information
and should not assume to have any. LLVM transforms this into:
%and = and i32 %a, 1
%tobool = icmp eq i32 %and, 0
So, in this case, the result of a bitwise and is compared against 0,
but nevertheless, we should not assume to have probability
information.
CodeGen/ARM/2013-10-11-select-stalls.ll started failing because the
changed probabilities changed the results of
ARMBaseInstrInfo::isProfitableToIfCvt() and led to an Ifcvt of the
diamond in the test. AFAICT, the test was never meant to test this and
thus changing the test input slightly to not change the probabilities
seems like the best way to preserve the meaning of the test.
llvm-svn: 234979
if ((a & 0x1) == 0x1) {
..
}
In this case we don't actually have any branch probability information and
should not assume to have any. LLVM transforms this into:
%and = and i32 %a, 1
%tobool = icmp eq i32 %and, 0
So, in this case, the result of a bitwise and is compared against 0,
but nevertheless, we should not assume to have probability
information.
llvm-svn: 234898
Combine something like:
(v8i8 concat_vectors (v2i8 bitcast (i16)) x4)
into:
(v8i8 (bitcast (v4i16 BUILD_VECTOR (i16) x4)))
If any of the scalars are floating point, use that throughout.
Differential Revision: http://reviews.llvm.org/D8948
llvm-svn: 234809
Currently, there's a single flag, checked by the pass itself.
It can't force-enable the pass (and is on by default), because it
might not even have been created, as that's the targets decision.
Instead, have separate explicit flags, so that the decision is
consistently made in the target.
Keep the flag as a last-resort "force-disable GlobalMerge" for now,
for backwards compatibility.
llvm-svn: 234666
For the most common ones (such as fadd), we already did the promotion.
Do the same thing for all the others.
Currently, we'll just crash/assert on all these operations, as
there's no hardware or libcall support whatsoever.
f16 (half) is specified as an interchange - not arithmetic - format,
and is expected to be promoted to single-precision for arithmetic
operations.
While there, teach the legalizer about promoting some of the (mostly
floating-point) operations that we never needed before.
Differential Revision: http://reviews.llvm.org/D8648
See related discussion on the thread for: http://reviews.llvm.org/D8755
llvm-svn: 234550
We already do:
concat_vectors(scalar, undef) -> scalar_to_vector(scalar)
When the scalar is legal.
When it's not, but is a truncated legal scalar, we can also do:
concat_vectors(trunc(scalar), undef) -> scalar_to_vector(scalar)
Which is equivalent, since the upper lanes are undef anyway.
While there, teach the combine to look at more than 2 operands.
Differential Revision: http://reviews.llvm.org/D8883
llvm-svn: 234530
The integer extend optimization tries to fold the extend into the load
instruction. This requires us to identify if the extend has already been
emitted or not and act accordingly on it.
The check that was originally performed for this was not sufficient. Besides
checking the ValueMap for a mapped register we also need to check if the
virtual register has already an associated machine instruction that defines it.
This fixes rdar://problem/20470788.
llvm-svn: 234529
restrictions when choosing a type for small-memcpy inlining in
SelectionDAGBuilder.
This ensures that the loads and stores output for the memcpy won't be further
expanded during legalization, which would cause the total number of instructions
for the memcpy to exceed (often significantly) the inlining thresholds.
<rdar://problem/17829180>
llvm-svn: 234462
The bug manifests when there are two loads and two stores chained as follows in
a DAG,
(ld v3f32) -> (st f32) -> (ld v3f32) -> (st f32)
and the stores' values are extracted from the preceding vector loads.
MergeConsecutiveStores would replace the first store in the chain with the
merged vector store, which would create a cycle between the merged store node
and the last load node that appears in the chain.
This commits fixes the bug by replacing the last store in the chain instead.
rdar://problem/20275084
Differential Revision: http://reviews.llvm.org/D8849
llvm-svn: 234430
This patch attempts to fold the shuffling of 'scalar source' inputs - BUILD_VECTOR and SCALAR_TO_VECTOR nodes - if the shuffle node is the only user. This folds away a lot of unnecessary shuffle nodes, and allows quite a bit of constant folding that was being missed.
Differential Revision: http://reviews.llvm.org/D8516
llvm-svn: 234004
extended loads.
Implement the related target lowering hook so that the optimization has a better
estimation of the cost of an extension.
rdar://problem/19267165
llvm-svn: 233753