Summary:
In PNaCl, most atomic instructions have their own @llvm.nacl.atomic.* function, each one, with a few exceptions, represents a consistent behaviour across all NaCl-supported targets. Unfortunately, the atomic RMW operations nand, [u]min, and [u]max aren't directly represented by any such @llvm.nacl.atomic.* function. This patch refines shouldExpandAtomicRMWInIR in TargetLowering so that a future `Le32TargetLowering` class can selectively inform the caller how the target desires the atomic RMW instruction to be expanded (ie via load-linked/store-conditional for ARM/AArch64, via cmpxchg for X86/others?, or not at all for Mips) if at all.
This does not represent a behavioural change and as such no tests were added.
Patch by: Richard Diamond.
Reviewers: jfb
Reviewed By: jfb
Subscribers: jfb, aemerson, t.p.northover, llvm-commits
Differential Revision: http://reviews.llvm.org/D7713
llvm-svn: 231250
a flag for now.
First off, thanks to Daniel Jasper for really pointing out the issue
here. It's been here forever (at least, I think it was there when
I first wrote this code) without getting really noticed or fixed.
The key problem is what happens when two reasonably common patterns
happen at the same time: we outline multiple cold regions of code, and
those regions in turn have diamonds or other CFGs for which we can't
just topologically lay them out. Consider some C code that looks like:
if (a1()) { if (b1()) c1(); else d1(); f1(); }
if (a2()) { if (b2()) c2(); else d2(); f2(); }
done();
Now consider the case where a1() and a2() are unlikely to be true. In
that case, we might lay out the first part of the function like:
a1, a2, done;
And then we will be out of successors in which to build the chain. We go
to find the best block to continue the chain with, which is perfectly
reasonable here, and find "b1" let's say. Laying out successors gets us
to:
a1, a2, done; b1, c1;
At this point, we will refuse to lay out the successor to c1 (f1)
because there are still un-placed predecessors of f1 and we want to try
to preserve the CFG structure. So we go get the next best block, d1.
... wait for it ...
Except that the next best block *isn't* d1. It is b2! d1 is waaay down
inside these conditionals. It is much less important than b2. Except
that this is exactly what we didn't want. If we keep going we get the
entire set of the rest of the CFG *interleaved*!!!
a1, a2, done; b1, c1; b2, c2; d1, f1; d2, f2;
So we clearly need a better strategy here. =] My current favorite
strategy is to actually try to place the block whose predecessor is
closest. This very simply ensures that we unwind these kinds of CFGs the
way that is natural and fitting, and should minimize the number of cache
lines instructions are spread across.
It also happens to be *dead simple*. It's like the datastructure was
specifically set up for this use case or something. We only push blocks
onto the work list when the last predecessor for them is placed into the
chain. So the back of the worklist *is* the nearest next block.
Unfortunately, a change like this is going to cause *soooo* many
benchmarks to swing wildly. So for now I'm adding this under a flag so
that we and others can validate that this is fixing the problems
described, that it seems possible to enable, and hopefully that it fixes
more of our problems long term.
llvm-svn: 231238
In a CFG with the edges A->B->C and A->C, B is an optional branch.
LLVM's default behavior is to lay the blocks out naturally, i.e. A, B,
C, in order to improve code locality and fallthroughs. However, if a
function contains many of those optional branches only a few of which
are taken, this leads to a lot of unnecessary icache misses. Moving B
out of line can work around this.
Review: http://reviews.llvm.org/D7719
llvm-svn: 231230
When trying to convert a BUILD_VECTOR into a shuffle, we try to split a single source vector that is twice as wide as the destination vector.
We can not do this when we also need the zero vector to create a blend.
This fixes PR22774.
Differential Revision: http://reviews.llvm.org/D8040
llvm-svn: 231219
(They are called emitDwarfDIE and emitDwarfAbbrevs in their new home)
llvm-dsymutil wants to reuse that code, but it doesn't have a DwarfUnit or
a DwarfDebug object to call those. It has access to an AsmPrinter though.
Having emitDIE in the AsmPrinter also removes the DwarfFile dependency
on DwarfDebug, and thus the patch drops that field.
Differential Revision: http://reviews.llvm.org/D8024
llvm-svn: 231210
GCC 4.7's libstdc++ doesn't have std::map::emplace, but it does have
std::unordered_map::emplace, and the use case here doesn't appear to
need ordering. The container has been changed in a separate/precursor
patch, and now this patch should hopefully build cleanly even with
GCC 4.7.
& then I realized the order of the container did matter, so extra
handling of ordering was added in r231189.
Original commit message:
This makes LiveRange non-copyable, and LiveInterval is already
non-movable (due to the explicit dtor), so now it's non-copyable and
non-movable.
Fix the one case where we were relying on the (deprecated in C++11)
implicit copy ctor of LiveInterval (which happened to work because the
ctor created an object with a null segmentSet, so double-deleting the
null pointer was fine).
llvm-svn: 231192
The order of this container was needed at one point - so, at that point
create a temporary array of pointers, sort those, then iterate them.
This keeps lookup efficient (& the lesser issue, of allowing the use of
emplace... ), object identity preserved, and ordered iteration in the
one place that requires it.
While this has no functional change, I realize it does mean allocating
an extra data structure and performing a sort - so if this looks suspect
to anyone regarding perf characteristics, I'm all ears.
llvm-svn: 231189
There is a known bug where the register coalescer fails to merge
subranges when multiple ranges end up in the "overflow" bit 32 of the
lanemasks. A proper fix for this is complicated so for now this is a
workaround which lets the register coalescer drop the subregister
liveness information (we just loose some precision by that) and
continue.
llvm-svn: 231186
Apparently something does care about ordering of LiveIntervals... so
revert all that stuff (r231175, r231176, r231177) & take some time to
re-evaluate.
llvm-svn: 231184
GCC 4.7's libstdc++ doesn't have std::map::emplace, but it does have
std::unordered_map::emplace, and the use case here doesn't appear to
need ordering. The container has been changed in a separate/precursor
patch, and now this patch should hopefully build cleanly even with
GCC 4.7.
Original commit message:
This makes LiveRange non-copyable, and LiveInterval is already
non-movable (due to the explicit dtor), so now it's non-copyable and
non-movable.
Fix the one case where we were relying on the (deprecated in C++11)
implicit copy ctor of LiveInterval (which happened to work because the
ctor created an object with a null segmentSet, so double-deleting the
null pointer was fine).
llvm-svn: 231176
This makes LiveRange non-copyable, and LiveInterval is already
non-movable (due to the explicit dtor), so now it's non-copyable and
non-movable.
Fix the one case where we were relying on the (deprecated in C++11)
implicit copy ctor of LiveInterval (which happened to work because the
ctor created an object with a null segmentSet, so double-deleting the
null pointer was fine).
llvm-svn: 231168
Ultimately, we'll need to leave something behind to indicate which
alloca will hold the exception, but we can figure that out when it comes
time to emit the __CxxFrameHandler3 catch handler table.
llvm-svn: 231164
From:
int M, total;
void foo() {
int i;
for (i = 0; i < M; i++) {
total = total + i / 2;
}
}
This is the kernel loop:
.LBB0_2: # %for.body
=>This Inner Loop Header: Depth=1
movl %edx, %esi
movl %ecx, %edx
shrl $31, %edx
addl %ecx, %edx
sarl %edx
addl %esi, %edx
incl %ecx
cmpl %eax, %ecx
jl .LBB0_2
--------------------------
The first mov insn "movl %edx, %esi" could be removed if we change "addl %esi, %edx" to "addl %edx, %esi".
The IR before TwoAddressInstructionPass is:
BB#2: derived from LLVM BB %for.body
Predecessors according to CFG: BB#1 BB#2
%vreg3<def> = COPY %vreg12<kill>; GR32:%vreg3,%vreg12
%vreg2<def> = COPY %vreg11<kill>; GR32:%vreg2,%vreg11
%vreg7<def,tied1> = SHR32ri %vreg3<tied0>, 31, %EFLAGS<imp-def,dead>; GR32:%vreg7,%vreg3
%vreg8<def,tied1> = ADD32rr %vreg3<tied0>, %vreg7<kill>, %EFLAGS<imp-def,dead>; GR32:%vreg8,%vreg3,%vreg7
%vreg9<def,tied1> = SAR32r1 %vreg8<kill,tied0>, %EFLAGS<imp-def,dead>; GR32:%vreg9,%vreg8
%vreg4<def,tied1> = ADD32rr %vreg9<kill,tied0>, %vreg2<kill>, %EFLAGS<imp-def,dead>; GR32:%vreg4,%vreg9,%vreg2
%vreg5<def,tied1> = INC64_32r %vreg3<kill,tied0>, %EFLAGS<imp-def,dead>; GR32:%vreg5,%vreg3
CMP32rr %vreg5, %vreg0, %EFLAGS<imp-def>; GR32:%vreg5,%vreg0
%vreg11<def> = COPY %vreg4; GR32:%vreg11,%vreg4
%vreg12<def> = COPY %vreg5<kill>; GR32:%vreg12,%vreg5
JL_4 <BB#2>, %EFLAGS<imp-use,kill>
Now TwoAddressInstructionPass will choose vreg9 to be tied with vreg4. However, it doesn't see that there is copy from vreg4 to vreg11 and another copy from vreg11 to vreg2 inside the loop body. To remove those copies, it is necessary to choose vreg2 to be tied with vreg4 instead of vreg9. This code pattern commonly appears when there is reduction operation in a loop.
So check for a reversed copy chain and if we encounter one then we can commute the add instruction so we can avoid a copy.
Patch by Wei Mi.
http://reviews.llvm.org/D7806
llvm-svn: 231148
Accidentally committed a few more of these cleanup changes than
intended. Still breaking these out & tidying them up.
This reverts commit r231135.
llvm-svn: 231136
There doesn't seem to be any need to assert that iterator assignment is
between iterators over the same node - if you want to reuse an iterator
variable to iterate another node, that's perfectly acceptable. Just
don't mix comparisons between iterators into disjoint sequences, as
usual.
llvm-svn: 231135
This type could be made copyable (= default a protected copy ctor in the
base class, and preferably make the derived class final to avoid risks
of providing a slicing copy operation to further derived classes) but it
seemed easier to avoid that complexity for a dump function that I assume
(by symmetry with ResourcePriorityQueue's dump, which was actively
buggy) not often used.
llvm-svn: 231133
frame register before checking if there is a DWARF register number for it.
Thanks to H.J. Lu for diagnosing this and providing the testcase!
llvm-svn: 231121
The cause of the issue is the interaction of two factors:
1) When generating a DW_TAG_imported_declaration DIE which imports another
imported declaration, the code in AsmPrinter/DwarfCompileUnit.cpp
asserts that the second imported declaration must already have a DIE.
2) There is a non-determinism in the order in which imported declarations
within the same scope are processed.
Because of the non-determinism (2), it is possible that an imported
declaration is processed before another one it depends on, breaking the
assumption in (1).
The source of the non-determinism is that the imported declaration
DIDescriptors are sorted by scope in DwarfDebug::beginModule(); however that
sort is not a stable_sort, therefore the order of the declarations within
the same scope is not preserved. The attached patch changes the std::sort to
a std::stable_sort and it fixes the problem.
Test omitted due to it being non-deterministic and depending on the
implementation of std::sort.
llvm-svn: 231100
This prevents the behavior observed in llvm.org/PR22369. I am not sure
whether I am reading the code correctly, but the early exit based on
isLiveOutPastPHIs() seems to make the wrong assumption that
RegisterCoalescer won't be able to coalesce those copies later.
This change hides the new behavior behind -no-phi-elim-live-out-early-exit
as it currently breaks four tests:
* Assertion in:
CodeGen/Hexagon/hwloop-cleanup.ll
* Worse code in:
CodeGen/X86/coalescer-commute4.ll
CodeGen/X86/phys_subreg_coalesce-2.ll
CodeGen/X86/zlib-longest-match.ll
The root cause here seems to be that the heuristic that determines
the visitation order in RegisterCoalescer gets less lucky.
llvm-svn: 231064
TargetRegisterInfo. DebugLocEntry now holds a buffer with the raw bytes
of the pre-calculated DWARF expression.
Ought to be NFC, but it does slightly alter the output format of the
textual assembly.
This reapplies 230930 without the assertion in DebugLocEntry::finalize()
because not all Machine registers can be lowered into DWARF register
numbers and floating point constants cannot be expressed.
llvm-svn: 231023
TargetRegisterInfo. DebugLocEntry now holds a buffer with the raw bytes
of the pre-calculated DWARF expression.
Ought to be NFC, but it does slightly alter the output format of the
textual assembly.
This reapplies 230930 with a relaxed assertion in DebugLocEntry::finalize()
that allows for empty DWARF expressions for constant FP values.
llvm-svn: 230975
TargetRegisterInfo. DebugLocEntry now holds a buffer with the raw bytes
of the pre-calculated DWARF expression.
Ought to be NFC, but it does slightly alter the output format of the
textual assembly.
llvm-svn: 230930
Such edges are zero matrix, and they bring no additional info to the
allocation problem, apart from contributing to nodes' degree. Removing
those edges is expected to improve allocation time.
Tune the spill cost comparison, as this gives better average performances
now that the nodes' degrees has changed.
llvm-svn: 230904
We were missing a check for the following fold in DAGCombiner:
// fold (fmul (fmul x, c1), c2) -> (fmul x, (fmul c1, c2))
If 'x' is also a constant, then we shouldn't do anything. Otherwise, we could end up swapping the operands back and forth forever.
This should fix:
http://llvm.org/bugs/show_bug.cgi?id=22698
Differential Revision: http://reviews.llvm.org/D7917
llvm-svn: 230884
