Allow LDRD to be formed from pairs with different LDR encodings. This was the original intention of the pass. Somewhere along the way, the LDR opcodes were refined which broke the optimization. We really don't care what the original opcodes are as long as they both map to the same LDRD and the immediate still fits.
Fixes rdar://10435045 ARMLoadStoreOptimization cannot handle mixed LDRi8/LDRi12
llvm-svn: 147922
Consider this code:
int h() {
int x;
try {
x = f();
g();
} catch (...) {
return x+1;
}
return x;
}
The variable x is undefined on the first edge to the landing pad, but it
has the f() return value on the second edge to the landing pad.
SplitAnalysis::getLastSplitPoint() would assume that the return value
from f() was live into the landing pad when f() throws, which is of
course impossible.
Detect these cases, and treat them as if the landing pad wasn't there.
This allows spill code to be inserted after the function call to f().
<rdar://problem/10664933>
llvm-svn: 147912
with other symbols.
An object in the __cfstring section is suppoed to be filled with CFString
objects, which have a pointer to ___CFConstantStringClassReference followed by a
pointer to a __cstring. If we allow the object in the __cstring section to be
merged with another global, then it could end up in any section. Because the
linker is going to remove these symbols in the final executable, we shouldn't
bother to merge them.
<rdar://problem/10564621>
llvm-svn: 147899
This function runs after all constant islands have been placed, and may
shrink some instructions to their 2-byte forms. This can actually cause
some constant pool entries to move out of range because of growing
alignment padding.
Treat instructions that may be shrunk the same as inline asm - they
erode the known alignment bits.
Also reinstate an old assertion in verify(). It is correct now that
basic block offsets include alignments.
Add a single large test case that will hopefully exercise many parts of
the constant island pass.
<rdar://problem/10670199>
llvm-svn: 147885
assembly source when it generates the TAG_subprogram dwarf debug info for
the labels that have nothing between them as in this bit of assembly source:
% cat ZeroLength.s
_func1:
_func2:
nop
One solution would be to not emit the subsequent labels with the same address
and use the next label with a different address or the end of the section for
the AT_high_pc value of the TAG_subprogram.
Turns out in llvm-mc it is not possible in all cases to determine of two
symbols have the same value at the point we put out the TAG_subprogram dwarf
debug info.
So we will have llvm-mc instead of putting out TAG_subprogram's put out
DW_TAG_label's. And the DW_TAG_label does not have a AT_high_pc value which
avoids the problem.
This commit is only the functional change to make the diffs clear as to what is
really being changed. The next commit will be to clean up the names of such
things like MCGenDwarfSubprogramEntry to something like MCGenDwarfLabelEntry.
rdar://10666925
llvm-svn: 147860
define physical registers. It's currently very restrictive, only catching
cases where the CE is in an immediate (and only) predecessor. But it catches
a surprising large number of cases.
rdar://10660865
llvm-svn: 147827
These heuristics are sufficient for enabling IV chains by
default. Performance analysis has been done for i386, x86_64, and
thumbv7. The optimization is rarely important, but can significantly
speed up certain cases by eliminating spill code within the
loop. Unrolled loops are prime candidates for IV chains. In many
cases, the final code could still be improved with more target
specific optimization following LSR. The goal of this feature is for
LSR to make the best choice of induction variables.
Instruction selection may not completely take advantage of this
feature yet. As a result, there could be cases of slight code size
increase.
Code size can be worse on x86 because it doesn't support postincrement
addressing. In fact, when chains are formed, you may see redundant
address plus stride addition in the addressing mode. GenerateIVChains
tries to compensate for the common cases.
On ARM, code size increase can be mitigated by using postincrement
addressing, but downstream codegen currently misses some opportunities.
llvm-svn: 147826
After collecting chains, check if any should be materialized. If so,
hide the chained IV users from the LSR solver. LSR will only solve for
the head of the chain. GenerateIVChains will then materialize the
chained IV users by computing the IV relative to its previous value in
the chain.
In theory, chained IV users could be exposed to LSR's solver. This
would be considerably complicated to implement and I'm not aware of a
case where we need it. In practice it's more important to
intelligently prune the search space of nontrivial loops before
running the solver, otherwise the solver is often forced to prune the
most optimal solutions. Hiding the chained users does this well, so
that LSR is more likely to find the best IV for the chain as a whole.
llvm-svn: 147801
LoopSimplify may not run on some outer loops, e.g. because of indirect
branches. SCEVExpander simply cannot handle outer loops with no preheaders.
Fixes rdar://10655343 SCEVExpander segfault.
llvm-svn: 147718
file error checking. Use that to error on an unfinished cfi_startproc.
The error is not nice, but is already better than a segmentation fault.
llvm-svn: 147717
opportunities that only present themselves after late optimizations
such as tail duplication .e.g.
## BB#1:
movl %eax, %ecx
movl %ecx, %eax
ret
The register allocator also leaves some of them around (due to false
dep between copies from phi-elimination, etc.)
This required some changes in codegen passes. Post-ra scheduler and the
pseudo-instruction expansion passes have been moved after branch folding
and tail merging. They were before branch folding before because it did
not always update block livein's. That's fixed now. The pass change makes
independently since we want to properly schedule instructions after
branch folding / tail duplication.
rdar://10428165
rdar://10640363
llvm-svn: 147716
This eliminates a lot of constant pool entries for -O0 builds of code
with many global variable accesses.
This speeds up -O0 codegen of consumer-typeset by 2x because the
constant island pass no longer has to look at thousands of constant pool
entries.
<rdar://problem/10629774>
llvm-svn: 147712
present in the bottom of the CFG triangle, as the transformation isn't
ever valuable if the branch can't be eliminated.
Also, unify some heuristics between SimplifyCFG's multiple
if-converters, for consistency.
This fixes rdar://10627242.
llvm-svn: 147630
code can incorrectly move the load across a store. This never
happens in practice today, but only because the current
heuristics accidentally preclude it.
llvm-svn: 147623
a combined-away node and the result of the combine isn't substantially
smaller than the input, it's just canonicalized. This is the first part
of a significant (7%) performance gain for Snappy's hot decompression
loop.
llvm-svn: 147604
Testing: passed 'make check' including LIT tests for all sequences being handled (both SSE and AVX)
Reviewers: Evan Cheng, David Blaikie, Bruno Lopes, Elena Demikhovsky, Chad Rosier, Anton Korobeynikov
llvm-svn: 147601
Now that canRealignStack() understands frozen reserved registers, it is
safe to use it for aligned spill instructions.
It will only return true if the registers reserved at the beginning of
register allocation allow for dynamic stack realignment.
<rdar://problem/10625436>
llvm-svn: 147579
(x > y) ? x : y
=>
(x >= y) ? x : y
So for something like
(x - y) > 0 : (x - y) ? 0
It will be
(x - y) >= 0 : (x - y) ? 0
This makes is possible to test sign-bit and eliminate a comparison against
zero. e.g.
subl %esi, %edi
testl %edi, %edi
movl $0, %eax
cmovgl %edi, %eax
=>
xorl %eax, %eax
subl %esi, $edi
cmovsl %eax, %edi
rdar://10633221
llvm-svn: 147512
This patch caused a miscompilation of oggenc because a frame pointer was
suddenly needed halfway through register allocation.
<rdar://problem/10625436>
llvm-svn: 147487
The failure seen on win32, when i64 type is illegal.
It happens on stage of conversion VECTOR_SHUFFLE to BUILD_VECTOR.
The failure message is:
llc: SelectionDAG.cpp:784: void VerifyNodeCommon(llvm::SDNode*): Assertion `(I->getValueType() == EltVT || (EltVT.isInteger() && I->getValueType().isInteger() && EltVT.bitsLE(I->getValueType()))) && "Wrong operand type!"' failed.
I added a special test that checks vector shuffle on win32.
llvm-svn: 147445
The failure seen on win32, when i64 type is illegal.
It happens on stage of conversion VECTOR_SHUFFLE to BUILD_VECTOR.
The failure message is:
llc: SelectionDAG.cpp:784: void VerifyNodeCommon(llvm::SDNode*): Assertion `(I->getValueType() == EltVT || (EltVT.isInteger() && I->getValueType().isInteger() && EltVT.bitsLE(I->getValueType()))) && "Wrong operand type!"' failed.
I added a special test that checks vector shuffle on win32.
llvm-svn: 147399
1. The ST*UX instructions that store and update the stack pointer did not set define/kill on R1. This became a problem when I activated post-RA scheduling (and had incorrectly adjusted the Frames-large test).
2. eliminateFrameIndex did not kill its scavenged temporary register, and this could cause the scavenger to exhaust all available registers (and its emergency spill slot) when there were a lot of CR values to spill. The 2010-02-12-saveCR test has been adjusted to check for this.
llvm-svn: 147359
captured. This allows the tracker to look at the specific use, which may be
especially interesting for function calls.
Use this to fix 'nocapture' deduction in FunctionAttrs. The existing one does
not iterate until a fixpoint and does not guarantee that it produces the same
result regardless of iteration order. The new implementation builds up a graph
of how arguments are passed from function to function, and uses a bottom-up walk
on the argument-SCCs to assign nocapture. This gets us nocapture more often, and
does so rather efficiently and independent of iteration order.
llvm-svn: 147327
This has the obvious advantage of being commutable and is always a win on x86 because
const - x wastes a register there. On less weird architectures this may lead to
a regression because other arithmetic doesn't fuse with it anymore. I'll address that
problem in a followup.
llvm-svn: 147254
LZCNT instructions are available. Force promotion to i32 to get
a smaller encoding since the fix-ups necessary are just as complex for
either promoted type
We can't do standard promotion for CTLZ when lowering through BSR
because it results in poor code surrounding the 'xor' at the end of this
instruction. Essentially, if we promote the entire CTLZ node to i32, we
end up doing the xor on a 32-bit CTLZ implementation, and then
subtracting appropriately to get back to an i8 value. Instead, our
custom logic just uses the knowledge of the incoming size to compute
a perfect xor. I'd love to know of a way to fix this, but so far I'm
drawing a blank. I suspect the legalizer could be more clever and/or it
could collude with the DAG combiner, but how... ;]
llvm-svn: 147251
my C-brain happy. Remove the unnecessary bits of pedantic IR fluff like
nounwind. Remove stray uses comments. Name things semantically rather
than tN so that adding a new test in the middle doesn't cause pain, and
so that new tests can be grouped semantically.
This exposes how little systematic testing is going on here. I noticed
this by finding several bugs via inspection and wondering why this test
wasn't catching any of them. =[
llvm-svn: 147248
'bsf' instructions here.
This one is actually debatable to my eyes. It's not clear that any chip
implementing 'tzcnt' would have a slow 'bsf' for any reason, and unless
EFLAGS or a zero input matters, 'tzcnt' is just a longer encoding.
Still, this restores the old behavior with 'tzcnt' enabled for now.
llvm-svn: 147246
X86ISelLowering C++ code. Because this is lowered via an xor wrapped
around a bsr, we want the dagcombine which runs after isel lowering to
have a chance to clean things up. In particular, it is very common to
see code which looks like:
(sizeof(x)*8 - 1) ^ __builtin_clz(x)
Which is trying to compute the most significant bit of 'x'. That's
actually the value computed directly by the 'bsr' instruction, but if we
match it too late, we'll get completely redundant xor instructions.
The more naive code for the above (subtracting rather than using an xor)
still isn't handled correctly due to the dagcombine getting confused.
Also, while here fix an issue spotted by inspection: we should have been
expanding the zero-undef variants to the normal variants when there is
an 'lzcnt' instruction. Do so, and test for this. We don't want to
generate unnecessary 'bsr' instructions.
These two changes fix some regressions in encoding and decoding
benchmarks. However, there is still a *lot* to be improve on in this
type of code.
llvm-svn: 147244
ARM targets with NEON units have access to aligned vector loads and
stores that are potentially faster than unaligned operations.
Add support for spilling the callee-saved NEON registers to an aligned
stack area using 16-byte aligned NEON loads and store.
This feature is off by default, controlled by an -align-neon-spills
command line option.
llvm-svn: 147211
time regressions. In general, it is beneficial to compile-time.
Original commit message:
Fix for bug #11429: Wrong behaviour for switches. Small improvement for code
size heuristics.
llvm-svn: 147175
probability wouldn't be considered "hot" in some weird loop structures
or other compounding probability patterns. This makes it much harder to
confuse, but isn't really a principled fix. I'd actually like it if we
could model a zero probability, as it would make this much easier to
reason about. Suggestions for how to do this better are welcome.
llvm-svn: 147142
performance regressions (both execution-time and compile-time) on our
nightly testers.
Original commit message:
Fix for bug #11429: Wrong behaviour for switches. Small improvement for code
size heuristics.
llvm-svn: 147131
Use the spill slot alignment as well as the local variable alignment to
determine when the stack needs to be realigned. This works now that the
ARM target can always realign the stack by using a base pointer.
Still respect the ARMBaseRegisterInfo::canRealignStack() function
vetoing a realigned stack. Don't use aligned spill code in that case.
llvm-svn: 146997
use the zero-undefined variants of CTTZ and CTLZ. These are just simple
patterns for now, there is more to be done to make real world code using
these constructs be optimized and codegen'ed properly on X86.
The existing tests are spiffed up to check that we no longer generate
unnecessary cmov instructions, and that we generate the very important
'xor' to transform bsr which counts the index of the most significant
one bit to the number of leading (most significant) zero bits. Also they
now check that when the variant with defined zero result is used, the
cmov is still produced.
llvm-svn: 146974
We used to rely on the *eh_sjlj_setjmp instructions to mark that a function
with setjmp/longjmp exception handling clobbers all the registers. But with
the recent reorganization of ARM EH, those eh_sjlj_setjmp instructions are
expanded away earlier, before PEI can see them to determine what registers to
save and restore. Mark the dispatchsetup instruction in the same way, since
that instruction cannot be expanded early. This also more accurately reflects
when the registers are clobbered.
llvm-svn: 146949
"mov r1, r2, lsl #0" should assemble as "mov r1, r2" even though it's
not strictly legal UAL syntax. It's a common extension and the friendly
thing to do.
rdar://10604663
llvm-svn: 146937
merging types by name when we can. We still don't guarantee type name linkage
but we do it when obviously the right thing to do. This makes LTO type names
easier to read, for example.
llvm-svn: 146932
This change reduces the number of instructions generated.
For example,
(load (add (sub $n0, $n1), (MipsLo got(s))))
results in the following sequence of instructions:
1. sub $n2, $n0, $n1
2. lw got(s)($n2)
Previously, three instructions were needed.
1. sub $n2, $n0, $n1
2. addiu $n3, $n2, got(s)
3. lw 0($n3)
llvm-svn: 146888
internal nightly testers. Original commit message:
By popular demand, link up types by name if they are isomorphic and one is an
autorenamed version of the other. This makes the IR easier to read, because
we don't end up with random renamed versions of the types after LTO'ing a large
app.
llvm-svn: 146838
autorenamed version of the other. This makes the IR easier to read, because
we don't end up with random renamed versions of the types after LTO'ing a large app.
llvm-svn: 146728
into Analysis as a standalone function, since there's no need for
it to be in VMCore. Also, update it to use isKnownNonZero and
other goodies available in Analysis, making it more precise,
enabling more aggressive optimization.
llvm-svn: 146610
On ARM, peephole optimization for ABS creates a trivial cfg triangle which tempts machine sink to sink instructions in code which is really straight line code. Sometimes this sinking may alter register allocator input such that use and def of a reg is divided by a branch in between, which may result in extra spills. Now mahine sink avoids sinking if final sink destination is post dominator.
Radar 10266272.
llvm-svn: 146604
Work in progress. Parsing for non-writeback, single spaced register lists
works now. The rest have the representations better factored, but still
need more to be able to parse properly.
llvm-svn: 146579
When 'cmp rn #imm' doesn't match due to the immediate not being representable,
but 'cmn rn, #-imm' does match, use the latter in place of the former, as
it's equivalent.
rdar://10552389
llvm-svn: 146567
to finalize MI bundles (i.e. add BUNDLE instruction and computing register def
and use lists of the BUNDLE instruction) and a pass to unpack bundles.
- Teach more of MachineBasic and MachineInstr methods to be bundle aware.
- Switch Thumb2 IT block to MI bundles and delete the hazard recognizer hack to
prevent IT blocks from being broken apart.
llvm-svn: 146542
test cases where there were a lot of relocations applied relative to a large
rodata section. Gas would create a symbol for each of these whereas we would
be relative to the beginning of the rodata section. This change mimics what
gas does.
Patch by Jack Carter.
llvm-svn: 146468
These modifiers simply select either the low or high D subregister of a Neon
Q register. I've also removed the unimplemented 'p' modifier, which turns out
to be a bit different than the comment here suggests and as far as I can tell
was only intended for internal use in Apple's version of gcc.
llvm-svn: 146417
I followed three heuristics for deciding whether to set 'true' or
'false':
- Everything target independent got 'true' as that is the expected
common output of the GCC builtins.
- If the target arch only has one way of implementing this operation,
set the flag in the way that exercises the most of codegen. For most
architectures this is also the likely path from a GCC builtin, with
'true' being set. It will (eventually) require lowering away that
difference, and then lowering to the architecture's operation.
- Otherwise, set the flag differently dependending on which target
operation should be tested.
Let me know if anyone has any issue with this pattern or would like
specific tests of another form. This should allow the x86 codegen to
just iteratively improve as I teach the backend how to differentiate
between the two forms, and everything else should remain exactly the
same.
llvm-svn: 146370
intrinsic syntax.
Now that this is explicitly covered, I plan to upgrade the existing test
suite to use an explicit immediate. Note that I plan to specify 'true'
in most places rather than the auto-upgraded value as that is the far
more common value to end up here as that is the value coming from GCC's
builtins. The only place I'm likely to put a 'false' in is when testing
x86 which actually has different instructions for the two variants.
llvm-svn: 146369
does. The _GLOBAL_OFFSET_TABLE_ is still magical in that we get a R_386_GOTPC,
but it doesn't change the immediate in the same way as when the expression
has no right hand side symbol.
llvm-svn: 146311
When the immediate operand of an AND or BIC instruction isn't representable
in the immediate field of the instruction, but the bitwise negation of the
immediate is, assemble the instruction as the inverse operation instead
with the inverted immediate as the operand.
rdar://10550057
llvm-svn: 146283
Refactor the instructions into fixed writeback and register-stride
writeback variants to simplify the offset operand (no more optional
register operand using reg0). This is a simpler representation and allows
the assembly parser to more easily handle these instructions.
Add tests for the instruction variants now supported.
llvm-svn: 146278
generates the dwarf Compile Unit DIE and a dwarf subprogram DIE for each
non-temporary label.
The next part will be to get the clang driver to enable this when assembling
a .s file. rdar://9275556
llvm-svn: 146262
Patch by Brendon Cahoon!
This extends the existing LoopUnroll and LoopUnrollPass. Brendon
measured no regressions in the llvm test suite with -unroll-runtime
enabled. This implementation works by using the existing loop
unrolling code to unroll the loop by a power-of-two (default 8). It
generates an if-then-else sequence of code prior to the loop to
execute the extra iterations before entering the unrolled loop.
llvm-svn: 146245
We must not issue a bitcast operation for integer-promotion of vector types, because the
location of the values in the vector may be different.
llvm-svn: 146150
When the file isn't being built with subsections-via-symbols, symbol
differences involving non-local symbols can be resolved more aggressively.
Needed for gas compatibility.
llvm-svn: 146054
It's always good to prune early, but formulae that are unsatisfactory
in their own right need to be removed before running any other pruning
heuristics. We easily avoid generating such formulae, but we need them
as an intermediate basis for forming other good formulae.
llvm-svn: 145906
Previously, all ARM::CONSTPOOL_ENTRY instructions had a hardwired
alignment of 4 bytes emitted by ARMAsmPrinter. Now the same alignment
is set on the basic block.
This is in preparation of supporting ARM constant pool islands with
different alignments.
llvm-svn: 145890
libgcc sets the stack limit field in TCB to 256 bytes above the actual
allocated stack limit. This means if the function's stack frame needs
less than 256 bytes, we can just compare the stack pointer with the
stack limit. This should result in lesser calls to __morestack.
llvm-svn: 145766
Currently LLVM pads the call to __morestack with a add and sub of 8
bytes to esp. This isn't correct since __morestack expects the call
to be followed directly by a ret.
This commit also adjusts the relevant test-case.
llvm-svn: 145765
argument value type. Otherwise, the sign/zero-extend has no effect on arguments
passed via the stack (i.e., undefined high-order bits).
rdar://10515467
llvm-svn: 145701
Like V_SET0, these instructions are expanded by ExpandPostRA to xorps /
vxorps so they can participate in execution domain swizzling.
This also makes the AVX variants redundant.
llvm-svn: 145440
weak variable are compiled by different compilers, such as GCC and LLVM, while
LLVM may increase the alignment to the preferred alignment there is no reason to
think that GCC will use anything more than the ABI alignment. Since it is the
GCC version that might end up in the final program (as the linkage is weak), it
is wrong to increase the alignment of loads from the global up to the preferred
alignment as the alignment might only be the ABI alignment.
Increasing alignment up to the ABI alignment might be OK, but I'm not totally
convinced that it is. It seems better to just leave the alignment of weak
globals alone.
llvm-svn: 145413
as MC is the only assembler we support.
This splits MS/Windows and GNU/Windows ASM infos into two seperate classes.
While there is currently only one difference, full MS C++ ABI support will
require many more.
llvm-svn: 145409
Conservatively returns zero when the GV does not specify an alignment nor is it
initialized. Previously it returns ABI alignment for type of the GV. However, if
the type is a "packed" type, then the under-specified alignments is attached to
the load / store instructions. In that case, the alignment of the type cannot be
trusted.
rdar://10464621
llvm-svn: 145300
than ABI alignment. These are loads / stores from / to "packed" data structures.
Their alignments are intentionally under-specified.
rdar://10301431
llvm-svn: 145273
was centered around the premise of laying out a loop in a chain, and
then rotating that chain. This is good for preserving contiguous layout,
but bad for actually making sane rotations. In order to keep it safe,
I had to essentially make it impossible to rotate deeply nested loops.
The information needed to correctly reason about a deeply nested loop is
actually available -- *before* we layout the loop. We know the inner
loops are already fused into chains, etc. We lose information the moment
we actually lay out the loop.
The solution was the other alternative for this algorithm I discussed
with Benjamin and some others: rather than rotating the loop
after-the-fact, try to pick a profitable starting block for the loop's
layout, and then use our existing layout logic. I was worried about the
complexity of this "pick" step, but it turns out such complexity is
needed to handle all the important cases I keep teasing out of benchmarks.
This is, I'm afraid, a bit of a work-in-progress. It is still
misbehaving on some likely important cases I'm investigating in Olden.
It also isn't really tested. I'm going to try to craft some interesting
nested-loop test cases, but it's likely to be extremely time consuming
and I don't want to go there until I'm sure I'm testing the correct
behavior. Sadly I can't come up with a way of getting simple, fine
grained test cases for this logic. We need complex loop structures to
even trigger much of it.
llvm-svn: 145183
heavily on AnalyzeBranch. That routine doesn't behave as we want given
that rotation occurs mid-way through re-ordering the function. Instead
merely check that there are not unanalyzable branching constructs
present, and then reason about the CFG via successor lists. This
actually simplifies my mental model for all of this as well.
The concrete result is that we now will rotate more loop chains. I've
added a test case from Olden highlighting the effect. There is still
a bit more to do here though in order to regain all of the performance
in Olden.
llvm-svn: 145179
pass. This is designed to achieve one of the important optimizations
that the old code placement pass did, but more simply.
This is a somewhat rough and *very* conservative version of the
transform. We could get a lot fancier here if there are profitable cases
to do so. In particular, this only looks for a single pattern, it
insists that the loop backedge being rotated away is the last backedge
in the chain, and it doesn't provide any means of doing better in-loop
placement due to the rotation. However, it appears that it will handle
the important loops I am finding in the LLVM test suite.
llvm-svn: 145158
was returning incorrect values in rare cases, and incorrectly marking
exact conversions as inexact in some more common cases. Fixes PR11406, and a
missed optimization in test/CodeGen/X86/fp-stack-O0.ll.
llvm-svn: 145141
need lots of fanciness around retaining a reference to a Chain's slot in
the BlockToChain map, but that's all gone now. We can just go directly
to allocating the new chain (which will update the mapping for us) and
using it.
Somewhat gross mechanically generated test case replicates the issue
Duncan spotted when actually testing this out.
llvm-svn: 145120
conflicts, we should only be adding the first block of the chain to the
list, lest we try to merge into the middle of that chain. Most of the
places we were doing this we already happened to be looking at the first
block, but there is no reason to assume that, and in some cases it was
clearly wrong.
I've added a couple of tests here. One already worked, but I like having
an explicit test for it. The other is reduced from a test case Duncan
reduced for me and used to crash. Now it is handled correctly.
llvm-svn: 145119
and positive: positive, because it could be directly computed to be positive;
negative, because the nsw flags means it is either negative or undefined (the
multiplication always overflowed).
llvm-svn: 145104
Before:
movabsq $4294967296, %rax ## encoding: [0x48,0xb8,0x00,0x00,0x00,0x00,0x01,0x00,0x00,0x00]
testq %rax, %rdi ## encoding: [0x48,0x85,0xf8]
jne LBB0_2 ## encoding: [0x75,A]
After:
btq $32, %rdi ## encoding: [0x48,0x0f,0xba,0xe7,0x20]
jb LBB0_2 ## encoding: [0x72,A]
btq is usually slower than testq because it doesn't fuse with the jump, but here we're better off
saving one register and a giant movabsq.
llvm-svn: 145103
further. This invariant just wasn't going to work in the face of
unanalyzable branches; we need to be resillient to the phenomenon of
chains poking into a loop and poking out of a loop. In fact, we already
were, we just needed to not assert on it.
This was found during a bootstrap with block placement turned on.
llvm-svn: 145100
successors, they just are all landing pad successors. We handle this the
same way as no successors. Comments attached for the next person to wade
through here and another lovely test case courtesy of Benjamin Kramer's
bugpoint reduction.
llvm-svn: 145098
This was a bug in keeping track of the available domains when merging
domain values.
The wrong domain mask caused ExecutionDepsFix to try to move VANDPSYrr
to the integer domain which is only available in AVX2.
Also add an assertion to catch future attempts at emitting AVX2
instructions.
llvm-svn: 145096
reversed in the function's original ordering, and we happened to
encounter it while handling an outer unnatural CFG structure.
Thanks to the test case reduced from GCC's source by Benjamin Kramer.
This may also fix a crasher in gzip that Duncan reduced for me, but
I haven't yet gotten to testing that one.
llvm-svn: 145094
updateTerminator code didn't correctly handle EH terminators in one very
specific case. AnalyzeBranch would find no terminator instruction, and
so the fallback in updateTerminator is to assume fallthrough. This is
correct, but the destination of the fallthrough was assumed to be the
first successor.
This is *almost always* true, but in certain cases the loop
transformations will cause the landing pad to be the first successor!
Instead of this brittle logic, actually look through the successors for
a non-landing-pad accessor, and to assert if more than one is found.
This will hopefully fix some (if not all) of the self host miscompiles
with block placement. Thanks to Benjamin Kramer for reporting, Nick
Lewycky for an initial stab at a reduction, and Duncan for endless
advice on EH (which I know nothing about) as well as reviewing the
actual fix.
llvm-svn: 145062
properly account for the *global* probability of the edge being taken.
This manifested as a very large number of unconditional branches to
blocks being merged against the CFG even though they weren't
particularly hot within the CFG.
The fix is to check whether the edge being merged is both locally hot
relative to other successors for the source block, and globally hot
compared to other (unmerged) predecessors of the destination block.
This introduces a new crasher on GCC single-source, but it's currently
behind a flag, and Ben has offered to work on the reduction. =]
llvm-svn: 145010
is actually being tested. Also add some FileCheck goodness to much more
carefully ensure that the result is the desired result. Before this test
would only have failed through an assert failure if the underlying fix
were reverted.
Also, add some weight metadata and a comment explaining exactly what is
going on to a trick section of the test case. Originally, we were
getting very unlucky and trying to form a block chain that isn't
actually profitable. I'm working on a fix to avoid forming these
unprofitable chains, and that would also have masked any failure from
this test case. The easy solution is to add some metadata that makes it
*really* profitable to form the bad chain here.
llvm-svn: 145006
formation phase and into the initial walk of the basic blocks. We
essentially pre-merge all blocks where unanalyzable fallthrough exists,
as we won't be able to update the terminators effectively after any
reorderings. This is quite a bit more principled as there may be CFGs
where the second half of the unanalyzable pair has some analyzable
predecessor that gets placed first. Then it may get placed next,
implicitly breaking the unanalyzable branch even though we never even
looked at the part that isn't analyzable. I've included a test case that
triggers this (thanks Benjamin yet again!), and I'm hoping to synthesize
some more general ones as I dig into related issues.
Also, to make this new scheme work we have to be able to handle branches
into the middle of a chain, so add this check. We always fallback on the
incoming ordering.
Finally, this starts to really underscore a known limitation of the
current implementation -- we don't consider broken predecessors when
merging successors. This can caused major missed opportunities, and is
something I'm planning on looking at next (modulo more bug reports).
llvm-svn: 144994
The loop tree's inclusive block lists are painful and expensive to
update. (I have no idea why they're inclusive). The design was
supposed to handle this case but the implementation missed it and my
unit tests weren't thorough enough.
Fixes PR11335: loop unroll update.
llvm-svn: 144970
The right way to check for a binary operation is
cast<BinaryOperator>. The original check: cast<Instruction> &&
numOperands() == 2 would match phi "instructions", leading to an
infinite loop in extreme corner case: a useless phi with operands
[self, constant] that prior optimization passes failed to remove,
being used in the loop by another useless phi, in turn being used by an
lshr or udiv.
Fixes PR11350: runaway iteration assertion.
llvm-svn: 144935
ADDs. MaxOffs is used as a threshold to limit the size of the offset. Tradeoffs
being: (1) If we can't materialize the large constant then we'll cause fast-isel
to bail. (2) Too large of an offset can't be directly encoded in the ADD
resulting in a MOV+ADD. Generally not a bad thing because otherwise we would
have had ADD+ADD, but on Thumb this turns into a MOVS+MOVT+ADD. Working on a fix
for that. (3) Conversely, too low of a threshold we'll miss opportunities to
coalesce ADDs.
rdar://10412592
llvm-svn: 144886
We don't (yet) have the granularity in the fixups to be specific about which
bitranges are affected. That's a future cleanup, but we're not there yet.
llvm-svn: 144852
For example,
vld1.f64 {d2-d5}, [r2,:128]!
Should be equivalent to:
vld1.f64 {d2,d3,d4,d5}, [r2,:128]!
It's not documented syntax in the ARM ARM, but it is consistent with what's
accepted for VLDM/VSTM and is unambiguous in meaning, so it's a good thing to
support.
rdar://10451128
llvm-svn: 144727
When the 3rd operand is not a low-register, and the first two operands are
the same low register, the parser was incorrectly trying to use the 16-bit
instruction encoding.
rdar://10449281
llvm-svn: 144679
has a reference to it. Unfortunately, that doesn't work for codegen passes
since we don't get notified of MBB's being deleted (the original BB stays).
Use that fact to our advantage and after printing a function, check if
any of the IL BBs corresponds to a symbol that was not printed. This fixes
pr11202.
llvm-svn: 144674
block sequence when recovering from unanalyzable control flow
constructs, *always* use the function sequence. I'm not sure why I ever
went down the path of trying to use the loop sequence, it is
fundamentally not the correct sequence to use. We're trying to preserve
the incoming layout in the cases of unreasonable control flow, and that
is only encoded at the function level. We already have a filter to
select *exactly* the sub-set of blocks within the function that we're
trying to form into a chain.
The resulting code layout is also significantly better because of this.
In several places we were ending up with completely unreasonable control
flow constructs due to the ordering chosen by the loop structure for its
internal storage. This change removes a completely wasteful vector of
basic blocks, saving memory allocation in the common case even though it
costs us CPU in the fairly rare case of unnatural loops. Finally, it
fixes the latest crasher reduced out of GCC's single source. Thanks
again to Benjamin Kramer for the reduction, my bugpoint skills failed at
it.
llvm-svn: 144627
Two new TargetInstrInfo hooks lets the target tell ExecutionDepsFix
about instructions with partial register updates causing false unwanted
dependencies.
The ExecutionDepsFix pass will break the false dependencies if the
updated register was written in the previoius N instructions.
The small loop added to sse-domains.ll runs twice as fast with
dependency-breaking instructions inserted.
llvm-svn: 144602
and stores capture) to permit the caller to see each capture point and decide
whether to continue looking.
Use this inside memdep to do an analysis that basicaa won't do. This lets us
solve another devirtualization case, fixing PR8908!
llvm-svn: 144580
instructions of the two-address operands) in order to avoid inserting copies.
This fixes the few regressions introduced when the two-address hack was
disabled (without regressing the improvements).
rdar://10422688
llvm-svn: 144559
the sum of the edge weights not overflowing uint32, and crashed when
they did. This is generally safe as BranchProbabilityInfo tries to
provide this guarantee. However, the CFG can get modified during codegen
in a way that grows the *sum* of the edge weights. This doesn't seem
unreasonable (imagine just adding more blocks all with the default
weight of 16), but it is hard to come up with a case that actually
triggers 32-bit overflow. Fortuately, the single-source GCC build is
good at this. The solution isn't very pretty, but its no worse than the
previous code. We're already summing all of the edge weights on each
query, we can sum them, check for an overflow, compute a scale, and sum
them again.
I've included a *greatly* reduced test case out of the GCC source that
triggers it. It's a pretty lame test, as it clearly is just barely
triggering the overflow. I'd like to have something that is much more
definitive, but I don't understand the fundamental pattern that triggers
an explosion in the edge weight sums.
The buggy code is duplicated within this file. I'll colapse them into
a single implementation in a subsequent commit.
llvm-svn: 144526
get loop info structures associated with them, and so we need some way
to make forward progress selecting and placing basic blocks. The
technique used here is pretty brutal -- it just scans the list of blocks
looking for the first unplaced candidate. It keeps placing blocks like
this until the CFG becomes tractable.
The cost is somewhat unfortunate, it requires allocating a vector of all
basic block pointers eagerly. I have some ideas about how to simplify
and optimize this, but I'm trying to get the logic correct first.
Thanks to Benjamin Kramer for the reduced test case out of GCC. Sadly
there are other bugs that GCC is tickling that I'm reducing and working
on now.
llvm-svn: 144516
second algorithm, but only loosely. It is more heavily based on the last
discussion I had with Andy. It continues to walk from the inner-most
loop outward, but there is a key difference. With this algorithm we
ensure that as we visit each loop, the entire loop is merged into
a single chain. At the end, the entire function is treated as a "loop",
and merged into a single chain. This chain forms the desired sequence of
blocks within the function. Switching to a single algorithm removes my
biggest problem with the previous approaches -- they had different
behavior depending on which system triggered the layout. Now there is
exactly one algorithm and one basis for the decision making.
The other key difference is how the chain is formed. This is based
heavily on the idea Andy mentioned of keeping a worklist of blocks that
are viable layout successors based on the CFG. Having this set allows us
to consistently select the best layout successor for each block. It is
expensive though.
The code here remains very rough. There is a lot that needs to be done
to clean up the code, and to make the runtime cost of this pass much
lower. Very much WIP, but this was a giant chunk of code and I'd rather
folks see it sooner than later. Everything remains behind a flag of
course.
I've added a couple of tests to exercise the issues that this iteration
was motivated by: loop structure preservation. I've also fixed one test
that was exhibiting the broken behavior of the previous version.
llvm-svn: 144495
NAKAMURA Takumi