Summary:
I don't think rL309080 is the right fix for PR33494 -- caching ExitLimit only
hides the problem[0]. The real issue is that because of how we forget SCEV
expressions ScalarEvolution::getBackedgeTakenInfo, in the test case for PR33494
computing the backedge for any loop invalidates the trip count for every other
loop. This effectively makes the SCEV cache useless.
I've instead made the SCEV expression invalidation in
ScalarEvolution::getBackedgeTakenInfo less aggressive to fix this issue.
[0]: One way to think about this is that rL309080 essentially augmented the
backedge-taken-count cache with another equivalent exit-limit cache. The bug
went away because we were explicitly not clearing the exit-limit cache in
getBackedgeTakenInfo. But instead of doing all of that, we can just avoid
clearing the backedge-taken-count cache.
Reviewers: mkazantsev, mzolotukhin
Subscribers: mcrosier, llvm-commits
Differential Revision: https://reviews.llvm.org/D39361
llvm-svn: 319678
(This reapplies r314253. r314253 was reverted on r314482 because of a
correctness regression on P100, but that regression was identified to be
something else.)
Summary:
Don't bail out on constant divisors for divisions that can be narrowed without
introducing control flow . This gives us a 32 bit multiply instead of an
emulated 64 bit multiply in the generated PTX assembly.
Reviewers: jlebar
Subscribers: jholewinski, mcrosier, llvm-commits
Differential Revision: https://reviews.llvm.org/D38265
llvm-svn: 319677
As part of the unification of the debug format and the MIR format, print
MBB references as '%bb.5'.
The MIR printer prints the IR name of a MBB only for block definitions.
* find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g'
* find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g'
* find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g'
* grep -nr 'BB#' and fix
Differential Revision: https://reviews.llvm.org/D40422
llvm-svn: 319665
Summary:
The compiler fails with the following error message:
fatal error: error in backend: ran out of registers during
register allocation
Tail call optimization for Armv8-M.base fails to meet all the required
constraints when handling calls to function pointers where the
arguments take up r0-r3. This is because the pointer to the
function to be called can only be stored in r0-r3, but these are
all occupied by arguments. This patch makes sure that tail call
optimization does not try to handle this type of calls.
Reviewers: chill, MatzeB, olista01, rengolin, efriedma
Reviewed By: olista01, efriedma
Subscribers: efriedma, aemerson, javed.absar, llvm-commits, kristof.beyls
Differential Revision: https://reviews.llvm.org/D40706
llvm-svn: 319664
This reverts commit r319533 as it broke llvm-config --system-libs output
and everything that depends on it (which is mostly out of tree or
downstream folks, but includes a couple of llvm buildbots as well).
I think I have a fix for this in D40779, but I want someone to look
review it first. In the mean time, I am reverting this change, as it
seems to break a lot of people.
llvm-svn: 319663
Summary:
Currently, we only support predication for forward loops with step
of 1. This patch enables loop predication for reverse or
countdownLoops, which satisfy the following conditions:
1. The step of the IV is -1.
2. The loop has a singe latch as B(X) = X <pred>
latchLimit with pred as s> or u>
3. The IV of the guard is the decrement
IV of the latch condition (Guard is: G(X) = X-1 u< guardLimit).
This patch was downstream for a while and is the last series of patches
that's from our LP implementation downstream.
Reviewers: apilipenko, mkazantsev, sanjoy
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D40353
llvm-svn: 319659
PTX requires that identifiers consist only of [a-zA-Z0-9_$]. The
existing pass already ensured this for globals and this patch adds
the cleanup for functions with local linkage.
However, there was a different problem in the case of collisions
of the adjusted name: The ValueSymbolTable then automatically
appended ".N" with increasing Ns to get a unique name while helping
the ABI demangling. Special case this behavior to omit the dots and
append N directly. This will always give us legal names according
to the PTX requirements.
Differential Revision: https://reviews.llvm.org/D40573
llvm-svn: 319657
This is causing a failure in the llvm-clang-x86_64-expensive-checks-win
buildbot, and I can't reproduce it locally, so reverting until I can work out
what is wrong.
llvm-svn: 319654
Summary:
The pass that inserts s_waitcnt instructions where needed propagated
info used to track dependencies for each block by iterating over the
predecessor blocks. The iteration was terminated when a predecessor
that had not yet been processed was encountered. Any info in blocks
later in the list was therefore not processed, leading to the
possiblility of a required s_waitcnt not being inserted.
The fix is simply to change the "break" to "continue" for the
relevant loops, so that all visited blocks are processed. This
is likely what was intended when the code was written.
There is no test case provided for this fix because:
1) the only example that reproduces this is large and resistant to
being reduced
2) the change is trivial
Subscribers: arsenm, kzhuravl, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye
Differential Revision: https://reviews.llvm.org/D40544
llvm-svn: 319651
This adds a "invalid operands for instruction" diagnostic for
instructions where there is an instruction encoding with the correct
mnemonic and which is available for this target, but where multiple
operands do not match those which were provided. This makes it clear
that there is some combination of operands that is valid for the current
target, which the default diagnostic of "invalid instruction" does not.
Since this is a very general error, we only emit it if we don't have a
more specific error.
Differential revision: https://reviews.llvm.org/D36747
llvm-svn: 319649
An instruction returned by TII->convertToThreeAddress() may contain a %noreg
(undef) operand, which is not expected by tryInstructionTransform(). So if
this MI is sunk to a lower point in MBB, it must be skipped when later
encountered.
A new set SunkInstrs is used for this purpose.
Note: there is no test supplied here, as this was triggered on SystemZ while
working on a review of instruction flags. A test case for this bugfix will be
included in the upcoming SystemZ commit.
Review: Quentin Colombet
https://reviews.llvm.org/D40711
llvm-svn: 319646
Both LoadedVT and NarrowLoad are passed as references and neither
of them are used by any of its callers.
Differential Revision: https://reviews.llvm.org/D40713
llvm-svn: 319645
This matches how it is done on X86.
This allows using emulated tls on windows; in MinGW environments,
native tls isn't supported at the moment.
Set the right Data*bitsDirective for windows to match the existing
tests for other platforms. Make parts of the existing tests a regex,
to allow matching .section .rdata for windows, to avoid having to
duplicate the rest of the tests for windows.
Differential Revision: https://reviews.llvm.org/D40770
llvm-svn: 319644
This matches how it is done on X86.
This allows using emulated tls on windows; in MinGW environments,
native tls isn't supported at the moment.
Differential Revision: https://reviews.llvm.org/D40769
llvm-svn: 319643
These instructions can be used by widening to 512-bits and extracting back to 128/256. We do similar to several other instructions already.
llvm-svn: 319641
We already do this as a DAG combine. The version during lowering can only trigger if known bits changes something that improves known bits analysis. But this means we should be improving known bits analysis to work on the unlowered form instead.
llvm-svn: 319640
If we have a non-splat constant shift amount, the minimum shift amount can be used to infer the number of zero upper bits of the result. There's probably a lot more that we can do here, but this
fixes a case where I wanted to infer the sign bit as zero when all the shift amounts are non-zero.
llvm-svn: 319639
SelectionDAGISel::LowerArguments assumes sret addr space is 0, which is
not true for amdgcn---amdgiz target.
This patch fixes that.
Differential Revision: https://reviews.llvm.org/D40255
llvm-svn: 319630
Original change was rL319488.
This was reverted rL319602 due to a gcc 7.1 warning.
Differential Revision: https://reviews.llvm.org/D40772
llvm-svn: 319626
Two issues found when doing codegen for splitting vector with non-zero alloca addr space:
DAGTypeLegalizer::SplitVecRes_INSERT_VECTOR_ELT/SplitVecOp_EXTRACT_VECTOR_ELT uses dummy pointer info for creating
SDStore. Since one pointer operand contains multiply and add, InferPointerInfo is unable to
infer the correct pointer info, which ends up with a dummy pointer info for the target to lower
store and results in isel failure. The fix is to introduce MachinePointerInfo::getUnknownStack to
represent MachinePointerInfo which is known in alloca address space but without other information.
TargetLowering::getVectorElementPointer uses value type of pointer in addr space 0 for
multiplication of index and then add it to the pointer. However the pointer may be in an addr
space which has different size than addr space 0. The fix is to use the pointer value type for
index multiplication.
Differential Revision: https://reviews.llvm.org/D39758
llvm-svn: 319622
Makes it easier to grok where each is supposed to be used, mainly useful for adding to the AVX512 instructions but hopefully can be used more in SSE/AVX as well.
llvm-svn: 319614
Currently, the outliner considers candidates that intersect with themselves in
the candidate pruning step. That is, candidates of the form "AA" in ranges like
"AAAAAA". In that range, it looks like there are 5 instances of "AA" that could
possibly be outlined, and that's considered in the benefit calculation.
However, only at most 3 instances of "AA" could ever be outlined in "AAAAAA".
Thus, it's possible to pass through "AA" to the candidate selection step even
though it's *never* the case that "AA" could be outlined. This makes it so that
when we find candidates, we consider only non-overlapping occurrences of that
candidate.
llvm-svn: 319588
r230670 introduced a step to map EH register numbers to standard
DWARF register numbers. This failed to consider the case when a
user .cfi_* directive uses an integer literal rather than a
register name, to specify a DWARF register number that has no
corresponding LLVM register number (e.g. a special register that
the compiler and assembler have no name for).
Fixes PR34028.
Patch by Roland McGrath
Differential Revision: https://reviews.llvm.org/D36493
llvm-svn: 319586
Turns out we can have comparisons which are indirect users of the induction variable that we can make invariant. In this case, there is no loop invariant value contributing and we'd fail an assert.
The test case was found by a java fuzzer and reduced. It's a real cornercase. You have to have a static loop which we've already proven only executes once, but haven't broken the backedge on, and an inner phi whose result can be constant folded by SCEV using exit count reasoning but not proven by isKnownPredicate. To my knowledge, only the fuzzer has hit this case.
llvm-svn: 319583
These are blocks that haven't not been executed during training. For large
projects this could make a significant difference. For the project, I was
looking at, I got an order of magnitude decrease in the size of the total YAML
files with this and r319235.
Differential Revision: https://reviews.llvm.org/D40678
Re-commit after fixing the failing testcase in rL319576, rL319577 and
rL319578.
llvm-svn: 319581
Summary:
Adding support for -print-module-scope similar to how it is
being done for function passes. This option causes loop-pass printer
to emit a whole-module IR instead of just a loop itself.
Reviewers: sanjoy, silvas, weimingz
Reviewed By: sanjoy
Subscribers: apilipenko, skatkov, llvm-commits
Differential Revision: https://reviews.llvm.org/D40247
llvm-svn: 319566
Summary:
When debugging function passes it happens to be rather useful to dump
the whole module before the transformation and then use this dump
to analyze this single transformation by running it separately
on that particular module state.
Introducing
-print-module-scope
debugging option that forces all the function-level IR dumps
to become whole-module dumps.
This option builds on top of normal dumping controls like
-print-before/after
-filter-print-funcs
The plan is to eventually extend this option to cover other local passes
(at least loop passes) but that should go as a separate change.
Reviewers: sanjoy, weimingz, silvas, fedor.sergeev
Reviewed By: weimingz
Subscribers: apilipenko, skatkov, llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D40245
llvm-svn: 319561
These are blocks that haven't not been executed during training. For large
projects this could make a significant difference. For the project, I was
looking at, I got an order of magnitude decrease in the size of the total YAML
files with this and r319235.
Differential Revision: https://reviews.llvm.org/D40678
llvm-svn: 319556
It causes builds to fail with "Instruction does not dominate all uses" (PR35497).
> Patch tries to improve vectorization of the following code:
>
> void add1(int * __restrict dst, const int * __restrict src) {
> *dst++ = *src++;
> *dst++ = *src++ + 1;
> *dst++ = *src++ + 2;
> *dst++ = *src++ + 3;
> }
> Allows to vectorize even if the very first operation is not a binary add, but just a load.
>
> Fixed issues related to previous commit.
>
> Reviewers: spatel, mzolotukhin, mkuper, hfinkel, RKSimon, filcab, ABataev
>
> Reviewed By: ABataev, RKSimon
>
> Subscribers: llvm-commits, RKSimon
>
> Differential Revision: https://reviews.llvm.org/D28907
llvm-svn: 319550
Summary:
1/ Operand folding during complex pattern matching for LEAs has been extended, such that it promotes Scale to
accommodate similar operand appearing in the DAG e.g.
T1 = A + B
T2 = T1 + 10
T3 = T2 + A
For above DAG rooted at T3, X86AddressMode will now look like
Base = B , Index = A , Scale = 2 , Disp = 10
2/ During OptimizeLEAPass down the pipeline factorization is now performed over LEAs so that if there is an opportunity
then complex LEAs (having 3 operands) could be factored out e.g.
leal 1(%rax,%rcx,1), %rdx
leal 1(%rax,%rcx,2), %rcx
will be factored as following
leal 1(%rax,%rcx,1), %rdx
leal (%rdx,%rcx) , %edx
3/ Aggressive operand folding for AM based selection for LEAs is sensitive to loops, thus avoiding creation of any complex LEAs within a loop.
4/ Simplify LEA converts (lea (BASE,1,INDEX,0) --> add (BASE, INDEX) which offers better through put.
PR32755 will be taken care of by this pathc.
Previous patch revisions : r313343 , r314886
Reviewers: lsaba, RKSimon, craig.topper, qcolombet, jmolloy, jbhateja
Reviewed By: lsaba, RKSimon, jbhateja
Subscribers: jmolloy, spatel, igorb, llvm-commits
Differential Revision: https://reviews.llvm.org/D35014
llvm-svn: 319543
Summary:
A true or false result is expected from a comparison, but it seems the possibility of undef was overlooked, which could lead to a failed assert. This is fixed by this patch by bailing out if we encounter undef.
The bug is old and the assert has been there since the end of 2014, so it seems this is unusual enough to forego optimization.
Patch by: JesperAntonsson
Reviewers: spatel, eeckstein, hans
Reviewed By: hans
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D40639
llvm-svn: 319537
Summary:
zlib support was hard-wired to off for (non-cygwin) windows targets.
This disables some features, such as reading debug info from compressed
dwarf sections.
This has been this way since zlib support was added in 2013 (r180083),
but there is no obvious reason for that. Zlib is perfectly capable of
being compiled for windows (it even has a cmake file that works out of
the box).
This enables one to turn on zlib support on windows, if one has zlib
avaliable.
Reviewers: rnk, beanz
Subscribers: mgorny, aprantl, llvm-commits
Differential Revision: https://reviews.llvm.org/D40655
llvm-svn: 319533
Patch tries to improve vectorization of the following code:
void add1(int * __restrict dst, const int * __restrict src) {
*dst++ = *src++;
*dst++ = *src++ + 1;
*dst++ = *src++ + 2;
*dst++ = *src++ + 3;
}
Allows to vectorize even if the very first operation is not a binary add, but just a load.
Fixed issues related to previous commit.
Reviewers: spatel, mzolotukhin, mkuper, hfinkel, RKSimon, filcab, ABataev
Reviewed By: ABataev, RKSimon
Subscribers: llvm-commits, RKSimon
Differential Revision: https://reviews.llvm.org/D28907
llvm-svn: 319531
Summary: LegalizerInfo assumes all G_MERGE_VALUES and G_UNMERGE_VALUES instructions are legal, so it is not possible to legalize vector operations on illegal vector types. This patch fixes the problem by removing the related check and adding default actions for G_MERGE_VALUES and G_UNMERGE_VALUES.
Reviewers: qcolombet, ab, dsanders, aditya_nandakumar, t.p.northover, kristof.beyls
Reviewed By: dsanders
Subscribers: rovka, javed.absar, igorb, llvm-commits
Differential Revision: https://reviews.llvm.org/D39823
llvm-svn: 319524
The default legalization for v2i32 is promotion to v2i64. This results in a gather that reads 64-bit elements rather than 32. If one of the elements is near a page boundary this can cause an illegal access that can fault.
We also miscalculate the scale for the gather which is an even worse problem, but we probably could have found a separate way to fix that.
llvm-svn: 319521
Type promotion makes no guarantee about the contents of the promoted bits. Since the gather/scatter instruction will use the bits to calculate addresses, we need to ensure they aren't garbage.
llvm-svn: 319520
Even with the sparse file optimizations the SYM64 test can still be painfully
slow. This unnecessarily slows down devs. It's critical that we test that the
switch to the SYM64 format occurs at 4GB but there isn't any better of a way to
fake the size of the file than sparse files. This change introduces a flag that
allows the cutoff to be arbitrarily set to whatever power of two is desired.
The flag is hidden as it really isn't meant to be used outside this one test.
This is unfortunate but appears necessary, at least until the average hard
drive is much faster.
The changes to the test require some explanation. Prior to this change we knew
that the SYM64 format was being used because the file was simply too large to
have validly handled this case if the SYM64 format were not used. To ensure
that the SYM64 format is still being used I am grepping the file for "SYM64".
Without changing the filename however this would be pointless because "SYM64"
would occur in the file either way. So the filename of the test is also changed
in order to avoid this issue.
Differential Revision: https://reviews.llvm.org/D40632
llvm-svn: 319507
These command line options are not intended for public use, and often
don't even make sense in the context of a particular tool anyway. About
90% of them are already hidden, but when people add new options they
forget to hide them, so if you were to make a brand new tool today, link
against one of LLVM's libraries, and run tool -help you would get a
bunch of junk that doesn't make sense for the tool you're writing.
This patch hides these options. The real solution is to not have
libraries defining command line options, but that's a much larger effort
and not something I'm prepared to take on.
Differential Revision: https://reviews.llvm.org/D40674
llvm-svn: 319505
If the thin module has no references to an internal global in the
merged module, we need to make sure to preserve that property if the
global is a member of a comdat group, as otherwise promotion can end
up adding global symbols to the comdat, which is not allowed.
This situation can arise if the external global in the thin module
has dead constant users, which would cause use_empty() to return
false and would cause us to try to promote it. To prevent this from
happening, discard the dead constant users before asking whether a
global is empty.
Differential Revision: https://reviews.llvm.org/D40593
llvm-svn: 319494
This was storing the hash alongside the key so that the hash
doesn't need to be re-computed every time, but in doing so it
was allocating a structure to keep the key size small in the
DenseMap. This is a noble goal, but it also leads to a pointer
indirection on every probe, and this cost of this pointer
indirection ends up being higher than the cost of having a
slightly larger entry in the hash table. Removing this not only
simplifies the code, but yields a small but noticeable
performance improvement in the type merging algorithm.
llvm-svn: 319493
The LLVM "hidden" flag needs to be passed through the Wasm
intermediate objects in order for the linker to apply
it to the final Wasm object.
The corresponding change in LLD is here: https://github.com/WebAssembly/lld/pull/14
Patch by Nicholas Wilson
Differential Revision: https://reviews.llvm.org/D40442
llvm-svn: 319488
This teaches memcpyopt to make a non-local memdep query when a local query
indicates that the dependency is non-local. This notably allows it to
eliminate many more llvm.memcpy calls in common Rust code, often by 20-30%.
Fixes PR28958.
Differential Revision: https://reviews.llvm.org/D38374
llvm-svn: 319482
G_ATOMICRMW_* is generally legal on AArch64. The exception is G_ATOMICRMW_NAND.
G_ATOMIC_CMPXCHG_WITH_SUCCESS needs to be lowered to G_ATOMIC_CMPXCHG with an
external comparison.
Note that IRTranslator doesn't generate these instructions yet.
llvm-svn: 319466
output
As part of the unification of the debug format and the MIR format,
always use `printReg` to print all kinds of registers.
Updated the tests using '_' instead of '%noreg' until we decide which
one we want to be the default one.
Differential Revision: https://reviews.llvm.org/D40421
llvm-svn: 319445
In rare cases we can receive request to inject into completelly empty basic block. In the normal case
all basic blocks contain at least terminator instruction, but it is possible that the only instruction is
catchpad instruction which is not part of the instruction iterator. This case seems rare enough to not care
about it.
Submiting without review, since it seems almost NFC. I couldn't come up with any reasonable way to test this.
llvm-svn: 319444
This re-commits everything that was pulled in r314244. The transformation
is off by default (patch to enable it to follow). The code is refactored
to have a single entry-point and provide fine-grained control over patterns
that it selects. This patch also fixes the bugs in the original code.
Everything that failed with the original patch has been re-tested with this
patch (with the transformation turned on). So the patch to turn this on is
soon to follow.
Differential Revision: https://reviews.llvm.org/D38575
llvm-svn: 319434
Re applying after fixing issues in the diff, sorry for any painful conflicts/merges!
Original RFC: http://lists.llvm.org/pipermail/llvm-dev/2017-August/117028.html
This change adds a '.stack-size' section containing metadata on function stack sizes to output ELF files behind the new -stack-size-section flag. The section contains pairs of function symbol references (8 byte) and stack sizes (unsigned LEB128).
The contents of this section can be used to measure changes to stack sizes between different versions of the compiler or a source base. The advantage of having a section is that we can extract this information when examining binaries that we didn't build, and it allows users and tools easy access to that information just by referencing the binary.
There is a follow up change to add an option to clang.
Thanks.
Reviewers: hfinkel, MatzeB
Reviewed By: MatzeB
Subscribers: thegameg, asb, llvm-commits
Differential Revision: https://reviews.llvm.org/D39788
llvm-svn: 319430
As part of the unification of the debug format and the MIR format, avoid
printing "vreg" for virtual registers (which is one of the current MIR
possibilities).
Basically:
* find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E "s/%vreg([0-9]+)/%\1/g"
* grep -nr '%vreg' . and fix if needed
* find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E "s/ vreg([0-9]+)/ %\1/g"
* grep -nr 'vreg[0-9]\+' . and fix if needed
Differential Revision: https://reviews.llvm.org/D40420
llvm-svn: 319427
Summary:
Original RFC: http://lists.llvm.org/pipermail/llvm-dev/2017-August/117028.html
I wasn't sure who to put as reviewers, so please add/remove people as appropriate.
This change adds a '.stack-size' section containing metadata on function stack sizes to output ELF files behind the new -stack-size-section flag. The section contains pairs of function symbol references (8 byte) and stack sizes (unsigned LEB128).
The contents of this section can be used to measure changes to stack sizes between different versions of the compiler or a source base. The advantage of having a section is that we can extract this information when examining binaries that we didn't build, and it allows users and tools easy access to that information just by referencing the binary.
There is a follow up change to add an option to clang.
Thanks.
Reviewers: hfinkel, MatzeB
Reviewed By: MatzeB
Subscribers: thegameg, asb, llvm-commits
Differential Revision: https://reviews.llvm.org/D39788
llvm-svn: 319423
visitAND attempts to narrow the width of extending loads that are
then masked off. ReduceLoadWidth already exists for a similar purpose
and handles shifts, so I've moved the code to handle AND nodes there.
Differential Revision: https://reviews.llvm.org/D39595
llvm-svn: 319421
Csmith generated a program where a store after load to the same address did
not get chained after the new load created during DAG legalizing, and so
performed an illegal overwrite of the expected value.
When the new zero-extending load is created, the chain users of the original
load must be updated, which was not done previously.
A similar case was also found and handled in lowerBITCAST.
Review: Ulrich Weigand
https://reviews.llvm.org/D40542
llvm-svn: 319409
Currently, SROA splits loads and stores only when they are accessing the whole alloca.
This patch relaxes this limitation to allow splitting a load/store if all other loads and stores to the alloca are disjoint to or fully included in the current load/store. If there is no other load or store that crosses the boundary of the current load/store, the current splitting implementation works as is.
The whole-alloca loads and stores meet this new condition and so they are still splittable.
Here is a simplified motivating example.
struct record {
long long a;
int b;
int c;
};
int func(struct record r) {
for (int i = 0; i < r.c; i++)
r.b++;
return r.b;
}
When updating r.b (or r.c as well), LLVM generates redundant instructions on some platforms (such as x86_64, ppc64); here, r.b and r.c are packed into one 64-bit GPR when the struct is passed as a method argument.
With this patch, the above example is compiled into only few instructions without loop.
Without the patch, unnecessary loop-carried dependency is introduced by SROA and the loop cannot be eliminated by the later optimizers.
Differential Revision: https://reviews.llvm.org/D32998
llvm-svn: 319407
Normal type legalization will widen everything. This requires forcing 0s into the mask register. We can instead choose the form that only reads 2 elements without zeroing the mask.
llvm-svn: 319406
Support for outlining multiple regions of each function is added, as well as some basic heuristics to determine which regions are good to outline. Outline candidates limited to regions that are single-entry & single-exit. We also avoid outlining regions that produce live-exit variables, which may inhibit some forms of code motion (like commoning).
Fallback to the regular partial inlining scheme is retained when either i) no regions are identified for outlining in the function, or ii) the outlined function could not be inlined in any of its callers.
Differential Revision: https://reviews.llvm.org/D38190
llvm-svn: 319398
GFX9 does not enable bounds checking for the resource descriptors
used for private access, so it should be OK to use vaddr with
a potentially negative value.
llvm-svn: 319393
This is a fix for the coverage segment builder.
If multiple regions must be popped off the active stack at once, and
more than one of them end at the same location, emit a segment using the
count from the most-recent completed region.
Fixes PR35437, rdar://35760630
Testing: invoked llvm-cov on a stage2 build of clang, additional unit
tests, check-profile
llvm-svn: 319391
- add -ppc-reg-with-percent-prefix option to use %r3 etc as register
names
- split off logic for Darwinish verbose conditional codes into a helper
function
- be explicit about Darwin vs AIX vs GNUish assembler flavors
Based on the patch from Alexandre Yukio Yamashita
Differential Revision: https://reviews.llvm.org/D39016
llvm-svn: 319381
This class had some code that would automatically remap type
indices before hashing and serializing. The only caller of
this method was the TypeStreamMerger anyway, and the method
doesn't make general sense, and prevents making certain future
improvements to the class. So, factoring this up one level
into the TypeStreamMerger where it belongs.
llvm-svn: 319377
If we put in an assertsext/zext here, we're able to generate better truncate code using pack on pre-avx512 targets.
Similar is already done during type legalization. This is the equivalent for op legalization
Differential Revision: https://reviews.llvm.org/D40591
llvm-svn: 319368
A couple of places in LLD were passing references to
TypeTableCollections around, which makes it hard to change the
implementation at runtime. However, these cases only needed to
iterate over the types in the collection, and TypeCollection
already provides a handy abstract interface for this purpose.
By implementing this interface, we can get rid of the need to
pass TypeTableBuilder references around, which should allow us
to swap the implementation at runtime in subsequent patches.
llvm-svn: 319345
Partially reverting enabling of post-legalization store merge
(r319036) for just ARM backend as it is causing incorrect code
in some Thumb2 cases.
llvm-svn: 319331
When lowering a G_BRCOND, we generate a TSTri of the condition against
1, which sets the flags, and then a Bcc which branches based on the
value of the flags.
Unfortunately, we were using the wrong condition code to check whether
we need to branch (EQ instead of NE), which caused all our branches to
do the opposite of what they were intended to do. This patch fixes the
issue by using the correct condition code.
llvm-svn: 319313
When translating pseudo to MC, v_add/sub/subrev_u32 shall be mapped via a separate table as GFX8 has opcodes with the same names.
These instructions shall also be labelled as renamed for pseudoToMCOpcode to handle them correctly.
Reviewers: arsenm
Differential Revision: https://reviews.llvm.org/D40550
llvm-svn: 319311
This will allow compilation of assembly files targeting armv7e-m without having
to specify the Tag_CPU_arch attribute as a workaround.
Differential revision: https://reviews.llvm.org/D40370
Patch by Ian Tessier!
llvm-svn: 319303
If common type is different we should bail out due to we will not be
able to create a select or Phi of these values.
Basically it is done in ExtAddrMode::compare however it does not work
if we handle the null first and then two values of different types.
so add a check in initializeMap as well. The check in ExtAddrMode::compare
is used as earlier bail out.
Reviewers: reames, john.brawn
Reviewed By: john.brawn
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D40479
llvm-svn: 319292
Separate the handling of AND/AND8 out from PHI/OR/ISEL checking. The reasoning
is the others need all their operands to be sign/zero extended for their output
to also be sign/zero extended. This is true for AND and sign-extension, but for
zero-extension we only need at least one of the input operands to be zero
extended for the result to also be zero extended.
Differential Revision: https://reviews.llvm.org/D39078
llvm-svn: 319289
The object can't straddle the address space
wrap around, so I think it's OK to assume any
offsets added to the base object pointer can't
overflow. Similar logic already appears to be
applied in SelectionDAGBuilder when lowering
aggregate returns.
llvm-svn: 319272
GFX9 stopped using m0 for most DS instructions. Select
a different instruction without the use. I think this will
be less error prone than trying to manually maintain m0
uses as needed.
llvm-svn: 319270
Previously we had an isel pattern to add the truncate. Instead use Promote to add the truncate to the DAG before isel.
The Promote legalization code had to be updated to prevent an infinite loop if promotion took multiple steps because it wasn't remembering the previously tried value.
llvm-svn: 319259
An alloca may be larger than a variable that is described to be stored
there. Don't create a dbg.value for fragments that are outside of the
variable.
This fixes PR35447.
https://bugs.llvm.org/show_bug.cgi?id=35447
llvm-svn: 319230
The IRTranslator cannot generate these instructions at the moment so there's no
issue with not having implemented ISel for them yet. D40092 will add
G_ATOMIC_CMPXCHG_WITH_SUCCESS and G_ATOMICRMW_* to the IRTranslator and a
further patch will add support for lowering G_ATOMIC_CMPXCHG_WITH_SUCCESS into
G_ATOMIC_CMPXCHG with an external success check via the `Lower` action.
The separation of G_ATOMIC_CMPXCHG_WITH_SUCCESS and G_ATOMIC_CMPXCHG is
to import SelectionDAG rules while still supporting targets that prefer to
custom lower the original LLVM-IR-like operation.
llvm-svn: 319216
Summary:
Recommitting this with the correct sorting predicate. The Low field of Clusters is a ConstantInt and
cannot be directly compared. So we needed to invoke slt (signed less than) to compare correctly.
This fixes failures in the following tests uncovered by D39245:
LLVM :: CodeGen/ARM/ifcvt3.ll
LLVM :: CodeGen/ARM/switch-minsize.ll
LLVM :: CodeGen/X86/switch.ll
LLVM :: CodeGen/X86/switch-bt.ll
LLVM :: CodeGen/X86/switch-density.ll
Reviewers: hans, fhahn
Reviewed By: hans
Subscribers: aemerson, llvm-commits, kristof.beyls
Differential Revision: https://reviews.llvm.org/D40541
llvm-svn: 319210
Summary:
They're not always mutually exclusive. read-modify-write atomics are both
at the same time. One example of this is the SWP instructions on AArch64.
Another example is GlobalISel's G_ATOMICRMW_* generic instructions which
will be added in a later patch.
Reviewers: arphaman, aemerson
Reviewed By: aemerson
Subscribers: aemerson, javed.absar, kristof.beyls, llvm-commits
Differential Revision: https://reviews.llvm.org/D40157
llvm-svn: 319202
The motivation behind this patch is that future directions require us to
be able to compute the hash value of records independently of actually
using them for de-duplication.
The current structure of TypeSerializer / TypeTableBuilder being a
single entry point that takes an unserialized type record, and then
hashes and de-duplicates it is not flexible enough to allow this.
At the same time, the existing TypeSerializer is already extremely
complex for this very reason -- it tries to be too many things. In
addition to serializing, hashing, and de-duplicating, ti also supports
splitting up field list records and adding continuations. All of this
functionality crammed into this one class makes it very complicated to
work with and hard to maintain.
To solve all of these problems, I've re-written everything from scratch
and split the functionality into separate pieces that can easily be
reused. The end result is that one class TypeSerializer is turned into 3
new classes SimpleTypeSerializer, ContinuationRecordBuilder, and
TypeTableBuilder, each of which in isolation is simple and
straightforward.
A quick summary of these new classes and their responsibilities are:
- SimpleTypeSerializer : Turns a non-FieldList leaf type into a series of
bytes. Does not do any hashing. Every time you call it, it will
re-serialize and return bytes again. The same instance can be re-used
over and over to avoid re-allocations, and in exchange for this
optimization the bytes returned by the serializer only live until the
caller attempts to serialize a new record.
- ContinuationRecordBuilder : Turns a FieldList-like record into a series
of fragments. Does not do any hashing. Like SimpleTypeSerializer,
returns references to privately owned bytes, so the storage is
invalidated as soon as the caller tries to re-use the instance. Works
equally well for LF_FIELDLIST as it does for LF_METHODLIST, solving a
long-standing theoretical limitation of the previous implementation.
- TypeTableBuilder : Accepts sequences of bytes that the user has already
serialized, and inserts them by de-duplicating with a hash table. For
the sake of convenience and efficiency, this class internally stores a
SimpleTypeSerializer so that it can accept unserialized records. The
same is not true of ContinuationRecordBuilder. The user is required to
create their own instance of ContinuationRecordBuilder.
Differential Revision: https://reviews.llvm.org/D40518
llvm-svn: 319198
As part of the unification of the debug format and the MIR format,
always print registers as lowercase.
* Only debug printing is affected. It now follows MIR.
Differential Revision: https://reviews.llvm.org/D40417
llvm-svn: 319187
Generalize FixFunctionBitcasts to handle varargs functions. This in
particular fixes the case where clang bitcasts away a varargs when
calling a K&R-style function.
This avoids interacting with tricky ABI details because it operates
at the LLVM IR level before varargs ABI details are exposed.
This fixes PR35385.
llvm-svn: 319186
Looking through Agner, FTST is very similar to generic float compare behaviour, so I've added them to the existing IIC_FCOMI (WriteFAdd) tags.
llvm-svn: 319184
Atom's FABS/FCHS/FSQRT latencies taken from Agner.
Note: I just added FSIN and FCOS to the existing IIC_FSINCOS itinerary, which is actually a more costly instruction.
llvm-svn: 319175
This is needed for cases when the memory access is not as big as the width of
the data type. For instance, storing i1 (1 bit) would be done in a byte (8
bits).
Using 'BitSize >> 3' (or '/ 8') would e.g. give the memory access of an i1 a
size of 0, which for instance makes alias analysis return NoAlias even when
it shouldn't.
There are no tests as this was done as a follow-up to the bugfix for the case
where this was discovered (r318824). This handles more similar cases.
Review: Björn Petterson
https://reviews.llvm.org/D40339
llvm-svn: 319173
LLVM Coding Standards:
Function names should be verb phrases (as they represent actions), and
command-like function should be imperative. The name should be camel
case, and start with a lower case letter (e.g. openFile() or isFoo()).
Differential Revision: https://reviews.llvm.org/D40416
llvm-svn: 319168
Certain ARM implementations treat icache clear instruction as a memory read,
and CPU segfaults on trying to clear cache on !PROT_READ page.
We workaround this in Memory::protectMappedMemory by adding
PROT_READ to affected pages, clearing the cache, and then setting
desired protection.
This fixes "AllocationTests/MappedMemoryTest.***/3" unit-tests on
affected hardware.
Reviewers: psmith, zatrazz, kristof.beyls, lhames
Reviewed By: lhames
Subscribers: llvm-commits, krytarowski, peter.smith, jgreenhalgh, aemerson,
rengolin
Patch by maxim-kuvrykov!
Differential Revision: https://reviews.llvm.org/D40423
llvm-svn: 319166
The core idea is to (re-)introduce some redundancies where their cost is
hidden by the cost of materializing immediates for constant operands of
PHI nodes. When the cost of the redundancies is covered by this,
avoiding materializing the immediate has numerous benefits:
1) Less register pressure
2) Potential for further folding / combining
3) Potential for more efficient instructions due to immediate operand
As a motivating example, consider the remarkably different cost on x86
of a SHL instruction with an immediate operand versus a register
operand.
This pattern turns up surprisingly frequently, but is somewhat rarely
obvious as a significant performance problem.
The pass is entirely target independent, but it does rely on the target
cost model in TTI to decide when to speculate things around the PHI
node. I've included x86-focused tests, but any target that sets up its
immediate cost model should benefit from this pass.
There is probably more that can be done in this space, but the pass
as-is is enough to get some important performance on our internal
benchmarks, and should be generally performance neutral, but help with
more extensive benchmarking is always welcome.
One awkward part is that this pass has to be scheduled after
*everything* that can eliminate these kinds of redundancies. This
includes SimplifyCFG, GVN, etc. I'm open to suggestions about better
places to put this. We could in theory make it part of the codegen pass
pipeline, but there doesn't really seem to be a good reason for that --
it isn't "lowering" in any sense and only relies on pretty standard cost
model based TTI queries, so it seems to fit well with the "optimization"
pipeline model. Still, further thoughts on the pipeline position are
welcome.
I've also only implemented this in the new pass manager. If folks are
very interested, I can try to add it to the old PM as well, but I didn't
really see much point (my use case is already switched over to the new
PM).
I've tested this pretty heavily without issue. A wide range of
benchmarks internally show no change outside the noise, and I don't see
any significant changes in SPEC either. However, the size class
computation in tcmalloc is substantially improved by this, which turns
into a 2% to 4% win on the hottest path through tcmalloc for us, so
there are definitely important cases where this is going to make
a substantial difference.
Differential revision: https://reviews.llvm.org/D37467
llvm-svn: 319164
Summary:
I think we do not need to analyze debug intrinsics here, as they should
not impact codegen. This has 2 benefits: 1) slightly less work to do and
2) avoiding generating optimization remarks for converting calls to
debug intrinsics to tail calls, which are not really helpful for users.
Based on work by Sander de Smalen.
Reviewers: davide, trentxintong, aprantl
Reviewed By: aprantl
Subscribers: llvm-commits, JDevlieghere
Tags: #debug-info
Differential Revision: https://reviews.llvm.org/D40440
llvm-svn: 319158
Summary:
The entire algorithm operates per basic-block, so for cache locality
it should be better to re-optimize a basic-block immediately rather than
in a separate loop.
I don't have performance measurements.
Change-Id: I85106570bd623c4ff277faaa50ee43258e1ddcc5
Reviewers: arsenm, rampitec
Subscribers: kzhuravl, wdng, yaxunl, dstuttard, tpr, llvm-commits, t-tye
Differential Revision: https://reviews.llvm.org/D40344
llvm-svn: 319156
Summary:
The PeepholeOptimizer pass calls this function solely based on checking
DefMI->isMoveImmediate(), which only checks the MoveImm bit of the
instruction description. So it's up to FoldImmediate itself to properly
check that DefMI *actually* moves from an immediate.
I don't have a separate test case for this, but the next patch introduces
a test case which happens to crash without this change.
This error is caught by the assertion in MachineOperand::getImm().
Change-Id: I88e7cdbcf54d75e1a296822e6fe5f9a5f095bbf8
Reviewers: arsenm, rampitec
Subscribers: kzhuravl, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits
Differential Revision: https://reviews.llvm.org/D40342
llvm-svn: 319155
Currently, we use a set of pairs to cache responces like `CompareValueComplexity(X, Y) == 0`. If we had
proved that `CompareValueComplexity(S1, S2) == 0` and `CompareValueComplexity(S2, S3) == 0`,
this cache does not allow us to prove that `CompareValueComplexity(S1, S3)` is also `0`.
This patch replaces this set with `EquivalenceClasses` that merges Values into equivalence sets so that
any two values from the same set are equal from point of `CompareValueComplexity`. This, in particular,
allows us to prove the fact from example above.
Differential Revision: https://reviews.llvm.org/D40429
llvm-svn: 319153
The priorities in the section name suffixes are zero padded,
allowing the linker to just do a lexical sort.
Add zero padding for .ctors sections in ELF as well.
Differential Revision: https://reviews.llvm.org/D40407
llvm-svn: 319150
Currently, we use a set of pairs to cache responces like `CompareSCEVComplexity(X, Y) == 0`. If we had
proved that `CompareSCEVComplexity(S1, S2) == 0` and `CompareSCEVComplexity(S2, S3) == 0`,
this cache does not allow us to prove that `CompareSCEVComplexity(S1, S3)` is also `0`.
This patch replaces this set with `EquivalenceClasses` any two values from the same set are equal from
point of `CompareSCEVComplexity`. This, in particular, allows us to prove the fact from example above.
Differential Revision: https://reviews.llvm.org/D40428
llvm-svn: 319149
This is to address a problem similar to those in D37460 for Scalar PRE. We should not
PRE across an instruction that may not pass execution to its successor unless it is safe
to speculatively execute it.
Differential Revision: https://reviews.llvm.org/D38619
llvm-svn: 319147
Unoptimized IR can have linear sequences of stores to an array, where the
initial GEP for the first store is formed from the pointer to the array, and the
GEP for each store after the first is formed from the previous GEP with some
offset in an inductive fashion.
The (large) resulting DAG when analyzed by DAGCombine undergoes an excessive
number of combines as each store node is examined every time its' offset node
is combined with any child of the offset. One of the transformations is
findBetterNeighborChains which assists MergeConsecutiveStores. The former
relies on repeated chain walking to do its' work, however MergeConsecutiveStores
is disabled at O0 which makes the transformation redundant.
Any optimization level other than O0 would invoke InstCombine which would
resolve the chain of GEPs into flat base + offset GEP for each store which
does not exhibit the repeated examination of each store to the array.
Disabling this optimization fixes an excessive compile time issue (30~ minutes
for the test case provided) at O0.
Reviewers: niravd, craig.topper, t.p.northover
Differential Revision: https://reviews.llvm.org/D40193
llvm-svn: 319142
This fixes cases where we wouldn't perform various register operand
checks just because we didn't happen to have a definition in the
MCInstrDesc. This changes the code to only skip the tests that actually
depend on the MCInstrDesc definition.
This makes the machine verifier spot the problem from
https://llvm.org/PR33071 after the pass that actually caused it.
llvm-svn: 319141
Additional checks for phi operands:
- first operand should be a virtual register def. It should not be
tied, implicit, internalread, earlyclobber or a read.
- The other operands should be register/mbb operands next to each other
- The register operands should not be implicit, internalread,
earlyclobber, debug or tied.
- We can perform most of the PHI checks even for unreachable blocks.
llvm-svn: 319140
Revert "[SROA] Propagate !range metadata when moving loads."
Revert "[Mem2Reg] Clang-format unformatted parts of this file. NFCI."
Davide says they broke a bot.
llvm-svn: 319131
This adds code to protect WebAssembly's `trunc_s` family of opcodes
from values outside their domain. Even though such conversions have
full undefined behavior in C/C++, LLVM IR's `fptosi` and `fptoui` do
not, and only return undef.
This also implements the proposed non-trapping float-to-int conversion
feature and uses that instead when available.
llvm-svn: 319128
These lines all exist identically either under SSE2, AVX2 or AVX512. Given that VLX implies all of those, these aren't providing anything new.
llvm-svn: 319124
With AVX512 vXi1 types are legal so we shouldn't be extending them.
This change is similar to existing code in the zext(setcc) combine.
llvm-svn: 319120
Which VTs are considered simple is determined by the superset of the legal types of all targets in LLVM. If we're looking at VTs that are going to be split down to 512-bits we should allow any VT not just simple ones since the simple list changes over time as new targets are added.
llvm-svn: 319110
The previous implementation would only look 1 DW_AT_specification or DW_AT_abstract_origin deep. This means DWARFDie::getName() would fail in certain cases. I ran into such a case while creating a tool that used the LLVM DWARF parser to generate a symbolication format so I have seen this in the wild.
Differential Revision: https://reviews.llvm.org/D40156
llvm-svn: 319104
We don't do this for narrow vectors under AVX or SSE features. We also don't set them to Expand like we do for many vectors op. Nor does TargetLoweringBase.cpp. This leads me to believe these default to Legal.
llvm-svn: 319103
This tries to propagate !range metadata to a pre-existing load
when a load is optimized out. This is done instead of adding an
assume because converting loads to and from assumes creates a
lot of IR.
Patch by Ariel Ben-Yehuda.
Differential Revision: https://reviews.llvm.org/D37216
llvm-svn: 319096
This patch adds a peep hole optimization to remove any redundant toc save
instructions added as part of the call sequence for indirect calls. It removes
any toc saves within a function that are dominated by another toc save.
Differential Revision: https://reviews.llvm.org/D39736
llvm-svn: 319087
enum TailCallKind { TCK_None = 0, TCK_Tail = 1, TCK_MustTail = 2,
TCK_NoTail = 3 };
TCK_NoTail is greater than TCK_Tail so taking the min does not do the
correct thing.
rdar://35639547
llvm-svn: 319075
The existing library assumed that a stream's length would never
change. This makes some things simpler, but it's not flexible
enough for what we need, especially for writable streams where
what you really want is for each call to write to actually append.
llvm-svn: 319070
I don't have a good test case for this at the moment. I was playing around with a change in legalizing and triggered this code to produce a PSHUFD with sse1 only.
llvm-svn: 319066
Similar for vXi16/vXi8 with BWI.
Any vector larger than 512 bits will be split to 512 bits during legalization. But without this we will fold sexts with them before that making it difficult to recover leading to scalarization.
llvm-svn: 319059
This patch extends on to rL307174 to not use the power9 vector extract with
variable index instructions when extracting word element 1. For such cases,
the existing selection of MFVSRWZ provides a better sequence.
Differential Revision: https://reviews.llvm.org/D38287
llvm-svn: 319049
DWARF4 relative DW_AT_high_pc values are now displayed as absolute
addresses. The relative value is only shown when explicitly dumping the
forms, i.e. in show-form or verbose mode.
```
DW_AT_low_pc (0x0000000000000049)
DW_AT_high_pc (0x00000019)
```
becomes
```
DW_AT_low_pc (0x0000000000000049)
DW_AT_high_pc (0x0000000000000062)
```
Differential revision: https://reviews.llvm.org/D40317
rdar://35416943
llvm-svn: 319044
Summary:
Now that store-merge is only generates type-safe stores, do a second
pass just before instruction selection to allow lowered intrinsics to
be merged as well.
Reviewers: jyknight, hfinkel, RKSimon, efriedma, rnk, jmolloy
Subscribers: javed.absar, llvm-commits
Differential Revision: https://reviews.llvm.org/D33675
llvm-svn: 319036
Make the print format consistent with other assembler instructions.
Adding a tab character instead of space in asmstring of Ext and Ins
instructions.
Removing space around the tab character for JALRC and replacing space with
tab in JRC.
Patch by Milos Stojanovic.
Differential Revision: https://reviews.llvm.org/D38144
llvm-svn: 319030
AMDGPU backend errors with "unsupported call to function" upon
encountering a call to llvm.log{,10}.{f16,f32} intrinsics. This patch
adds custom lowering to avoid that error on both R600 and SI.
Reviewers: arsenm, jvesely
Subscribers: tstellar
Differential Revision: https://reviews.llvm.org/D29942
llvm-svn: 319025
The current way that trivial addressing modes are detected incorrectly thinks
that null pointers are non-trivial, leading to an infinite loop where we keep
duplicating the same select. Fix this by aware of null when deciding if an
addressing mode is trivial.
Differential Revision: https://reviews.llvm.org/D40447
llvm-svn: 319019
As mentioned on PR17367, many instructions are missing scheduling tags preventing us from setting 'CompleteModel = 1' for better instruction analysis. This patch deals with FMA/FMA4 which is one of the bigger offenders (along with AVX512 in general).
Annoyingly all scheduler models need to define WriteFMA (now that its actually used), even for older targets without FMA/FMA4 support, but that is an existing problem shared by other schedule classes.
Differential Revision: https://reviews.llvm.org/D40351
llvm-svn: 319016
The commit https://reviews.llvm.org/rL318143 computes incorrectly to offset to
restore LR from.
The number of tPOP operands is 2 (condition) + 2 (implicit def and use of SP) +
count of the popped registers. We need to load LR from just past the last
register, hence the correct offset should be either getNumOperands() - 4 and
getNumExplicitOperands() - 2 (multiplied by 4).
Differential revision: https://reviews.llvm.org/D40305
llvm-svn: 319014
Sanjoy Das