Summary:
This is a relatively simple scheme: we use the index emitted in the
bitcode to avoid loading all the global metadata. Instead we load
the index with their position in the bitcode so that we can load each
of them individually. Materializing the global metadata block in this
condition only triggers loading the named metadata, and the ones
referenced from there (transitively). When materializing a function,
metadata from the global block are loaded lazily as they are
referenced.
Two main current limitations are:
1) Global values other than functions are not materialized on demand,
so we need to eagerly load METADATA_GLOBAL_DECL_ATTACHMENT records
(and their transitive dependencies).
2) When we load a single metadata, we don't recurse on the operands,
instead we use a placeholder or a temporary metadata. Unfortunately
tepmorary nodes are very expensive. This is why we don't have it
always enabled and only for importing.
These two limitations can be lifted in a subsequent improvement if
needed.
With this change, the total link time of opt with ThinLTO and Debug
Info enabled is going down from 282s to 224s (~20%).
Reviewers: pcc, tejohnson, dexonsmith
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D28113
llvm-svn: 291027
If this is a problem for anyone (shared_ptr is two pointers in size,
whereas IntrusiveRefCntPtr is 1 - and the ref count control block that
make_shared adds is probably larger than the one int in RefCountedBase)
I'd prefer to address this by adding a lower-overhead version of
shared_ptr (possibly refactoring IntrusiveRefCntPtr into such a thing)
to avoid the intrusiveness - this allows memory ownership to remain
orthogonal to types and at least to me, seems to make code easier to
understand (since no implicit ownership acquisition can happen).
This recommits 291006, reverted in r291007.
llvm-svn: 291016
Summary:
When promoting fp-to-uint16 to fp-to-sint32, the result is actually zero
extended. For example, given double 65534.0, without legalization:
fp-to-uint16: 65534.0 -> 0xfffe
With the legalization:
fp-to-sint32: 65534.0 -> 0x0000fffe
Without this patch, legalization wrongly emits a signed extend assertion,
which is consumed by later icmp instruction, and cause miscompile.
Note that the floating point value must be in [0, 65535), otherwise the
behavior is undefined.
This patch reverts r279223 behavior and adds more tests and
documentations.
In PR29041's context, James Molloy mentioned that:
We don't need to mask because conversion from float->uint8_t is
undefined if the integer part of the float value is not representable in
uint8_t. Therefore we can assume this doesn't happen!
which is totally true and good, because fptoui is documented clearly to
have undefined behavior when overflow/underflow happens. We should take
the advantage of this behavior so that we can save unnecessary mask
instructions.
Reviewers: jmolloy, nadav, echristo, kbarton
Subscribers: mehdi_amini, nemanjai, llvm-commits
Differential Revision: https://reviews.llvm.org/D28284
llvm-svn: 291015
Summary:
Instead of matching:
(a + i) + 1 -> (a + i, undef, 1)
Now it matches:
(a + i) + 1 -> (a, i, 1)
Reviewers: rengolin
Differential Revision: http://reviews.llvm.org/D26367
From: Evgeny Stupachenko <evstupac@gmail.com>
llvm-svn: 291012
If this is a problem for anyone (shared_ptr is two pointers in size,
whereas IntrusiveRefCntPtr is 1 - and the ref count control block that
make_shared adds is probably larger than the one int in RefCountedBase)
I'd prefer to address this by adding a lower-overhead version of
shared_ptr (possibly refactoring IntrusiveRefCntPtr into such a thing)
to avoid the intrusiveness - this allows memory ownership to remain
orthogonal to types and at least to me, seems to make code easier to
understand (since no implicit ownership acquisition can happen).
llvm-svn: 291006
This change aims to unify and correct our logic for when we need to allow for
the possibility of the linker adding a TOC restoration instruction after a
call. This comes up in two contexts:
1. When determining tail-call eligibility. If we make a tail call (i.e.
directly branch to a function) then there is no place for the linker to add
a TOC restoration.
2. When determining when we need to add a nop instruction after a call.
Likewise, if there is a possibility that the linker might need to add a
TOC restoration after a call, then we need to put a nop after the call
(the bl instruction).
First problem: We were using similar, but different, logic to decide (1) and
(2). This is just wrong. Both the resideInSameModule function (used when
determining tail-call eligibility) and the isLocalCall function (used when
deciding if the post-call nop is needed) were supposed to be determining the
same underlying fact (i.e. might a TOC restoration be needed after the call).
The same logic should be used in both places.
Second problem: The logic in both places was wrong. We only know that two
functions will share the same TOC when both functions come from the same
section of the same object. Otherwise the linker might cause the functions to
use different TOC base addresses (unless the multi-TOC linker option is
disabled, in which case only shared-library boundaries are relevant). There are
a number of factors that can cause functions to be placed in different sections
or come from different objects (-ffunction-sections, explicitly-specified
section names, COMDAT, weak linkage, etc.). All of these need to be checked.
The existing logic only checked properties of the callee, but the properties of
the caller must also be checked (for example, calling from a function in a
COMDAT section means calling between sections).
There was a conceptual error in the resideInSameModule function in that it
allowed tail calls to functions with weak linkage and protected/hidden
visibility. While protected/hidden visibility does prevent the function
implementation from being replaced at runtime (via interposition), it does not
prevent the linker from using an alternate implementation at link time (i.e.
using some strong definition to replace the provided weak one during linking).
If this happens, then we're still potentially looking at a required TOC
restoration upon return.
Otherwise, in general, the post-call nop is needed wherever ELF interposition
needs to be supported. We don't currently support ELF interposition at the IR
level (see http://lists.llvm.org/pipermail/llvm-dev/2016-November/107625.html
for more information), and I don't think we should try to make it appear to
work in the backend in spite of that fact. Unfortunately, because of the way
that the ABI works, we need to generate code as if we supported interposition
whenever the linker might insert stubs for the purpose of supporting it.
Differential Revision: https://reviews.llvm.org/D27231
llvm-svn: 291003
This reapplies r289828 (reverted in r289833 as it broke the address sanitizer). The
debugloc is now only set when the instruction is not a call, as this causes the
verifier to assert (the inliner requires an inlinable callsite to have a debug loc
if the caller and callee have debug info).
Original commit message:
Simplify CFG will try to sink the last instruction in a series of basic blocks,
creating a "common" instruction in the successor block (sinkLastInstruction).
When it does this, the debug location of the single instruction should be the
merged debug locations of the commoned instructions.
Original review: https://reviews.llvm.org/D27590
llvm-svn: 290973
This CPU type was not previously recognized by LLVM which led to emitting
poor (and sometimes incorrect) code in some JIT workloads on such a machine.
llvm-svn: 290961
Summary:
In mergeSPUpdates, debug values need to be ignored when getting the
previous element, otherwise debug data could have an impact on codegen.
In eliminateCallFramePseudoInstr, debug values after the erased element
could have an impact on codegen and should be skipped.
Closes PR31319 (https://llvm.org/bugs/show_bug.cgi?id=31319)
Reviewers: aprantl, MatzeB, mkuper
Subscribers: gbedwell, llvm-commits
Differential Revision: https://reviews.llvm.org/D27688
llvm-svn: 290955
Summary:
The InlineSpiller was accessing the DominatorTreeBase directly
through the public data member DT in the MachineDominatorTree.
This is not a good idea as the "cached" information in
SplitCriticalEdges is not applied before the access.
The DominatorTreeBase must be accessed through the member
function getBase() in MachineDominatorTree.
The fault was introduced in r266162.
I think the public data member DT in the MachineDominatorTree
should have been made private in the original code (r215576)
that introduced the concept of lazily updating the
MachineDominatorTree information from
MachineBasicBlock::SplitCriticalEdge().
Patch by Karl-Johan Karlsson <karl-johan.karlsson@ericsson.com>
Reviewers: wmi, qcolombet
Subscribers: llvm-commits, bjope, uabelho
Differential Revision: https://reviews.llvm.org/D27983
llvm-svn: 290950
Replacing the memory operand in the intrinsic versions of the comis/ucomis instrucions from f128mem to ssmem/sdmem accordingly.
Differential Revision: https://reviews.llvm.org/D28138
llvm-svn: 290948
In some cases its more efficient to combine TRUNC( BINOP( X, Y ) ) --> BINOP( TRUNC( X ), TRUNC( Y ) ) if the binop is legal for the truncated types.
This is true for vector integer multiplication (especially vXi64), as well as ADD/AND/XOR/OR in cases where we only need to truncate one of the inputs at runtime (e.g. a duplicated input or an one use constant we can fold).
Further work could be done here - scalar cases (especially i64) could often benefit (if we avoid partial registers etc.), other opcodes, and better analysis of when truncating the inputs reduces costs.
I have considered implementing this for all targets within the DAGCombiner but wasn't sure we could devise a suitable cost model system that would give us the range we need.
Differential Revision: https://reviews.llvm.org/D28219
llvm-svn: 290947
We can perform the following:
(add (zext (add nuw X, C1)), C2) -> (zext (add nuw X, C1+C2))
This is only possible if C2 is negative and C2 is greater than or equal to negative C1.
llvm-svn: 290927
As per post-commit review for r289993 (D27775), we can only safely
import a type as a decl if it has an Identifier, as the Name alone
is not enough to be unique across modules.
llvm-svn: 290915
I wrote this patch before seeing the comment in:
https://reviews.llvm.org/D27114
...that suggests we should actually be canonicalizing the other way.
So just in case we decide this is the right way, we might as well
have a cleaner implementation.
llvm-svn: 290912
Use getReturnedArgOperand() instead of rolling our own. Note that it's
equivalent because there can only be one 'returned' operand.
The existing code was also incorrect: there already was awkward logic to
ignore callee/EH blocks, but operands can now also be operand bundles,
in which case we'll look for non-existent parameter attributes.
Unfortunately, this isn't observable in-tree, as it only crashes when
exercising the regular call lowering logic with operand bundles.
Still, this is a nice small cleanup anyway.
llvm-svn: 290905
Provide a distinct contents for semBogus and semPPCDoubleDouble in order
to prevent compilers from collapsing them to a single memory address,
while we heavily rely on every semantic having distinct address.
This happens if insecure optimization collapsing identical values is
enabled. As a result, APFloats of semBogus are indistinguishable from
semPPCDoubleDouble -- and whenever the move constructor is used, the old
value beings being incorrectly recognized as a semPPCDoubleDouble.
Since the values in semPPCDoubleDouble are not used anywhere,
we can easily solve this issue via altering the value of one of the
fields and therefore ensuring that the collapse can not occur.
Differential Revision: https://reviews.llvm.org/D28112
llvm-svn: 290896
Summary:
No need to have this per-architecture. While there, unify 32-bit ARM's
behaviour with what changed elsewhere and start function names lowercase
as per the coding standards. Individual entry emission code goes to the
entry's own class.
Fully tested on amd64, cross-builds on both ARMs and PowerPC.
Reviewers: dberris
Subscribers: aemerson, llvm-commits
Differential Revision: https://reviews.llvm.org/D28209
llvm-svn: 290858
Summary:
Regardless how the loop body weight is distributed, we should preserve
total loop body weight. i.e. we should have same weight reaching the body of the loop
or its duplicates in peeled and unpeeled case.
Reviewers: mkuper, davidxl, anemet
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D28179
llvm-svn: 290833
Apparently my suggestion of using ternary doesn't really work
as clang complains about incompatible types on LHS and RHS. Some
GCC versions happen to accept the code but clang behaviour is
correct here.
llvm-svn: 290822
Add an explicit LLVM_ENABLE_DIA_SDK option to control building support
for DIA SDK-based debugging. Control its value to match whether DIA SDK
support was found and expose it in LLVMConfig (alike LLVM_ENABLE_ZLIB).
Its value is needed for LLDB to determine whether to run tests requiring
DIA support. Currently it is obtained from llvm/Config/config.h;
however, this file is not available for standalone builds. Following
this change, LLDB will be modified to use the value from LLVMConfig.
Differential Revision: https://reviews.llvm.org/D26255
llvm-svn: 290818
GNU as rejects input where .cfi_sections is used after .cfi_startproc,
if the new section differs from the old. Adjust our output to always
emit .cfi_sections before the first .cfi_startproc to minimize necessary
code.
Differential Revision: https://reviews.llvm.org/D28011
llvm-svn: 290817
Summary:
This avoids the very fragile code for null expressions. We could also use a denseset that tracks which things have null expressions instead, but that seems pretty fragile and premature optimization.
This resolves a number of infinite loop cases, test reductions coming.
Reviewers: davide
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D28193
llvm-svn: 290816
Summary: Previously, we tried to fix up the equivalences during symbolic evaluation. This does not work. Now, we change the equivalences during congruence finding, where it belongs. We also initialize the equivalence table to give a maximal answer.
Reviewers: davide
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D28192
llvm-svn: 290815
X86 target does not provide any target specific cost calculation for interleave patterns.It uses the common target-independent calculation, which gives very high numbers. As a result, the scalar version is chosen in many cases. The situation on AVX-512 is even worse, since we have 3-src shuffles that significantly reduce the cost.
In this patch I calculate the cost on AVX-512. It will allow to compare interleave pattern with gather/scatter and choose a better solution (PR31426).
* Shiffle-broadcast cost will be changed in Simon's upcoming patch.
Differential Revision: https://reviews.llvm.org/D28118
llvm-svn: 290810
Summary:
`PromotedFloats` needs to be checked in
`DAGTypeLegalizer::PerformExpensiveChecks`. This patch fixes a few type
legalization failures with expansive checks for ARM fp16 tests.
Reviewers: baldrick, bogner, arsenm
Subscribers: arsenm, aemerson, llvm-commits
Differential Revision: https://reviews.llvm.org/D28187
llvm-svn: 290796
I'm not sure if this was intentional, but today
isGuaranteedToTransferExecutionToSuccessor returns true for readonly and
argmemonly calls that may throw. This commit changes the function to
not implicitly infer nounwind this way.
Even if we eventually specify readonly calls as not throwing,
isGuaranteedToTransferExecutionToSuccessor is not the best place to
infer that. We should instead teach FunctionAttrs or some other such
pass to tag readonly functions / calls as nounwind instead.
llvm-svn: 290794
I don't think this hole is currently exposed, but I crashed regression tests for
jump-threading and loop-vectorize after I added calls to isKnownNonNullAt() in
InstSimplify as part of trying to solve PR28430:
https://llvm.org/bugs/show_bug.cgi?id=28430
That's because they call into value tracking with a context instruction, but no
other parts of the query structure filled in.
For more background, see the discussion in:
https://reviews.llvm.org/D27855
llvm-svn: 290786
This was originally motivated by a compile time problem I've since figured out how to solve differently, but the cleanup seemed useful. We had the same logic - which essentially implemented find - in several places. By commoning them out, I can implement find and allow erase to be inlined at the call sites if profitable.
Differential Revision: https://reviews.llvm.org/D28183
llvm-svn: 290779
Summary:
gep 0, 0 is equivalent to bitcast. LLVM canonicalizes it
to getelementptr because it make SROA can then handle it.
Simple case like
void g(A &a) {
z(a);
if (glob)
a.foo();
}
void testG() {
A a;
g(a);
}
was not devirtualized with -fstrict-vtable-pointers because luck of
handling for gep 0 in Memory Dependence Analysis
Reviewers: dberlin, nlewycky, chandlerc
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D28126
llvm-svn: 290763
CVP doesn't care about the order of blocks visited, but by using a pre-order traversal over the graph we can a) not visit unreachable blocks and b) optimize as we go so that analysis of later blocks produce slightly more precise results.
I noticed this via inspection and don't have a concrete example which points to the issue.
llvm-svn: 290760
This is similar to the allocfn case - if an alloca is not captured, then it's
necessarily thread-local.
Differential Revision: https://reviews.llvm.org/D28170
llvm-svn: 290738
Summary:
The current loop complete unroll algorithm checks if unrolling complete will reduce the runtime by a certain percentage. If yes, it will apply a fixed boosting factor to the threshold (by discounting cost). The problem for this approach is that the threshold abruptly. This patch makes the boosting factor a function of runtime reduction percentage, capped by a fixed threshold. In this way, the threshold changes continuously.
The patch also simplified the code by reducing one parameter in UP.
The patch only affects code-gen of two speccpu2006 benchmark:
445.gobmk binary size decreases 0.08%, no performance change.
464.h264ref binary size increases 0.24%, no performance change.
Reviewers: mzolotukhin, chandlerc
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D26989
llvm-svn: 290737
"Changed" doesn't actually change within the loop, so there's
no reason to keep track of it - we always return false during
analysis and true after the transformation is made.
llvm-svn: 290735
We correctly canonicalized (add (sext x), (sext y)) to (sext (add x, y))
where possible. However, we didn't perform the same canonicalization
for zexts or for muls.
llvm-svn: 290733
This moves the exit block and insertion point computation to be eager,
instead of after seeing the first scalar we can promote.
The cost is relatively small (the computation happens anyway, see discussion
on D28147), and the code is easier to follow, and can bail out earlier
if there's a catchswitch present.
llvm-svn: 290729
We would check whether we have a prehader *or* dedicated exit blocks,
and go into the promotion loop. Then, for each alias set we'd check
if we have a preheader *and* dedicated exit blocks, and bail if not.
Instead, bail immediately if we don't have both.
llvm-svn: 290728
We want to recompute LCSSA only when we actually promoted a value.
This means we only need to look at changes made by promotion when
deciding whether to recompute it or not, not at regular sinking/hoisting.
(This was what the code was documented as doing, just not what it did)
Hopefully NFC.
llvm-svn: 290726
Summary:
This class is unnecessary.
Its comment indicated that it was a compile error to allocate an
instance of a class that inherits from RefCountedBaseVPTR on the stack.
This may have been true at one point, but it's not today.
Moreover you really do not want to allocate *any* refcounted object on
the stack, vptrs or not, so if we did have a way to prevent these
objects from being stack-allocated, we'd want to apply it to regular
RefCountedBase too, obviating the need for a separate RefCountedBaseVPTR
class.
It seems that the main way RefCountedBaseVPTR provides safety is by
making its subclass's destructor virtual. This may have been helpful at
one point, but these days clang will emit an error if you define a class
with virtual functions that inherits from RefCountedBase but doesn't
have a virtual destructor.
Reviewers: compnerd, dblaikie
Subscribers: cfe-commits, klimek, llvm-commits, mgorny
Differential Revision: https://reviews.llvm.org/D28162
llvm-svn: 290717
This reverts commit r290694. It broke sanitizer tests on Win64. I'll
probably bring this back, but the jump tables will just live in .text
like they do for MSVC.
llvm-svn: 290714
This fixes the issue exposed in PR31393, where we weren't trying
sufficiently hard to diagnose bad TBAA metadata.
This does reduce the variety in the error messages we print out, but I
think the tradeoff of verifying more, simply and quickly overrules the
need for more helpful error messags here.
llvm-svn: 290713
Among other stuff, this allows to use predefined .option.machine_version_major
/minor/stepping symbols in the directive.
Relevant test expanded at once (also file renamed for clarity).
Differential Revision: https://reviews.llvm.org/D28140
llvm-svn: 290710
This change adds a new intrinsic which is intended to provide memcpy functionality
with additional atomicity guarantees. Please refer to the review thread
or language reference for further details.
Differential Revision: https://reviews.llvm.org/D27133
llvm-svn: 290708
We bypassed the intrinsic and returned the passthru operand, but we should also add the intrinsic to the worklist since its now dead. This can allow DCE to find it sooner and remove it. Similar was done for InsertElement when the inserted element isn't demanded.
llvm-svn: 290704
Summary:
The optimal iteration order for this problem is RPO order. We want to
process as many preds of a backedge as we can before we process the
backedge.
At the same time, as we add predicate handling, we want to be able to
touch instructions that are dominated by a given block by
ranges (because a change in value numbering a predicate possibly
affects all users we dominate that are using that predicate).
If we don't do it this way, we can't do value inference over
backedges (the paper covers this in depth).
The newgvn branch currently overshoots the last part, and guarantees
that it will touch *at least* the right set of instructions, but it
does touch more. This is because the bitvector instruction ranges are
currently generated in RPO order (so we take the max and the min of
the ranges of dominated blocks, which means there are some in the
middle we didn't have to touch that we did).
We can do better by sorting the dominator tree, and then just using
dominator tree order.
As a preliminary, the dominator tree has some RPO guarantees, but not
enough. It guarantees that for a given node, your idom must come
before you in the RPO ordering. It guarantees no relative RPO ordering
for siblings. We add siblings in whatever order they appear in the module.
So that is what we fix.
We sort the children array of the domtree into RPO order, and then use
the dominator tree for ordering, instead of RPO, since the dominator
tree is now a valid RPO ordering.
Note: This would help any other pass that iterates a forward problem
in dominator tree order. Most of them are single pass. It will still
maximize whatever result they compute. We could also build the
dominator tree in this order, but our incremental updates would still
put it out of sort order, and recomputing the sort order is almost as
hard as general incremental updates of the domtree.
Also note that the sorting does not affect any tests, etc. Nothing
depends on domtree order, including the verifier, the equals
functions for domtree nodes, etc.
How much could this matter, you ask?
Here are the current numbers.
This is generated by running NewGVN over all files in LLVM.
Note that once we propagate equalities, the differences go up by an
order of magnitude or two (IE instead of 29, the max ends up in the
thousands, since the worst case we add a factor of N, where N is the
number of branch predicates). So while it doesn't look that stark for
the default ordering, it gets *much much* worse. There are also
programs in the wild where the difference is already pretty stark
(2 iterations vs hundreds).
RPO ordering:
759040 Number of iterations is 1
112908 Number of iterations is 2
Default dominator tree ordering:
755081 Number of iterations is 1
116234 Number of iterations is 2
603 Number of iterations is 3
27 Number of iterations is 4
2 Number of iterations is 5
1 Number of iterations is 7
Dominator tree sorted:
759040 Number of iterations is 1
112908 Number of iterations is 2
<yay!>
Really bad ordering (sort domtree siblings in postorder. not quite the
worst possible, but yeah):
754008 Number of iterations is 1
21 Number of iterations is 10
8 Number of iterations is 11
6 Number of iterations is 12
5 Number of iterations is 13
2 Number of iterations is 14
2 Number of iterations is 15
3 Number of iterations is 16
1 Number of iterations is 17
2 Number of iterations is 18
96642 Number of iterations is 2
1 Number of iterations is 20
2 Number of iterations is 21
1 Number of iterations is 22
1 Number of iterations is 29
17266 Number of iterations is 3
2598 Number of iterations is 4
798 Number of iterations is 5
273 Number of iterations is 6
186 Number of iterations is 7
80 Number of iterations is 8
42 Number of iterations is 9
Reviewers: chandlerc, davide
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D28129
llvm-svn: 290699
I added one for Value back in r262045, and I'm starting to think we
should have these for any class with bitfields whose memory efficiency
really matters.
llvm-svn: 290698
Summary:
We were already using 32-bit jump table entries, but this was a
consequence of the default PIC model on Win64, and not an intentional
design decision. This patch ensures that we always use 32-bit label
difference jump table entries on Win64 regardless of the PIC model. This
is a good idea because it saves executable size and object file size.
Moving the jump tables to .rdata cleans up the disassembled object code
and reduces the available ROP targets, but it requires adding one more
RIP-relative lea to the code. COFF doesn't have relocations to express
the difference between two arbitrary symbols, so we can't use the jump
table label in the label difference like we do elsewhere.
Fixes PR31488
Reviewers: majnemer, compnerd
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D28141
llvm-svn: 290694
The Bitstream reader and writer are limited to handle a "size_t" at
most, which means that we can't backpatch and read back a 64bits
value on 32 bits platform.
llvm-svn: 290693
Summary:
Previously isIntrinsic() called getName(). This involves a hashtable
lookup, so is nontrivially expensive. And isIntrinsic() is called
frequently, particularly by dyn_cast<IntrinsicInstr>.
This patch steals a bit of IntID and uses that to store whether or not
getName() starts with "llvm."
Reviewers: bogner, arsenm, joker-eph
Subscribers: sanjoy, llvm-commits
Differential Revision: https://reviews.llvm.org/D22949
llvm-svn: 290691
This index record the position for each metadata record in
the bitcode, so that the reader will be able to lazy-load
on demand each individual record.
We also make sure that every abbrev is emitted upfront so
that the block can be skipped while reading.
I don't plan to commit this before having the reader
counterpart, but I figured this can be reviewed mostly
independently.
Recommit r290684 (was reverted in r290686 because a test
was broken) after adding a threshold to avoid emitting
the index when unnecessary (little amount of metadata).
This optimization "hides" a limitation of the ability
to backpatch in the bitstream: we can only backpatch
safely when the position has been flushed. So if we emit
an index for one metadata, it is possible that (part of)
the offset placeholder hasn't been flushed and the backpatch
will fail.
Differential Revision: https://reviews.llvm.org/D28083
llvm-svn: 290690
emplace_back is not faster if it is equivalent to push_back. In this cases emplaced value had the
same type that the one stored in container. It is ugly and it might be even slower (see
Scott Meyers presentation about emplacement).
llvm-svn: 290685
Summary:
This index record the position for each metadata record in
the bitcode, so that the reader will be able to lazy-load
on demand each individual record.
We also make sure that every abbrev is emitted upfront so
that the block can be skipped while reading.
I don't plan to commit this before having the reader
counterpart, but I figured this can be reviewed mostly
independently.
Reviewers: pcc, tejohnson
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D28083
llvm-svn: 290684
Jump table emission can switch to .rdata before
WinException::endFunction gets called. Just remember the appropriate
text section we started in and reset back to it when we end the
function. We were already switching sections back from .xdata anyway.
Fixes the first problem in PR31488, so that now COFF switch tables can
live in .rdata if we want them to.
llvm-svn: 290678
This is an orthogonal and separated layer instead of being embedded
inside the pass manager. While it adds a small amount of complexity, it
is fairly minimal and the composability and control seems worth the
cost.
The logic for this ends up being nicely isolated and targeted. It should
be easy to experiment with different iteration strategies wrapped around
the CGSCC bottom-up walk using this kind of facility.
The mechanism used to track devirtualization is the simplest one I came
up with. I think it handles most of the cases the existing iteration
machinery handles, but I haven't done a *very* in depth analysis. It
does however match the basic intended semantics, and we can tweak or
tune its exact behavior incrementally as necessary. One thing that we
may want to revisit is freshly building the value handle set on each
iteration. While I don't think this will be a significant cost (it is
strictly fewer value handles but more churn of value handes than the old
call graph), it is conceivable that we'll want a somewhat more clever
tracking mechanism. My hope is to layer that on as a follow up patch
with data supporting any implementation complexity it adds.
This code also provides for a basic count heuristic: if the number of
indirect calls decreases and the number of direct calls increases for
a given function in the SCC, we assume devirtualization is responsible.
This matches the heuristics currently used in the legacy pass manager.
Differential Revision: https://reviews.llvm.org/D23114
llvm-svn: 290665
analyses when we're about to break apart an SCC.
We can't wait until after breaking apart the SCC to invalidate things:
1) Which SCC do we then invalidate? All of them?
2) Even if we invalidate all of them, a newly created SCC may not have
a proxy that will convey the invalidation to functions!
Previously we only invalidated one of the SCCs and too late. This led to
stale analyses remaining in the cache. And because the caching strategy
actually works, they would get used and chaos would ensue.
Doing invalidation early is somewhat pessimizing though if we *know*
that the SCC structure won't change. So it turns out that the design to
make the mutation API force the caller to know the *kind* of mutation in
advance was indeed 100% correct and we didn't do enough of it. So this
change also splits two cases of switching a call edge to a ref edge into
two separate APIs so that callers can clearly test for this and take the
easy path without invalidating when appropriate. This is particularly
important in this case as we expect most inlines to be between functions
in separate SCCs and so the common case is that we don't have to so
aggressively invalidate analyses.
The LCG API change in turn needed some basic cleanups and better testing
in its unittest. No interesting functionality changed there other than
more coverage of the returned sequence of SCCs.
While this seems like an obvious improvement over the current state, I'd
like to revisit the core concept of invalidating within the CG-update
layer at all. I'm wondering if we would be better served forcing the
callers to handle the invalidation beforehand in the cases that they
can handle it. An interesting example is when we want to teach the
inliner to *update and preserve* analyses. But we can cross that bridge
when we get there.
With this patch, the new pass manager an build all of the LLVM test
suite at -O3 and everything passes. =D I haven't bootstrapped yet and
I'm sure there are still plenty of bugs, but this gives a nice baseline
so I'm going to increasingly focus on fleshing out the missing
functionality, especially the bits that are just turned off right now in
order to let us establish this baseline.
llvm-svn: 290664
There are cases of AVX-512 instructions that have two possible encodings. This is the case with instructions that use vector registers with low indexes of 0 - 15 and do not use the zmm registers or the mask k registers.
The EVEX encoding prefix requires 4 bytes whereas the VEX prefix can take only up to 3 bytes. Consequently, using the VEX encoding for these instructions results in a code size reduction of ~2 bytes even though it is compiled with the AVX-512 features enabled.
Reviewers: Craig Topper, Zvi Rackoover, Elena Demikhovsky
Differential Revision: https://reviews.llvm.org/D27901
llvm-svn: 290663
when they are call edges at the leaf but may (transitively) be reached
via ref edges.
It turns out there is a simple rule: insert everything as a ref edge
which is a safe conservative default. Then we let the existing update
logic handle promoting some of those to call edges.
Note that it would be fairly cheap to make these call edges right away
if that is desirable by testing whether there is some existing call path
from the source to the target. It just seemed like slightly more
complexity in this code path that isn't strictly necessary. If anyone
feels strongly about handling this differently I'm happy to change it.
llvm-svn: 290649
due to a call cycle.
This actually crashed the ref removal before.
I've added a unittest that covers this kind of interesting graph
structure and mutation.
llvm-svn: 290645
currenty relies on the old PM's dependency system forming LCSSA.
The new PM will require a different design for this, and for now this is
causing most of the issues I'm currently seeing in testing. I'd like to
get to a testable baseline and then work on re-enabling things one at
a time.
llvm-svn: 290644
This adds a combine that canonicalizes a chain of inserts which broadcasts
a value into a single insert + a splat shufflevector.
This fixes PR31286.
Differential Revision: https://reviews.llvm.org/D27992
llvm-svn: 290641
There is no need to do this within an analysis. That method shouldn't
even be reached if this predicate holds as the actual useful
optimization is in the analysis manager itself.
llvm-svn: 290614
The effect of the bug was that we would incorrectly create summaries
for global and weak values defined in module asm (since we were
essentially testing for bit 1 which is SF_Undefined, and the
RecordStreamer ignores local undefined references). This would have
resulted in conservatively disabling importing of anything referencing
globals and weaks defined in module asm. Added these cases to the test
which now fails without this bug fix.
Fixes PR31459.
llvm-svn: 290610
The feature allows for conditional assembly, filling the entries
of .amd_kernel_code_t etc.
Symbols are defined with value 0 at the beginning of each kernel scope.
After each register usage, the respective symbol is set to:
value = max( value, ( register index + 1 ) )
Thus, at the end of scope the value represents a count of used registers.
Kernel scopes begin at .amdgpu_hsa_kernel directive, end at the
next .amdgpu_hsa_kernel (or EOF, whichever comes first). There is also
dummy scope that lies from the beginning of source file til the
first .amdgpu_hsa_kernel.
Test added.
Differential Revision: https://reviews.llvm.org/D27859
llvm-svn: 290608
Because operand was not marked as seen it was visited twice.
It doesn't change behavior of optimization, it just saves redudant
visit, so no test changes.
llvm-svn: 290607
This requires custom handling because BasicAA caches handles to other
analyses and so it needs to trigger indirect invalidation.
This fixes one of the common crashes when using the new PM in real
pipelines. I've also tweaked a regression test to check that we are at
least handling the most immediate case.
I'm going to work at re-structuring this test some to both scale better
(rather than all being in one file) and check more invalidation paths in
a follow-up commit, but I wanted to get the basic bug fix in place.
llvm-svn: 290603
not really wired into the loop pass manager in a way that will let us
productively use these passes yet.
This lets the new PM get farther in basic testing which is useful for
establishing a good baseline of "doesn't explode". There are still
plenty of crashers in basic testing though, this just gets rid of some
noise that is well understood and not representing a specific or narrow
bug.
llvm-svn: 290601
inter-analysis dependencies) to use the new invalidation infrastructure.
This teaches it to invalidate itself when any of the peer function
AA results that it uses become invalid. We do this by just tracking the
originating IDs. I've kept it in a somewhat clunky API since some users
of AAResults are outside the new PM right now. We can clean this API up
if/when those users go away.
Secondly, it uses the registration on the outer analysis manager proxy
to trigger deferred invalidation when a module analysis result becomes
invalid.
I've included test cases that specifically try to trigger use-after-free
in both of these cases and they would crash or hang pretty horribly for
me even without ASan. Now they work nicely.
The `InvalidateAnalysis` utility pass required some tweaking to be
useful in this context and it still is pretty garbage. I'd like to
switch it back to the previous implementation and teach the explicit
invalidate method on the AnalysisManager to take care of correctly
triggering indirect invalidation, but I wanted to go ahead and send this
out so folks could see how all of this stuff works together in practice.
And, you know, that it does actually work. =]
Differential Revision: https://reviews.llvm.org/D27205
llvm-svn: 290595
that require deferred invalidation.
This handles the other real-world invalidation scenario that we have
cases of: a function analysis which caches references to a module
analysis. We currently do this in the AA aggregation layer and might
well do this in other places as well.
Since this is relative rare, the technique is somewhat more cumbersome.
Analyses need to register themselves when accessing the outer analysis
manager's proxy. This proxy is already necessarily present to allow
access to the outer IR unit's analyses. By registering here we can track
and trigger invalidation when that outer analysis goes away.
To make this work we need to enhance the PreservedAnalyses
infrastructure to support a (slightly) more explicit model for "sets" of
analyses, and allow abandoning a single specific analyses even when
a set covering that analysis is preserved. That allows us to describe
the scenario of preserving all Function analyses *except* for the one
where deferred invalidation has triggered.
We also need to teach the invalidator API to support direct ID calls
instead of always going through a template to dispatch so that we can
just record the ID mapping.
I've introduced testing of all of this both for simple module<->function
cases as well as for more complex cases involving a CGSCC layer.
Much like the previous patch I've not tried to fully update the loop
pass management layer because that layer is due to be heavily reworked
to use similar techniques to the CGSCC to handle updates. As that
happens, we'll have a better testing basis for adding support like this.
Many thanks to both Justin and Sean for the extensive reviews on this to
help bring the API design and documentation into a better state.
Differential Revision: https://reviews.llvm.org/D27198
llvm-svn: 290594
skipping indirectly recursive inline chains.
To do this, we implicitly build an inline stack for each callsite and
check prior to inlining that doing so would not form a cycle. This uses
the exact same technique and even shares some code with the legacy PM
inliner.
This solution remains deeply unsatisfying to me because it means we
cannot actually iterate the inliner externally. Doing so would not be
able to easily detect and avoid such cycles. Some day I would very much
like to have a solution that works without this internal state to detect
cycles, but this is not that day.
llvm-svn: 290590
We currently ignore the `allocsize` attribute on functions calls with
the `nobuiltin` attribute when trying to lower `@llvm.objectsize`. We
shouldn't care about `nobuiltin` here: `allocsize` is explicitly added
by the user, not inferred based on a function's symbol.
llvm-svn: 290588
This also makes us no longer check for `allocsize` on intrinsic calls.
This shouldn't matter, since intrinsics should provide the information
we get from `allocsize` on their own.
llvm-svn: 290585
PMULDQ/PMULUDQ vXi64 instructions only use the even numbered v2Xi32 input elements which SimplifyDemandedVectorElts should try and use.
This builds on r290554 which added supported for 128 and 256-bit.
llvm-svn: 290582
The 128 and 256 bit masked intrinsics are currently unused by clang. The sse and avx2 unmasked intrinsics are used instead. The new 512-bit intrinsic will be used to do the same. Then all masked versions will removed and autoupgraded.
llvm-svn: 290573
An earlier commit added support for unmasked scalar operations. At that time isel wouldn't generate an optimal sequence for masked operations, but that has now been fixed.
llvm-svn: 290566
inside of `InlineFunction`. Prior to this, call instructions are
specifically being rewritten and replaced within the inlined region,
invalidating some of the call sites.
Several of these regions are using the same technique to walk the
inlined region so this seems clearly safe up to this point.
I've also added a short circuit to the scan for call sites based on what
other code is doing.
With this, the most common crash I've found in the new inliner code is
fixed. I've turned it on for another test case that covers this
scenario.
I'll make my way through most of the other inliner test cases
just to get some easy coverage next.
llvm-svn: 290562
removing fully-dead comdats without removing dead entries in comdats
with live members.
This factors the core logic out of the current inliner's internals to
a reusable utility and leverages that in both places. The factored out
code should also be (minorly) more efficient in cases where we have very
few dead functions or dead comdats to consider.
I've added a test case to cover this behavior of the always inliner.
This is the last significant bug in the new PM's always inliner I've
found (so far).
llvm-svn: 290557
PMULDQ/PMULUDQ vXi64 instructions only use the even numbered v2Xi32 input elements which SimplifyDemandedVectorElts should try and use.
Differential Revision: https://reviews.llvm.org/D28119
llvm-svn: 290554
Mostly use a bit more idiomatic C++ where we can,
so we can combine some things later.
Reviewers: davide
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D28111
llvm-svn: 290550
Summary:
I only do this for unmasked cases for now because isel is failing to fold the mask. I'll try to fix that soon.
I'll do the same thing for packed add/sub/mul/div in a future patch.
Reviewers: delena, RKSimon, zvi, craig.topper
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D27879
llvm-svn: 290535
Summary:
This patch adds support for converting the masked vpermv intrinsics into shufflevector instructions if the indices are constants.
We also need to wrap a select instruction around the shuffle to take care of the masking part. InstCombine will take care of optimizing the select if the mask is constant so I didn't bother checking for that.
Reviewers: zvi, delena, spatel, RKSimon
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D27825
llvm-svn: 290530
multiple asynchronous RPC calls.
ParallelCallGroup allows multiple asynchronous calls to be dispatched,
and provides a wait method that blocks until all asynchronous calls have
been executed on the remote and all return value handlers run on the
local machine.
This will allow, for example, the JIT client to issue memory allocation calls
for all sections in parallel, then block until all memory has been allocated
on the remote and the allocated addresses registered with the client, at which
point the JIT client can proceed to applying relocations.
llvm-svn: 290523
This makes it explicit what is the exact list to handle, and it
looks much more easy to manipulate and understand that the
previous custom tracking of min/max to express the range where
to look for.
Differential Revision: https://reviews.llvm.org/D28089
llvm-svn: 290507
whether functions are removed, and fix the new PM's always inliner to
actually pass this test.
Without this, the new PM's always inliner leaves all the functions
kicking around which won't work out very well given the semantics of
always inline.
Doing this really highlights how frustrating the current alwaysinline
semantic contract is though -- why can we put it on *external*
functions, etc?
Also I've added a number of tricky and interesting test cases for
removing functions with the always inliner. There is one remaining case
not handled -- fully removing comdats -- and I've left a FIXME about
this.
llvm-svn: 290457
Pretty boring and lame as-is but necessary. This is definitely a place
we'll end up with extension hooks longer term. =]
Differential Revision: https://reviews.llvm.org/D28076
llvm-svn: 290449
The pass creates some state which expects to be cleaned up by
a later instance of the same pass. opt-bisect happens to expose
this not ideal design because calling skipLoop() will result in
this state not being cleaned up at times and an assertion firing
in `doFinalization()`. Chandler tells me the new pass manager will
give us options to avoid these design traps, but until it's not ready,
we need a workaround for the current pass infrastructure. Fix provided
by Andy Kaylor, see the review for a complete discussion.
Differential Revision: https://reviews.llvm.org/D25848
llvm-svn: 290427
According to the Cortex-A57 doc, FDIV/FSQRT instructions should use F0 unit
(W-unit in AArch64SchedA57.td, the same as cryptography instructions),
not F1 unit (X-unit in td, like ASIMD absolute diff accum SABA/UABA).
This patch changes FDIV/FSQRT scheduling declarations to use A57UnitW
instead of A57UnitX. Also, latencies for those instructions are
corrected.
Patch by Andrew Zhogin.
llvm-svn: 290426
Summary:
In mergeSPUpdates, debug values need to be ignored when getting the
previous element, otherwise debug data could have an impact on codegen.
In eliminateCallFramePseudoInstr, debug values after the erased element
could have an impact on codegen and should be skipped.
Closes PR31319 (https://llvm.org/bugs/show_bug.cgi?id=31319)
Reviewers: mkuper, MatzeB, aprantl
Subscribers: gbedwell, llvm-commits
Differential Revision: https://reviews.llvm.org/D27688
llvm-svn: 290423
1.Fix pessimized case in FIXME.
2.Add tests for it.
3.The canonicalisation on shifts results in different sequence for
tests of machine-licm.Correct some check lines.
Differential Revision: https://reviews.llvm.org/D27916
llvm-svn: 290410
This patch fixes some ASAN unittest failures on FreeBSD. See the
cfe-commits email thread for r290169 for more on those.
According to the LangRef, the allocsize attribute only tells us about
the number of bytes that exist at the memory location pointed to by the
return value of a function. It does not necessarily mean that the
function will only ever allocate. So, we need to be very careful about
treating functions with allocsize as general allocation functions. This
patch makes us fully conservative in this regard, though I suspect that
we have room to be a bit more aggressive if we want.
This has a FIXME that can be fixed by a relatively straightforward
refactor; I just wanted to keep this patch minimal. If this sticks, I'll
come back and fix it in a few days.
llvm-svn: 290397
Summary:
This change rewrites a core component in the ImplicitNullChecks pass for
greater simplicity since the original design was over-complicated for no
good reason. Please review this as essentially a new pass. The change
is almost NFC and I've added a test case for a scenario that this new
code handles that wasn't handled earlier.
The implicit null check pass, at its core, is a code hoisting transform.
It differs from "normal" code transforms in that it speculates
potentially faulting instructions (by design), but a lot of the usual
hazard detection logic (register read-after-write etc.) still applies.
We previously detected hazards by keeping track of registers defined and
used by machine instructions over an instruction range, but that was
unwieldy and did not actually confer any performance benefits. The
intent was to have linear time complexity over the number of machine
instructions considered, but it ended up being N^2 is practice.
This new version is more obviously O(N^2) (with N capped to 8 by
default) in hazard detection. It does not attempt to be clever in
tracking register uses or defs (the previous cleverness here was a
source of bugs).
Once this is checked in, I'll extract out the `IsSuitableMemoryOp` and
`CanHoistLoadInst` lambda into member functions (they're too complicated
to be inline lambdas) and do some other related NFC cleanups.
Reviewers: reames, anna, atrick
Subscribers: mcrosier, llvm-commits
Differential Revision: https://reviews.llvm.org/D27592
llvm-svn: 290394
This patch adds support for YAML<->DWARF for debug_info sections.
This re-lands r290147, reverted in 290148, re-landed in r290204 after fixing the issue that caused bots to fail (thank you UBSan!), and reverted again in r290209 due to failures on big endian systems.
After adding support for preserving endianness, this should be good now.
llvm-svn: 290386
Use a dummy private function with inline asm calls instead of module
level asm blocks for CFI jumptables.
The main advantage is that now jumptable codegen can be affected by
the function attributes (like target_cpu on ARM). Module level asm
gets the default subtarget based on the target triple, which is often
not good enough.
This change also uses asm constraints/arguments to reference
jumptable targets and aliases directly. We no longer do asm name
mangling in an IR pass.
Differential Revision: https://reviews.llvm.org/D28012
llvm-svn: 290384
We used to not check generic vregs, but that is actually a mistake given
nothing in the GlobalISel pipeline is going to fix the constraints on
target specific instructions. Therefore, the target has to have them
right from the start.
llvm-svn: 290380
The InstructionSelect pass will not look at target specific instructions
since they are already selected. As a result, the operands of target
specific instructions must be properly constrained, because it is not
going to fix them.
This fixes invalid register classes on call instruction.
llvm-svn: 290377
When generic virtual registers get constrained, because of a use on a
target specific operation for instance, we end up with regular virtual
registers with a type and that's perfectly fine.
llvm-svn: 290376
This is going to be needed to be able to constraint register class on
target specific instruction while the RegBankSelect pass did not run
yet.
llvm-svn: 290375
Move the logic to constraint register from InstructionSelector to a
utility function. It will be required by other passes in the GlobalISel
pipeline.
llvm-svn: 290374
Canonicalize a select with a constant to the false side. This
enables more instruction shrinking opportunities since an
inline immediate can be used for the false side of v_cndmask_b32_e32.
This seems to usually be better but causes some code size regressions
in some tests.
llvm-svn: 290372
This is a succeeding patch of https://reviews.llvm.org/D22840 to address the
issue when a value to be merged into an int64 pair is in a different BB. Redoing
the store splitting in CodeGenPrepare so we can match the pattern across multiple
BBs and move some instructions into the same BB. We still keep the code in dag
combine so that we can catch cases that show up after DAG combining runs.
Differential Revision: https://reviews.llvm.org/D25914
llvm-svn: 290365
This is for splitMergedValStore in DAG Combine to share the target query interface
with similar logic in CodeGenPrepare.
Differential Revision: https://reviews.llvm.org/D24707
llvm-svn: 290363
Follow up to D27209 fix, this patch now properly handles single transient
instruction in basic block.
Patch by Aleksandar Beserminji.
Differential Revision: https://reviews.llvm.org/D27856
llvm-svn: 290361
When the pipeliner is renaming phi values, it may need to iterate through
the phi operands to check for other phis. However, the pipeliner should
stop once it reaches a phi that is outside the pipelined loop.
Also, when the generateExistingPhis code is unable to reuse an existing
phi, the default code that computes the PhiOp2 is only to be used when
the pipeliner is generating the kernel. Otherwise, the phi may be a value
computed earlier in the same epilog.
Patch by Brendon Cahoon.
llvm-svn: 290355
The code have been developed by Daniel Berlin over the years, and
the new implementation goal is that of addressing shortcomings of
the current GVN infrastructure, i.e. long compile time for large
testcases, lack of phi predication, no load/store value numbering
etc...
The current code just implements the "core" GVN algorithm, although
other pieces (load coercion, phi handling, predicate system) are
already implemented in a branch out of tree. Once the core is stable,
we'll start adding pieces on top of the base framework.
The test currently living in test/Transform/NewGVN are a copy
of the ones in GVN, with proper `XFAIL` (missing features in NewGVN).
A flag will be added in a future commit to enable NewGVN, so that
interested parties can exercise this code easily.
Differential Revision: https://reviews.llvm.org/D26224
llvm-svn: 290346
Summary: This is needed for later SDWA support in CodeGen.
Reviewers: vpykhtin, tstellarAMD
Subscribers: arsenm, kzhuravl, wdng, nhaehnle, yaxunl, tony-tye
Differential Revision: https://reviews.llvm.org/D27412
llvm-svn: 290338
Summary: Real instruction should copy constraints from real instruction. This allows auto-generated disassembler to correctly process tied operands.
Reviewers: nhaustov, vpykhtin, tstellarAMD
Subscribers: arsenm, kzhuravl, wdng, nhaehnle, yaxunl, tony-tye
Differential Revision: https://reviews.llvm.org/D27847
llvm-svn: 290336
Replacing the memory operand in the ymm version of VPMADDWD from i128mem to i256mem.
Differential Revision: https://reviews.llvm.org/D28024
llvm-svn: 290333
declarations.
We're using a custom class here instead of the helper template, these
bits just didn't get deleted when the other bits did get deleted. This
was found by a really nice MSVC warning about explicitly instantiating
a template where some member functions aren't defined and thus can't be
instantiatied.
llvm-svn: 290327
from the old pass manager in the new one.
I'm not trying to support (initially) the numerous options that are
currently available to customize the pass pipeline. If we end up really
wanting them, we can add them later, but I suspect many are no longer
interesting. The simplicity of omitting them will help a lot as we sort
out what the pipeline should look like in the new PM.
I've also documented to the best of my ability *why* each pass or group
of passes is used so that reading the pipeline is more helpful. In many
cases I think we have some questionable choices of ordering and I've
left FIXME comments in place so we know what to come back and revisit
going forward. But for now, I've left it as similar to the current
pipeline as I could.
Lastly, I've had to comment out several places where passes are not
ported to the new pass manager or where the loop pass infrastructure is
not yet ready. I did at least fix a few bugs in the loop pass
infrastructure uncovered by running the full pipeline, but I didn't want
to go too far in this patch -- I'll come back and re-enable these as the
infrastructure comes online. But I'd like to keep the comments in place
because I don't want to lose track of which passes need to be enabled
and where they go.
One thing that seemed like a significant API improvement was to require
that we don't build pipelines for O0. It seems to have no real benefit.
I've also switched back to returning pass managers by value as at this
API layer it feels much more natural to me for composition. But if
others disagree, I'm happy to go back to an output parameter.
I'm not 100% happy with the testing strategy currently, but it seems at
least OK. I may come back and try to refactor or otherwise improve this
in subsequent patches but I wanted to at least get a good starting point
in place.
Differential Revision: https://reviews.llvm.org/D28042
llvm-svn: 290325
When DwarfExpression is emitting a fragment that is located in a
register and that fragment is smaller than the register, and the
register must be composed from sub-registers (are you still with me?)
the last DW_OP_piece operation must not be larger than the size of the
fragment itself, since the last piece of the fragment could be smaller
than the last subregister that is being emitted.
rdar://problem/29779065
llvm-svn: 290324
There are helpers for testing for constant or constant build_vector,
and for splat ConstantFP vectors, but not for a constantfp or
non-splat ConstantFP vector.
llvm-svn: 290317
This is to put the vector into a well defined state. Apparently the state of a
vector after being moved from is valid but unspecified. Found with clang-tidy.
llvm-svn: 290298
This adds a basic tablegen backend that analyzes the SelectionDAG
patterns to find simple ones that are eligible for GlobalISel-emission.
That's similar to FastISel, with one notable difference: we're not fed
ISD opcodes, so we need to map the SDNode operators to generic opcodes.
That's done using GINodeEquiv in TargetGlobalISel.td.
Otherwise, this is mostly boilerplate, and lots of filtering of any kind
of "complicated" pattern. On AArch64, this is sufficient to match G_ADD
up to s64 (to ADDWrr/ADDXrr) and G_BR (to B).
Differential Revision: https://reviews.llvm.org/D26878
llvm-svn: 290284
Each function summary has an attached list of type identifier GUIDs. The
idea is that during the regular LTO phase we would match these GUIDs to type
identifiers defined by the regular LTO module and store the resolutions in
a top-level "type identifier summary" (which will be implemented separately).
Differential Revision: https://reviews.llvm.org/D27967
llvm-svn: 290280
In order for the llvm DWARF parser to be used in LLDB we will need to be able to get the parent of a DIE. This patch adds that functionality by changing the DWARFDebugInfoEntry class to store a depth field instead of a sibling index. Using a depth field allows us to easily calculate the sibling and the parent without increasing the size of DWARFDebugInfoEntry.
I tested llvm-dsymutil on a debug version of clang where this fully parses DWARF in over 1200 .o files to verify there was no serious regression in performance.
Added a full suite of unit tests to test this functionality.
Differential Revision: https://reviews.llvm.org/D27995
llvm-svn: 290274
RTDyldMemoryManager.cpp describes the differing __register_frame
API between libunwind and libgcc, with a mailing list posting URL.
The original link was 404; replace it with what I believe is the
intended post, as well as a reference to the "OS X" implementation in
libunwind.
Differential Revision: https://reviews.llvm.org/D27965
llvm-svn: 290269
The constantexpr parsing was too constrained and rejected legal vector GEPs.
This relaxes it to be similar to the ones for instruction parsing.
This fixes PR30816.
Differential Revision: https://reviews.llvm.org/D28013
llvm-svn: 290261
For vector GEPs, CastGEPIndices can end up in an infinite recursion, because
we compare the vector type to the scalar pointer type, find them different,
and then try to cast a type to itself.
Differential Revision: https://reviews.llvm.org/D28009
llvm-svn: 290260
I added API for creation a target specific memory node in DAG. Today, all memory nodes are common for all targets and their constructors are located in SelectionDAG.cpp.
There are some cases in X86 where we need to create a special node - truncation-with-saturation store, float-to-half-store.
In the current patch I added truncation-with-saturation nodes and I'm using them for intrinsics. In the future I plan to implement DAG lowering for truncation-with-saturation pattern.
Differential Revision: https://reviews.llvm.org/D27899
llvm-svn: 290250
The vectorcall calling convention specifies that arguments to functions are to be passed in registers, when possible.
vectorcall uses more registers for arguments than fastcall or the default x64 calling convention use.
The vectorcall calling convention is only supported in native code on x86 and x64 processors that include Streaming SIMD Extensions 2 (SSE2) and above.
The current implementation does not handle Homogeneous Vector Aggregates (HVAs) correctly and this review attempts to fix it.
This aubmit also includes additional lit tests to cover better HVAs corner cases.
Differential Revision: https://reviews.llvm.org/D27392
llvm-svn: 290240
In r267672, where the loop distribution pragma was introduced, I tried
it hard to keep the old behavior for opt: when opt is invoked
with -loop-distribute, it should distribute the loop (it's off by
default when ran via the optimization pipeline).
As MichaelZ has discovered this has the unintended consequence of
breaking a very common developer work-flow to reproduce compilations
using opt: First you print the pass pipeline of clang
with -debug-pass=Arguments and then invoking opt with the returned
arguments.
clang -debug-pass will include -loop-distribute but the pass is invoked
with default=off so nothing happens unless the loop carries the pragma.
While through opt (default=on) we will try to distribute all loops.
This changes opt's default to off as well to match clang. The tests are
modified to explicitly enable the transformation.
llvm-svn: 290235
Mehdi Amini