This test code previously caused a failure in the module verifier,
because SimplifyCFG created this invalid instruction, which tries to
take the address of inline asm:
%.sink = select i1 %1, i64 ()* asm "mov $0, #1", "=r", i64 ()* asm %"mov $0, #2", "=r"
This has been fixed recently, presumably by James Molloy's patches that
re-wrote and changed parts of SimplifyCFG, so this patch just adds a
regression test for it.
Differential Revision: https://reviews.llvm.org/D24231
llvm-svn: 280660
Previously we were extending to copying the whole ZMM register. The register allocator shouldn't use XMM16-31 or YMM16-31 in this configuration as the instructions to spill them aren't available.
llvm-svn: 280648
Summary:
A frontend may designate a particular suspend to be final, by setting the second argument of the coro.suspend intrinsic to true. Such a suspend point has two properties:
* it is possible to check whether a suspended coroutine is at the final suspend point via coro.done intrinsic;
* a resumption of a coroutine stopped at the final suspend point leads to undefined behavior. The only possible action for a coroutine at a final suspend point is destroying it via coro.destroy intrinsic.
This patch adds final suspend handling logic to CoroEarly and CoroSplit passes.
Now, the final suspend point example from docs\Coroutines.rst compiles and produces expected result (see test/Transform/Coroutines/ex5.ll).
Reviewers: majnemer
Subscribers: mehdi_amini, llvm-commits
Differential Revision: https://reviews.llvm.org/D24068
llvm-svn: 280646
memcpy with ld/st.
When InstCombine replaces a memcpy with loads+stores it does not copy over the
llvm.mem.parallel_loop_access from the memcpy instruction. This patch fixes
that.
Differential Revision: https://reviews.llvm.org/D23499
llvm-svn: 280617
As it turns out, whether we zero-extend or sign-extend i8/i16 constants, which
are illegal types promoted to i32 on PowerPC, is a choice constrained by
assumptions within the infrastructure. Specifically, the logic in
FunctionLoweringInfo::ComputePHILiveOutRegInfo assumes that constant PHI
operands will be zero extended, and so, at least when materializing constants
that are PHI operands, we must do the same.
The rest of our fast-isel implementation does not appear to depend on the fact
that we were sign-extending i8/i16 constants, and all other targets also appear
to zero-extend small-bitwidth constants in fast-isel; we'll now do the same (we
had been doing this only for i1 constants, and sign-extending the others).
Fixes PR27721.
llvm-svn: 280614
Summary:
The inliner may need to determine where a given funclet unwinds to,
and this determination may depend on other funclets throughout the
funclet tree. The code that performs this walk in getUnwindDestToken
memoizes results to avoid redundant computations. In the case that
a funclet's unwind destination is derived from its ancestor, there's
code to walk back down the tree from the ancestor updating the memo
map of its descendants to record the unwind destination. This change
fixes that code to account for the case that some descendant has a
different unwind destination, which can happen if that unwind dest
is a descendant of the EHPad being queried and thus didn't determine
its unwind destination.
Also update test inline-funclets.ll, which is supposed to cover such
scenarios, to include a case that fails an assertion without this fix
but passes with it.
Fixes PR29151.
Reviewers: majnemer
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D24117
llvm-svn: 280610
CGP currently drops select's MD_prof profile data when
generating conditional branch which can lead to bad
code layout. The patch fixes the issue.
Differential Revision: http://reviews.llvm.org/D24169
llvm-svn: 280600
Because the recent change about ODR type uniquing in the context,
we can reach types defined in another module during IR linking.
This triggered some assertions in case we IR link without starting
from an empty module. To alleviate that, we can self-map metadata
defined in the destination module so that they won't be visited.
Differential Revision: https://reviews.llvm.org/D23841
llvm-svn: 280599
Summary:
This contains two changes that reduce the time spent in WQM, with the
intention of reducing bandwidth required by VMEM loads:
1. Sampling instructions by themselves don't need to run in WQM, only their
coordinate inputs need it (unless of course there is a dependent sampling
instruction). The initial scanInstructions step is modified accordingly.
2. When switching back from WQM to Exact, switch back as soon as possible.
This affects the logic in processBlock.
This should always be a win or at best neutral.
There are also some cleanups (e.g. remove unused ExecExports) and some new
debugging output.
Reviewers: arsenm, tstellarAMD, mareko
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: http://reviews.llvm.org/D22092
llvm-svn: 280590
Summary:
This fixes a rare bug in polygon stippling with non-monolithic pixel shaders.
The underlying problem is as follows: the prolog part contains the polygon
stippling sequence, i.e. a kill. The main part then enables WQM based on the
_reduced_ exec mask, effectively undoing most of the polygon stippling.
Since we cannot know whether polygon stippling will be used, the main part
of a non-monolithic shader must always return to exact mode to fix this
problem.
Reviewers: arsenm, tstellarAMD, mareko
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: https://reviews.llvm.org/D23131
llvm-svn: 280589
readlane/writelane do not support using m0 as the output/input.
Constrain the register class of spill vregs to try to avoid this,
but also handle spilling of the physreg when necessary by inserting
an additional copy to a normal SGPR.
llvm-svn: 280584
PowerPC assembly code in the wild, so it seems, has things like this:
bc+ 12, 28, .L9
This is a bit odd because the '+' here becomes part of the BO field, and the BO
field is otherwise the first operand. Nevertheless, the ISA specification does
clearly say that the +- hint syntax applies to all conditional-branch mnemonics
(that test either CTR or a condition register, although not the forms which
check both), both basic and extended, so this is supposed to be valid.
This introduces some asm-parser-only definitions which take only the upper
three bits from the specified BO value, and the lower two bits are implied by
the +- suffix (via some associated aliases).
Fixes PR23646.
llvm-svn: 280571
dcbf has an optional hint-like field, add support for the extended form and the
associated mnemonics (dcbfl and dcbflp).
Partially fixes PR24796.
llvm-svn: 280559
Before we were kind of imitating the behavior of a Yaml sequence
by outputting each record one after the other. This makes it a
little cumbersome when we want to go the other direction -- from
Yaml to Pdb. So this treats FieldList records as no different than
any other list of records, by printing them as a Yaml sequence with
the exact same format.
llvm-svn: 280549
When we have an offset into a global, etc. that is accessed relative to the TOC
base pointer, and the offset is larger than the minimum alignment of the global
itself and the TOC base pointer (which is 8-byte aligned), we can still fold
the @toc@ha into the memory access, but we must update the addis instruction's
symbol reference with the offset as the symbol addend. When there is only one
use of the addi to be folded and only one use of the addis that would need its
symbol's offset adjusted, then we can make the adjustment and fold the @toc@l
into the memory access.
llvm-svn: 280545
Subregister definitions are considered uses for the purpose of tracking
liveness of the whole register. At the same time, when calculating live
interval subranges, subregister defs should not be treated as uses.
Differential Revision: https://reviews.llvm.org/D24190
llvm-svn: 280532
Previously we were splitting our records at 0xFFFF bytes, which the
Microsoft tools don't like.
Should fix failure on the new Windows self-host buildbot.
This length appears in microsoft-pdb/PDB/dbi/dbiimpl.h
llvm-svn: 280522
For the store of a wide value merged from a pair of values, especially int-fp pair,
sometimes it is more efficent to split it into separate narrow stores, which can
remove the bitwise instructions or sink them to colder places.
Now the feature is only enabled on x86 target, and only store of int-fp pair is
splitted. It is possible that the application scope gets extended with perf evidence
support in the future.
Differential Revision: https://reviews.llvm.org/D22840
llvm-svn: 280505
The motivating case occurs with SSE/AVX scalar intrinsics, so this is a first step towards
shrinking that to a single shufflevector.
Note that the transform is intentionally limited to shuffles that are equivalent to vector
selects to avoid creating arbitrary shuffle masks that may not lower well.
This should solve PR29126:
https://llvm.org/bugs/show_bug.cgi?id=29126
Differential Revision: https://reviews.llvm.org/D23886
llvm-svn: 280504
For uniform instructions, we're only required to generate a scalar value for
the first vector lane of each unroll iteration. Thus, if we have a reverse
interleaved group, computing the member index off the scalar GEP corresponding
to the last vector lane of its pointer operand technically makes the GEP
non-uniform. We should compute the member index off the first scalar GEP
instead.
I've added the updated member index computation to the existing reverse
interleaved group test.
llvm-svn: 280497
This patch fixes a crash caused by an incorrect folding of an ordered comparison
between a packed floating point vector and a splat vector of NaN.
An ordered comparison between a vector and a constant vector of NaN, should
always be folded into a constant vector where each element is i1 false.
Since revision 266175, SimplifyFCmpInst folds the ordered fcmp into a scalar
'false'. Later on, this would cause an assertion failure, since the value type
of the folded value doesn't match the expected value type of the uses of the
original instruction: "Assertion failed: New->getType() == getType() &&
"replaceAllUses of value with new value of different type!".
This patch fixes the issue and adds a test case to the already existing test
InstSimplify/floating-point-compares.ll.
Differential Revision: https://reviews.llvm.org/D24143
llvm-svn: 280488
This fixes a regression introduced by revision 268094.
Revision 268094 added the following dag combine rule:
// trunc (shl x, K) -> shl (trunc x), K => K < vt.size / 2
That rule converts a truncate of a shift-by-constant into a shift of a truncated
value. We do this only if the shift count is less than half the size in bits of
the truncated value (K < vt.size / 2).
The problem is that the constraint on the shift count is incorrect, so the rule
doesn't work well in some cases involving vector types. The combine rule should
have been written instead like this:
// trunc (shl x, K) -> shl (trunc x), K => K < vt.getScalarSizeInBits()
Basically, if K is smaller than the "scalar size in bits" of the truncated value
then we know that by "sinking" the truncate into the operand of the shift we
would never accidentally make the shift undefined.
This patch fixes the check on the shift count, and adds test cases to make sure
that we don't regress the behavior.
Differential Revision: https://reviews.llvm.org/D24154
llvm-svn: 280482
We're sinking stores, which is a good thing, but in the process creating selects for the store address operand, which SROA/Mem2Reg can't look through, which caused serious regressions.
The real fix is in SROA, which I'll be looking into.
llvm-svn: 280470
As Sanjay suggested when he added the hook, PPC should return true from
hasAndNotCompare. We have an efficient negated 'and' on PPC (which can feed a
compare).
Fixes PR27203.
llvm-svn: 280457
Following a suggestion by Sanjay, we should lower:
%shl = shl i32 1, %y
%and = and i32 %x, %shl
%cmp = icmp eq i32 %and, %shl
ret i1 %cmp
into:
subfic r4, r4, 32
rlwnm r3, r3, r4, 31, 31
Add this pattern and some associated patterns for the 64-bit case and the
not-equal case. Fixes PR27356.
llvm-svn: 280454
When applying our address-formation PPC64 peephole, we are reusing the @ha TOC
addis value with the low parts associated with different offsets (i.e.
different effective symbol addends). We were assuming this was okay so long as
the offsets were less than the alignment of the global variable being accessed.
This ignored the fact, however, that the TOC base pointer itself need only be
8-byte aligned. As a result, what we were doing is legal only for offsets less
than 8 regardless of the alignment of the object being accessed.
Fixes PR28727.
llvm-svn: 280441
The logic in this function assumes that the P8 supports fusion of addis/addi,
but it does not. As a result, there is no advantage to restricting our peephole
application, merging addi instructions into dependent memory accesses, even
when the addi has multiple users, regardless of whether or not we're optimizing
for size.
We might need something like this again for the P9; I suspect we'll revisit
this code when we work on P9 tuning.
llvm-svn: 280440
When expanding a SETCC for which the low half is known to evaluate to false,
we can only throw it away for LT/GT comparisons, not LE/GE.
This fixes PR29170.
Differential Revision: https://reviews.llvm.org/D24151
llvm-svn: 280424
Prior to this, we could generate a vector_shuffle from an IR shuffle when the
size of the result was exactly the sum of the sizes of the input vectors.
If the output vector was narrower - e.g. a <12 x i8> being formed by a shuffle
with two <8 x i8> inputs - we would lower the shuffle to a sequence of extracts
and inserts.
Instead, we can form a larger vector_shuffle, and then extract a subvector
of the right size - e.g. shuffle the two <8 x i8> inputs into a <16 x i8>
and then extract a <12 x i8>.
This also includes a target-specific X86 combine that in the presence of
AVX2 combines:
(vector_shuffle <mask> (concat_vectors t1, undef)
(concat_vectors t2, undef))
into:
(vector_shuffle <mask> (concat_vectors t1, t2), undef)
in cases where this allows us to form VPERMD/VPERMQ.
(This is not a separate commit, as that pattern does not appear without
the DAGBuilder change.)
llvm-svn: 280418
Summary: This patch adds asm.js-style setjmp/longjmp handling support for WebAssembly. It also uses JavaScript's try and catch mechanism.
Reviewers: jpp, dschuff
Subscribers: jfb, dschuff
Differential Revision: https://reviews.llvm.org/D24121
llvm-svn: 280415
They're another source of generic vregs, which are going to need a type on the
definition when we remove the register width from MachineRegisterInfo.
llvm-svn: 280412
According to spec cvtdq2pd and cvtps2pd instructions don't require memory operand to be aligned
to 16 bytes. This patch removes this requirement from the memory folding table.
Differential Revision: https://reviews.llvm.org/D23919
llvm-svn: 280402
Add runtime metdata for pointee alignment of pointer type kernel argument. The key is KeyArgPointeeAlign and the value is a 32 bit unsigned integer.
Differential Revision: https://reviews.llvm.org/D24145
llvm-svn: 280399
Legalization tends to create anyext(trunc) patterns. This should always be
combined - into either a single trunc, a single ext, or nothing if the
types match exactly. But if we happen to combine the trunc first, we may pull
the trunc away from the anyext or make it implicit (e.g. the truncate(extract)
-> extract(bitcast) fold).
To prevent this, we can avoid doing the fold, similarly to how we already handle
fpround(fpextend).
Differential Revision: https://reviews.llvm.org/D23893
llvm-svn: 280386
This scheduler describes a processor which covers all MIPS ISAs based
around the interAptiv and P5600 timings.
Reviewers: vkalintiris, dsanders
Differential Revision: https://reviews.llvm.org/D23551
llvm-svn: 280374
While removing a scalar shackle from an icmp fold, I noticed that I couldn't find any tests to trigger
this code path.
The 'and' shrinking transform should be handled by InstCombiner::foldCastedBitwiseLogic()
or eliminated with InstSimplify. The icmp narrowing is part of InstCombiner::foldICmpWithCastAndCast().
Differential Revision: https://reviews.llvm.org/D24031
llvm-svn: 280370
This was a real restriction in the original version of SinkIfThenCodeToEnd. Now it's been rewritten, the restriction can be lifted.
As part of this, we handle a very common and useful case where one of the incoming branches is actually conditional. Consider:
if (a)
x(1);
else if (b)
x(2);
This produces the following CFG:
[if]
/ \
[x(1)] [if]
| | \
| | \
| [x(2)] |
\ | /
[ end ]
[end] has two unconditional predecessor arcs and one conditional. The conditional refers to the implicit empty 'else' arc. This same pattern can also be caused by an empty default block in a switch.
We can't sink the call to x() down to end because no call to x() happens on the third incoming arc (assume that x() has sideeffects for the sake of argument; if something is safe to speculate we could indeed sink nevertheless but this cannot happen in the general case and causes many extra selects).
We are now able to detect this case and split off the unconditional arcs to a common successor:
[if]
/ \
[x(1)] [if]
| | \
| | \
| [x(2)] |
\ / |
[sink.split] |
\ /
[ end ]
Now we can sink the call to x() into %sink.split. This can cause significant code simplification in many testcases.
llvm-svn: 280364
r279460 rewrote this function to be able to handle more than two incoming edges and took pains to ensure this didn't regress anything.
This time we change the logic for determining if an instruction should be sunk. Previously we used a single pass greedy algorithm - sink instructions until one requires more than one PHI node or we run out of instructions to sink.
This had the problem that sinking instructions that had non-identical but trivially the same operands needed extra logic so we sunk them aggressively. For example:
%a = load i32* %b %d = load i32* %b
%c = gep i32* %a, i32 0 %e = gep i32* %d, i32 1
Sinking %c and %e would naively require two PHI merges as %a != %d. But the loads are obviously equivalent (and maybe can't be hoisted because there is no common predecessor).
This is why we implemented the fairly complex function areValuesTriviallySame(), to look through trivial differences like this. However it's just not clever enough.
Instead, throw areValuesTriviallySame away, use pointer equality to check equivalence of operands and switch to a two-stage algorithm.
In the "scan" stage, we look at every sinkable instruction in isolation from end of block to front. If it's sinkable, we keep track of all operands that required PHI merging.
In the "sink" stage, we iteratively sink the last non-terminator in the source blocks. But when calculating how many PHIs are actually required to be inserted (to work out if we should stop or not) we remove any values that have already been sunk from the set of PHI-merges required, which allows us to be more aggressive.
This turns an algorithm with potentially recursive lookahead (looking through GEPs, casts, loads and any other instruction potentially not CSE'd) to two linear scans.
llvm-svn: 280351
LLVM has an @llvm.eh.dwarf.cfa intrinsic, used to lower the GCC-compatible
__builtin_dwarf_cfa() builtin. As pointed out in PR26761, this is currently
broken on PowerPC (and likely on ARM as well). Currently, @llvm.eh.dwarf.cfa is
lowered using:
ADD(FRAMEADDR, FRAME_TO_ARGS_OFFSET)
where FRAME_TO_ARGS_OFFSET defaults to the constant zero. On x86,
FRAME_TO_ARGS_OFFSET is lowered to 2*SlotSize. This setup, however, does not
work for PowerPC. Because of the way that the stack layout works, the canonical
frame address is not exactly (FRAMEADDR + FRAME_TO_ARGS_OFFSET) on PowerPC
(there is a lower save-area offset as well), so it is not just a matter of
implementing FRAME_TO_ARGS_OFFSET for PowerPC (unless we redefine its
semantics -- We can do that, since it is currently used only for
@llvm.eh.dwarf.cfa lowering, but the better to directly lower the CFA construct
itself (since it can be easily represented as a fixed-offset FrameIndex)). Mips
currently does this, but by using a custom lowering for ADD that specifically
recognizes the (FRAMEADDR, FRAME_TO_ARGS_OFFSET) pattern.
This change introduces a ISD::EH_DWARF_CFA node, which by default expands using
the existing logic, but can be directly lowered by the target. Mips is updated
to use this method (which simplifies its implementation, and I suspect makes it
more robust), and updates PowerPC to do the same.
Fixes PR26761.
Differential Revision: https://reviews.llvm.org/D24038
llvm-svn: 280350
As discussed in https://reviews.llvm.org/D22666, our current mechanism to
support -pg profiling, where we insert calls to mcount(), or some similar
function, is fundamentally broken. We insert these calls in the frontend, which
means they get duplicated when inlining, and so the accumulated execution
counts for the inlined-into functions are wrong.
Because we don't want the presence of these functions to affect optimizaton,
they should be inserted in the backend. Here's a pass which would do just that.
The knowledge of the name of the counting function lives in the frontend, so
we're passing it here as a function attribute. Clang will be updated to use
this mechanism.
Differential Revision: https://reviews.llvm.org/D22825
llvm-svn: 280347
Summary:
This change promotes the 'isTailCall(...)' member function to
TargetInstrInfo as a query interface for determining on a per-target
basis whether a given MachineInstr is a tail call instruction. We build
upon this in the XRay instrumentation pass to emit special sleds for
tail call optimisations, where we emit the correct kind of sled.
The tail call sleds look like a mix between the function entry and
function exit sleds. Form-wise, the sled comes before the "jmp"
instruction that implements the tail call similar to how we do it for
the function entry sled. Functionally, because we know this is a tail
call, it behaves much like an exit sled -- i.e. at runtime we may use
the exit trampolines instead of a different kind of trampoline.
A follow-up change to recognise these sleds will be done in compiler-rt,
so that we can start intercepting these initially as exits, but also
have the option to have different log entries to more accurately reflect
that this is actually a tail call.
Reviewers: echristo, rSerge, majnemer
Subscribers: mehdi_amini, dberris, llvm-commits
Differential Revision: https://reviews.llvm.org/D23986
llvm-svn: 280334
-fprofile-dir=path allows the user to specify where .gcda files should be
emitted when the program is run. In particular, this is the first flag that
causes the .gcno and .o files to have different paths, LLVM is extended to
support this. -fprofile-dir= does not change the file name in the .gcno (and
thus where lcov looks for the source) but it does change the name in the .gcda
(and thus where the runtime library writes the .gcda file). It's different from
a GCOV_PREFIX because a user can observe that the GCOV_PREFIX_STRIP will strip
paths off of -fprofile-dir= but not off of a supplied GCOV_PREFIX.
To implement this we split -coverage-file into -coverage-data-file and
-coverage-notes-file to specify the two different names. The !llvm.gcov
metadata node grows from a 2-element form {string coverage-file, node dbg.cu}
to 3-elements, {string coverage-notes-file, string coverage-data-file, node
dbg.cu}. In the 3-element form, the file name is already "mangled" with
.gcno/.gcda suffixes, while the 2-element form left that to the middle end
pass.
llvm-svn: 280306
Summary: This patch adds asm.js-style setjmp/longjmp handling support for WebAssembly. It also uses JavaScript's try and catch mechanism.
Reviewers: jpp, dschuff
Subscribers: jfb, dschuff
Differential Revision: https://reviews.llvm.org/D23928
llvm-svn: 280302
This reverts commit r280268, it causes all MSVC 2013 to ICE. This
appears to have been fixed in a later MSVC 2013 update, because I cannot
reproduce it locally. That said, all upstream LLVM bots are broken right
now, so I am reverting.
Also reverts dependent change r280275, "[Hexagon] Deal with undefs when
extending live intervals".
llvm-svn: 280301
More preparation for dropping source types from MachineInstrs: regsters coming
out of already-selected code (i.e. non-generic instructions) don't have a type,
but that information is needed so we must add it manually.
This is done via a new G_TYPE instruction.
llvm-svn: 280292
Summary:
If the register has a negative value then unsigned overflow will occur;
this case is sometimes even created intentionally by LSR. For now
disable GA+reg folding. Fixes PR29127
Differential Revision: https://reviews.llvm.org/D24053
llvm-svn: 280285
Summary:
Use MemorySSA, if requested, to do less conservative memory dependency
checking.
This change doesn't enable the MemorySSA enhanced EarlyCSE in the
default pipelines, so should be NFC.
Reviewers: dberlin, sanjoy, reames, majnemer
Subscribers: mcrosier, llvm-commits
Differential Revision: http://reviews.llvm.org/D19821
llvm-svn: 280279
Summary: This fixes some OpenCV tests that were broken by libclc commit r276443.
Reviewers: arsenm, jvesely
Subscribers: arsenm, wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D24051
llvm-svn: 280274
that use the Mach::linkedit_data_command type for the load commands that are
currently used in the MachOObjectFile constructor.
This contains the missing checks for LC_DATA_IN_CODE and
LC_LINKER_OPTIMIZATION_HINT load commands and the fields for the
Mach::linkedit_data_command type. Checking for other load commands that
use this type will be added later.
Also fixed a couple of places that was using sizeof(MachOObjectFile::LoadCommandInfo)
that should have been using sizeof(MachO::load_command).
llvm-svn: 280267
Adjust the test to reflect the changes to common handling in r279911.
This test wasn't running due to an incorrect REQUIRES and thus missed
being modified for r279911 before. It was changed to XFAIL when the
bad REQUIRES was discovered.
Remove the XFAIL and move to a new X86 subdirectory that will properly
disable on non-X86.
llvm-svn: 280256
The shape of the vtable is passed down as the size of the
__vtbl_ptr_type. This special pointer type appears both as the pointee
type of the vptr type, and by itself in every dynamic class. For classes
with multiple vtables, only the shape of the primary vftable is
included, as the shape of all secondary vftables will be the same as in
the base class.
Fixes PR28150
llvm-svn: 280254
This is a first step towards supporting deopt value lowering and reporting entirely with the register allocator. I hope to build on this in the near future to support live-on-return semantics, but I have a use case which allows me to test and investigate code quality with just the live-in semantics so I've chosen to start there. For those curious, my use cases is our implementation of the "__llvm_deoptimize" function we bind to @llvm.deoptimize. I'm choosing not to hard code that fact in the patch and instead make it configurable via function attributes.
The basic approach here is modelled on what is done for the "Live In" values on stackmaps and patchpoints. (A secondary goal here is to remove one of the last barriers to merging the pseudo instructions.) We start by adding the operands directly to the STATEPOINT SDNode. Once we've lowered to MI, we extend the remat logic used by the register allocator to fold virtual register uses into StackMap::Indirect entries as needed. This does rely on the fact that the register allocator rematerializes. If it didn't along some code path, we could end up with more vregs than physical registers and fail to allocate.
Today, we *only* fold in the register allocator. This can create some weird effects when combined with arguments passed on the stack because we don't fold them appropriately. I have an idea how to fix that, but it needs this patch in place to work on that effectively. (There's some weird interaction with the scheduler as well, more investigation needed.)
My near term plan is to land this patch off-by-default, experiment in my local tree to identify any correctness issues and then start fixing codegen problems one by one as I find them. Once I have the live-in lowering fully working (both correctness and code quality), I'm hoping to move on to the live-on-return semantics. Note: I don't have any *known* miscompiles with this patch enabled, but I'm pretty sure I'll find at least a couple. Thus, the "experimental" tag and the fact it's off by default.
Differential Revision: https://reviews.llvm.org/D24000
llvm-svn: 280250
types. This is the LLVM counterpart and it adds options that map onto FP
exceptions and denormal build attributes allowing better fp math library
selections.
Differential Revision: https://reviews.llvm.org/D24070
llvm-svn: 280246
We check that a sinking candidate is used by only one PHI node during our legality checks. However for instructions that are used by other sinking candidates our heuristic is less conservative. This can result in a candidate actually being illegal when we come to sink it because of how we sunk a predecessor. Do the used-by-only-one-PHI checks again during sinking to ensure we don't crash.
llvm-svn: 280228
Summary:
Simply replace usage of aliases to functions with aliasee.
This came up when bitcode linking to builtin library and
calls to aliases not being resolved.
Also made minor improvements to existing test.
Reviewers: tstellarAMD, alex-t, vpykhtin
Subscribers: arsenm, wdng, rampitec
Differential Revision: https://reviews.llvm.org/D24023
llvm-svn: 280221
We're sinking stores, which is a good thing, but in the process creating selects for the store address operand, which SROA/Mem2Reg can't look through, which caused serious regressions.
The real fix is in SROA, which I'll be looking into.
llvm-svn: 280219
This was a real restriction in the original version of SinkIfThenCodeToEnd. Now it's been rewritten, the restriction can be lifted.
As part of this, we handle a very common and useful case where one of the incoming branches is actually conditional. Consider:
if (a)
x(1);
else if (b)
x(2);
This produces the following CFG:
[if]
/ \
[x(1)] [if]
| | \
| | \
| [x(2)] |
\ | /
[ end ]
[end] has two unconditional predecessor arcs and one conditional. The conditional refers to the implicit empty 'else' arc. This same pattern can also be caused by an empty default block in a switch.
We can't sink the call to x() down to end because no call to x() happens on the third incoming arc (assume that x() has sideeffects for the sake of argument; if something is safe to speculate we could indeed sink nevertheless but this cannot happen in the general case and causes many extra selects).
We are now able to detect this case and split off the unconditional arcs to a common successor:
[if]
/ \
[x(1)] [if]
| | \
| | \
| [x(2)] |
\ / |
[sink.split] |
\ /
[ end ]
Now we can sink the call to x() into %sink.split. This can cause significant code simplification in many testcases.
llvm-svn: 280217
r279460 rewrote this function to be able to handle more than two incoming edges and took pains to ensure this didn't regress anything.
This time we change the logic for determining if an instruction should be sunk. Previously we used a single pass greedy algorithm - sink instructions until one requires more than one PHI node or we run out of instructions to sink.
This had the problem that sinking instructions that had non-identical but trivially the same operands needed extra logic so we sunk them aggressively. For example:
%a = load i32* %b %d = load i32* %b
%c = gep i32* %a, i32 0 %e = gep i32* %d, i32 1
Sinking %c and %e would naively require two PHI merges as %a != %d. But the loads are obviously equivalent (and maybe can't be hoisted because there is no common predecessor).
This is why we implemented the fairly complex function areValuesTriviallySame(), to look through trivial differences like this. However it's just not clever enough.
Instead, throw areValuesTriviallySame away, use pointer equality to check equivalence of operands and switch to a two-stage algorithm.
In the "scan" stage, we look at every sinkable instruction in isolation from end of block to front. If it's sinkable, we keep track of all operands that required PHI merging.
In the "sink" stage, we iteratively sink the last non-terminator in the source blocks. But when calculating how many PHIs are actually required to be inserted (to work out if we should stop or not) we remove any values that have already been sunk from the set of PHI-merges required, which allows us to be more aggressive.
This turns an algorithm with potentially recursive lookahead (looking through GEPs, casts, loads and any other instruction potentially not CSE'd) to two linear scans.
llvm-svn: 280216
This was deliberately disabled during my rewrite of SinkIfThenToEnd to keep behaviour
at least vaguely consistent with the previous version and keep it as close to NFC as
I could.
There's no real reason not to merge sideeffect calls though, so let's do it! Small fixup
along the way to ensure we don't create indirect calls.
Should fix PR28964.
llvm-svn: 280215
When a function contains something, such as inline asm, which explicitly
clobbers the register used as the frame pointer, don't spill it twice. If we
need a frame pointer, it will be saved/restored in the prologue/epilogue code.
Explicitly spilling it again will reuse the same spill slot used by the
prologue/epilogue code, thus clobbering the saved value. The same applies
to the base-pointer or PIC-base register.
Partially fixes PR26856. Thanks to Ulrich for his analysis and the small
inline-asm reproducer.
llvm-svn: 280188
Summary:
1) CoroEarly now lowers llvm.coro.promise intrinsic that allows to obtain
a coroutine promise pointer from a coroutine frame and vice versa.
2) CoroFrame now interprets Promise argument of llvm.coro.begin to
place CoroutinPromise alloca at a deterministic offset from the coroutine frame.
Now, the coroutine promise example from docs\Coroutines.rst compiles and produces expected result (see test/Transform/Coroutines/ex4.ll).
Reviewers: majnemer
Subscribers: llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D23993
llvm-svn: 280184
Summary:
LSV was using two vector sets (heads and tails) to track pairs of adjiacent position to vectorize.
A recent optimization is trying to obtain the longest chain to vectorize and assumes the positions
in heads(H) and tails(T) match, which is not the case is there are multiple tails for the same head.
e.g.:
i1: store a[0]
i2: store a[1]
i3: store a[1]
Leads to:
H: i1
T: i2 i3
Instead of:
H: i1 i1
T: i2 i3
So the positions for instructions that follow i3 will have different indexes in H/T.
This patch resolves PR29148.
This issue also surfaced the fact that if the chain is too long, and TLI
returns a "not-fast" answer, the whole chain will be abandoned for
vectorization, even though a smaller one would be beneficial.
Added a testcase and FIXME for this.
Reviewers: tstellarAMD, arsenm, jlebar
Subscribers: mzolotukhin, wdng, llvm-commits
Differential Revision: https://reviews.llvm.org/D24057
llvm-svn: 280179
Legalization ends up creating many G_SEQUENCE/G_EXTRACT pairs which leads to
inefficient codegen (even for -O0), so add a quick pass over the function to
remove them again.
llvm-svn: 280155
We don't need to limit predication to blocks that have a single incoming
edge, we just need to use the right mask.
This fixes PR30172.
Differential Revision: https://reviews.llvm.org/D24009
llvm-svn: 280148
We're intending to move to a world where the type of a register is determined
by its (unique) def. This is incompatible with physregs, which are untyped.
It also means the other passes don't have to worry quite so much about
register-class compatibility and inserting COPYs appropriately.
llvm-svn: 280132
Add support for printing the GNU Notes. This allows an easy way to view the
build id for a binary built with the build id. Currently, this only handles the
GNU notes, though it would be easy to extend for other note types (default,
FreeBSD, NetBSD, etc). Only the GNU style is supported currently.
llvm-svn: 280131
The coverage reports contain the source or binary file paths. On Windows,
the file path might contain the seperators of both '/' and '\'. This patch
uses the native path in the coverage reports. For example, on Windows,
all '/' are converted to '\'.
Differential Revision: https://reviews.llvm.org/D23922
llvm-svn: 280061
Implement Bill's suggested fix for 32-bit targets for PR22711 (for the
alignment of each entry). As pointed out in the bug report, we could just force
the section alignment, since we only add pointer-sized things currently, but
this fix is somewhat more future-proof.
llvm-svn: 280049
The "long call" option forces the use of the indirect calling sequence for all
calls (even those that don't really need it). GCC provides this option; This is
helpful, under certain circumstances, for building very-large binaries, and
some other specialized use cases.
Fixes PR19098.
llvm-svn: 280040
Summary:
Fix a couple issues limiting the application of indirect call promotion
in ThinLTO mode:
- Invoke indirect call promotion before globalopt, since it may
eliminate imported functions which appear unreferenced.
- Invoke indirect call promotion with InLTO=true so that the PGOFuncName
metadata is used to get the name for locals which would have been
renamed during promotion.
Reviewers: davidxl, mehdi_amini
Subscribers: Prazek, llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D24004
llvm-svn: 280024
For little-Endian PowerPC, we generally target only P8 and later by default.
However, generic (older) 64-bit configurations are still an option, and in that
case, partword atomics are not available (e.g. stbcx.). To lower i8/i16 atomics
without true i8/i16 atomic operations, we emulate using i32 atomics in
combination with a bunch of shifting and masking, etc. The amount by which to
shift in little-Endian mode is different from the amount in big-Endian mode (it
is inverted -- meaning we can leave off the xor when computing the amount).
Fixes PR22923.
llvm-svn: 280022
According to the arm arm specifications, 4 bytes are needed for a shift instead
of 8, this was causing the movt instruction to write to a different register
sometimes.
Patch by Walter Erquinigo!
llvm-svn: 280005
After r279649 when getting a vector value from VectorLoopValueMap, we create an
insertelement sequence on-demand if the value has been scalarized instead of
vectorized. We previously inserted this insertelement sequence before the
value's first vector user. However, this insert location is problematic if that
user is the phi node of a first-order recurrence. With this patch, we move the
insertelement sequence after the last scalar instruction we created when
scalarizing the value. Thus, the value's vector definition in the new loop will
immediately follow its scalar definitions. This should fix PR30183.
Reference: https://llvm.org/bugs/show_bug.cgi?id=30183
llvm-svn: 280001
Summary:
The SILoadStoreOptimizer can now look ahead more then one instruction when
looking for instructions to merge, which greatly improves the number of
loads/stores that we are able to merge.
Moving the pass before scheduling avoids increasing register pressure after
the scheduler, so that the scheduler's register pressure estimates will be
more accurate. It also gives more consistent results, since it is no longer
affected by minor scheduling changes.
Reviewers: arsenm
Subscribers: arsenm, kzhuravl, llvm-commits
Differential Revision: https://reviews.llvm.org/D23814
llvm-svn: 279991
This bug shows up with diamonds that share unpredicable, unanalyzable branches.
There's an included test case from Hexagon. What was happening was that we were
attempting to predicate the branch instruction despite the fact that it was
checked to be the same. Now for unanalyzable branches we skip over the branch
instructions when predicating the block.
Differential Revision: https://reviews.llvm.org/D23939
llvm-svn: 279985
Summary:
Calling __asan_poison_stack_memory and __asan_unpoison_stack_memory for small
variables is too expensive.
Code is disabled by default and can be enabled by -asan-experimental-poisoning.
PR27453
Reviewers: eugenis
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23947
llvm-svn: 279984
With r279911 I accidentally regressed the gold/X86/start-lib-common.ll
test for newer golds (v1.12+) that honor the --start-lib/--end-lib.
Remove the alignment which should not be there to make this work with
both old and new gold linkers.
Additionally, now that we have a subdirectory for v1.12+ gold tests,
copy this test there and check specifically for the v1.12+ behavior.
llvm-svn: 279977
Summary:
While walking the use chain for identifying rematerializable values in RS4GC,
add the case where the current value and base value are the same PHI nodes.
This will aid rematerialization of geps and casts instead of relocating.
Reviewers: sanjoy, reames, igor
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23920
llvm-svn: 279975
Assuming the default FP env, we should not treat fdiv and frem any differently in terms of
trapping behavior than any other FP op. Ie, FP ops do not trap with the default FP env.
This matches how we treat the fdiv/frem in IR with isSafeToSpeculativelyExecute() and in
the backend after:
https://reviews.llvm.org/rL279970
llvm-svn: 279973
Summary:
[Coroutines] Part 9: Add cleanup subfunction.
This patch completes coroutine heap allocation elision. Now, the heap elision example from docs\Coroutines.rst compiles and produces expected result (see test/Transform/Coroutines/ex3.ll)
Intrinsic Changes:
* coro.free gets a token parameter tying it to coro.id to allow reliably discovering all coro.frees associated with a particular coroutine.
* coro.id gets an extra parameter that points back to a coroutine function. This allows to check whether a coro.id describes the enclosing function or it belongs to a different function that was later inlined.
CoroSplit now creates three subfunctions:
# f$resume - resume logic
# f$destroy - cleanup logic, followed by a deallocation code
# f$cleanup - just the cleanup code
CoroElide pass during devirtualization replaces coro.destroy with either f$destroy or f$cleanup depending whether heap elision is performed or not.
Other fixes, improvements:
* Fixed buglet in Shape::buildFrame that was not creating coro.save properly if coroutine has more than one suspend point.
* Switched to using variable width suspend index field (no longer limited to 32 bit index field can be as little as i1 or as large as i<whatever-size_t-is>)
Reviewers: majnemer
Subscribers: llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D23844
llvm-svn: 279971
Assuming the default FP env, we should not treat fdiv and frem any differently in terms of
trapping behavior than any other FP op. Ie, FP ops do not trap with the default FP env.
This matches how we treat these ops in IR with isSafeToSpeculativelyExecute(). There's a
similar bug in Constant::canTrap().
This bug manifests in PR29114:
https://llvm.org/bugs/show_bug.cgi?id=29114
...as a sequence of scalar divisions instead of a vector division on x86 for a <3 x float>
type.
Differential Revision: https://reviews.llvm.org/D23974
llvm-svn: 279970
MRI::getMaxLaneMaskForVReg does not always cover the whole register.
For example, on X86 the upper 16 bits of EAX cannot be accessed via
any subregister. Consequently, there is no lane mask that only covers
that part of EAX. The getMaxLaneMaskForVReg will return the union of
the lane masks for all subregisters, and in case of EAX, that union
will not cover the upper 16 bits.
This fixes https://llvm.org/bugs/show_bug.cgi?id=29132
llvm-svn: 279969
Summary:
For shrinking SOPK instructions, we were creating a hint to tell the
register allocator to use the register allocated for src0 for the dst
operand as well. However, this seems to not work sometimes depending
on the order virtual registers are assigned physical registers.
To fix this, I've added a second allocation hint which does the reverse,
asks that the register allocated for dst is used for src0.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: https://reviews.llvm.org/D23862
llvm-svn: 279968
The problem occurs when the Node doesn't updated in place , UpdateNodeOperation() return the node that already exist.
In this case assert fail in PromoteIntegerOperand() , N have 2 results ( val + chain).
Differential Revision: http://reviews.llvm.org/D23756
llvm-svn: 279961
Over eager combing prevents the correct folding of writemasks.
At the moment this occurs for ALL EVEX shuffles, in the future we need to check that the user of the root shuffle is a VSELECT that can fold to a writemask.
llvm-svn: 279934
Fixed a bug in run-time checks for possible memory conflicts inside loop.
The bug is in Low <-> High boundaries calculation. The High boundary should be calculated as "last memory access pointer + element size".
Differential revision: https://reviews.llvm.org/D23176
llvm-svn: 279930
Previously we weren't creating masked logical operations if bitcasts appeared between the logic operation and the select. The IR optimizers can move bitcasts across logic operations and create these cases. To minimize the number of cases we need to handle, this change promotes all logic ops to an i64 vector type just like when only SSE or AVX is available.
Unfortunately, this also has the consequence of making it difficult to select unmasked VPANDD/VPORD/VPXORD in all the cases it was previously used. This is the cause of most of the test change. This shouldn't result in any functional change though.
llvm-svn: 279929
Summary:
This addresses a regression in common handling from the new LTO
API in r278338. Only create a new common if the size is different.
The type comparison against an array type fails when the size is
different but not an array. GlobalMerge does not handle the
array types as well and we lose some global merging opportunities.
Reviewers: mehdi_amini
Subscribers: junbuml, llvm-commits, mehdi_amini
Differential Revision: https://reviews.llvm.org/D23955
llvm-svn: 279911
There's only one use of this for the convenience
of a pattern. I think v_mov_b64_pseudo should also be
moved, but SIFoldOperands does currently make use of it.
llvm-svn: 279901
When global-isel fails on a MachineFunction MF, MF will be cleaned up
and given to SDISel.
Thanks to this fallback, we can already perform correctness test even if
we support only a small portion of the functions in a test.
llvm-svn: 279891
Adds a baseline test for lowering shuffles where the width of the output
vector is not twice the size of the input vectors. Many of those sequences
are suboptimal, and will hopefully be improved in follow-up patches.
llvm-svn: 279888
Summary:
If the scheduler clusters the loads, then the offsets will be sorted,
but it is possible for the scheduler to scheduler loads together
without out explicitly clustering them, which would give us non-sorted
offsets.
Also, we will want to do this if we move the load/store optimizer before
the scheduler.
Reviewers: arsenm
Subscribers: arsenm, llvm-commits, kzhuravl
Differential Revision: https://reviews.llvm.org/D23776
llvm-svn: 279870
Summary:
This is obviously an interesting case because it may motivate code
restructuring or LTO.
Reporting this requires instantiation of ORE in the loop where the call
sites are first gathered. I've checked compile-time
overhead *with* -Rpass-with-hotness and the worst slow-down was 6% in
mcf and quickly tailing off. As before without -Rpass-with-hotness
there is no overhead.
Because this could be a pretty noisy diagnostics, it is currently
qualified as 'verbose'. As of this patch, 'verbose' diagnostics are
only emitted with -Rpass-with-hotness, i.e. when the output is expected
to be filtered.
Reviewers: eraman, chandlerc, davidxl, hfinkel
Subscribers: tejohnson, Prazek, davide, llvm-commits
Differential Revision: https://reviews.llvm.org/D23415
llvm-svn: 279860
In the code to detect fixed-point conversions and make use of AArch64's special
instructions, we weren't prepared for weird types. The fptosi direction got
fixed recently, but not the similar sitofp code.
llvm-svn: 279852
It's unclear how the old
%res(32) = G_ICMP { s32, s32 } intpred(eq), %0, %1
is actually different from an s1 verison
%res(1) = G_ICMP { s1, s32 } intpred(eq), %0, %1
so we'll remove it for now.
llvm-svn: 279843
Summary:
In fuctions that contained debug info but were empty otherwise,
the ARM load/store optimizer could abort. This was because
function MergeReturnIntoLDM handled the special case where a
Machine Basic BLock is empty by calling MBB.empty(). However, this
returns false in presence of debug info, although the function
should be considered empty in the eyes of the load/store optimizer.
This has been fixed by handling the case where searching through the
block finds only debug instructions.
Reviewers: rengolin, dexonsmith, llvm-commits, jmolloy
Subscribers: t.p.northover, aemerson, rengolin, samparker
Differential Revision: https://reviews.llvm.org/D23847
llvm-svn: 279820
This was for some reason skipping operands that are subregisters
instead of keeping the same subregister index.
v_movreld_b32 expects src0 to be the subregister of the tied
super register use/def.
e.g.
v_movreld_b32 v0, v9, <imp-def, tied3> v[0:3], <imp-use, tied2> v[0:3]
was being replaced with
v[4:7] = copy v[0:3]
v_movreld_b32 v0, v9, <imp-def, tied3> v[4:7], <imp-use, tied2> v[4:7],
which really writes to v[0:3]
llvm-svn: 279804
This reverts most of r274613 (AKA r274626) and its follow-ups (r276347, r277289),
due to miscompiles in the test suite. The FastISel change was left in, because
it apparently fixes an unrelated issue.
(Recommit of r279782 which was broken due to a bad merge.)
This fixes 4 out of the 5 test failures in PR29112.
llvm-svn: 279788
This reverts most of r274613 and its follow-ups (r276347, r277289), due to
miscompiles in the test suite. The FastISel change was left in, because it
apparently fixes an unrelated issue.
This fixes 4 out of the 5 test failures in PR29112.
llvm-svn: 279782
Its existence is largely historical, apparently we tried to make ARM object
files look maybe-almost-possibly runnable by putting our best guess at the
actual value into relocated locations. Of course, the real linker then comes
along and can completely change things.
But it should only be there for word-sized and movw/movt relocations. It can't
be encoded in branch relocations, and I've seen it mess up validity
calculations twice in the last couple of weeks so the default is clearly problematic.
llvm-svn: 279773
Summary:
This fixes pr29105. The reason is that lifetime marks creates new
aliasing pointers the original ones, but before this patch aliases
were not checked in performMemCpyToMemSetOptzn.
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D23846
llvm-svn: 279769
The 32-bit variants of these operations don't depend on the bits not being
operated on, so they also naturally model operations narrower than the actual
register width.
llvm-svn: 279760
Fix VPAVG detection to require AVX512BW, not AVX512F for 512-bit widths,
and change associated asserts to assert in the right direction...
This fixes PR29111.
llvm-svn: 279755
when unroll runtime iteration loop.
In llvm::UnrollRuntimeLoopRemainder, if the loop to be unrolled is the inner
loop inside a loop nest, the scalar evolution needs to be dropped for its
parent loop which is done by ScalarEvolution::forgetLoop. However, we can
postpone forgetLoop to the end of UnrollRuntimeLoopRemainder so TripCountSC
expansion can still reuse existing value.
Differential Revision: https://reviews.llvm.org/D23572
llvm-svn: 279748
It is invalid to hoist stores or loads if they are not executed on all paths
from the hoisting point to the exit of the function. In the testcase, there are
paths in the loop that do not execute the stores or the loads, and so hoisting
them within the loop is unsafe.
The problem is that the current implementation of hoistingFromAllPaths is
incomplete: it walks all blocks dominated by the hoisting point, and does not
return false when the loop contains a path on which the hoisted ld/st is
not executed.
Differential Revision: https://reviews.llvm.org/D23843
llvm-svn: 279732
Rename AllVRegsAllocated to NoVRegs. This avoids the connotation of
running after register and simply describes that no vregs are used in
a machine function. With that we can simply compute the property and do
not need to dump/parse it in .mir files.
Differential Revision: http://reviews.llvm.org/D23850
llvm-svn: 279698
tracksSubRegLiveness only depends on the Subtarget and a cl::opt, there
is not need to change it or save/parse it in a .mir file.
Make the field const and move the initialization LiveIntervalAnalysis to the
MachineRegisterInfo constructor. Also cleanup some code and fix some
instances which better use MachineRegisterInfo::subRegLivenessEnabled() instead
of TargetSubtargetInfo::enableSubRegLiveness().
llvm-svn: 279676
The cost of predicating a diamond is only the instructions that are not shared
between the two branches. Additionally If a predicate clobbering instruction
occurs in the shared portion of the branches (e.g. a cond move), it may still
be possible to if convert the sub-cfg. This change handles these two facts by
rescanning the non-shared portion of a diamond sub-cfg to recalculate both the
predication cost and whether both blocks are pred-clobbering.
Fixed 2 bugs before recommitting. Branch instructions must be compared and found
identical before diamond conversion. Also, predicate-clobbering instructions in
the shared prefix disqualifies a potential diamond conversion. Includes tests
for both.
llvm-svn: 279670
A branch-distance to a Thumb function shouldn't be forced to be odd for
CBZ/CBNZ instructions because (assuming it's within range), it's going to be a
valid, even offset.
llvm-svn: 279665
Summary:
This patch implements readlane/readfirstlane intrinsics.
TODO: need to define a new register class to consider the case
that the source could be a vector register or M0.
Reviewed by:
arsenm and tstellarAMD
Differential Revision:
http://reviews.llvm.org/D22489
llvm-svn: 279660
These are no different in load behaviour to the existing ADD/SUB/MUL/DIV scalar ops but were missing from isNonFoldablePartialRegisterLoad
llvm-svn: 279652
In cases where .dwo/.dwp files are guaranteed to be available, skipping
the extra online (in the .o file) inline info can save a substantial
amount of space - see the original r221306 for more details there.
llvm-svn: 279650
This patch unifies the data structures we use for mapping instructions from the
original loop to their corresponding instructions in the new loop. Previously,
we maintained two distinct maps for this purpose: WidenMap and ScalarIVMap.
WidenMap maintained the vector values each instruction from the old loop was
represented with, and ScalarIVMap maintained the scalar values each scalarized
induction variable was represented with. With this patch, all values created
for the new loop are maintained in VectorLoopValueMap.
The change allows for several simplifications. Previously, when an instruction
was scalarized, we had to insert the scalar values into vectors in order to
maintain the mapping in WidenMap. Then, if a user of the scalarized value was
also scalar, we had to extract the scalar values from the temporary vector we
created. We now aovid these unnecessary scalar-to-vector-to-scalar conversions.
If a scalarized value is used by a scalar instruction, the scalar value is used
directly. However, if the scalarized value is needed by a vector instruction,
we generate the needed insertelement instructions on-demand.
A common idiom in several locations in the code (including the scalarization
code), is to first get the vector values an instruction from the original loop
maps to, and then extract a particular scalar value. This patch adds
getScalarValue for this purpose along side getVectorValue as an interface into
VectorLoopValueMap. These functions work together to return the requested
values if they're available or to produce them if they're not.
The mapping has also be made less permissive. Entries can be added to
VectorLoopValue map with the new initVector and initScalar functions.
getVectorValue has been modified to return a constant reference to the mapped
entries.
There's no real functional change with this patch; however, in some cases we
will generate slightly different code. For example, instead of an insertelement
sequence following the definition of an instruction, it will now precede the
first use of that instruction. This can be seen in the test case changes.
Differential Revision: https://reviews.llvm.org/D23169
llvm-svn: 279649
I'm not sure if the `!isa<CallInst>(Inst) &&
!isa<TerminatorInst>(Inst))` bit is correct either, but this fixes the
case we know is broken.
llvm-svn: 279647
Includes adding more general support for the pattern: VZEXT_MOVL(VZEXT_LOAD(ptr)) -> VZEXT_LOAD(ptr)
This has unearthed a couple of latent poor codegen issues (MINSS/MAXSS scalar load folding and MOVDDUP/BROADCAST load folding patterns), which will be fixed shortly.
Its also reduced a couple of tests so that they no longer reach the instruction threshold necessary to be combined to PSHUFB (see PR26183).
llvm-svn: 279646
Summary:
With support now in the new LTO API for caching (r279576), add
optional ThinLTO caching in the gold-plugin.
Reviewers: mehdi_amini
Subscribers: mehdi_amini, llvm-commits
Differential Revision: https://reviews.llvm.org/D23836
llvm-svn: 279631
This patch includes the following changes:
- Included header "Code coverage report" and include the date that the report was created.
- Included title (as specified in a command line option, (i.e llvm-cov -project-title="Simple Test")
- In the summary, list the elf files that the source code file has contributed to.
- Used column heading for "Line No.", "Count No.", Source".
Differential Revision: https://reviews.llvm.org/D23345
llvm-svn: 279628
The register allocator can split a live interval of a register into a set
of smaller intervals. After the allocation of registers is complete, the
rewriter will modify the IR to replace virtual registers with the corres-
ponding physical registers. At this stage, if a register corresponding
to a subregister of a virtual register is used, the rewriter will check
if that subregister is undefined, and if so, it will add the <undef> flag
to the machine operand. The function verifying liveness of the subregis-
ter would assume that it is undefined, unless any of the subranges of the
live interval proves otherwise.
The problem is that the live intervals created during splitting do not
have any subranges, even if the original parent interval did. This could
result in the <undef> flag placed on a register that is actually defined.
Differential Revision: http://reviews.llvm.org/D21189
llvm-svn: 279625
Extend instruction definitions from nearly all ISAs to include
appropriate instruction itineraries. Change MIPS16s gp prologue
generation to use real instructions instead of using a pseudo
instruction.
Reviewers: dsanders, vkalintiris
Differential Review: https://reviews.llvm.org/D23548
llvm-svn: 279623
div/rem instructions in basic blocks that require predication currently prevent
vectorization. This patch extends the existing mechanism for predicating stores
to handle other instructions and leverages it to predicate divs and rems.
Differential Revision: https://reviews.llvm.org/D22918
llvm-svn: 279620
manager, including both plumbing and logic to handle function pass
updates.
There are three fundamentally tied changes here:
1) Plumbing *some* mechanism for updating the CGSCC pass manager as the
CG changes while passes are running.
2) Changing the CGSCC pass manager infrastructure to have support for
the underlying graph to mutate mid-pass run.
3) Actually updating the CG after function passes run.
I can separate them if necessary, but I think its really useful to have
them together as the needs of #3 drove #2, and that in turn drove #1.
The plumbing technique is to extend the "run" method signature with
extra arguments. We provide the call graph that intrinsically is
available as it is the basis of the pass manager's IR units, and an
output parameter that records the results of updating the call graph
during an SCC passes's run. Note that "...UpdateResult" isn't a *great*
name here... suggestions very welcome.
I tried a pretty frustrating number of different data structures and such
for the innards of the update result. Every other one failed for one
reason or another. Sometimes I just couldn't keep the layers of
complexity right in my head. The thing that really worked was to just
directly provide access to the underlying structures used to walk the
call graph so that their updates could be informed by the *particular*
nature of the change to the graph.
The technique for how to make the pass management infrastructure cope
with mutating graphs was also something that took a really, really large
number of iterations to get to a place where I was happy. Here are some
of the considerations that drove the design:
- We operate at three levels within the infrastructure: RefSCC, SCC, and
Node. In each case, we are working bottom up and so we want to
continue to iterate on the "lowest" node as the graph changes. Look at
how we iterate over nodes in an SCC running function passes as those
function passes mutate the CG. We continue to iterate on the "lowest"
SCC, which is the one that continues to contain the function just
processed.
- The call graph structure re-uses SCCs (and RefSCCs) during mutation
events for the *highest* entry in the resulting new subgraph, not the
lowest. This means that it is necessary to continually update the
current SCC or RefSCC as it shifts. This is really surprising and
subtle, and took a long time for me to work out. I actually tried
changing the call graph to provide the opposite behavior, and it
breaks *EVERYTHING*. The graph update algorithms are really deeply
tied to this particualr pattern.
- When SCCs or RefSCCs are split apart and refined and we continually
re-pin our processing to the bottom one in the subgraph, we need to
enqueue the newly formed SCCs and RefSCCs for subsequent processing.
Queuing them presents a few challenges:
1) SCCs and RefSCCs use wildly different iteration strategies at
a high level. We end up needing to converge them on worklist
approaches that can be extended in order to be able to handle the
mutations.
2) The order of the enqueuing need to remain bottom-up post-order so
that we don't get surprising order of visitation for things like
the inliner.
3) We need the worklists to have set semantics so we don't duplicate
things endlessly. We don't need a *persistent* set though because
we always keep processing the bottom node!!!! This is super, super
surprising to me and took a long time to convince myself this is
correct, but I'm pretty sure it is... Once we sink down to the
bottom node, we can't re-split out the same node in any way, and
the postorder of the current queue is fixed and unchanging.
4) We need to make sure that the "current" SCC or RefSCC actually gets
enqueued here such that we re-visit it because we continue
processing a *new*, *bottom* SCC/RefSCC.
- We also need the ability to *skip* SCCs and RefSCCs that get merged
into a larger component. We even need the ability to skip *nodes* from
an SCC that are no longer part of that SCC.
This led to the design you see in the patch which uses SetVector-based
worklists. The RefSCC worklist is always empty until an update occurs
and is just used to handle those RefSCCs created by updates as the
others don't even exist yet and are formed on-demand during the
bottom-up walk. The SCC worklist is pre-populated from the RefSCC, and
we push new SCCs onto it and blacklist existing SCCs on it to get the
desired processing.
We then *directly* update these when updating the call graph as I was
never able to find a satisfactory abstraction around the update
strategy.
Finally, we need to compute the updates for function passes. This is
mostly used as an initial customer of all the update mechanisms to drive
their design to at least cover some real set of use cases. There are
a bunch of interesting things that came out of doing this:
- It is really nice to do this a function at a time because that
function is likely hot in the cache. This means we want even the
function pass adaptor to support online updates to the call graph!
- To update the call graph after arbitrary function pass mutations is
quite hard. We have to build a fairly comprehensive set of
data structures and then process them. Fortunately, some of this code
is related to the code for building the cal graph in the first place.
Unfortunately, very little of it makes any sense to share because the
nature of what we're doing is so very different. I've factored out the
one part that made sense at least.
- We need to transfer these updates into the various structures for the
CGSCC pass manager. Once those were more sanely worked out, this
became relatively easier. But some of those needs necessitated changes
to the LazyCallGraph interface to make it significantly easier to
extract the changed SCCs from an update operation.
- We also need to update the CGSCC analysis manager as the shape of the
graph changes. When an SCC is merged away we need to clear analyses
associated with it from the analysis manager which we didn't have
support for in the analysis manager infrsatructure. New SCCs are easy!
But then we have the case that the original SCC has its shape changed
but remains in the call graph. There we need to *invalidate* the
analyses associated with it.
- We also need to invalidate analyses after we *finish* processing an
SCC. But the analyses we need to invalidate here are *only those for
the newly updated SCC*!!! Because we only continue processing the
bottom SCC, if we split SCCs apart the original one gets invalidated
once when its shape changes and is not processed farther so its
analyses will be correct. It is the bottom SCC which continues being
processed and needs to have the "normal" invalidation done based on
the preserved analyses set.
All of this is mostly background and context for the changes here.
Many thanks to all the reviewers who helped here. Especially Sanjoy who
caught several interesting bugs in the graph algorithms, David, Sean,
and others who all helped with feedback.
Differential Revision: http://reviews.llvm.org/D21464
llvm-svn: 279618
Summary:
This patch adds coroutine frame building algorithm. Now, simple coroutines such as ex0.ll and ex1.ll (first examples from docs\Coroutines.rst can be compiled).
Documentation and overview is here: http://llvm.org/docs/Coroutines.html.
Upstreaming sequence (rough plan)
1.Add documentation. (https://reviews.llvm.org/D22603)
2.Add coroutine intrinsics. (https://reviews.llvm.org/D22659)
...
7. Split coroutine into subfunctions. (https://reviews.llvm.org/D23461)
8. Coroutine Frame Building algorithm <= we are here
9. Add f.cleanup subfunction.
10+. The rest of the logic
Reviewers: majnemer
Subscribers: mehdi_amini, llvm-commits
Differential Revision: https://reviews.llvm.org/D23586
llvm-svn: 279609
Re-apply this patch, hopefully I will get away without any warnings
in the constructor now.
This patch removes the MachineFunctionAnalysis. Instead we keep a
map from IR Function to MachineFunction in the MachineModuleInfo.
This allows the insertion of ModulePasses into the codegen pipeline
without breaking it because the MachineFunctionAnalysis gets dropped
before a module pass.
Peak memory should stay unchanged without a ModulePass in the codegen
pipeline: Previously the MachineFunction was freed at the end of a codegen
function pipeline because the MachineFunctionAnalysis was dropped; With
this patch the MachineFunction is freed after the AsmPrinter has
finished.
Differential Revision: http://reviews.llvm.org/D23736
llvm-svn: 279602
Simon Pilgrim