The VM layout on iOS is not stable between releases. On 64-bit iOS and
its derivatives we use a dynamic shadow offset that enables ASan to
search for a valid location for the shadow heap on process launch rather
than hardcode it.
This commit extends that approach for 32-bit iOS plus derivatives and
their simulators.
rdar://50645192
rdar://51200372
rdar://51767702
Reviewed By: delcypher
Differential Revision: https://reviews.llvm.org/D63586
llvm-svn: 364105
128/256 bit scalar_to_vectors are canonicalized to (insert_subvector undef, (scalar_to_vector), 0). We have isel patterns that try to match this pattern being used by a vzmovl to use a 128-bit instruction and a subreg_to_reg.
This patch detects the insert_subvector undef portion of this and pulls it through the vzmovl, creating a narrower vzmovl and an insert_subvector allzeroes. We can then match the insertsubvector into a subreg_to_reg operation by itself. Then we can fall back on existing (vzmovl (scalar_to_vector)) patterns.
Note, while the scalar_to_vector case is the motivating case I didn't restrict to just that case. I'm also wondering about shrinking any 256/512 vzmovl to an extract_subvector+vzmovl+insert_subvector(allzeros) but I fear that would have bad implications to shuffle combining.
I also think there is more canonicalization we can do with vzmovl with loads or scalar_to_vector with loads to create vzload.
Differential Revision: https://reviews.llvm.org/D63512
llvm-svn: 364095
With this we can now fully code generate jump tables, which is important for code size.
Differential Revision: https://reviews.llvm.org/D63223
llvm-svn: 364086
This change makes use of the newly refactored SwitchLoweringUtils code from
SelectionDAG to in order to generate jump tables and range checks where appropriate.
Much of this code is ported from SDAG with some modifications. We generate
G_JUMP_TABLE and G_BRJT instructions when JT opportunities are found. This means
that targets which previously relied on the naive one MBB per case stmt
translation will now start falling back until they add support for the new opcodes.
For range checks, we don't generate any previously unused operations. This
just recognizes contiguous ranges of case values and generates a single block per
range. Single case value blocks are just a special case of ranges so we get that
support almost for free.
There are still some optimizations missing that I haven't ported over, and
bit-tests are also unimplemented. This patch series is already complex enough.
Actual arm64 support for selection of jump tables is coming in a later patch.
Differential Revision: https://reviews.llvm.org/D63169
llvm-svn: 364085
This patch introduces a new heuristic for guiding operand reordering. The new "look-ahead" heuristic can look beyond the immediate predecessors. This helps break ties when the immediate predecessors have identical opcodes (see lit test for an example).
Committed on behalf of @vporpo (Vasileios Porpodas)
Differential Revision: https://reviews.llvm.org/D60897
llvm-svn: 364084
We already use vmovq for v2i64/v2f64 vzmovl. But we were using a
blendpd+xorpd for v4i64/v4f64/v8i64/v8f64 under opt speed. Or
movsd+xorpd under optsize.
I think the blend with 0 or movss/d is only needed for
vXi32 where we don't have an instruction that can move 32
bits from one xmm to another while zeroing upper bits.
movq is no worse than blendpd on any known CPUs.
llvm-svn: 364079
We sometimes get poor code size because constants of types < 32b are legalized
as 32 bit G_CONSTANTs with a truncate to fit. This works but means that the
localizer can no longer sink them (although it's possible to extend it to do so).
On AArch64 however s8 and s16 constants can be selected in the same way as s32
constants, with a mov pseudo into a W register. If we make s8 and s16 constants
legal then we can avoid unnecessary truncates, they can be CSE'd, and the
localizer can sink them as normal.
There is a caveat: if the user of a smaller constant has to widen the sources,
we end up with an anyext of the smaller typed G_CONSTANT. This can cause
regressions because of the additional extend and missed pattern matching. To
remedy this, there's a new artifact combiner to generate the wider G_CONSTANT
if it's legal for the target.
Differential Revision: https://reviews.llvm.org/D63587
llvm-svn: 364075
We do not have to spread using the precompiled binaries in the tests,
when we can use YAML. This patch removes the dynrel.elf binary and adds
a few comments to the test cases.
Differential revision: https://reviews.llvm.org/D63641
llvm-svn: 364052
Summary:
The motivation for this was to propagate fast-math flags like nnan and
ninf on vector floating point operations to the corresponding scalar
operations to take advantage of follow-on optimizations. But I think
the same argument applies to all of our IR flags: if they apply to the
vector operation then they also apply to all the individual scalar
operations, and they might enable follow-on optimizations.
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D63593
llvm-svn: 364051
There are some test that are splitted into main part + input yaml for no visible reason.
This patch inines the yaml part for the 3 test cases I found.
Differential revision: https://reviews.llvm.org/D63644
llvm-svn: 364049
Summary:
LLVM Allows Targets to provide information that guides optimisations
made to LLVM IR. This is done with callbacks on a TargetTransformInfo object.
This patch adds a TargetTransformInfo class for RISC-V. This will allow us to
implement RISC-V specific callbacks as they become necessary.
This commit also adds the getIntImmCost callbacks, and tests them with a simple
constant hoisting test. Our immediate costs are on the conservative side, for
the moment, but we prevent hoisting in most circumstances anyway.
Previous review was on D63007
Reviewers: asb, luismarques
Reviewed By: asb
Subscribers: ributzka, MaskRay, llvm-commits, Jim, benna, psnobl, jocewei, PkmX, rkruppe, the_o, brucehoult, MartinMosbeck, rogfer01, edward-jones, zzheng, jrtc27, shiva0217, kito-cheng, niosHD, sabuasal, apazos, simoncook, johnrusso, rbar, hiraditya, mgorny
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D63433
llvm-svn: 364046
This patch teaches the bottleneck analysis how to identify and print the most
expensive sequence of instructions according to the simulation. Fixes PR37494.
The goal is to help users identify the sequence of instruction which is most
critical for performance.
A dependency graph is internally used by the bottleneck analysis to describe
data dependencies and processor resource interferences between instructions.
There is one node in the graph for every instruction in the input assembly
sequence. The number of nodes in the graph is independent from the number of
iterations simulated by the tool. It means that a single node of the graph
represents all the possible instances of a same instruction contributed by the
simulated iterations.
Edges are dynamically "discovered" by the bottleneck analysis by observing
instruction state transitions and "backend pressure increase" events generated
by the Execute stage. Information from the events is used to identify critical
dependencies, and materialize edges in the graph. A dependency edge is uniquely
identified by a pair of node identifiers plus an instance of struct
DependencyEdge::Dependency (which provides more details about the actual
dependency kind).
The bottleneck analysis internally ranks dependency edges based on their impact
on the runtime (see field DependencyEdge::Dependency::Cost). To this end, each
edge of the graph has an associated cost. By default, the cost of an edge is a
function of its latency (in cycles). In practice, the cost of an edge is also a
function of the number of cycles where the dependency has been seen as
'contributing to backend pressure increases'. The idea is that the higher the
cost of an edge, the higher is the impact of the dependency on performance. To
put it in another way, the cost of an edge is a measure of criticality for
performance.
Note how a same edge may be found in multiple iteration of the simulated loop.
The logic that adds new edges to the graph checks if an equivalent dependency
already exists (duplicate edges are not allowed). If an equivalent dependency
edge is found, field DependencyEdge::Frequency of that edge is incremented by
one, and the new cost is cumulatively added to the existing edge cost.
At the end of simulation, costs are propagated to nodes through the edges of the
graph. The goal is to identify a critical sequence from a node of the root-set
(composed by node of the graph with no predecessors) to a 'sink node' with no
successors. Note that the graph is intentionally kept acyclic to minimize the
complexity of the critical sequence computation algorithm (complexity is
currently linear in the number of nodes in the graph).
The critical path is finally computed as a sequence of dependency edges. For
edges describing processor resource interferences, the view also prints a
so-called "interference probability" value (by dividing field
DependencyEdge::Frequency by the total number of iterations).
Examples of critical sequence computations can be found in tests added/modified
by this patch.
On output streams that support colored output, instructions from the critical
sequence are rendered with a different color.
Strictly speaking the analysis conducted by the bottleneck analysis view is not
a critical path analysis. The cost of an edge doesn't only depend on the
dependency latency. More importantly, the cost of a same edge may be computed
differently by different iterations.
The number of dependencies is discovered dynamically based on the events
generated by the simulator. However, their number is not fixed. This is
especially true for edges that model processor resource interferences; an
interference may not occur in every iteration. For that reason, it makes sense
to also print out a "probability of interference".
By construction, the accuracy of this analysis (as always) is strongly dependent
on the simulation (and therefore the quality of the information available in the
scheduling model).
That being said, the critical sequence effectively identifies a performance
criticality. Instructions from that sequence are expected to have a very big
impact on performance. So, users can take advantage of this information to focus
their attention on specific interactions between instructions.
In my experience, it works quite well in practice, and produces useful
output (in a reasonable amount time).
Differential Revision: https://reviews.llvm.org/D63543
llvm-svn: 364045
These instructions let you load half a vector register at once from
two general-purpose registers, or vice versa.
The assembly syntax for these instructions mentions the vector
register name twice. For the move _into_ a vector register, the MC
operand list also has to mention the register name twice (once as the
output, and once as an input to represent where the unchanged half of
the output register comes from). So we can conveniently assign one of
the two asm operands to be the output $Qd, and the other $QdSrc, which
avoids confusing the auto-generated AsmMatcher too much. For the move
_from_ a vector register, there's no way to get round the fact that
both instances of that register name have to be inputs, so we need a
custom AsmMatchConverter to avoid generating two separate output MC
operands. (And even that wouldn't have worked if it hadn't been for
D60695.)
Reviewers: dmgreen, samparker, SjoerdMeijer, t.p.northover
Subscribers: javed.absar, kristof.beyls, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62679
llvm-svn: 364041
This adds the `MVE_qDest_rSrc` superclass and all its instances, plus
a few other instructions that also take a scalar input register or two.
I've also belatedly added custom diagnostic messages to the operand
classes for odd- and even-numbered GPRs, which required matching
changes in two of the existing MVE assembly test files.
Reviewers: dmgreen, samparker, SjoerdMeijer, t.p.northover
Subscribers: javed.absar, kristof.beyls, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62678
llvm-svn: 364040
llvm-mc or clang with -g normally produces debug info describing the
assembler source itself; however, if that source already contains some
.file/.loc directives, we should instead emit the debug info described
by those directives. For certain assembler sources seen in the wild
(particularly in the Chrome build) this was causing a crash due to
incorrect assumptions about legal sequences of assembler source text.
Fixes PR38994.
Differential Revision: https://reviews.llvm.org/D63573
llvm-svn: 364039
The sat add/sub tests still have unnecessary extract_subvector((vandnps ymm, ymm), 0) uses that should be split to (vandnps (extract_subvector(ymm, 0), extract_subvector(ymm, 0)), but its getting better.
llvm-svn: 364038
Summary:
This adds the `MVE_qDest_qSrc` superclass and all instructions that
inherit from it. It's not the complete class of _everything_ with a
q-register as both destination and source; it's a subset of them that
all have similar encodings (but it would have been hopelessly unwieldy
to call it anything like MVE_111x11100).
This category includes add/sub with carry; long multiplies; halving
multiplies; multiply and accumulate, and some more complex
instructions.
Reviewers: dmgreen, samparker, SjoerdMeijer, t.p.northover
Subscribers: javed.absar, kristof.beyls, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62677
llvm-svn: 364037
Many LLVM-based tools already support response files (i.e. files
containing a list of options, specified with '@'). This change simply
updates the documentation and help text for some of these tools to
include it. I haven't attempted to fix all tools, just a selection that
I am interested in.
I've taken the opportunity to add some tests for --help behaviour, where
they were missing. We could expand these tests, but I don't think that's
within scope of this patch.
This fixes https://bugs.llvm.org/show_bug.cgi?id=42233 and
https://bugs.llvm.org/show_bug.cgi?id=42236.
Reviewed by: grimar, MaskRay, jkorous
Differential Revision: https://reviews.llvm.org/D63597
llvm-svn: 364036
--help and -h are automatically supported by the command-line parser,
unless overridden by the tool. The behaviour of the PrintHelpMessage
being used for -h prior to this patch is subtly different to that
provided by --help automatically (it omits certain elements of help text
and options, such as --help-list), so overriding the default is not
desirable, without good reason. This patch removes the explicit
specification of -h and its behaviour, so that the default behaviour is
used.
Reviewed by: hintonda
Differential Revision: https://reviews.llvm.org/D63565
llvm-svn: 364029
Summary:
These take a pair of vector register to compare, and a comparison type
(written in the form of an Arm condition suffix); they output a vector
of booleans in the VPR register, where predication can conveniently
use them.
Reviewers: dmgreen, samparker, SjoerdMeijer, t.p.northover
Subscribers: javed.absar, kristof.beyls, hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D62676
llvm-svn: 364027
G_INTTOPTR can prevent the localizer from moving G_CONSTANTs, but since it's
essentially a side effect free cast instruction we can remat both instructions.
This patch changes the localizer to enable localization of the chains by
iterating over the entry block instructions in reverse order. That way, uses will
localized first, and then the defs are free to be localized as well.
This also changes the previous SmallPtrSet of localized instructions to use a
SetVector instead. We're dealing with pointers and need deterministic iteration
order.
Overall, this change improves ARM64 -O0 CTMark code size by around 0.7% geomean.
Differential Revision: https://reviews.llvm.org/D63630
llvm-svn: 364001
Also, add a FIXME for the unsafe transform on a unary FNeg. A unary FNeg can only be transformed to a FMul by -1.0 when the nnan flag is present. The unary FNeg project is a WIP, so the unsafe transformation is acceptable until that work is complete.
The bogus assert with introduced in D63445.
llvm-svn: 363998
As discussed in PR42314:
https://bugs.llvm.org/show_bug.cgi?id=42314
Improving the canonicalization for these patterns:
rL363956
...means we should adjust/enhance the related simplification.
https://rise4fun.com/Alive/w1cp
Name: isPow2 or zero
%x = and i32 %xx, 2048
%a = add i32 %x, -1
%r = and i32 %a, %x
=>
%r = i32 0
llvm-svn: 363997
Every called function could possibly need this to calculate the
absolute address of stack objectst, and this avoids inserting a copy
around every call site in the kernel. It's also somewhat cleaner to
keep this in a callee saved SGPR.
llvm-svn: 363990
Currently, many profiling tests on Solaris FAIL like
Command Output (stderr):
--
Undefined first referenced
symbol in file
__llvm_profile_register_names_function /tmp/lit_tmp_Nqu4eh/infinite_loop-9dc638.o
__llvm_profile_register_function /tmp/lit_tmp_Nqu4eh/infinite_loop-9dc638.o
Solaris 11.4 ld supports the non-standard GNU ld extension of adding
__start_SECNAME and __stop_SECNAME labels to sections whose names are valid
as C identifiers. Given that we already use Solaris 11.4-only features
like ld -z gnu-version-script-compat and fully working .preinit_array
support in compiler-rt, we don't need to worry about older versions of
Solaris ld.
The patch documents that support (although the comment in
lib/Transforms/Instrumentation/InstrProfiling.cpp
(needsRuntimeRegistrationOfSectionRange) is quite cryptic what it's
actually about), and adapts the affected testcase not to expect the
alternativeq __llvm_profile_register_functions and __llvm_profile_init.
It fixes all affected tests.
Tested on amd64-pc-solaris2.11.
Differential Revision: https://reviews.llvm.org/D41111
llvm-svn: 363984
Summary:
The getClobberingMemoryAccess API checks for clobbering accesses in a loop by walking the backedge. This may check if a memory access is being
clobbered by the loop in a previous iteration, depending how smart AA got over the course of the updates in MemorySSA (it does not occur when built from scratch).
If no clobbering access is found inside the loop, it will optimize to an access outside the loop. This however does not mean that access is safe to sink.
Given:
```
for i
load a[i]
store a[i]
```
The access corresponding to the load can be optimized to outside the loop, and the load can be hoisted. But it is incorrect to sink it.
In order to sink the load, we'd need to check no Def clobbers the Use in the same iteration. With this patch we currently restrict sinking to either
Defs not existing in the loop, or Defs preceding the load in the same block. An easy extension is to ensure the load (Use) post-dominates all Defs.
Caught by PR42294.
This issue also shed light on the converse problem: hoisting stores in this same scenario would be illegal. With this patch we restrict
hoisting of stores to the case when their corresponding Defs are dominating all Uses in the loop.
Reviewers: george.burgess.iv
Subscribers: jlebar, Prazek, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D63582
llvm-svn: 363982
I added a canonicalization to create this general pattern in:
rL363956
But as noted in PR42314:
https://bugs.llvm.org/show_bug.cgi?id=42314#c11
...we have a (potentially expensive) simplification for the version
of the code that we just canonicalized away from, so we should
add/adjust that code to match.
llvm-svn: 363981
This fixes CodeGen/available-externally-suppress.c when the new pass manager is
turned on by default. available_externally was not emitted during -O2 -flto
runs when it should still be retained for link time inlining purposes. This can
be fixed by checking that we aren't LTOPrelinking when adding the
EliminateAvailableExternallyPass.
Differential Revision: https://reviews.llvm.org/D63580
llvm-svn: 363971
