Davide Italiano [Sat, 29 Apr 2017 00:18:26 +0000 (00:18 +0000)]
[LoopUnswitch] Make DEBUG output more readable (part 2).
I fixed my miscompile in r301722 and I hope I don't have to take
a look at this code again now that Chandler has a new LoopUnswitch
pass, but maybe this could be of use for somebody else in the
meanwhile.
Fuzzer: Mark test/cxxstring.test UNSUPPORTED: windows
This has been mysteriously failing since r301593, which cleaned up the
types of things like size_t and SIZE_MAX for freestanding targets. Reid
and Kostya suggested marking it as UNSUPPORTED on windows, given that no
one has been able to reproduce locally.
Hans Wennborg [Fri, 28 Apr 2017 23:01:32 +0000 (23:01 +0000)]
Revert r301697 "[IR] Make add/remove Attributes use AttrBuilder instead of AttributeList"
This broke the Clang build. (Clang-side patch missing?)
Original commit message:
> [IR] Make add/remove Attributes use AttrBuilder instead of
> AttributeList
>
> This change cleans up call sites and avoids creating temporary
> AttributeList objects.
>
> NFC
Adrian Prantl [Fri, 28 Apr 2017 22:25:46 +0000 (22:25 +0000)]
Remove line and file from DINamespace.
Fixes the issue highlighted in
http://lists.llvm.org/pipermail/cfe-dev/2014-June/037500.html.
The DW_AT_decl_file and DW_AT_decl_line attributes on namespaces can
prevent LLVM from uniquing types that are in the same namespace. They
also don't carry any meaningful information.
Matt Arsenault [Fri, 28 Apr 2017 22:18:19 +0000 (22:18 +0000)]
InferAddressSpaces: Avoid looking up deleted values
While looking at pure addressing expressions, it's possible
for the value to appear later in Postorder.
I haven't been able to come up with a testcase where this
exhibits an actual issue, but if you insert a dump before
the value map lookup, a few testcases crash.
Matt Arsenault [Fri, 28 Apr 2017 22:18:08 +0000 (22:18 +0000)]
InferAddressSpaces: Infer from just addrspacecasts
Eliminates some more cases where some subset of the addressing
computation remains flat. Some cases with addrspacecasts
in nested constant expressions are still left behind however.
[ConstantRange] Use APInt::isNullValue rather than APInt::isMinValue where it would make more sense to thing of 0 as 0 rather than the minimum unsigned value. NFC
[APInt] Add an isNullValue method to check for all bits being zero. Use it in a couple internal methods where it makes more sense than isMinValue or !getBoolValue. NFC
I used Null rather than Zero to match the getNullValue method name.
There are some other places outside APInt where isNullValue would be more readable than isMinValue even though they do the same thing. I'll update those in future patches.
Davide Italiano [Fri, 28 Apr 2017 21:30:50 +0000 (21:30 +0000)]
[LoopUnswitch] Make DEBUG output more readable.
While debugging a miscompile I realized loopunswitch doesn't
put newlines when printing the instruction being replacement.
Ending up with a single line with many instruction replaced isn't
the best for readability and/or mental sanity.
Matt Arsenault [Fri, 28 Apr 2017 21:01:46 +0000 (21:01 +0000)]
TableGen: Add IntrHasSideEffects property for intrinsics
The IntrNoMem, IntrReadMem, IntrWriteMem, and IntrArgMemOnly intrinsic
properties differ from their corresponding LLVM IR attributes by specifying
that the intrinsic, in addition to its memory properties, has no other side
effects.
The IntrHasSideEffects flag used in combination with one of the memory flags
listed above, makes it possible to define an intrinsic such that its
properties at the CodeGen layer match its properties at the IR layer.
The method is called "get *Param* Alignment", and is only used for
return values exactly once, so it should take argument indices, not
attribute indices.
Adds a new method finalizeLowering to TargetLoweringBase. This is in
preparation for an upcoming commit.
This function is meant for target specific adjustments to
MachineFrameInfo or register reservations.
Move the freezeRegisters() and the hasCopyImplyingStackAdjustment()
handling into the new function to prove the concept. As an added bonus
GlobalISel no longer missed the hasCopyImplyingStackAdjustment()
handling with this.
Bob Haarman [Fri, 28 Apr 2017 20:17:15 +0000 (20:17 +0000)]
limit to 2 parallel links when using thinlto
Summary:
When using ThinLTO, the linker performs its own parallelism. This
change limits the number of parallel link jobs that Ninja will issue
to keep the total number of threads reasonable when linking with
ThinLTO.
Use Argument::hasAttribute and AttributeList::ReturnIndex more
This eliminates many extra 'Idx' induction variables in loops over
arguments in CodeGen/ and Target/. It also reduces the number of places
where we assume that ReturnIndex is 0 and that we should add one to
argument numbers to get the corresponding attribute list index.
Bitcode: Do not remove empty summary entries when reading a per-module summary.
This became no longer necessary after D19462 landed, and will be incompatible
with an upcoming change to the summary data structures that changes how we
represent references.
. swap 4-bit register encoding, 16-bit offset and 32-bit imm to support big endian archs
. add a test
Reported-by: David S. Miller <davem@davemloft.net> Signed-off-by: Alexei Starovoitov <ast@kernel.org>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@301653 91177308-0d34-0410-b5e6-96231b3b80d8
[InlineCost] Improve the cost heuristic for Switch
Summary:
The motivation example is like below which has 13 cases but only 2 distinct targets
```
lor.lhs.false2: ; preds = %if.then
switch i32 %Status, label %if.then27 [
i32 -7012, label %if.end35
i32 -10008, label %if.end35
i32 -10016, label %if.end35
i32 15000, label %if.end35
i32 14013, label %if.end35
i32 10114, label %if.end35
i32 10107, label %if.end35
i32 10105, label %if.end35
i32 10013, label %if.end35
i32 10011, label %if.end35
i32 7008, label %if.end35
i32 7007, label %if.end35
i32 5002, label %if.end35
]
```
which is compiled into a balanced binary tree like this on AArch64 (similar on X86)
```
.LBB853_9: // %lor.lhs.false2
mov w8, #10012
cmp w19, w8
b.gt .LBB853_14
// BB#10: // %lor.lhs.false2
mov w8, #5001
cmp w19, w8
b.gt .LBB853_18
// BB#11: // %lor.lhs.false2
mov w8, #-10016
cmp w19, w8
b.eq .LBB853_23
// BB#12: // %lor.lhs.false2
mov w8, #-10008
cmp w19, w8
b.eq .LBB853_23
// BB#13: // %lor.lhs.false2
mov w8, #-7012
cmp w19, w8
b.eq .LBB853_23
b .LBB853_3
.LBB853_14: // %lor.lhs.false2
mov w8, #14012
cmp w19, w8
b.gt .LBB853_21
// BB#15: // %lor.lhs.false2
mov w8, #-10105
add w8, w19, w8
cmp w8, #9 // =9
b.hi .LBB853_17
// BB#16: // %lor.lhs.false2
orr w9, wzr, #0x1
lsl w8, w9, w8
mov w9, #517
and w8, w8, w9
cbnz w8, .LBB853_23
.LBB853_17: // %lor.lhs.false2
mov w8, #10013
cmp w19, w8
b.eq .LBB853_23
b .LBB853_3
.LBB853_18: // %lor.lhs.false2
mov w8, #-7007
add w8, w19, w8
cmp w8, #2 // =2
b.lo .LBB853_23
// BB#19: // %lor.lhs.false2
mov w8, #5002
cmp w19, w8
b.eq .LBB853_23
// BB#20: // %lor.lhs.false2
mov w8, #10011
cmp w19, w8
b.eq .LBB853_23
b .LBB853_3
.LBB853_21: // %lor.lhs.false2
mov w8, #14013
cmp w19, w8
b.eq .LBB853_23
// BB#22: // %lor.lhs.false2
mov w8, #15000
cmp w19, w8
b.ne .LBB853_3
```
However, the inline cost model estimates the cost to be linear with the number
of distinct targets and the cost of the above switch is just 2 InstrCosts.
The function containing this switch is then inlined about 900 times.
This change use the general way of switch lowering for the inline heuristic. It
etimate the number of case clusters with the suitability check for a jump table
or bit test. Considering the binary search tree built for the clusters, this
change modifies the model to be linear with the size of the balanced binary
tree. The model is off by default for now :
-inline-generic-switch-cost=false
This change was originally proposed by Haicheng in D29870.
Reviewers: hans, bmakam, chandlerc, eraman, haicheng, mcrosier
[ARM] GlobalISel: Get rid of ARMInstructionSelector.h. NFC.
Declare the ARMInstructionSelector in an anonymous namespace, to make it
more in line with the other targets which were migrated to this in
r299637 in order to avoid TableGen'erated headers being included in
non-GlobalISel builds.
Andrew Ng [Fri, 28 Apr 2017 08:44:30 +0000 (08:44 +0000)]
[DebugInfo][X86] Improve X86 Optimize LEAs handling of debug values.
This is a follow up to the fix in r298360 to improve the handling of debug
values when redundant LEAs are removed. The fix in r298360 effectively
discarded the debug values. This patch now attempts to preserve the debug
values by using the DWARF DW_OP_stack_value operation via prependDIExpr.
Moved functions appendOffset and prependDIExpr from Local.cpp to
DebugInfoMetadata.cpp and made them available as static member functions of
DIExpression.
Max Kazantsev [Fri, 28 Apr 2017 06:05:48 +0000 (06:05 +0000)]
[EarlyCSE] Remove guards with conditions known to be true
If a condition is calculated only once, and there are multiple guards on this condition, we should be able
to remove all guards dominated by the first of them. This patch allows EarlyCSE to try to find the condition
of a guard among the known values, and if it is true, remove the guard. Otherwise we keep the guard and
mark its condition as 'true' for future consideration.
[SelectionDAG] Use KnownBits struct in DAG's computeKnownBits and simplifyDemandedBits
This patch replaces the separate APInts for KnownZero/KnownOne with a single KnownBits struct. This is similar to what was done to ValueTracking's version recently.
This is largely a mechanical transformation from KnownZero to Known.Zero.
[SelectionDAG] Use various APInt methods to reduce temporary APInt creation
This patch uses various APInt methods to reduce the number of temporary APInts. These were all found while working through converting SelectionDAG's computeKnownBits to also use the KnownBits struct recently added to the ValueTracking version.
Sanjoy Das [Fri, 28 Apr 2017 04:48:42 +0000 (04:48 +0000)]
[StackMaps] Increase the size of the "location size" field
Summary:
In some cases LLVM (especially the SLP vectorizer) will create vectors
that are 256 bytes (or larger). Given that this is intentional[0] is
likely to get more common, this patch updates the StackMap binary
format to deal with the spill locations for said vectors.
This change also bumps the stack map version from 2 to 3.
COFF Import libraries which use the obsolete CONSTANT export are
supposed to get two symbols, one with the `_imp_` prefix and one
without. Ensure that we expose both for iteration. This is necessary
to fix the librarian with COFF CONSTANT exports.
[llvm-pdbdump] Allow printing only a portion of a stream.
When dumping raw data from a stream, you might know the offset
of a certain record you're interested in, as well as how long
that record is. Previously, you had to dump the entire stream
and wade through the bytes to find the interesting record.
This patch allows you to specify an offset and length on the
command line, and it will only dump the requested range.
Use a combination of !associated, comdat, @llvm.compiler.used and
custom sections to allow dead stripping of globals and their asan
metadata. Sometimes.
Currently this works on LLD, which supports SHF_LINK_ORDER with
sh_link pointing to the associated section.
This also works on BFD, which seems to treat comdats as
all-or-nothing with respect to linker GC. There is a weird quirk
where the "first" global in each link is never GC-ed because of the
section symbols.
At this moment it does not work on Gold (as in the globals are never
stripped).
This is a second re-land of r298158. This time, this feature is
limited to -fdata-sections builds.
The only reason not to is global registration, which can be
TU-specific. This is not the case when there are no instrumented
globals. This is also limited to ELF targets, because MachO does
not have comdat, and COFF linkers may GC comdat constructors.
The benefit of this is a lot less __asan_init() calls: one per DSO
instead of one per TU. It's also necessary for the upcoming
gc-sections-for-globals change on Linux, where multiple references to
section start symbols trigger quadratic behaviour in gold linker.
This is a second re-land of r298756. This time with a flag to disable
the whole thing to avoid a bug in the gold linker:
https://sourceware.org/bugzilla/show_bug.cgi?id=19002
This patch dumps the raw bytes of the .rsrc sections that
are present in COFF object and executable files. Subsequent
patches will parse this information and dump in a more human
readable format.
Differential Revision: https://reviews.llvm.org/D32463
Patch By: Eric Beckmann
[PM/LoopUnswitch] Introduce a new, simpler loop unswitch pass.
Currently, this pass only focuses on *trivial* loop unswitching. At that
reduced problem it remains significantly better than the current loop
unswitch:
- Old pass is worse than cubic complexity. New pass is (I think) linear.
- New pass is much simpler in its design by focusing on full unswitching. (See
below for details on this).
- New pass doesn't carry state for thresholds between pass iterations.
- New pass doesn't carry state for correctness (both miscompile and
infloop) between pass iterations.
- New pass produces substantially better code after unswitching.
- New pass can handle more trivial unswitch cases.
- New pass doesn't recompute the dominator tree for the entire function
and instead incrementally updates it.
I've ported all of the trivial unswitching test cases from the old pass
to the new one to make sure that major functionality isn't lost in the
process. For several of the test cases I've worked to improve the
precision and rigor of the CHECKs, but for many I've just updated them
to handle the new IR produced.
My initial motivation was the fact that the old pass carried state in
very unreliable ways between pass iterations, and these mechansims were
incompatible with the new pass manager. However, I discovered many more
improvements to make along the way.
This pass makes two very significant assumptions that enable most of these
improvements:
1) Focus on *full* unswitching -- that is, completely removing whatever
control flow construct is being unswitched from the loop. In the case
of trivial unswitching, this means removing the trivial (exiting)
edge. In non-trivial unswitching, this means removing the branch or
switch itself. This is in opposition to *partial* unswitching where
some part of the unswitched control flow remains in the loop. Partial
unswitching only really applies to switches and to folded branches.
These are very similar to full unrolling and partial unrolling. The
full form is an effective canonicalization, the partial form needs
a complex cost model, cannot be iterated, isn't canonicalizing, and
should be a separate pass that runs very late (much like unrolling).
2) Leverage LLVM's Loop machinery to the fullest. The original unswitch
dates from a time when a great deal of LLVM's loop infrastructure was
missing, ineffective, and/or unreliable. As a consequence, a lot of
complexity was added which we no longer need.
With these two overarching principles, I think we can build a fast and
effective unswitcher that fits in well in the new PM and in the
canonicalization pipeline. Some of the remaining functionality around
partial unswitching may not be relevant today (not many test cases or
benchmarks I can find) but if they are I'd like to add support for them
as a separate layer that runs very late in the pipeline.
Purely to make reviewing and introducing this code more manageable, I've
split this into first a trivial-unswitch-only pass and in the next patch
I'll add support for full non-trivial unswitching against a *fixed*
threshold, exactly like full unrolling. I even plan to re-use the
unrolling thresholds, as these are incredibly similar cost tradeoffs:
we're cloning a loop body in order to end up with simplified control
flow. We should only do that when the total growth is reasonably small.
One of the biggest changes with this pass compared to the previous one
is that previously, each individual trivial exiting edge from a switch
was unswitched separately as a branch. Now, we unswitch the entire
switch at once, with cases going to the various destinations. This lets
us unswitch multiple exiting edges in a single operation and also avoids
numerous extremely bad behaviors, where we would introduce 1000s of
branches to test for thousands of possible values, all of which would
take the exact same exit path bypassing the loop. Now we will use
a switch with 1000s of cases that can be efficiently lowered into
a jumptable. This avoids relying on somehow forming a switch out of the
branches or getting horrible code if that fails for any reason.
Another significant change is that this pass actively updates the CFG
based on unswitching. For trivial unswitching, this is actually very
easy because of the definition of loop simplified form. Doing this makes
the code coming out of loop unswitch dramatically more friendly. We
still should run loop-simplifycfg (at the least) after this to clean up,
but it will have to do a lot less work.
Finally, this pass makes much fewer attempts to simplify instructions
based on the unswitch. Something like loop-instsimplify, instcombine, or
GVN can be used to do increasingly powerful simplifications based on the
now dominating predicate. The old simplifications are things that
something like loop-instsimplify should get today or a very, very basic
loop-instcombine could get. Keeping that logic separate is a big
simplifying technique.
Most of the code in this pass that isn't in the old one has to do with
achieving specific goals:
- Updating the dominator tree as we go
- Unswitching all cases in a switch in a single step.
I think it is still shorter than just the trivial unswitching code in
the old pass despite having this functionality.
Sanjoy Das [Thu, 27 Apr 2017 17:17:16 +0000 (17:17 +0000)]
Use a pointer type for target frame indices during statepoint lowering
Summary:
The type of the target frame index is intptr, not the type of the value we're
going to store into it. Without this change we crash in the attached test case
when trying to type-legalize a TargetFrameIndex.
Patchpoint lowering types the target frame index as intptr as well.
Refactor DynamicLibrary so searching for a symbol will have a defined order and
libraries are properly unloaded when llvm_shutdown is called.
Summary:
This was mostly affecting usage of the JIT, where storing the library handles in
a set made iteration unordered/undefined. This lead to disagreement between the
JIT and native code as to what the address and implementation of particularly on
Windows with stdlib functions:
JIT: putenv_s("TEST", "VALUE") // called msvcrt.dll, putenv_s
JIT: getenv("TEST") -> "VALUE" // called msvcrt.dll, getenv
Native: getenv("TEST") -> NULL // called ucrt.dll, getenv
Also fixed is the issue of DynamicLibrary::getPermanentLibrary(0,0) on Windows
not giving priority to the process' symbols as it did on Unix.
[CodeView] Isolate Debug Info Fragments into standalone classes.
Previously parsing of these were all grouped together into a
single master class that could parse any type of debug info
fragment.
With writing forthcoming, the complexity of each individual
fragment is enough to warrant them having their own classes so
that reading and writing of each fragment type can be grouped
together, but isolated from the code for reading and writing
other fragment types.
In doing so, I found a place where parsing code was duplicated
for the FileChecksums fragment, across llvm-readobj and the
CodeView library, and one of the implementations had a bug.
Now that the codepaths are merged, the bug is resolved.
We have a lot of very similarly named classes related to
dealing with module debug info. This patch has NFC, it just
renames some classes to be more descriptive (albeit slightly
more to type). The mapping from old to new class names is as
follows:
Old | New
ModInfo | DbiModuleDescriptor
ModuleSubstream | ModuleDebugFragment
ModStream | ModuleDebugStream
With the corresponding Builder classes renamed accordingly.
[mips][microMIPS] Adding code size reduction pass for MicroMIPS
Author: milena.vujosevic.janicic
Reviewers: sdardis
The code implements size reduction pass for MicroMIPS.
Load and store instructions are examined and transformed, if possible.
lw32 instruction is transformed into 16-bit instruction lwsp
sw32 instruction is transformed into 16-bit instruction swsp
Arithmetic instrcutions are examined and transformed, if possible.
addu32 instruction is transformed into 16-bit instruction addu16
subu32 instruction is transformed into 16-bit instruction subu16
Differential Revision: https://reviews.llvm.org/D15144
Jonas Paulsson [Thu, 27 Apr 2017 11:01:18 +0000 (11:01 +0000)]
[SystemZ] Remove incorrect assert in SystemZTTIImpl
In getCmpSelInstrCost(), CondTy may actually be scalar while ValTy is a
vector when LoopVectorizer is the caller. Therefore the assert that CondTy
must be a vector type if ValTy is was wrong and is now removed.
projects / llvm / log