OSDN Git Service
Lang Hames [Fri, 17 Aug 2018 21:18:18 +0000 (21:18 +0000)]
[ORC] Rename VSO to JITDylib.
VSO was a little close to VDSO (an acronym on Linux for Virtual Dynamic Shared
Object) for comfort. It also risks giving the impression that instances of this
class could be shared between ExecutionSessions, which they can not.
JITDylib seems moderately less confusing, while still hinting at how this
class is intended to be used, i.e. as a JIT-compiled stand-in for a dynamic
library (code that would have been a dynamic library if you had wanted to
compile it ahead of time).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340084
91177308-0d34-0410-b5e6-
96231b3b80d8
Zachary Turner [Fri, 17 Aug 2018 21:18:05 +0000 (21:18 +0000)]
[MS Demangler] Demangle all remaining types of operators.
This demangles all remaining special operators including thunks,
RTTI Descriptors, and local static guard variables.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340083
91177308-0d34-0410-b5e6-
96231b3b80d8
Krzysztof Parzyszek [Fri, 17 Aug 2018 21:12:37 +0000 (21:12 +0000)]
[Hexagon] Remove unused functions from HexagonInstPrinter, NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340081
91177308-0d34-0410-b5e6-
96231b3b80d8
Michael Kruse [Fri, 17 Aug 2018 19:31:41 +0000 (19:31 +0000)]
[AST] Adapt Polly to AnalysisSetTracker changes. NFC.
The method AliasSetTracker::getAliasSetForPointer was removed and replaced by AliasSetTracker::getAliasSetFor for the restructuring in r339930.
Since Polly uses AliasSetTracker::getAliasSetForPointer, a temporary fix has been committed in r339937 with a comment:
Can someone from polly please migrate usage and then delete the wrapper?
This commit is doing exactly that.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340072
91177308-0d34-0410-b5e6-
96231b3b80d8
Jordan Rupprecht [Fri, 17 Aug 2018 19:18:20 +0000 (19:18 +0000)]
Fix windows buildbots by removing : from filenames
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340071
91177308-0d34-0410-b5e6-
96231b3b80d8
Jordan Rupprecht [Fri, 17 Aug 2018 18:51:11 +0000 (18:51 +0000)]
[llvm-objcopy] Add support for -I binary -B <arch>.
Summary:
The -I (--input-target) and -B (--binary-architecture) flags exist but are currently silently ignored. This adds support for -I binary for architectures i386, x86-64 (and alias i386:x86-64), arm, aarch64, sparc, and ppc (powerpc:common64). This is largely based on D41687.
This is done by implementing an additional subclass of Reader, BinaryReader, which works by interpreting the input file as contents for .data field, sets up a synthetic header, and adds additional sections/symbols (e.g. _binary__tmp_data_txt_start).
Reviewers: jakehehrlich, alexshap, jhenderson, javed.absar
Reviewed By: jhenderson
Subscribers: jyknight, nemanjai, kbarton, fedor.sergeev, jrtc27, kristof.beyls, paulsemel, llvm-commits
Differential Revision: https://reviews.llvm.org/D50343
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340070
91177308-0d34-0410-b5e6-
96231b3b80d8
Jun Lim [Fri, 17 Aug 2018 18:40:41 +0000 (18:40 +0000)]
Test commit
I just removed a blank space.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340069
91177308-0d34-0410-b5e6-
96231b3b80d8
Vedant Kumar [Fri, 17 Aug 2018 18:39:19 +0000 (18:39 +0000)]
Remove a hardcoded address in test/DebugInfo/X86/vla-multi.ll
This relaxes a test to make it less brittle.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340068
91177308-0d34-0410-b5e6-
96231b3b80d8
Alina Sbirlea [Fri, 17 Aug 2018 18:37:15 +0000 (18:37 +0000)]
[IDF] Make GD const.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340067
91177308-0d34-0410-b5e6-
96231b3b80d8
Matt Davis [Fri, 17 Aug 2018 18:06:01 +0000 (18:06 +0000)]
[llvm-mca] Reformat a few lines (fix spacing). NFC.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340065
91177308-0d34-0410-b5e6-
96231b3b80d8
Reka Kovacs [Fri, 17 Aug 2018 18:05:38 +0000 (18:05 +0000)]
[Support] NFC: Fix docstring in FileSystem.h.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340063
91177308-0d34-0410-b5e6-
96231b3b80d8
Simon Pilgrim [Fri, 17 Aug 2018 18:03:11 +0000 (18:03 +0000)]
[X86][SSE] Lower constant vXi8 ISD::SRL/ISD::SRA using PMULLW
Extending the concept introduced in D49562, this patch lowers constant vXi8 ISD::SRL/ISD::SRA by zero/sign extending to vXi16 and using PMULLW and then truncating the high 8 bits of the result.
Differential Revision: https://reviews.llvm.org/D50781
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340062
91177308-0d34-0410-b5e6-
96231b3b80d8
Evandro Menezes [Fri, 17 Aug 2018 17:59:53 +0000 (17:59 +0000)]
[InstCombine] Refactor the simplification of pow() (NFC)
Refactor all cases dealing with `exp{,2,10}()` into one function in
preparation for D49273. Otherwise, NFC.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340061
91177308-0d34-0410-b5e6-
96231b3b80d8
Evandro Menezes [Fri, 17 Aug 2018 17:59:38 +0000 (17:59 +0000)]
[NFC] Expand test cases for simplifying pow()
In prepatration for the improvements that D49273 enables.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340060
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Fri, 17 Aug 2018 17:57:25 +0000 (17:57 +0000)]
[X86] Use hasOneUse instead of isOnlyUserOf. NFCI
isOnlyUserOf is a little heavier because it allows the node to be used multiple times by the other node. In this case we are looking at a truncate which only has one operand so we know it can only use it once. Thus hasOneUse is better.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340059
91177308-0d34-0410-b5e6-
96231b3b80d8
Simon Pilgrim [Fri, 17 Aug 2018 17:45:15 +0000 (17:45 +0000)]
[TableGen] Don't separately search for DefaultMode when we're going to iterate the set anyway. NFCI.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340055
91177308-0d34-0410-b5e6-
96231b3b80d8
Alina Sbirlea [Fri, 17 Aug 2018 17:39:15 +0000 (17:39 +0000)]
[IDF] Teach Iterated Dominance Frontier to use a snapshot CFG based on a GraphDiff.
Summary:
Create the ability to compute IDF using a CFG View.
For this, we'll need a new DT created using a list of Updates (to be refactored later to a GraphDiff), and the GraphTraits based on the same GraphDiff.
Reviewers: kuhar, george.burgess.iv, mzolotukhin
Subscribers: sanjoy, jlebar, llvm-commits
Differential Revision: https://reviews.llvm.org/D50675
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340052
91177308-0d34-0410-b5e6-
96231b3b80d8
Teresa Johnson [Fri, 17 Aug 2018 16:53:47 +0000 (16:53 +0000)]
[ThinLTO] Add option for printing import failure reasons
Summary:
Adds the option for the printing of summary information about functions
considered but rejected for importing during the thin link.
Reviewers: davidxl
Subscribers: mehdi_amini, inglorion, eraman, steven_wu, dexonsmith, llvm-commits
Differential Revision: https://reviews.llvm.org/D50881
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340047
91177308-0d34-0410-b5e6-
96231b3b80d8
Zachary Turner [Fri, 17 Aug 2018 16:14:05 +0000 (16:14 +0000)]
[MS Demangler] Rework the way operators are demangled.
Previously, some of the code for actually parsing mangled
operator names was more like formatting code in nature,
and was interspersed with the demangling code which builds
the AST. This means that by the time we got to the printing
code, we had lost all information about what type of operator
we had, and all we were left with was a string that we just
had to print. However, not all operators are actually even
operators. it's basically just a catch-all mangling for
"special names", and for some of the other types it helps
to know when we're actually doing the printing what it is.
This patch changes the way things work by introducing an
OperatorInfo structure and corresponding enumeration. When
we demangle we store the enumeration value and demangled
components separately. This gives more flexibility during
printing.
In doing so, some demanglings of special names which we didn't
previously support come out of this for free, so we now demangle
those.
A few are more complex and are better left for a followup patch
though.
An exhaustive test of every possible operator code is included,
with the ones that don't yet work commented out.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340046
91177308-0d34-0410-b5e6-
96231b3b80d8
Simon Pilgrim [Fri, 17 Aug 2018 15:54:07 +0000 (15:54 +0000)]
[TableGen] TypeInfer - Cache the legal types as TypeSetByHwMode
We were just caching the MVT set of legal types, then every call creating a new TypeSetByHwMode with it and passing it back on the stack. There's no need to do this - we can create and cache the whole TypeSetByHwMode once and return a const reference to it each time.
Additionally, TypeInfer::expandOverloads wasn't making use of the fact that the cache just contains a default mode containing all the types.
Saves up to 30secs in debug builds of x86 -gen-dag-isel.
Differential Revision: https://reviews.llvm.org/D50903
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340042
91177308-0d34-0410-b5e6-
96231b3b80d8
Hsiangkai Wang [Fri, 17 Aug 2018 15:22:04 +0000 (15:22 +0000)]
[DebugInfo] Generate DWARF debug information for labels. (Fix leak problems)
There are two forms for label debug information in DWARF format.
1. Labels in a non-inlined function:
DW_TAG_label
DW_AT_name
DW_AT_decl_file
DW_AT_decl_line
DW_AT_low_pc
2. Labels in an inlined function:
DW_TAG_label
DW_AT_abstract_origin
DW_AT_low_pc
We will collect label information from DBG_LABEL. Before every DBG_LABEL,
we will generate a temporary symbol to denote the location of the label.
The symbol could be used to get DW_AT_low_pc afterwards. So, we create a
mapping between 'inlined label' and DBG_LABEL MachineInstr in DebugHandlerBase.
The DBG_LABEL in the mapping is used to query the symbol before it.
The AbstractLabels in DwarfCompileUnit is used to process labels in inlined
functions.
We also keep a mapping between scope and labels in DwarfFile to help to
generate correct tree structure of DIEs.
It also generates label debug information under global isel.
Differential Revision: https://reviews.llvm.org/D45556
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340039
91177308-0d34-0410-b5e6-
96231b3b80d8
Stefan Pintilie [Fri, 17 Aug 2018 15:15:26 +0000 (15:15 +0000)]
[PowerPC] Generate lxsd instead of the ld->mtvsrd sequence for vector loads
This patch addresses:
- Implementation within PPCISelLowering.cpp to check if we should use direct
load into vector instructions (such as lxsd/lfd ) when the scalar_to_vector
function is used; which will allow us to catch as many cases of the
scalar_to_vector uses as possible to translate the ld->mtvsrd sequence into
lxsd.
- Test cases to exhibit the behaviour of emitting lxsd/lfd.
Patch by amyk
Differential revision: https://reviews.llvm.org/D49698
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340037
91177308-0d34-0410-b5e6-
96231b3b80d8
Andrea Di Biagio [Fri, 17 Aug 2018 15:01:37 +0000 (15:01 +0000)]
[llvm-mca] Removed references to HWStallEvent in Scheduler.h. NFCI
class Scheduler should not know anything of hardware event listeners and
hardware stall events (HWStallEvent). HWStallEvent objects should only be
constructed by pipeline stages to notify listeners of hardware events.
No functional change intended.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340036
91177308-0d34-0410-b5e6-
96231b3b80d8
Francis Visoiu Mistrih [Fri, 17 Aug 2018 14:47:01 +0000 (14:47 +0000)]
[x86] Fix test breaking on Darwin after r339962
* -march=x86-64 -> -mtriple=x86_64-unknown-linux to avoid _ prefixes to
symbols
* add -start-before to avoid running the whole codegen on the IR. I
assumed it is meant to be running after X86SpeculativeLoadHardening.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340034
91177308-0d34-0410-b5e6-
96231b3b80d8
Francis Visoiu Mistrih [Fri, 17 Aug 2018 14:46:56 +0000 (14:46 +0000)]
[X86] Fix liveness information when expanding X86::EH_SjLj_LongJmp64
test/CodeGen/X86/shadow-stack.ll has the following machine verifier
errors:
```
*** Bad machine code: Using a killed virtual register ***
- function: bar
- basic block: %bb.6 entry (0x7fdc81857818)
- instruction: %3:gr64 = MOV64rm killed %2:gr64, 1, $noreg, 8, $noreg
- operand 1: killed %2:gr64
*** Bad machine code: Using a killed virtual register ***
- function: bar
- basic block: %bb.6 entry (0x7fdc81857818)
- instruction: $rsp = MOV64rm killed %2:gr64, 1, $noreg, 16, $noreg
- operand 1: killed %2:gr64
*** Bad machine code: Virtual register killed in block, but needed live out. ***
- function: bar
- basic block: %bb.2 entry (0x7fdc818574f8)
Virtual register %2 is used after the block.
```
The fix here is to only copy the machine operand's register without the
kill flags for all the instructions except the very last one of the
sequence.
I had to insert dummy PHIs in the test case to force the NoPHI function
property to be set to false. More on this here: https://llvm.org/PR38439
Differential Revision: https://reviews.llvm.org/D50260
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340033
91177308-0d34-0410-b5e6-
96231b3b80d8
Florian Hahn [Fri, 17 Aug 2018 14:39:53 +0000 (14:39 +0000)]
[NewGVN] Add tests for r340031.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340032
91177308-0d34-0410-b5e6-
96231b3b80d8
Florian Hahn [Fri, 17 Aug 2018 14:39:04 +0000 (14:39 +0000)]
[InstrSimplify,NewGVN] Add option to ignore additional instr info when simplifying.
NewGVN uses InstructionSimplify for simplifications of leaders of
congruence classes. It is not guaranteed that the metadata or other
flags/keywords (like nsw or exact) of the leader is available for all members
in a congruence class, so we cannot use it for simplification.
This patch adds a InstrInfoQuery struct with a boolean field
UseInstrInfo (which defaults to true to keep the current behavior as
default) and a set of helper methods to get metadata/keywords for a
given instruction, if UseInstrInfo is true. The whole thing might need a
better name, to avoid confusion with TargetInstrInfo but I am not sure
what a better name would be.
The current patch threads through InstrInfoQuery to the required
places, which is messier then it would need to be, if
InstructionSimplify and ValueTracking would share the same Query struct.
The reason I added it as a separate struct is that it can be shared
between InstructionSimplify and ValueTracking's query objects. Also,
some places do not need a full query object, just the InstrInfoQuery.
It also updates some interfaces that do not take a Query object, but a
set of optional parameters to take an additional boolean UseInstrInfo.
See https://bugs.llvm.org/show_bug.cgi?id=37540.
Reviewers: dberlin, davide, efriedma, sebpop, hiraditya
Reviewed By: hiraditya
Differential Revision: https://reviews.llvm.org/D47143
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340031
91177308-0d34-0410-b5e6-
96231b3b80d8
Krzysztof Parzyszek [Fri, 17 Aug 2018 14:24:24 +0000 (14:24 +0000)]
[Hexagon] Expand vgather pseudos during packetization
This will allow packetizing the vgather expansion with other instructions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340028
91177308-0d34-0410-b5e6-
96231b3b80d8
Alex Bradbury [Fri, 17 Aug 2018 14:03:37 +0000 (14:03 +0000)]
[AtomicExpandPass] Widen partword atomicrmw or/xor/and before tryExpandAtomicRMW
This patch performs a widening transformation of bitwise atomicrmw
{or,xor,and} and applies it prior to tryExpandAtomicRMW. This operates
similarly to convertCmpXchgToIntegerType. For these operations, the i8/i16
atomicrmw can be implemented in terms of the 32-bit atomicrmw by appropriately
manipulating the operands. There is no functional change for the handling of
partword or/xor, but the transformation for partword 'and' is new.
The advantage of performing this transformation early is that the same
code-path can be used regardless of the approach used to expand the atomicrmw
(AtomicExpansionKind). i.e. the same logic is used for
AtomicExpansionKind::CmpXchg and can also be used by the intrinsic-based
expansion in D47882.
Differential Revision: https://reviews.llvm.org/D48129
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340027
91177308-0d34-0410-b5e6-
96231b3b80d8
Anna Thomas [Fri, 17 Aug 2018 13:44:00 +0000 (13:44 +0000)]
[LICM] Add a diagnostic analysis for identifying alias information
Summary:
Currently, in LICM, we use the alias set tracker to identify if the
instruction (we're interested in hoisting) aliases with instruction that
modifies that memory location.
This patch adds an LICM alias analysis diagnostic tool that checks the
mod ref info of the instruction we are interested in hoisting/sinking,
with every instruction in the loop. Because of O(N^2) complexity this
is now only a diagnostic tool to show the limitation we have with the
alias set tracker and is OFF by default.
Test cases show the difference with the diagnostic analysis tool, where
we're able to hoist out loads and readonly + argmemonly calls from the
loop, where the alias set tracker analysis is not able to hoist these
instructions out.
Reviewers: reames, mkazantsev, fedor.sergeev, hfinkel
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D50854
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340026
91177308-0d34-0410-b5e6-
96231b3b80d8
Roger Ferrer Ibanez [Fri, 17 Aug 2018 13:40:03 +0000 (13:40 +0000)]
[RISCV] Remove unused function
This function is not virtual, it is private and it is not called anywhere. No
regression is introduced by removing it.
I think we can safely remove it.
Differential Revision: https://reviews.llvm.org/D50836
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340024
91177308-0d34-0410-b5e6-
96231b3b80d8
Sanjay Patel [Fri, 17 Aug 2018 13:23:44 +0000 (13:23 +0000)]
[ConstantFolding] add simplifications for funnel shift intrinsics
This is another step towards being able to canonicalize to the funnel shift
intrinsics in IR (see D49242 for the initial patch).
We should not have any loss of simplification power in IR between these and
the equivalent IR constructs.
Differential Revision: https://reviews.llvm.org/D50848
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340022
91177308-0d34-0410-b5e6-
96231b3b80d8
Simon Pilgrim [Fri, 17 Aug 2018 13:03:17 +0000 (13:03 +0000)]
[TableGen] TypeSetByHwMode::insert - cache the default MVT. NFCI.
Avoids repeated count()/find() calls that we've already have the default values for.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340020
91177308-0d34-0410-b5e6-
96231b3b80d8
Luke Cheeseman [Fri, 17 Aug 2018 12:53:22 +0000 (12:53 +0000)]
[AArch64] - Generate pointer authentication instructions
- Generate pointer authentication instructions
- The functions instrumented depend on function attribtues:
all (all functions instrumentent)
non-leaf (only those that spill LR)
none
- Function epilogues sign the LR before spilling to the stack and authenticate
the LR once restored
- If the target is v8.3a or greater than can use the combined authenticate and
return instruction
Differential revision: https://reviews.llvm.org/D49793
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340018
91177308-0d34-0410-b5e6-
96231b3b80d8
Nemanja Ivanovic [Fri, 17 Aug 2018 12:41:49 +0000 (12:41 +0000)]
Revert extraneous directory added by accident in rL340016
It appears that the way this patch was produced ended up creating an
extra 'llvm' directory where the test was placed. When I committed the
patch, that directory ended up being created upstream. This commit should
revert that.
Sorry for the noise.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340017
91177308-0d34-0410-b5e6-
96231b3b80d8
Nemanja Ivanovic [Fri, 17 Aug 2018 12:35:44 +0000 (12:35 +0000)]
[PowerPC] Generate Power9 extswsli extend sign and shift immediate instruction
Add a DAG combine for the PowerPC code generator to generate the Power9 extswsli
extend sign and shift immediate instruction.
Patch by RolandF.
Differential revision: https://reviews.llvm.org/D49879
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340016
91177308-0d34-0410-b5e6-
96231b3b80d8
Simon Pilgrim [Fri, 17 Aug 2018 12:25:18 +0000 (12:25 +0000)]
[DAGCombiner] extractShiftForRotate - fix out of range shift issue
Don't just check for negative shift amounts.
Fixes OSS Fuzz #9935
https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=9935
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340015
91177308-0d34-0410-b5e6-
96231b3b80d8
Andrea Di Biagio [Fri, 17 Aug 2018 11:33:48 +0000 (11:33 +0000)]
[InstCombine] Remove unused method FAddCombine::createFDiv(). NFC
This commit fixes a (gcc 7.3.0) [-Wunused-function] warning caused by the
presence of unused method FaddCombine::createFDiv().
The last use of that method was removed at r339519.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340014
91177308-0d34-0410-b5e6-
96231b3b80d8
Bernard Ogden [Fri, 17 Aug 2018 11:29:49 +0000 (11:29 +0000)]
[ARM/AArch64] Support FP16 +fp16fml instructions
Add +fp16fml feature for new FP16 instructions, which are a
mandatory part of FP16 from v8.4-A and an optional part of FP16
from v8.2-A. It doesn't seem to be possible to model this in
LLVM, but the relationship between the options is handled by
the related clang patch.
In keeping with what I think is the usual practice, the fp16fml
extension is accepted regardless of base architecture version.
Builds on/replaces Sjoerd Meijer's patch to add these instructions at
https://reviews.llvm.org/D49839.
Differential Revision: https://reviews.llvm.org/D50228
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340013
91177308-0d34-0410-b5e6-
96231b3b80d8
Bernard Ogden [Fri, 17 Aug 2018 11:26:57 +0000 (11:26 +0000)]
[ARM/AArch64] TargetParserTest fixes
Adds some missing tests for the FP16 extension,
fixes an existing test that misnames it.
Differential Revision: https://reviews.llvm.org/D50227
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340012
91177308-0d34-0410-b5e6-
96231b3b80d8
Simon Pilgrim [Fri, 17 Aug 2018 10:52:49 +0000 (10:52 +0000)]
[DAGCombine] Improve (sra (sra x, c1), c2) -> (sra x, (add c1, c2)) folding
Add support for cases where only some c1+c2 results exceed the max bitshift, clamping accordingly.
Differential Revision: https://reviews.llvm.org/D35722
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340010
91177308-0d34-0410-b5e6-
96231b3b80d8
Daniel Cederman [Fri, 17 Aug 2018 10:40:00 +0000 (10:40 +0000)]
[Sparc] Get sret arg size from CallLoweringInfo.getArgs()
Summary:
Looking at the callee argument list, as is done now, might not work if
the function has been typecasted into one that is expected to return
a struct. This change also simplifies the code.
The isFP128ABICall() function can be removed as it is no longer needed.
The test in fp128.ll has been updated to verify this.
Reviewers: jyknight, venkatra
Reviewed By: jyknight
Subscribers: fedor.sergeev, jrtc27, llvm-commits
Differential Revision: https://reviews.llvm.org/D48117
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340008
91177308-0d34-0410-b5e6-
96231b3b80d8
Simon Pilgrim [Fri, 17 Aug 2018 09:47:52 +0000 (09:47 +0000)]
Fix "control reaches end of non-void function" -Wreturn-type warning. NFCI.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340006
91177308-0d34-0410-b5e6-
96231b3b80d8
Daniel Cederman [Fri, 17 Aug 2018 09:18:31 +0000 (09:18 +0000)]
[Sparc] Flush register windows for @llvm.returnaddress(1)
Summary: When @llvm.returnaddress is called with a value higher than 0
it needs to read from the call stack to get the return address. This
means that the register windows needs to be flushed to the stack to
guarantee that the data read is valid. For values higher than 1 this
is done indirectly by the call to getFRAMEADDR(), but not for the value 1.
Reviewers: jyknight, venkatra
Reviewed By: jyknight
Subscribers: fedor.sergeev, jrtc27, llvm-commits
Differential Revision: https://reviews.llvm.org/D48636
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340003
91177308-0d34-0410-b5e6-
96231b3b80d8
Chen Zheng [Fri, 17 Aug 2018 07:51:01 +0000 (07:51 +0000)]
[MISC]Fix wrong usage of std::equal()
Differential Revision: https://reviews.llvm.org/D49958
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@340000
91177308-0d34-0410-b5e6-
96231b3b80d8
Sjoerd Meijer [Fri, 17 Aug 2018 07:34:01 +0000 (07:34 +0000)]
[ARM][NFC] ARMCodeGenPrepare: some refactoring and algorithm description
Differential Revision: https://reviews.llvm.org/D50846
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339997
91177308-0d34-0410-b5e6-
96231b3b80d8
Max Kazantsev [Fri, 17 Aug 2018 06:19:17 +0000 (06:19 +0000)]
[MustExecute] Fix algorithmic bug in isGuaranteedToExecute. PR38514
The description of `isGuaranteedToExecute` does not correspond to its implementation.
According to description, it should return `true` if an instruction is executed under the
assumption that its loop is *entered*. However there is a sophisticated alrogithm inside
that tries to prove that the instruction is executed if the loop is *exited*, which is not the
same thing for infinite loops. There is an attempt to protect from dealing with infinite loops
by prohibiting loops without exit blocks, however an infinite loop can have exit blocks.
As result of that, MustExecute can falsely consider some blocks that are never entered as
mustexec, and LICM can hoist dangerous instructions out of them basing on this fact.
This may introduce UB to programs which did not contain it initially.
This patch removes the problematic algorithm and replaced it with a one which tries to
prove what is required in description.
Differential Revision: https://reviews.llvm.org/D50558
Reviewed By: reames
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339984
91177308-0d34-0410-b5e6-
96231b3b80d8
Max Kazantsev [Fri, 17 Aug 2018 05:20:25 +0000 (05:20 +0000)]
[NFC] Add tests to ensure that improvement of MustThrow analysis will not lead to problems in future
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339983
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Fri, 17 Aug 2018 04:47:16 +0000 (04:47 +0000)]
Revert r339977: [GISel]: Add Opcodes for a few LLVM Intrinsics
This is breaking ~all the bots.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339982
91177308-0d34-0410-b5e6-
96231b3b80d8
Brian Cain [Fri, 17 Aug 2018 04:38:41 +0000 (04:38 +0000)]
[llvm-mc-assemble-fuzzer] Update API - Pass MCObjectWriter instead of a stream
Fixes build breakage of llvm-mc-assemble-fuzzer introduced by r332749.
Fix provided by pbhatu (Pratik Bhatu)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339981
91177308-0d34-0410-b5e6-
96231b3b80d8
Graydon Hoare [Fri, 17 Aug 2018 04:13:19 +0000 (04:13 +0000)]
[Support] Add a public API to allow clearing all (static) timer groups.
Summary:
Formerly, all timer groups were automatically cleared when printed out. In
https://reviews.llvm.org/rL324788 this behaviour was changed to not-clearing
timers on printout, to allow printing timers more than once, but as a result
clients (specifically Swift) that relied on the clear-on-print behaviour to
inhibit duplicate timer printing on shutdown were broken.
Rather than revert that change, this change adds a new API that enables
clients that _want_ to clear all timers to do so explicitly.
Reviewers: george.karpenkov, thegameg
Reviewed By: george.karpenkov
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D50874
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339980
91177308-0d34-0410-b5e6-
96231b3b80d8
Aditya Nandakumar [Fri, 17 Aug 2018 01:41:56 +0000 (01:41 +0000)]
[GISel]: Add Opcodes for a few LLVM Intrinsics
https://reviews.llvm.org/D50401
Add opcodes for llvm.intrinsic.trunc, round, and update the IRTranslator
for the same.
Reviewed by: dsanders.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339977
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Fri, 17 Aug 2018 01:10:33 +0000 (01:10 +0000)]
[ADT] Replace a member initializer of a union with an explicit
constructor.
This breaking an old/weird host compiler is my best bet for the current
crashes I'm getting from bots since this functionality was added to this
ADT.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339975
91177308-0d34-0410-b5e6-
96231b3b80d8
Heejin Ahn [Fri, 17 Aug 2018 00:12:04 +0000 (00:12 +0000)]
[WebAssembly] Modify LateEHPrepare one-line description (NFC)
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339972
91177308-0d34-0410-b5e6-
96231b3b80d8
David Blaikie [Thu, 16 Aug 2018 23:57:15 +0000 (23:57 +0000)]
DebugInfo: Remove command line (& target-based) disabling of pubnames in favor of metadata
Now that Clang disables NVPTX pubnames via metadata there's no need for
this fallback to target detection in the backend.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339970
91177308-0d34-0410-b5e6-
96231b3b80d8
Heejin Ahn [Thu, 16 Aug 2018 23:50:59 +0000 (23:50 +0000)]
[WebAssembly] CFG stackify support for exception handling
Summary:
This adds support for exception handling to CFGStackify pass. This only
adds TRY / END_TRY markers and DOES NOT yet fix unwind mismatches that
can be created by the linearization of the CFG into the structural wasm
format. The mismatch fix will be added by following patches.
In detail, this patch
- Added support for TRY / END_TRY markers to support EH
- Changed many static functions into class member functions as they take
too many arguments now
- Added several more bookeeping data structures
- Refactored routines that decide where to insert markers, because
without refactoring this got too complicated as we added support for new
kinds of markers (TRY/END_TRY).
- Rewrote rethrow instructions' BB arguments to relative depths in EH
pad stack.
Reviewers: dschuff, sunfish
Subscribers: sbc100, jgravelle-google, llvm-commits
Differential Revision: https://reviews.llvm.org/D48273
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339967
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Thu, 16 Aug 2018 23:11:05 +0000 (23:11 +0000)]
[x86/MIR] Implement support for pre- and post-instruction symbols, as
well as MIR parsing support for `MCSymbol` `MachineOperand`s.
The only real way to test pre- and post-instruction symbol support is to
use them in operands, so I ended up implementing that within the patch
as well. I can split out the operand support if folks really want but it
doesn't really seem worth it.
The functional implementation of pre- and post-instruction symbols is
now *completely trivial*. Two tiny bits of code in the (misnamed)
AsmPrinter. It should be completely target independent as well. We emit
these exactly the same way as we emit basic block labels. Most of the
code here is to give full dumping, MIR printing, and MIR parsing support
so that we can write useful tests.
The MIR parsing of MC symbol operands still isn't 100%, as it forces the
symbols to be non-temporary and non-local symbols with names. However,
those names often can encode most (if not all) of the special semantics
desired, and unnamed symbols seem especially annoying to serialize and
de-serialize. While this isn't perfect or full support, it seems plenty
to write tests that exercise usage of these kinds of operands.
The MIR support for pre-and post-instruction symbols was quite
straightforward. I chose to print them out in an as-if-operand syntax
similar to debug locations as this seemed the cleanest way and let me
use nice introducer tokens rather than inventing more magic punctuation
like we use for memoperands.
However, supporting MIR-based parsing of these symbols caused me to
change the design of the symbol support to allow setting arbitrary
symbols. Without this, I don't see any reasonable way to test things
with MIR.
Differential Revision: https://reviews.llvm.org/D50833
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339962
91177308-0d34-0410-b5e6-
96231b3b80d8
Sanjay Patel [Thu, 16 Aug 2018 22:46:20 +0000 (22:46 +0000)]
[InstCombine] add reflection fold for tan(-x)
This is a follow-up suggested with rL339604.
For tan(), we don't have a corresponding LLVM
intrinsic -- unlike sin/cos -- so this is the
only way/place that we can do this fold currently.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339958
91177308-0d34-0410-b5e6-
96231b3b80d8
Vedant Kumar [Thu, 16 Aug 2018 22:24:47 +0000 (22:24 +0000)]
[InstrProf] Use atomic profile counter updates for TSan
Thread sanitizer instrumentation fails to skip all loads and stores to
profile counters. This can happen if profile counter updates are merged:
%.sink = phi i64* ...
%pgocount5 = load i64, i64* %.sink
%27 = add i64 %pgocount5, 1
%28 = bitcast i64* %.sink to i8*
call void @__tsan_write8(i8* %28)
store i64 %27, i64* %.sink
To suppress TSan diagnostics about racy counter updates, make the
counter updates atomic when TSan is enabled. If there's general interest
in this mode it can be surfaced as a clang/swift driver option.
Testing: check-{llvm,clang,profile}
rdar://
40477803
Differential Revision: https://reviews.llvm.org/D50867
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339955
91177308-0d34-0410-b5e6-
96231b3b80d8
Sanjay Patel [Thu, 16 Aug 2018 22:05:51 +0000 (22:05 +0000)]
[InstCombine] add tests for tan with negated arg; NFC
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339953
91177308-0d34-0410-b5e6-
96231b3b80d8
Alina Sbirlea [Thu, 16 Aug 2018 21:58:44 +0000 (21:58 +0000)]
Update MemorySSA in Local utils removing blocks.
Summary: Extend Local utils to update MemorySSA.
Subscribers: sanjoy, jlebar, Prazek, george.burgess.iv, llvm-commits
Differential Revision: https://reviews.llvm.org/D48790
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339951
91177308-0d34-0410-b5e6-
96231b3b80d8
Justin Bogner [Thu, 16 Aug 2018 21:55:09 +0000 (21:55 +0000)]
[docs] Try to clarify the FuzzingLLVM docs
Try to improve these docs based on some recent questions that were
sent to llvm-dev:
http://lists.llvm.org/pipermail/llvm-dev/2018-August/125329.html
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339949
91177308-0d34-0410-b5e6-
96231b3b80d8
Alina Sbirlea [Thu, 16 Aug 2018 21:54:33 +0000 (21:54 +0000)]
[DomTree] Add constructor to create a new DT based on current DT/CFG and a set of Updates.
Summary:
Add the posibility of creating a new DT using a set of Updates.
This will essentially create a DT based on a CFG snapshot/view.
Additional refactoring for either this patch or follow-ups:
- create an utility for building BUI.
- replace BUI with a GraphDiff.
Reviewers: kuhar
Subscribers: sanjoy, jlebar, llvm-commits
Differential Revision: https://reviews.llvm.org/D50671
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339947
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Thu, 16 Aug 2018 21:54:05 +0000 (21:54 +0000)]
[DAGCombiner] Don't reassociate operations that have the vector reduction flag set.
When nodes are reassociated the vector-reduction flag gets lost.
The test case is here is what would happen if you had a sum of absolute differences loop that started with a non-zero but contant sum and that loop was unrolled. The vectorizer will generate a constant vector for the initial value. And DAGCombiner reassociate tries to move it down the addition tree erasing the vector-reduction flag. Interestingly this moves constants the opposite direction of the reassociate IR pass.
I've chosen to just punt on the reassociate, but I suppose we could maybe preserve the flag if both nodes have it set.
Differential Revision: https://reviews.llvm.org/D50827
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339946
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Thu, 16 Aug 2018 21:54:02 +0000 (21:54 +0000)]
[X86] In EFLAGS copy pass, don't emit EXTRACT_SUBREG instructions since we're after peephole
Normally the peephole pass converts EXTRACT_SUBREG to COPY instructions. But we're after peephole so we can't rely on it to clean these up.
To fix this, the eflags pass now emits a COPY with a subreg input.
I also noticed that in 32-bit mode we need to constrain the input to the copy to ensure the subreg is valid. Otherwise we'll fail verify-machineinstrs
Differential Revision: https://reviews.llvm.org/D50656
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339945
91177308-0d34-0410-b5e6-
96231b3b80d8
Richard Smith [Thu, 16 Aug 2018 21:40:57 +0000 (21:40 +0000)]
Factor Node creation out of the demangler. No functionality change
intended.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339944
91177308-0d34-0410-b5e6-
96231b3b80d8
Reid Kleckner [Thu, 16 Aug 2018 21:34:41 +0000 (21:34 +0000)]
[MC] Improve COFF associative section lookup
Handle the case when the symbol is private. Private symbols are not in
the COFF object file symbol table, so they aren't inserted into
SymbolMap. We can't look up the section of the symbol that way. Instead,
get the MCSection from the MCSymbol and map that to the object file
section.
Print a better error message when the symbol has no section, like when
the symbol is undefined.
Fixes PR38607
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339942
91177308-0d34-0410-b5e6-
96231b3b80d8
Chandler Carruth [Thu, 16 Aug 2018 21:30:05 +0000 (21:30 +0000)]
[MI] Change the array of `MachineMemOperand` pointers to be
a generically extensible collection of extra info attached to
a `MachineInstr`.
The primary change here is cleaning up the APIs used for setting and
manipulating the `MachineMemOperand` pointer arrays so chat we can
change how they are allocated.
Then we introduce an extra info object that using the trailing object
pattern to attach some number of MMOs but also other extra info. The
design of this is specifically so that this extra info has a fixed
necessary cost (the header tracking what extra info is included) and
everything else can be tail allocated. This pattern works especially
well with a `BumpPtrAllocator` which we use here.
I've also added the basic scaffolding for putting interesting pointers
into this, namely pre- and post-instruction symbols. These aren't used
anywhere yet, they're just there to ensure I've actually gotten the data
structure types correct. I'll flesh out support for these in
a subsequent patch (MIR dumping, parsing, the works).
Finally, I've included an optimization where we store any single pointer
inline in the `MachineInstr` to avoid the allocation overhead. This is
expected to be the overwhelmingly most common case and so should avoid
any memory usage growth due to slightly less clever / dense allocation
when dealing with >1 MMO. This did require several ergonomic
improvements to the `PointerSumType` to reasonably support the various
usage models.
This also has a side effect of freeing up 8 bits within the
`MachineInstr` which could be repurposed for something else.
The suggested direction here came largely from Hal Finkel. I hope it was
worth it. ;] It does hopefully clear a path for subsequent extensions
w/o nearly as much leg work. Lots of thanks to Reid and Justin for
careful reviews and ideas about how to do all of this.
Differential Revision: https://reviews.llvm.org/D50701
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339940
91177308-0d34-0410-b5e6-
96231b3b80d8
David Blaikie [Thu, 16 Aug 2018 21:29:55 +0000 (21:29 +0000)]
DebugInfo: Add metadata support for disabling DWARF pub sections
In cases where the debugger load time is a worthwhile tradeoff (or less
costly - such as loading from a DWP instead of a variety of DWOs
(possibly over a high-latency/distributed filesystem)) against object
file size, it can be reasonable to disable pubnames and corresponding
gdb-index creation in the linker.
A backend-flag version of this was implemented for NVPTX in
D44385/r327994 - which was fine for NVPTX which wouldn't mix-and-match
CUs. Now that it's going to be a user-facing option (likely powered by
"-gno-pubnames", the same as GCC) it should be encoded in the
DICompileUnit so it can vary per-CU.
After this, likely the NVPTX support should be migrated to the metadata
& the previous flag implementation should be removed.
Reviewers: aprantl
Differential Revision: https://reviews.llvm.org/D50213
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339939
91177308-0d34-0410-b5e6-
96231b3b80d8
Michael Berg [Thu, 16 Aug 2018 20:59:45 +0000 (20:59 +0000)]
add a missed case for binary op FMF propagation under select folds
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339938
91177308-0d34-0410-b5e6-
96231b3b80d8
Philip Reames [Thu, 16 Aug 2018 20:58:48 +0000 (20:58 +0000)]
[AST] Speculative build fix for a polly buildbot
I don't have polly setup to bulld locally and don't plan to. This should let the old API adapt to the new one. Can someone from polly please migrate usage and then delete the wrapper?
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339937
91177308-0d34-0410-b5e6-
96231b3b80d8
Philip Reames [Thu, 16 Aug 2018 20:48:55 +0000 (20:48 +0000)]
[MemLoc] Fix a bug causing any use of invariant.end to crash in LICM
The fix is fairly simple, but is says something unpleasant about the usage and testing of invariant.start/end scopes that this went undetected. To put this in perspective, *any* invariant.end in a loop flowing through LICM crashed. I haven't bothered to figure out just how far back this goes, but it's not caused by any of the recent changes. We're probably talking months if not years.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339936
91177308-0d34-0410-b5e6-
96231b3b80d8
Krzysztof Parzyszek [Thu, 16 Aug 2018 20:12:15 +0000 (20:12 +0000)]
[SystemZ] Require asserts in subregliveness-06.mir
The option -misched=shuffle is only available with !NDEBUG builds.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339931
91177308-0d34-0410-b5e6-
96231b3b80d8
Philip Reames [Thu, 16 Aug 2018 20:11:15 +0000 (20:11 +0000)]
[LICM][NFC] Restructure pointer invalidation API in terms of MemoryLocation
Main value is just simplifying code. I'll further simply the argument handling case in a bit, but that involved a slightly orthogonal change so I went with the mildy ugly intermediate for this patch.
Note that the isSized check in the old LICM code was not carried across. It turns out that check was dead. a) no test exercised it, and b) langref and verifier had been updated to disallow unsized types used in loads.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339930
91177308-0d34-0410-b5e6-
96231b3b80d8
Andrea Di Biagio [Thu, 16 Aug 2018 19:45:13 +0000 (19:45 +0000)]
[llvm-mca] Fix -Wpessimizing-move warnings introduced by r339923.
Reported by buildbot `clang-with-lto-ubuntu` ( build #9858 ).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339928
91177308-0d34-0410-b5e6-
96231b3b80d8
Peter Collingbourne [Thu, 16 Aug 2018 19:29:01 +0000 (19:29 +0000)]
Add missing test file from r339799.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339927
91177308-0d34-0410-b5e6-
96231b3b80d8
Craig Topper [Thu, 16 Aug 2018 19:27:43 +0000 (19:27 +0000)]
[X86] Pre-commit test case for D50827.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339926
91177308-0d34-0410-b5e6-
96231b3b80d8
Jacob Gravelle [Thu, 16 Aug 2018 19:24:31 +0000 (19:24 +0000)]
[WebAssembly] Remove temporary workaround for function bitcasts
Summary:
EM_ASM no longer is lowered as varargs in C, so this workaround is
obsolete.
Reviewers: dschuff, sunfish
Subscribers: sbc100, aheejin, llvm-commits
Differential Revision: https://reviews.llvm.org/D50859
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339925
91177308-0d34-0410-b5e6-
96231b3b80d8
Krzysztof Parzyszek [Thu, 16 Aug 2018 19:13:28 +0000 (19:13 +0000)]
[MachineVerifier] Check if predecessor is jointly dominated by undefs
Each use of a value should be jointly dominated by the union of defs and
undefs. It can happen that it will only be jointly dominated by undefs,
and that is still legal. Make sure that the verifier is aware of that.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339924
91177308-0d34-0410-b5e6-
96231b3b80d8
Andrea Di Biagio [Thu, 16 Aug 2018 19:00:48 +0000 (19:00 +0000)]
[llvm-mca] Refactor how execution is orchestrated by the Pipeline.
This patch changes how instruction execution is orchestrated by the Pipeline.
In particular, this patch makes it more explicit how instructions transition
through the various pipeline stages during execution.
The main goal is to simplify both the stage API and the Pipeline execution. At
the same time, this patch fixes some design issues which are currently latent,
but that are likely to cause problems in future if people start defining custom
pipelines.
The new design assumes that each pipeline stage knows the "next-in-sequence".
The Stage API has gained three new methods:
- isAvailable(IR)
- checkNextStage(IR)
- moveToTheNextStage(IR).
An instruction IR can be executed by a Stage if method `Stage::isAvailable(IR)`
returns true.
Instructions can move to next stages using method moveToTheNextStage(IR).
An instruction cannot be moved to the next stage if method checkNextStage(IR)
(called on the current stage) returns false.
Stages are now responsible for moving instructions to the next stage in sequence
if necessary.
Instructions are allowed to transition through multiple stages during a single
cycle (as long as stages are available, and as long as all the calls to
`checkNextStage(IR)` returns true).
Methods `Stage::preExecute()` and `Stage::postExecute()` have now become
redundant, and those are removed by this patch.
Method Pipeline::runCycle() is now simpler, and it correctly visits stages
on every begin/end of cycle.
Other changes:
- DispatchStage no longer requires a reference to the Scheduler.
- ExecuteStage no longer needs to directly interact with the
RetireControlUnit. Instead, executed instructions are now directly moved to the
next stage (i.e. the retire stage).
- RetireStage gained an execute method. This allowed us to remove the
dependency with the RCU in ExecuteStage.
- FecthStage now updates the "program counter" during cycleBegin() (i.e.
before we start executing new instructions).
- We no longer need Stage::Status to be returned by method execute(). It has
been dropped in favor of a more lightweight llvm::Error.
Overally, I measured a ~11% performance gain w.r.t. the previous design. I also
think that the Stage interface is probably easier to read now. That being said,
code comments have to be improved, and I plan to do it in a follow-up patch.
Differential revision: https://reviews.llvm.org/D50849
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339923
91177308-0d34-0410-b5e6-
96231b3b80d8
Eli Friedman [Thu, 16 Aug 2018 18:39:39 +0000 (18:39 +0000)]
[SelectionDAG] Improve the legalisation lowering of UMULO.
There is no way in the universe, that doing a full-width division in
software will be faster than doing overflowing multiplication in
software in the first place, especially given that this same full-width
multiplication needs to be done anyway.
This patch replaces the previous implementation with a direct lowering
into an overflowing multiplication algorithm based on half-width
operations.
Correctness of the algorithm was verified by exhaustively checking the
output of this algorithm for overflowing multiplication of 16 bit
integers against an obviously correct widening multiplication. Baring
any oversights introduced by porting the algorithm to DAG, confidence in
correctness of this algorithm is extremely high.
Following table shows the change in both t = runtime and s = space. The
change is expressed as a multiplier of original, so anything under 1 is
“better” and anything above 1 is worse.
+-------+-----------+-----------+-------------+-------------+
| Arch | u64*u64 t | u64*u64 s | u128*u128 t | u128*u128 s |
+-------+-----------+-----------+-------------+-------------+
| X64 | - | - | ~0.5 | ~0.64 |
| i686 | ~0.5 | ~0.6666 | ~0.05 | ~0.9 |
| armv7 | - | ~0.75 | - | ~1.4 |
+-------+-----------+-----------+-------------+-------------+
Performance numbers have been collected by running overflowing
multiplication in a loop under `perf` on two x86_64 (one Intel Haswell,
other AMD Ryzen) based machines. Size numbers have been collected by
looking at the size of function containing an overflowing multiply in
a loop.
All in all, it can be seen that both performance and size has improved
except in the case of armv7 where code size has regressed for 128-bit
multiply. u128*u128 overflowing multiply on 32-bit platforms seem to
benefit from this change a lot, taking only 5% of the time compared to
original algorithm to calculate the same thing.
The final benefit of this change is that LLVM is now capable of lowering
the overflowing unsigned multiply for integers of any bit-width as long
as the target is capable of lowering regular multiplication for the same
bit-width. Previously, 128-bit overflowing multiply was the widest
possible.
Patch by Simonas Kazlauskas!
Differential Revision: https://reviews.llvm.org/D50310
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339922
91177308-0d34-0410-b5e6-
96231b3b80d8
Jordan Rupprecht [Thu, 16 Aug 2018 18:29:40 +0000 (18:29 +0000)]
[llvm-strip] Add support for -p/--preserve-dates
Summary: [llvm-strip] Preserve access/modification timestamps when -p is used.
Reviewers: jakehehrlich, jhenderson, alexshap
Reviewed By: jhenderson
Subscribers: llvm-commits
Differential Revision: https://reviews.llvm.org/D50744
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339921
91177308-0d34-0410-b5e6-
96231b3b80d8
Krzysztof Parzyszek [Thu, 16 Aug 2018 18:02:59 +0000 (18:02 +0000)]
[RegisterCoalescer] Shrink to uses if needed after removeCopyByCommutingDef
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339912
91177308-0d34-0410-b5e6-
96231b3b80d8
Zachary Turner [Thu, 16 Aug 2018 17:48:32 +0000 (17:48 +0000)]
Fix memory leak in demangling of string literals.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339909
91177308-0d34-0410-b5e6-
96231b3b80d8
Simon Pilgrim [Thu, 16 Aug 2018 17:44:33 +0000 (17:44 +0000)]
[TargetLowering] Add support for non-uniform vectors to BuildSDIV
This patch refactors the existing TargetLowering::BuildSDIV base implementation to support non-uniform constant vector denominators.
This is the last patch necessary to close PR36545
Differential Revision: https://reviews.llvm.org/D50765
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339908
91177308-0d34-0410-b5e6-
96231b3b80d8
Reid Kleckner [Thu, 16 Aug 2018 17:34:31 +0000 (17:34 +0000)]
[codeview] Use push_macro to avoid conflicts instead of a prefix
Summary:
This prefix was added in r333421, and it changed our dumper output to
say things like "CVRegEAX" instead of just "EAX". That's a functional
change that I'd rather avoid.
I tested GCC, Clang, and MSVC, and all of them support #pragma
push_macro. They don't issue warnings whem the macro is not defined
either.
I don't have a Mac so I can't test the real termios.h header, but I
looked at the termios.h sources online and looked for other conflicts.
I saw only the CR* macros, so those are the ones we work around.
Reviewers: zturner, JDevlieghere
Subscribers: hiraditya, llvm-commits
Differential Revision: https://reviews.llvm.org/D50851
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339907
91177308-0d34-0410-b5e6-
96231b3b80d8
Nirav Dave [Thu, 16 Aug 2018 17:22:31 +0000 (17:22 +0000)]
[MC] Cleanup noop default case spelling. NFC.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339906
91177308-0d34-0410-b5e6-
96231b3b80d8
Tom Stellard [Thu, 16 Aug 2018 17:15:03 +0000 (17:15 +0000)]
Revert "unittests: Don't install TestPlugin.so"
This reverts commit r339897.
This breaks the build on Windows and platforms where loadable modules
aren't supported.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339903
91177308-0d34-0410-b5e6-
96231b3b80d8
Matt Arsenault [Thu, 16 Aug 2018 17:07:52 +0000 (17:07 +0000)]
AMDGPU: Custom lower fexp
This will allow the library to just use __builtin_expf directly
without expanding this itself. Note f64 still won't work because
there is no exp instruction for it.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339902
91177308-0d34-0410-b5e6-
96231b3b80d8
Simon Pilgrim [Thu, 16 Aug 2018 17:07:41 +0000 (17:07 +0000)]
[X86][SSE] Add sdiv by nonuniform constant vector test containing -1/+1 and all-bits style constants
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339901
91177308-0d34-0410-b5e6-
96231b3b80d8
Evandro Menezes [Thu, 16 Aug 2018 17:03:22 +0000 (17:03 +0000)]
[NFC] Fix typo in test cases
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339900
91177308-0d34-0410-b5e6-
96231b3b80d8
Simon Pilgrim [Thu, 16 Aug 2018 16:54:06 +0000 (16:54 +0000)]
[TargetLowering] Refactor BuildSDIV in preparation for D50765. NFCI.
Pull out magic factor calculators into a helper function, use 0/+1/-1 multiplication factor to (optionally) add/sub the numerator.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339898
91177308-0d34-0410-b5e6-
96231b3b80d8
Tom Stellard [Thu, 16 Aug 2018 16:53:06 +0000 (16:53 +0000)]
unittests: Don't install TestPlugin.so
Summary:
add_llvm_loadable_module adds an install target by default, but this
module is only used for a unit test, so we don't need to install it.
Reviewers: philip.pfaffe, thakis
Subscribers: mgorny, llvm-commits
Differential Revision: https://reviews.llvm.org/D50668
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339897
91177308-0d34-0410-b5e6-
96231b3b80d8
Benjamin Kramer [Thu, 16 Aug 2018 16:50:23 +0000 (16:50 +0000)]
[MC] Remove unused variable
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339896
91177308-0d34-0410-b5e6-
96231b3b80d8
Nirav Dave [Thu, 16 Aug 2018 16:31:14 +0000 (16:31 +0000)]
[MC][X86] Enhance X86 Register expression handling to more closely match GCC.
Allow the comparison of x86 registers in the evaluation of assembler
directives. This generalizes and simplifies the extension from r334022
to catch another case found in the Linux kernel.
Reviewers: rnk, void
Reviewed By: rnk
Subscribers: hiraditya, nickdesaulniers, llvm-commits
Differential Revision: https://reviews.llvm.org/D50795
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339895
91177308-0d34-0410-b5e6-
96231b3b80d8
Zachary Turner [Thu, 16 Aug 2018 16:30:27 +0000 (16:30 +0000)]
Fix -Wmicrosoft-goto warnings.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339894
91177308-0d34-0410-b5e6-
96231b3b80d8
Zachary Turner [Thu, 16 Aug 2018 16:17:55 +0000 (16:17 +0000)]
Add support for AVX-512 CodeView registers.
When compiling with /arch:AVX512 and optimizations turned on,
we could crash while emitting debug info because we did not
have CodeView register constants for the AVX 512 register
set defined. This patch defines them.
Differential Revision: https://reviews.llvm.org/D50819
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339893
91177308-0d34-0410-b5e6-
96231b3b80d8
Zachary Turner [Thu, 16 Aug 2018 16:17:36 +0000 (16:17 +0000)]
[MS Demangler] Demangle string literals.
When demangling string literals, Microsoft's undname
simply prints 'string'. This patch implements string
literal demangling while doing a bit better than this
by decoding as much of the string as possible and
trying to faithfully reproduce the original string
literal definition.
This is a bit tricky because the different character
types char, char16_t, and char32_t are not uniquely
identified by the mangling, so we have to use a
heuristic to try to guess the character type. But
it works pretty well, and many tests are added to
illustrate the behavior.
Differential Revision: https://reviews.llvm.org/D50806
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339892
91177308-0d34-0410-b5e6-
96231b3b80d8
Zachary Turner [Thu, 16 Aug 2018 16:17:17 +0000 (16:17 +0000)]
[MS Demangler] Don't fail on MD5-mangled names.
When we have an MD5 mangled name, we shouldn't choke and say
that it's an invalid name. Even though it's impossible to demangle,
we should just output the original name.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339891
91177308-0d34-0410-b5e6-
96231b3b80d8
Simon Pilgrim [Thu, 16 Aug 2018 16:16:28 +0000 (16:16 +0000)]
[TableGen] TypeSetByHwMode::operator== optimization
This operator is called a great deal, by checking for the cheap isSimple equality cases first (a common occurrence) we can improve performance as we avoid a lot of std::map find/iteration in hasDefault.
isSimple also means that a default value is present, so we can avoid some hasDefault calls.
This also avoids a rather dodgy piece of logic that was checking for isSimple() && !VTS.isSimple() but not the inverse - it now uses the general hasDefault mode comparison test instead.
Saves around 15secs in debug builds of x86 -gen-dag-isel.
Differential Revision: https://reviews.llvm.org/D50841
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@339890
91177308-0d34-0410-b5e6-
96231b3b80d8