In a loop, create artificial dependences between the source of a
COPY/REG_SEQUENCE to the use in next iteration.
Eg:
SRC ----Data Dep--> COPY
COPY ---Anti Dep--> PHI (implies, to be used in next iteration)
PHI ----Data Dep--> USE
This patches creates
USE ----Artificial Dep---> SRC
This will effectively schedule the COPY late to eliminate additional copies.
Before this patch, the schedule can be
SRC, COPY, USE : The COPY is used in next iteration and it needs to be
preserved.
After this patch, the schedule can be
USE, SRC, COPY : The COPY is used in next iteration and the live interval is
reduced.
Differential Revision: https://reviews.llvm.org/D53303
llvm-svn: 344748
Adjusted the range check on a call instruction from 24 bits signed to
26 bits signed. While the instruction only encodes 24 bits, the target is
assumed to be 4 byte aligned, and the value that is encoded in the instruction
gets shifted left by 2 to form the offset. Also added a check that the offset is
indeed at least 4 byte aligned.
Differential Revision: https://reviews.llvm.org/D53401
llvm-svn: 344747
Summary:
The RefSlab::size can easily cause confusions, it returns the number of
different symbols, rahter than the number of all references.
- add numRefs() method and cache it, since calculating it everytime is nontrivial.
- clear misused places.
Reviewers: sammccall
Subscribers: ilya-biryukov, ioeric, MaskRay, jkorous, arphaman, kadircet, cfe-commits
Differential Revision: https://reviews.llvm.org/D53389
llvm-svn: 344745
Summary: They fail similarly to some of the other breakpoint tests on Windows, so I suspect the cause is the same. I've linked to the same bug.
Reviewers: asmith, zturner, jingham
Subscribers: abidh, lldb-commits
Differential Revision: https://reviews.llvm.org/D53331
llvm-svn: 344744
When optimizing for size, a loop is vectorized only if the resulting vector loop
completely replaces the original scalar loop. This holds if no runtime guards
are needed, if the original trip-count TC does not overflow, and if TC is a
known constant that is a multiple of the VF. The last two TC-related conditions
can be overcome by
1. rounding the trip-count of the vector loop up from TC to a multiple of VF;
2. masking the vector body under a newly introduced "if (i <= TC-1)" condition.
The patch allows loops with arbitrary trip counts to be vectorized under -Os,
subject to the existing cost model considerations. It also applies to loops with
small trip counts (under -O2) which are currently handled as if under -Os.
The patch does not handle loops with reductions, live-outs, or w/o a primary
induction variable, and disallows interleave groups.
(Third, final and main part of -)
Differential Revision: https://reviews.llvm.org/D50480
llvm-svn: 344743
Summary:
LSP is a slightly awkward map to C++ object lifetimes: the initialize request
is part of the protocol and provides information that doesn't change over the
lifetime of the server.
Until now, we handled this by initializing ClangdServer and ClangdLSPServer
right away, and making anything that can be set in the "initialize" request
mutable.
With this patch, we create ClangdLSPServer immediately, but defer creating
ClangdServer until "initialize". This opens the door to passing the relevant
initialize params in the constructor and storing them immutably.
(That change isn't actually done in this patch).
To make this safe, we have the MessageDispatcher enforce that the "initialize"
method is called before any other (as required by LSP). That way each method
handler can assume Server is initialized, as today.
As usual, while implementing this I found places where our test cases violated
the protocol.
Reviewers: ioeric
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, kadircet, cfe-commits
Differential Revision: https://reviews.llvm.org/D53398
llvm-svn: 344741
This patch exposes functionality added in rL344723 to the Clang driver/frontend
as a flag and adds appropriate metadata.
Driver tests pass:
```
ninja check-clang-driver
-snip-
Expected Passes : 472
Expected Failures : 3
Unsupported Tests : 65
```
Odd failure in CodeGen tests but unrelated to this:
```
ninja check-clang-codegen
-snip-
/SourceCache/llvm-trunk-8.0/tools/clang/test/CodeGen/builtins-wasm.c:87:10:
error: cannot compile this builtin function yet
-snip-
Failing Tests (1):
Clang :: CodeGen/builtins-wasm.c
Expected Passes : 1250
Expected Failures : 2
Unsupported Tests : 120
Unexpected Failures: 1
```
Original commit:
[X86] Support for the mno-tls-direct-seg-refs flag
Allows to disable direct TLS segment access (%fs or %gs). GCC supports a
similar flag, it can be useful in some circumstances, e.g. when a thread
context block needs to be updated directly from user space. More info and
specific use cases: https://bugs.llvm.org/show_bug.cgi?id=16145
Patch by nruslan (Ruslan Nikolaev).
Differential Revision: https://reviews.llvm.org/D53102
llvm-svn: 344739
Summary:
Most of its functionality is moved into ClangdLSPServer.
The decoupling between JSONRPCDispatcher, ProtocolCallbacks, ClangdLSPServer
was never real, and only served to obfuscate.
Some previous implicit/magic stuff is now explicit:
- the return type of LSP method calls are now in the signature
- no more reply() that gets the ID using global context magic
- arg tracing no longer relies on RequestArgs::stash context magic either
This is mostly refactoring, but some deliberate fixes while here:
- LSP method params are now by const reference
- notifications and calls are now distinct namespaces.
(some tests had protocol errors and needed updating)
- we now reply to calls we failed to decode
- outgoing calls use distinct IDs
A few error codes and message IDs changed in unimportant ways (see tests).
Reviewers: ioeric
Subscribers: mgorny, ilya-biryukov, javed.absar, MaskRay, jkorous, arphaman, jfb, kadircet, cfe-commits
Differential Revision: https://reviews.llvm.org/D53387
llvm-svn: 344737
Summary:
These are often not expected to be used directly e.g.
```
TEST_F(Fixture, X) {
^ // "Fixture_X_Test" expanded in the macro should be down ranked.
}
```
Only doing this for sema for now, as such symbols are mostly coming from sema
e.g. gtest macros expanded in the main file. We could also add a similar field
for the index symbol.
Reviewers: sammccall
Reviewed By: sammccall
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, kadircet, cfe-commits
Differential Revision: https://reviews.llvm.org/D53374
llvm-svn: 344736
Summary:
This would buy us more memory. Using a 32-bits integer is enough for
most human-readable source code (up to 4M lines and 4K columns).
Previsouly, we used 8 bytes for a position, now 4 bytes, it would save
us 8 bytes for each Ref and each Symbol instance.
For LLVM-project binary index file, we save ~13% memory.
| Before | After |
| 412MB | 355MB |
Reviewers: sammccall
Subscribers: ilya-biryukov, ioeric, MaskRay, jkorous, arphaman, kadircet, cfe-commits
Differential Revision: https://reviews.llvm.org/D53363
llvm-svn: 344735
Summary:
This is patch 2 of the new DivergenceAnalysis (https://reviews.llvm.org/D50433).
This patch contains a generic divergence analysis implementation for
unstructured, reducible Control-Flow Graphs. It contains two new classes.
The `SyncDependenceAnalysis` class lazily computes sync dependences, which
relate divergent branches to points of joining divergent control. The
`DivergenceAnalysis` class contains the generic divergence analysis
implementation.
Reviewers: nhaehnle
Reviewed By: nhaehnle
Subscribers: sameerds, kristina, nhaehnle, xbolva00, tschuett, mgorny, llvm-commits
Differential Revision: https://reviews.llvm.org/D51491
llvm-svn: 344734
Summary:
Previously, ptr.reset(new char[5]) will be replaced with `p =
make_unique<char[]>(5)`, the fix has side effect -- doing
default initialization, it may cause performace regression (we are
bitten by this rececntly)
The check should be conservative for these cases.
Reviewers: alexfh
Subscribers: xazax.hun, cfe-commits
Differential Revision: https://reviews.llvm.org/D53377
llvm-svn: 344733
Summary:
This patch makes Windows threads to compare by a thread ID, not by a handle.
It's because the same thread can have different handles on Windows
(for example, `GetCurrentThread` always returns the fake handle `-2`).
This leads to some incorrect behavior. For example, in `Process::GetRunLock`
always `m_public_run_lock` is returned without this patch.
Reviewers: zturner, clayborg, stella.stamenova
Reviewed By: stella.stamenova
Subscribers: stella.stamenova, lldb-commits
Tags: #lldb
Differential Revision: https://reviews.llvm.org/D53357
llvm-svn: 344729
For the Go bindings, this just removes the no longer useful "isa"-style
wrapper. If there is a user that is interested, they can add a wrapper
for `Instruction::isTerminator`.
For the OCaml bindings, this is just a documentation update.
llvm-svn: 344726
Summary:
In several places in the code we use the following pattern:
if (hasUnaryFloatFn(&TLI, Ty, LibFunc_tan, LibFunc_tanf, LibFunc_tanl)) {
[...]
Value *Res = emitUnaryFloatFnCall(X, TLI.getName(LibFunc_tan), B, Attrs);
[...]
}
In short, we check if there is a lib-function for a certain type, and then
we _always_ fetch the name of the "double" version of the lib function and
construct a call to the appropriate function, that we just checked exists,
using that "double" name as a basis.
This is of course a problem in cases where the target doesn't support the
"double" version, but e.g. only the "float" version.
In that case TLI.getName(LibFunc_tan) returns "", and
emitUnaryFloatFnCall happily appends an "f" to "", and we erroneously end
up with a call to a function called "f".
To solve this, the above pattern is changed to
if (hasUnaryFloatFn(&TLI, Ty, LibFunc_tan, LibFunc_tanf, LibFunc_tanl)) {
[...]
Value *Res = emitUnaryFloatFnCall(X, &TLI, LibFunc_tan, LibFunc_tanf,
LibFunc_tanl, B, Attrs);
[...]
}
I.e instead of first fetching the name of the "double" version and then
letting emitUnaryFloatFnCall() add the final "f" or "l", we let
emitUnaryFloatFnCall() fetch the right name from TLI.
Reviewers: eli.friedman, efriedma
Reviewed By: efriedma
Subscribers: efriedma, bjope, llvm-commits
Differential Revision: https://reviews.llvm.org/D53370
llvm-svn: 344725
Allows to disable direct TLS segment access (%fs or %gs). GCC supports
a similar flag, it can be useful in some circumstances, e.g. when a thread
context block needs to be updated directly from user space. More info
and specific use cases: https://bugs.llvm.org/show_bug.cgi?id=16145
There is another revision for clang as well.
Related: D53102
All X86 CodeGen tests appear to pass:
```
[46/47] Running lit suite /SourceCache/llvm-trunk-8.0/test/CodeGen
Testing Time: 23.17s
Expected Passes : 3801
Expected Failures : 15
Unsupported Tests : 8021
```
Reviewed by: Craig Topper.
Patch by nruslan (Ruslan Nikolaev).
Differential Revision: https://reviews.llvm.org/D53103
llvm-svn: 344723
This reverts commit r344647.
This causes build failures with [-Werror, -Wswitch]. Some cases where the newly
introduced enum value is not handled in particular are in:
lldb/source/Expression/REPL.cpp:350
lldb/source/Interpreter/CommandInterpreter.cpp:1529
(maybe there could be more)
As I don't understand lldb to make sure the likely trivial fixes are
correct and also as they might need additional tests, leaving to the
author to resolve.
llvm-svn: 344722
Returning the error to clients provides an opportunity to introduce readers to
the Expected and Error APIs and makes the tutorial more useful as a starting
point for a real JIT class, while only slightly complicating the code.
llvm-svn: 344720
Makes bitcode tests line up with what's actually called in the tool.
Should fix the failing bot.
Also fixes a warning that was being thrown about initialization braces.
Differential Revision: https://reviews.llvm.org/D53381
llvm-svn: 344707
Summary:
The original commit message was:
This uses CRTP (for performance reasons) to allow a user the override
demangler functions to implement custom parsing logic. The motivation
for this is LLDB, which needs to occasionaly modify the mangled names.
One such instance is already implemented via the TypeCallback member,
but this is very specific functionality which does not help with any
other use case. Currently we have a use case for modifying the
constructor flavours, which would require adding another callback. This
approach does not scale.
With CRTP, the user (LLDB) can override any function it needs without
any special support from the demangler library. After LLDB is ported to
use this instead of the TypeCallback mechanism, the callback can be
removed.
The only difference here is the addition of a unit test which exercises
the CRTP mechanism to override a function in the parser.
Reviewers: erik.pilkington, rsmith, EricWF
Subscribers: mgorny, kristina, llvm-commits
Differential Revision: https://reviews.llvm.org/D53300
llvm-svn: 344703
Summary:
Running the test suite with -a will now properly show all the executed
commands. The reports also include the environment under which the test
is being executed, which is helpful for reproducing issues.
Reviewers: EricWF
Subscribers: christof, dexonsmith, libcxx-commits
Differential Revision: https://reviews.llvm.org/D53215
llvm-svn: 344700
Summary:
To workaround a hardware issue in the (base + offset) calculation
when base is negative. The impact on code quality should be limited
since SILoadStoreOptimizer still runs afterwards and is able to
combine loads/stores based on known sign information.
This fixes visible corruption in Hitman on SI (easily reproducible
by running benchmark mode).
Change-Id: Ia178d207a5e2ac38ae7cd98b532ea2ae74704e5f
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99923
Reviewers: arsenm, mareko
Subscribers: jholewinski, kzhuravl, jvesely, wdng, yaxunl, dstuttard, tpr, t-tye, llvm-commits
Differential Revision: https://reviews.llvm.org/D53160
llvm-svn: 344698
Summary:
Moving SMRD to VMEM in SIFixSGPRCopies is rather bad for performance if
the load is really uniform. So select the scalar load intrinsics directly
to either VMEM or SMRD buffer loads based on divergence analysis.
If an offset happens to end up in a VGPR -- either because a floating
point calculation was involved, or due to other remaining deficiencies
in SIFixSGPRCopies -- we use v_readfirstlane.
There is some unrelated churn in tests since we now select MUBUF offsets
in a unified way with non-scalar buffer loads.
Change-Id: I170e6816323beb1348677b358c9d380865cd1a19
Reviewers: arsenm, alex-t, rampitec, tpr
Subscribers: kzhuravl, jvesely, wdng, yaxunl, dstuttard, t-tye, llvm-commits
Differential Revision: https://reviews.llvm.org/D53283
llvm-svn: 344696
Summary:
We try to recover gracefully on instructions that would crash the
program.
This includes some refactoring of runMeasurement() implementations.
Reviewers: gchatelet
Subscribers: tschuett, llvm-commits
Differential Revision: https://reviews.llvm.org/D53371
llvm-svn: 344695
Sumanth Gundapaneni