Previously, the MIPS backend would alwyas break down constant multiplications
into a series of shifts, adds, and subs. This patch changes that so the cost of
doing so is estimated.

The cost is estimated against worst case constant materialization and retrieving
the results from the HI/LO registers.

For cases where the value type of the multiplication is not legal, the cost of
legalization is estimated and is accounted for before performing the
optimization of breaking down the constant

This resolves PR36884.

Thanks to npl for reporting the issue!

Reviewers: abeserminji, smaksimovic

Differential Revision: https://reviews.llvm.org/D45316

llvm-svn: 330037

getVA was already implemented in the base class.

llvm-svn: 330036

llvm-svn: 330035

This adds code generation support for the FP16 vmaxnm/vminnm scalar
instructions.

Differential Revision: https://reviews.llvm.org/D44675

llvm-svn: 330034

The tests are still failing on some bots.

llvm-svn: 330033

[llvm-mca] Ensure that instructions with a schedule read-advance are always issued in the right order.

Normally, the Scheduler prioritizes older instructions over younger instructions
during the instruction issue stage. In one particular case where a dependent
instruction had a schedule read-advance associated to one of the input operands,
this rule was not correctly applied.

This patch fixes the issue and adds a test to verify that we don't regress that
particular case.

llvm-svn: 330032

llvm-svn: 330031

Similar to rL329834, don't rely on itinerary scheduler model to determine latencies for LEA thresholds, use the generic TargetSchedModel::computeInstrLatency call.

llvm-svn: 330030

Summary:
This was originally part of rL328132, and led to the discovery
of the issues addressed in rL328987. Re-landing.

Reviewers: xur, davidxl, bkramer

Reviewed By: bkramer

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D45545

llvm-svn: 330029

Many IDEs set breakpoints using absolute paths and this causes problems when the full path of the source file path doesn't match what is in the debug info. This can be due to different build systems and do or do not resolve symlinks. This patch allows relative breakpoint to be set correctly without needing to do any target.source-map tricks. If IDEs want to, they can send down relative paths like:

./main.c
./src/main.c
src/main.c
foo/bar/src/main.c

I used the breakpoint resolver to match on the file basename and then we weed out anything whose relative paths don't match. This will be a huge improvement for IDEs as they can specify as much of a relative path as desired to uniquely identify a source file in the current project.



Differential Revision: https://reviews.llvm.org/D45592

llvm-svn: 330028

llvm-svn: 330027

Reviewers: gchatelet

Subscribers: tschuett, llvm-commits

Differential Revision: https://reviews.llvm.org/D45625

llvm-svn: 330026

llvm-svn: 330025

Reviewers: fhahn, rengolin, javed.absar, huntergr, SjoerdMeijer, t.p.northover, echristo, evandro

Reviewed By: rengolin

Subscribers: tschuett, llvm-commits, kristof.beyls

Differential Revision: https://reviews.llvm.org/D45618

llvm-svn: 330024

llvm-svn: 330023

llvm-svn: 330022

llvm-svn: 330021

llvm-svn: 330020

llvm-svn: 330019

Summary:
This change extend the register dependency check for implicit operands in Copy instructions.
Fixes PR36902.

Reviewers: thegameg, sebpop, uweigand, jnspaulsson, gberry, mcrosier, qcolombet, MatzeB

Reviewed By: thegameg

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D44958

llvm-svn: 330018

llvm-svn: 330017

Summary:
This patch improves detection of incomplete code for protos and text protos.
This is especially important for text protos in raw string literals, since they
might be partial strings concatenated, and we'd like to disable formatting in
these cases.

Reviewers: sammccall

Reviewed By: sammccall

Subscribers: klimek, cfe-commits

Differential Revision: https://reviews.llvm.org/D44203

llvm-svn: 330016

Currently the test fails in sandboxed environnements.

llvm-svn: 330015

Summary:
Added instructions for contiguous stores, ST1, with scalar+imm addressing
modes and corresponding tests. The patch also adds parsing of
'mul vl' as needed for the VL-scaled immediate.

This is patch [6/6] in a series to add assembler/disassembler support for
SVE's contiguous ST1 (scalar+imm) instructions.

Reviewers: fhahn, rengolin, javed.absar, huntergr, SjoerdMeijer, t.p.northover, echristo, evandro

Reviewed By: rengolin

Subscribers: tschuett, llvm-commits, kristof.beyls

Differential Revision: https://reviews.llvm.org/D45432

llvm-svn: 330014

llvm-svn: 330013

Differential Revision: https://reviews.llvm.org/D45515

llvm-svn: 330012

Differential Revision: https://reviews.llvm.org/D45514

llvm-svn: 330011

llvm-svn: 330010

Patch by: Rafael Stahl!

Differential Revision: https://reviews.llvm.org/D45564

llvm-svn: 330009

Summary:
 - Target-independent tests are run all the time.
 - Tests that codegen X86 code are run when X86 is in build targets.
 - Tests that run X86 jitted code are run only on X86 hosts.

Reviewers: gchatelet

Subscribers: mgorny, llvm-commits, tschuett

Differential Revision: https://reviews.llvm.org/D45614

llvm-svn: 330008

A followup for rL330003
https://bugs.llvm.org/show_bug.cgi?id=6773

llvm-svn: 330007

"the the" -> "the", "we we" -> "we", etc

llvm-svn: 330006

When we're dealing with virtual (memory) threads created by the OS
plugins, there's no guarantee that the real thread and the backing
thread share a protocol ID. Instead, we should iterate over the memory
threads to find the virtual thread that is backed by the current real
thread.

Differential revision: https://reviews.llvm.org/D45497

rdar://36485830

The original revision (r329891) was reverted because the associated
tests ran into a deadlock on the Linux bots. That problem was resolved
by r330002.

llvm-svn: 330005

Summary:
Previsouly, class completions items from the index were missing
template parameters in both the snippet and the label.

Reviewers: sammccall, hokein

Reviewed By: sammccall

Subscribers: klimek, jkorous-apple, ioeric, MaskRay, cfe-commits

Differential Revision: https://reviews.llvm.org/D45482

llvm-svn: 330004

https://bugs.llvm.org/show_bug.cgi?id=6773

As discussed there, some backends may want to undo this fold
(x86+bmi for scalars, x86+sse for vectors, ...)
https://bugs.llvm.org/show_bug.cgi?id=37104

https://rise4fun.com/Alive/JXt

llvm-svn: 330003

Summary:
When performing a synchronous resume, the API mutex is held until the
process is stopped. This is fine, except for when the OS plugins are processing
an event before the main thread is aware of it, in which case we end up with a
deadlock because in the internal thread we acquire a resource lock first, and
then wait for the API lock, while in the main thread we do the opposite, we
already hold the API mutex but are now waiting for the event mutex to handle
the event.

This patch fixes this by relaxing the need for the API lock in the OS plugins.
We can get away with this because we now this code is executed in the main
thread. As stated in the comment above, we just want to ensure nobody else
messes with the API while we're making a change. In theory it's possible that
the main thread would release the lock while we're executing the function, but
prevent this would require a more structural solution (which we want, but do
not have today).

The same workaround was already present, but this patch generalizes it to the
whole file.

This will allow me to re-land r329891.

Reviewers: clayborg, jingham, labath

Subscribers: llvm-commits, lldb-commits

Differential Revision: https://reviews.llvm.org/D45586

llvm-svn: 330002

Summary:
The fold added in D45108 did not account for the fact that
the and instruction is commutative, and if the mask is a variable,
the mask variable and the fold variable may be swapped.

I have noticed this by accident when looking into [[ https://bugs.llvm.org/show_bug.cgi?id=6773 | PR6773 ]]

This extends/generalizes that fold, so it is handled too.

Reviewers: spatel, craig.topper

Reviewed By: spatel

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D45539

llvm-svn: 330001

Summary:
Added Z_(b|h|s|d) vector list RegisterOperands along with support to
add/print the vector lists.

This is patch [5/6] in a series to add assembler/disassembler support for
SVE's contiguous ST1 (scalar+imm) instructions.

Reviewers: fhahn, rengolin, javed.absar, huntergr, SjoerdMeijer, t.p.northover, echristo, evandro

Reviewed By: fhahn

Subscribers: tschuett, kristof.beyls, llvm-commits

Differential Revision: https://reviews.llvm.org/D45431

llvm-svn: 330000

llvm-svn: 329999

Summary:
- last change (+ the Apple support change) missed a lot of indentation
- shorten architecture SOURCES definitions as most fit 1 line/arch
- comment in English what's where, and where the different .a come from
  (using only the word "runtime" in the comment isn't useful, since the
  CMake primitive itself says "runtime" in its name)
- skip unsupported architectures quickly, to avoid extra indentation

Reviewers: dberris, eizan, kpw

Subscribers: mgorny, delcypher, #sanitizers, llvm-commits

Differential Revision: https://reviews.llvm.org/D45568

llvm-svn: 329998