llvm-svn: 174411

llvm-svn: 174410

llvm-svn: 174409

llvm-svn: 174408

llvm-svn: 174407

llvm-svn: 174406

This was fixed by r174402.

llvm-svn: 174405

 - This controls whether to execute against the locally built library or
   not. The default is currently True which maps to what was already being done
   by default.

 - I'd appreciate it if someone can implement the proper handling of this flag
   on linux, I no longer remember the details of its .so handling.

llvm-svn: 174404

The changed files are not yet clang-format clean, but we are getting close.

llvm-svn: 174403

llvm-svn: 174402

Currently, when a fragment is relaxed, its size is modified, but its
offset is not (it gets laid out as a side effect of checking whether
it needs relaxation), then all subsequent fragments are invalidated
because their offsets need to change. When bundling is enabled,
relaxed fragments need to get laid out again, because the increase in
size may push it over a bundle boundary. So instead of only
invalidating subsequent fragments, also invalidate the fragment that
gets relaxed, which causes it to get laid out again.

This patch also fixes some trailing whitespace and fixes the
bundling-related debug output of MCFragments.

llvm-svn: 174401

Something very strange is going on with the output registers in this
target. Its ISelLowering code is inserting dangling CopyToReg nodes,
hoping that those physregs won't get clobbered before the RETURN.

This patch adds the output registers as implicit uses on RETURN
instructions in the custom emission pass. I'd much prefer to have those
CopyToReg nodes glued to the RETURNs, but I don't see how.

llvm-svn: 174400

The liveout lists are about to be removed from MRI, this is the only
place they were used after register allocation.

Get the live out V registers directly from the return instructions
instead.

llvm-svn: 174399

This is a text file with Markdown-ish formatting because we haven't decided
where analyzer internal documents should go, but it's probably better to
have this in source control than sitting on my local drive forever.

llvm-svn: 174398

llvm-svn: 174397

llvm-svn: 174396

Patch by: Vincent Lejeune

Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 174395

Use one intrinsic for all sorts of interpolation.
Use two separate unexpanded instructions to represent INTERP_XY and _ZW -
this will allow to eliminate one part if it's not used.
Track liveness of special interpolation regs instead of reserving them -
this will allow to reuse those regs, lowering reg pressure.

Patch By: Vadim Girlin

v2[Vincent Lejeune]: Rebased against current llvm master

Signed-off-by: Vadim Girlin <vadimgirlin@gmail.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 174394

llvm-svn: 174393

Emitting the function name allows us to check for it in the FileCheck
tests so we can make sure FileCheck is checking the output of the
correct function.

llvm-svn: 174392

Fixes 37 piglit tests and allows e.g. FlightGear to run with radeonsi.

Patch by: Michel Dänzer

Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 174391

llvm-svn: 174390

for the existing instructions.

llvm-svn: 174389

llvm-svn: 174388

llvm-svn: 174387

rdar://13140508

llvm-svn: 174386

since only one of them is allowed in command-line, process them separately.

Otherwise, if more than one is specified in the command-line, one is processed normally
and the others are going to be treated and included as header files.

Related to radar://13140508

llvm-svn: 174385

llvm-svn: 174384

Nuke lib/interception/mach_override.

llvm-svn: 174383

Using -output-replacements will now output the replacements instead
of the changed code. This allows easier integration with tools that
need full control over what changed.

The format is an array of objects with the members "offset" (number),
"length" (number) and "replacement_text" (string), for example:

[
  {
    "offset": 42,
    "length": 5,
    "replacement_text": "  "
  },
  {
    "offset": 105,
    "length": 4,
    "replacement_text": ""
  }
]

llvm-svn: 174382

Some paths through the copy constructors for 'ErrorOr' were calling
'get' when 'HasError' and 'IsValid' were not properly initialized.
Depending on what happened to be in memory for those member variables
the asserts in 'get' might incorrectly fire.  Fixed by ensuring that
the member variables in question are always initialized before calling
'get'.

llvm-svn: 174381

llvm-svn: 174380

llvm-svn: 174379

This is backwards compatible with earlier integrations.
Also adds a basic test and a test for the ranges integration.

You can now run:
clang-format -offset=42 -length=15 -offset=150 -length=22
To re-format the ranges (42, +15) and (150, +22).

llvm-svn: 174378

In the loop vectorizer cost model, we used to ignore stores/loads of a pointer
type when computing the widest type within a loop. This meant that if we had
only stores/loads of pointers in a loop we would return a widest type of 8bits
(instead of 32 or 64 bit) and therefore a vector factor that was too big.

Now, if we see a consecutive store/load of pointers we use the size of a pointer
(from data layout).

This problem occured in SingleSource/Benchmarks/Shootout-C++/hash.cpp (reduced
test case is the first test in vector_ptr_load_store.ll).

radar://13139343

llvm-svn: 174377

llvm-svn: 174376

This should fix the valgrind buildbot failure.

llvm-svn: 174375

It caused hangups in compiling clang/lib/Parse/ParseDecl.cpp and clang/lib/Driver/Tools.cpp in stage2 on some hosts.

llvm-svn: 174374

In case of partial right OOB, ASan was reporting
  X is located 0 bytes to the right of [A, B)
where X was actually inside [A, B).

With this change, ASan will report B as the error address in such case.

llvm-svn: 174373

The sh_link in the ELF section header of .ARM.exidx should
be filled with the section index of the corresponding text
section.

llvm-svn: 174372