r189189 implemented AVX512 unpack by essentially performing a 256-bit unpack
between the low and the high 256 bits of src1 into the low part of the
destination and another unpack of the low and high 256 bits of src2 into the
high part of the destination.
I don't think that's how unpack works. AVX512 unpack simply has more 128-bit
lanes but other than it works the same way as AVX. So in each 128-bit lane,
we're always interleaving certain parts of both operands rather different
parts of one of the operands.
E.g. for this:
__v16sf a = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 };
__v16sf b = { 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 };
__v16sf c = __builtin_shufflevector(a, b, 0, 8, 1, 9, 4, 12, 5, 13, 16,
24, 17, 25, 20, 28, 21, 29);
we generated punpcklps (notice how the elements of a and b are not interleaved
in the shuffle). In turn, c was set to this:
0 16 1 17 4 20 5 21 8 24 9 25 12 28 13 29
Obviously this should have just returned the mask vector of the shuffle
vector.
I mostly reverted this change and made sure the original AVX code worked
for 512-bit vectors as well.
Also updated the tests because they matched the logic from the code.
llvm-svn: 217602
This is an extension of the change made with r215820:
http://llvm.org/viewvc/llvm-project?view=revision&revision=215820
That patch allowed combining of splatted vector FP constants that are multiplied.
This patch allows combining non-uniform vector FP constants too by relaxing the
check on the type of vector. Also, canonicalize a vector fmul in the
same way that we already do for scalars - if only one operand of the fmul is a
constant, make it operand 1. Otherwise, we miss potential folds.
This fold is also done by -instcombine, but it's possible that extra
fmuls may have been generated during lowering.
Differential Revision: http://reviews.llvm.org/D5254
llvm-svn: 217599
Refactored the R600_LDS_1A2D class a bit to get it to actually work.
It seemed to be previously unused and broken.
We also have to disable the conversion to the noret variant for now in
R600ISelLowering because the getLDSNoRetOp method only handles 1A1D LDS ops.
Someone can feel free to modify the AMDGPU::getLDSNoRetOp method to
work for more than 1A1D variants of LDS operations. It's being left as a
future TODO for now.
Signed-off-by: Aaron Watry <awatry at gmail.com>
Reviewed-by: Matt Arsenault <matthew.arsenault@amd.com>
llvm-svn: 217596
Now that the operations are all implemented, we can test this sub-arch here.
Signed-off-by: Aaron Watry <awatry@gmail.com>
Reviewed-by: Matt Arsenault <matthew.arsenault@amd.com>
llvm-svn: 217595
This was only present for SI before.
Cayman may still be missing, but I am unable to test that currently.
v2: Don't create atomicrmw max tests in separate file
Signed-off-by: Aaron Watry <awatry@gmail.com>
Reviewed-by: Matt Arsenault <matthew.arsenault@amd.com>
CC: Tom Stellard <thomas.stellard@amd.com>
llvm-svn: 217589
Summary:
http://llvm.org/bugs/show_bug.cgi?id=20892
Add support of C-style formatting enabling/disabling directives. Now the following two styles are supported:
// clang-format on
/* clang-format on */
The flexibility in comments (support of extra spaces and/or slashes, etc.) is deliberately avoided to simplify search in large code bases.
Reviewers: djasper
Reviewed By: djasper
Subscribers: cfe-commits, curdeius, klimek
Differential Revision: http://reviews.llvm.org/D5309
llvm-svn: 217588
This patch fixes the bad argument that GAS accepted but the IAS didn't,
ie. {#0x20}, moving it to {0x20} which both accept. It also makes the
ARMv7+ save/restore correct by using VFP instructions rather than old
co-processor ones.
Fixes PR20529.
llvm-svn: 217585
Summary:
They were used in the 'Module Structure' example but weren't otherwise
documented.
Credit to Reed Kotler for noticing.
Reviewers: hans
Reviewed By: hans
Subscribers: hans, llvm-commits
Differential Revision: http://reviews.llvm.org/D5191
llvm-svn: 217583
David Blaikie's commits r217563 & r217564, which added shared_ptr to the
CostPool have fixed some memory leak issues exposed by the PBQP with
coalescing constraints.
The sanitizer bot was failing because of those leaks. Now that the leaks
are gone, we can reenable the aarch64/pbqp test.
llvm-svn: 217580
The provided base must also be a multiple of the system's page size, which is a
reasonable enough demand.
Also check the other diagnostics more thoroughly.
llvm-svn: 217577
We used to crash processing any relevant @llvm.assume on a 32-bit target
(because we'd ask SE to subtract expressions of differing types). I've copied
our 'simple.ll' test, but with the data layout from arm-linux-gnueabihf to get
some meaningful test coverage here.
llvm-svn: 217574
The existing system linkers on Darwin and Linux are called "ld". We'd like to
eventually drop in lld as "ld" and have it just work. But lld is a universal
linker that requires the first option to be -flavor to know which command line
mode to emulate (gnu or darwin).
This change tests if argv[0] is "ld" and if so, if the tool was built on MacOSX
then assume the darwin flavor otherwise the gnu flavor. There are two test
cases which copy lld to "ld" and then run it. One for darwin and one for linux.
llvm-svn: 217566
Leveraging both intrusive shared_ptr-ing (std::enable_shared_from_this)
and shared_ptr<T>-owning-U (to allow external users to hold
std::shared_ptr<CostT> while keeping the underlying PoolEntry alive).
The intrusiveness could be removed if we had a weak_set that implicitly
removed items from the set when their underlying data went away.
This /might/ fix an existing memory leak reported by LeakSanitizer in
r217504.
llvm-svn: 217563
Need to convert the 64 element offset into bytes, not just the element
size like the normal case instructions.
Noticed by inspection. This can't be hit now because
st64 instructions aren't emitted during instruction selection,
and the post-RA scheduler isn't enabled.
llvm-svn: 217560
We may as well just use Symbolizer::GetOrInit() in all the cases.
Don't call Symbolizer::Get() early in tools initialization: these days
it doesn't do any important setup work, and we may as well create the
symbolizer the first time it's actually needed.
llvm-svn: 217558
Numerous changes, including:
* Changed the way variables and instructions are handled in basic blocks to
be more efficient.
* Eliminated SExprRef.
* Simplified futures.
* Fixed documentation.
* Compute dominator and post dominator trees.
llvm-svn: 217556
1. We were hitting the NextIsPrevious assertion because we were trying
to merge decl chains that were independent of each other because we had
no Sema object to allow them to find existing decls. This is fixed by
delaying loading the "preloaded" decls until Sema is available.
2. We were trying to get identifier info from an annotation token, which
asserts. The fix is to special-case the module annotations in the
preprocessed output printer.
Fixed in a single commit because when you hit 1 you almost invariably
hit 2 as well.
llvm-svn: 217550
Adam Nemet