Summary:
This patch introduces command-line support for the Armv8.6-a architecture and assembly support for BFloat16. Details can be found
https://community.arm.com/developer/ip-products/processors/b/processors-ip-blog/posts/arm-architecture-developments-armv8-6-a
in addition to the GCC patch for the 8..6-a CLI:
https://gcc.gnu.org/legacy-ml/gcc-patches/2019-11/msg02647.html
In detail this patch
- march options for armv8.6-a
- BFloat16 assembly
This is part of a patch series, starting with command-line and Bfloat16
assembly support. The subsequent patches will upstream intrinsics
support for BFloat16, followed by Matrix Multiplication and the
remaining Virtualization features of the armv8.6-a architecture.
Based on work by:
- labrinea
- MarkMurrayARM
- Luke Cheeseman
- Javed Asbar
- Mikhail Maltsev
- Luke Geeson
Reviewers: SjoerdMeijer, craig.topper, rjmccall, jfb, LukeGeeson
Reviewed By: SjoerdMeijer
Subscribers: stuij, kristof.beyls, hiraditya, dexonsmith, danielkiss, cfe-commits, llvm-commits
Tags: #clang, #llvm
Differential Revision: https://reviews.llvm.org/D76062
Summary:
In D76527, we stopped exporting symbols from clang, opt and llc unless
the `LLVM_ENABLE_PLUGINS` cmake variable is true (which causes clang's
own plugin collection to be built).
But another reasonable build configuration is to ask clang to export
its symbols for out-of-tree plugins to use, without building the
in-tree ones. That is, you might set `LLVM_EXPORT_SYMBOLS_FOR_PLUGINS`
without also setting `LLVM_ENABLE_PLUGINS` (at least if you're using
MSVC, where you need to ask explicitly for the symbols to be
exported).
In that situation, the symbols should still be exported, but after
D76527, they weren't being.
Reviewers: efriedma, john.brawn
Reviewed By: efriedma, john.brawn
Subscribers: mgorny, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D76760
Some MVE floating point instructions have gpr register variants that take
the scalar gpr value and splat them to all lanes. In order to accept
them in loops, the shuffle_vector and insert need to be sunk down into
the loop, next to the instruction so that ISel can see the whole
pattern.
This does that sinking for FAdd, FSub, FMul and FCmp. The patterns for
mul are slightly more constrained as there are no fms variants taking
register arguments.
Differential Revision: https://reviews.llvm.org/D76023
Summary:
Now Preferred is always the canonical (first) decl, Definition is always the def
if available.
In practice the index was already forcing this behaviour anyway, so there's no
change. (Unless you weren't using this index, in which case this patch makes
textDocument/declaration and toggling work as expected).
Reviewers: kadircet
Subscribers: ilya-biryukov, MaskRay, jkorous, arphaman, usaxena95, cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D73369
If an error happens which is related to a container the Container
Modeling checker adds note tags to all the container operations along
the bug path. This may be disturbing if there are other containers
beside the one which is affected by the bug. This patch restricts the
note tags to only the affected container and adjust the debug checkers
to be able to test this change.
Differential Revision: https://reviews.llvm.org/D75514
This allows conversion of a ParallelLoop from N induction variables to
some nuber of induction variables less than N.
The first intended use of this is for the GPUDialect to convert
ParallelLoops to iterate over 3 dimensions so they can be launched as
GPU Kernels.
To implement this:
- Normalize each iteration space of the ParallelLoop
- Use the same induction variable in a new ParallelLoop for multiple
original iterations.
- Split the new induction variable back into the original set of values
inside the body of the ParallelLoop.
Subscribers: mgorny, mehdi_amini, rriddle, jpienaar, burmako, shauheen, antiagainst, nicolasvasilache, arpith-jacob, mgester, lucyrfox, aartbik, liufengdb, Joonsoo, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D76363
Summary: Previously, we dropped the AST node for nonexistent member exprs.
Reviewers: sammccall
Subscribers: cfe-commits
Tags: #clang
Differential Revision: https://reviews.llvm.org/D76764
Container operations such as `push_back()`, `pop_front()`
etc. increment and decrement the abstract begin and end
symbols of containers. This patch introduces note tags
to `ContainerModeling` to track these changes. This helps
the user to better identify the source of errors related
to containers and iterators.
Differential Revision: https://reviews.llvm.org/D73720
- add method to get back an integer set from flat affine constraints;
this allows a round trip
- use this to complete the simplification of integer sets in
-simplify-affine-structures
- update FlatAffineConstraints::removeTrivialRedundancy to also do GCD
tightening and normalize by GCD (while still keeping it linear time).
Signed-off-by: Uday Bondhugula <uday@polymagelabs.com>
I want to extend D60940 to scalar FP which will prevent forming
masked instructions if the arithmetic op has another use. To
prepare for that, this patch updates tests to avoid repeating
the operation multiple times with different masking.
- UserParent = PN->getIncomingBlock(*I->use_begin());
+ UserParent = PN->getIncomingBlock(*SingleUse);
The first use of I may be droppable (llvm.assume).
When compiling llvm/lib/IR/AutoUpgrade.cpp with a bootstrapped clang
with ThinLTO with minimized bitcode files, I see such a case in
the function _ZN4llvm20UpgradeIntrinsicCallEPNS_8CallInstEPNS_8FunctionE
clang -c -fthinlto-index=AutoUpgrade.o.thinlto.bc AutoUpgrade.bc -O3
Unfortunately it is really difficult to get a minimized reproduce.
Previously, we would ignore alloca alignment when building the frame
and just use the natural alignment of the allocated type. If an alloca
is over-aligned for its IR type, this could lead to a frame entry with
inadequate alignment for the downstream uses of the alloca.
Since highly-aligned fields also tend to produce poor layouts under a
naive layout algorithm, I've also switched coroutine frames to use the
new optimal struct layout algorithm.
In order to communicate the frame size and alignment to later passes,
I needed to set align+dereferenceable attributes on the frame-pointer
parameter of the resume function. This is clearly the right thing to
do, but the align attribute currently seems to result in assumptions
being added during inlining that the optimizer cannot easily remove.
We can legalize the operation MUL for v8i16 with instruction (vmladduhm A, B, 0)
if altivec enabled. Now, it is set as custom and expand it later, which is not
the right way. And then, we can add the pattern to match the mul + add with (vmladduhm A, B, C)
Reviewed By: Nemanjai
Differential Revision: https://reviews.llvm.org/D76751
More mechanical splitting of tests so we can add a one use
check to the isel patterns for forming masked instructions.
In a few cases I changed immediates of instructions in
order to avoid needing to split.
Temporarily revert "tsan: fix leak of ThreadSignalContext for fibers"
because it breaks the LLDB bot on GreenDragon.
This reverts commit 93f7743851.
This reverts commit d8a0f76de7.
AMDGPUPropagateAttributes pass was skipping some of the functions
when cloning. Functions were added to root set and then skipped
on the next interation because they are already in the root set,
while were meant to be processed with different features.
Differential Revision: https://reviews.llvm.org/D76815
This particular debug shows all the time when I'm processing any MLIR
module:
discovered a new reachable node ^bb0
discovered a new reachable node ^bb0
discovered a new reachable node ^bb0
discovered a new reachable node ^bb0
...
(repeated x1875)
I think that printing all the basic blocks in the function is likely low
value enough that we can get away with removing this.
Differential Revision: https://reviews.llvm.org/D76813
We have several test directories whose names end in .spec. Ensure we
don't accidentally ignore them by removing a section of the .gitignore
that's irrelevant for libc++.
Differential Revision: https://reviews.llvm.org/D69195
Summary:
The test performs add on vector<16384xf32> with
number of workgroups = (128, 1, 1)
local workgroup size = (128, 1, 1)
On a NVIDIA Quadro P1000, I see the following results:
Command buffer submit time: 13us
Compute shader execution time: 19.616us
Differential Revision: https://reviews.llvm.org/D76799
AMDGPUPropagateAttributes can swap names while cloning a function.
Only do it if original symbol was not externally visible.
Differential Revision: https://reviews.llvm.org/D76789
Introduced by https://reviews.llvm.org/D72687. This condition can happen
when the tests are not being run at all, and we're only trying to generate
the libc++ headers.
Some tests do not fail at all when -verify is not supported, unless some
arbitrary warning flag is added to make them fail. We currently used
-Werror=unused-result to make them fail, but doing so makes the test
suite a lot more inscrutable. It seems better to just disable those
tests when -verify is not supported.
Differential Revision: https://reviews.llvm.org/D76256
Summary:
Some operations have custom syntax where an operand is always followed by a
specific token of streams if the operand is present. Parsing such operations
requires the ability to optionally parse an operand. Provide a relevant
function in the custom Op parser.
Differential Revision: https://reviews.llvm.org/D76779
Summary:
This patch allows code-sinking in InstCombine to be performed when instruction have uses in llvm.assume.
Use are considered droppable when it is preferable to modify the User such that the use disappears rather than to prevent a transformation because of the use.
for now uses are considered droppable if they are in an llvm.assume.
Reviewers: jdoerfert, nikic, spatel, lebedev.ri, sstefan1
Reviewed By: jdoerfert
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D73832
Summary:
Rename `pred_const_range` to `const_pred_range` to make it consistent with
the other pred/succ iterator definitions.
Reviewers: nicholas, dblaikie, nlewycky
Subscribers: llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D75962
Summary:
Rename `succ_const_iterator` to `const_succ_iterator` and
`succ_const_range` to `const_succ_range` for consistency with the
predecessor iterators, and the corresponding iterators in
MachineBasicBlock.
Reviewers: nicholas, dblaikie, nlewycky
Subscribers: hiraditya, bmahjour, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D75952
Summary:
The attribute parser fails to correctly parse unsigned 64 bit
attributes as the check `isNegative ? (int64_t)-val.getValue() >= 0
: (int64_t)val.getValue() < 0` will falsely detect an overflow for
unsigned values larger than 2^63-1.
This patch reworks the overflow logic to instead of doing arithmetic
on int64_t use APInt::isSignBitSet() and knowledge of the attribute
type.
Test-cases which verify the de-facto behavior of the parser and
triggered the previous faulty handing of unsigned 64 bit attrbutes are
also added.
Differential Revision: https://reviews.llvm.org/D76493
As reported in PR45298 and PR45299, vector_size type checking would
crash when done in a situation where the scalar is dependent, such as
a member of the current instantiation.
This is because the scalar checking ensures that you can implicitly
convert a value to a vector-type as long as it doesn't require
truncation. It does this by using the constant evaluator to get the
value as a float. Unfortunately, if the scalar is dependent (such as a
member of the current instantiation), we would hit the assert in the
evaluator.
This patch suppresses the truncation- of-value check in the first phase
of translation. All values are properly errored upon instantiation. This
has one minor regression, in that previously in a non-asserts build,
template<typename T>
struct S {
float4 f(float4 f) {
return k + f;
}
static constexpr k = 1.1; // causes a truncation on conversion.
};
would error immediately. Because 'k' is value dependent (as a
member-of-the-current-instantiation), this would still be evaluatable
(despite normally asserting). Due to this patch, this diagnostic is
delayed until instantiation time.
Because using -print-imports is not thread-safe, make the test rely on llvm-dis instead.
Also cover the ICALL-PROM part as intended originally.
Differential Revision: https://reviews.llvm.org/D76775
Sam McCall