As discussed on llvm-dev:
http://lists.llvm.org/pipermail/llvm-dev/2020-April/140729.html
This is hopefully the final remaining showstopper before we can remove
the 'experimental' from the reduction intrinsics.
No behavior was specified for the FP min/max reductions, so we have a
mess of different interpretations.
There are a few potential options for the semantics of these max/min ops.
I think this is the simplest based on current behavior/implementation:
make the reductions inherit from the existing llvm.maxnum/minnum intrinsics.
These correspond to libm fmax/fmin, and those are similar to the (now
deprecated?) IEEE-754 maxNum/minNum functions (NaNs are treated as missing
data). So the default expansion creates calls to libm functions.
Another option would be to inherit from llvm.maximum/minimum (NaNs propagate),
but most targets just crash in codegen when given those nodes because no
default expansion was ever implemented AFAICT.
We could also just assume 'nnan' semantics by default (we are already
assuming 'nsz' semantics in the maxnum/minnum intrinsics), but some targets
(AArch64, PowerPC) support the more defined behavior, so it doesn't make much
sense to not allow a tighter spec. Fast-math-flags (nnan) can be used to
loosen the semantics.
(Note that D67507 was proposed to update the LangRef to acknowledge the more
recent IEEE-754 2019 standard, but that patch seems to have stalled. If we do
update based on the new standard, the reduction instructions can seamlessly
inherit from whatever updates are made to the max/min intrinsics.)
x86 sees a regression here on 'nnan' tests because we have underlying,
longstanding bugs in FMF creation/propagation. Those need to be fixed apart
from this change (for example: https://llvm.org/PR35538). The expansion
sequence before this patch may not have been correct.
Differential Revision: https://reviews.llvm.org/D87391
This adjusts the description of `llvm.memcpy` to also allow operands
to be equal. This is in line with what Clang currently expects.
This change is intended to be temporary and followed by re-introduce
a variant with the non-overlapping guarantee for cases where we can
actually ensure that property in the front-end.
See the links below for more details:
http://lists.llvm.org/pipermail/cfe-dev/2020-August/066614.html
and PR11763.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D86815
The wording before this patch applies to llvm.mem.parallel_loop_access, not access groups.
Reviewed By: mppf, hfinkel
Differential Revision: https://reviews.llvm.org/D83781
This patch makes LangRef be explicit about the value of padding when storing an aggregate.
It states that when an aggregate is stored into memory, padding is filled with undef.
Here is a clue that supports this change (edited to reflect the discussion from llvm-dev):
- IPSCCP ignores padding and directly stores a constant aggregate if possible. It loses the data stored in the padding. https://godbolt.org/z/xzenYs Memcpyopt ignores (the preexisting value of) padding when copying an aggregate or storing a constant: https://godbolt.org/z/hY6ndd / https://godbolt.org/z/3WMP5a
The two items below are not relevant with this patch because Clang lowers load/store of individual field of struct into load/stores of the corresponding pointer with a primitive type. Also, when copy is needed, it uses memcpy instead of load/store of an aggregate, as discussed in the llvm-dev. However, this patch is still valid (as discussed) because it is needed to explain the two optimizations above.
- According to C17, the value of padding bytes when storing values in structures or unions is unspecified.
- I updated Alive2 and it did not find any problematic transformation from LLVM unit tests and while running translation validation of a few C programs.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D86189
We had already specified that second argument `n` of this intrinsic is `n > 0`,
but now add to this that the result is a poison value if this is not the case.
Differential Revision: https://reviews.llvm.org/D86637
According to the current LangRef, Memset/memcpy/memmove can take a
null/dangling pointer if the size is zero.
(Relevant thread: http://lists.llvm.org/pipermail/llvm-dev/2017-July/115665.html )
This patch expands it and allows the functions to take undef/poison pointers
too.
This required the updates in the align attribute since it isn't specified
what is the alignment of undef/poison pointers.
This patch states that their alignment is 1.
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D86643
A first version of get.active.lane.mask was committed in rG7fb8a40e5220. One of
the main purposes and uses of this intrinsic is to communicate information from
the middle-end to the back-end, but its current definition and semantics make
this actually very difficult. The intrinsic was defined as:
@llvm.get.active.lane.mask(%IV, %BTC)
where %BTC is the Backedge-Taken Count (variable names are different in the
LangRef spec). This allows to implicitly communicate the loop tripcount, which
can be reconstructed by calculating BTC + 1. But it has been very difficult to
prove that calculating BTC + 1 is safe and doesn't overflow. We need
complicated range and SCEV analysis, and thus the problem is that this
intrinsic isn't really doing what it was supposed to solve. Examples of the
overflow checks that are required in the (ARM) back-end are D79175 and D86074,
which aren't even complete/correct yet.
To solve this problem, we are revising the definitions/semantics for
get.active.lane.mask to avoid all the complicated overflow analysis. This means
that instead of communicating the BTC, we are now using the loop tripcount. Now
using LangRef's variable names, its semantics is changed from:
icmp ule (%base + i), %n
to:
icmp ult (%base + i), %n
with %n > 0 and corresponding to the loop tripcount. The intrinsic signature
remains the same.
Differential Revision: https://reviews.llvm.org/D86147
(Forgot to land this a couple of weeks back.)
In a recent series of changes, I've introduced support for using the respective operand bundle kinds on the statepoint. At the moment, code supports either/or, but there's no need to keep the old support around. For the moment, I am simply changing the specification and verifier to require zero length argument sets in the intrinsic.
The intrinsic itself is experimental. Given that, there's no forward serialization needed. The in tree uses and generation have already been updated to use the new operand bundle based forms, the only folks broken by the change will be those with frontends generating statepoints directly and the updates should be easy.
Why not go ahead and just remove the arguments entirely? Well, I plan to. But while working on this I've found that almost all of the arguments to the statepoint can be expressed via operand bundles or attributes. Given that, I'm planning a radical simplification of the arguments and figured I'd do one update not several small ones.
Differential Revision: https://reviews.llvm.org/D80892
Summary:
This support is needed for the Fortran array variables with pointer/allocatable
attribute. This support enables debugger to identify the status of variable
whether that is currently allocated/associated.
for pointer array (before allocation/association)
without DW_AT_associated
(gdb) pt ptr
type = integer (140737345375288:140737354129776)
(gdb) p ptr
value requires 35017956 bytes, which is more than max-value-size
with DW_AT_associated
(gdb) pt ptr
type = integer (:)
(gdb) p ptr
$1 = <not associated>
for allocatable array (before allocation)
without DW_AT_allocated
(gdb) pt arr
type = integer (140737345375288:140737354129776)
(gdb) p arr
value requires 35017956 bytes, which is more than max-value-size
with DW_AT_allocated
(gdb) pt arr
type = integer, allocatable (:)
(gdb) p arr
$1 = <not allocated>
Testing
- unit test cases added
- check-llvm
- check-debuginfo
Reviewed By: aprantl
Differential Revision: https://reviews.llvm.org/D83544
This allows tracking the in-memory type of a pointer argument to a
function for ABI purposes. This is essentially a stripped down version
of byval to remove some of the stack-copy implications in its
definition.
This includes the base IR changes, and some tests for places where it
should be treated similarly to byval. Codegen support will be in a
future patch.
My original attempt at solving some of these problems was to repurpose
byval with a different address space from the stack. However, it is
technically permitted for the callee to introduce a write to the
argument, although nothing does this in reality. There is also talk of
removing and replacing the byval attribute, so a new attribute would
need to take its place anyway.
This is intended avoid some optimization issues with the current
handling of aggregate arguments, as well as fixes inflexibilty in how
frontends can specify the kernel ABI. The most honest representation
of the amdgpu_kernel convention is to expose all kernel arguments as
loads from constant memory. Today, these are raw, SSA Argument values
and codegen is responsible for turning these into loads.
Background:
There currently isn't a satisfactory way to represent how arguments
for the amdgpu_kernel calling convention are passed. In reality,
arguments are passed in a single, flat, constant memory buffer
implicitly passed to the function. It is also illegal to call this
function in the IR, and this is only ever invoked by a driver of some
kind.
It does not make sense to have a stack passed parameter in this
context as is implied by byval. It is never valid to write to the
kernel arguments, as this would corrupt the inputs seen by other
dispatches of the kernel. These argumets are also not in the same
address space as the stack, so a copy is needed to an alloca. From a
source C-like language, the kernel parameters are invisible.
Semantically, a copy is always required from the constant argument
memory to a mutable variable.
The current clang calling convention lowering emits raw values,
including aggregates into the function argument list, since using
byval would not make sense. This has some unfortunate consequences for
the optimizer. In the aggregate case, we end up with an aggregate
store to alloca, which both SROA and instcombine turn into a store of
each aggregate field. The optimizer never pieces this back together to
see that this is really just a copy from constant memory, so we end up
stuck with expensive stack usage.
This also means the backend dictates the alignment of arguments, and
arbitrarily picks the LLVM IR ABI type alignment. By allowing an
explicit alignment, frontends can make better decisions. For example,
there's real no advantage to an aligment higher than 4, so a frontend
could choose to compact the argument layout. Similarly, there is a
high penalty to using an alignment lower than 4, so a frontend could
opt into more padding for small arguments.
Another design consideration is when it is appropriate to expose the
fact that these arguments are all really passed in adjacent
memory. Currently we have a late IR optimization pass in codegen to
rewrite the kernel argument values into explicit loads to enable
vectorization. In most programs, unrelated argument loads can be
merged together. However, exposing this property directly from the
frontend has some disadvantages. We still need a way to track the
original argument sizes and alignments to report to the driver. I find
using some side-channel, metadata mechanism to track this
unappealing. If the kernel arguments were exposed as a single buffer
to begin with, alias analysis would be unaware that the padding bits
betewen arguments are meaningless. Another family of problems is there
are still some gaps in replacing all of the available parameter
attributes with metadata equivalents once lowered to loads.
The immediate plan is to start using this new attribute to handle all
aggregate argumets for kernels. Long term, it makes sense to migrate
all kernel arguments, including scalars, to be passed indirectly in
the same manner.
Additional context is in D79744.
Make explicit that freeze does not touch paddings of an aggregate.
(Relevant comment: https://reviews.llvm.org/D83752#2152550)
This implies that `v = freeze(load p); store v, q` may still leave undef bits
or poison in memory if `v` is an aggregate, but it still happens for
non-byte integers such as i1.
Differential Revision: https://reviews.llvm.org/D83927
Some of the system registers readable on AArch64 and ARM platforms
return different values with each read (for example a timer counter),
these shouldn't be hoisted outside loops or otherwise interfered with,
but the normal @llvm.read_register intrinsic is only considered to read
memory.
This introduces a separate @llvm.read_volatile_register intrinsic and
maps all system-registers on ARM platforms to use it for the
__builtin_arm_rsr calls. Registers declared with asm("r9") or similar
are unaffected.
This changes the matrix load/store intrinsic definitions to load/store from/to
a pointer, and not from/to a pointer to a vector, as discussed in D83477.
This also includes the recommit of "[Matrix] Tighten LangRef definitions and
Verifier checks" which adds improved language reference descriptions of the
matrix intrinsics and verifier checks.
Differential Revision: https://reviews.llvm.org/D83785
Loop metadata nodes do not adhere to the documented property:
(a) LoopIDs are not unique: Any pass that duplicates IR will do it
including its metadata (e.g. LoopVersioning) such that multiple
loops are linked with the same LoopID. There is even a test case
(Transforms/LoopUnroll/unroll-pragmas-disabled.ll) for multiple
loops with the same LoopID.
(b) LoopIDs are not persistent: Adding or removing an item from a LoopID
can only be done by creating a new MDNode and assigning it to the
loop's branch(es). Passes such as LoopUnroll (llvm.loop.unroll.disable)
and LoopVectorize (llvm.loop.isvectorized) use this to mark loops to
not be transformed multiple times or to avoid that a LoopVersioned
original loop is transformed.
Update the documentation according to how llvm.loop is used in practice.
Differential Revision: https://reviews.llvm.org/D55290
This tightens the matrix intrinsic definitions in LLVM LangRef and adds
correspondings checks to the IR Verifier.
Differential Revision: https://reviews.llvm.org/D83477
fadd (fma A, B, (fmul C, D)), E --> fma A, B, (fma C, D, E)
This is only allowed when "reassoc" is present on the fadd.
As discussed in D80801, this transform goes beyond
what is allowed by "contract" FMF (-ffp-contract=fast).
That is because we are fusing the trailing add of 'E' with a
multiply, but without "reassoc", the code mandates that the
products A*B and C*D are added together before adding in 'E'.
I've added this example to the LangRef to try to clarify the
meaning of "contract". If that seems reasonable, we should
probably do something similar for the clang docs because
there does not appear to be any formal spec for the behavior
of -ffp-contract=fast.
Differential Revision: https://reviews.llvm.org/D82499
LLVM currently does not require function parameters or return values
to be fully initialized, and does not care if they are poison. This can
be useful if the frontend ABI makes no such demands, but may prevent
helpful backend transformations in case they do. Specifically, the C
and C++ languages require all scalar function operands to be fully
determined.
Introducing this attribute is of particular use to MemorySanitizer
today, although other transformations may benefit from it as well.
We can modify MemorySanitizer instrumentation to provide modest (17%)
space savings where `frozen` is present.
This commit only adds the attribute to the Language Reference, and
the actual implementation of the attribute will follow in a separate
commit.
Differential Revision: https://reviews.llvm.org/D82316
This cleans up the stack allocated by a @llvm.call.preallocated.setup.
Should either call the teardown or the preallocated call to clean up the
stack. Calling both is UB.
Add LangRef.
Add verifier check that the token argument is a @llvm.call.preallocated.setup.
Reviewed By: efriedma
Differential Revision: https://reviews.llvm.org/D83354
Summary:
D80831 changed part of the prefix usage for AIX.
But there are other places getting prefix from DataLayout.
This patch intends to make prefix usage consistent on AIX.
Reviewed by: hubert.reinterpretcast, daltenty
Differential Revision: https://reviews.llvm.org/D81270
Every other value parameter attribute uses parentheses, so accept this
as the preferred modern syntax. Updating everything to use the new
syntax is left for a future change.
Add a new builtin-function __builtin_expect_with_probability and
intrinsic llvm.expect.with.probability.
The interface is __builtin_expect_with_probability(long expr, long
expected, double probability).
It is mainly the same as __builtin_expect besides one more argument
indicating the probability of expression equal to expected value. The
probability should be a constant floating-point expression and be in
range [0.0, 1.0] inclusive.
It is similar to builtin-expect-with-probability function in GCC
built-in functions.
Differential Revision: https://reviews.llvm.org/D79830
This patch adjust the load/store matrix intrinsics, formerly known as
llvm.matrix.columnwise.load/store, to improve the naming and allow
passing of extra information (volatile).
The patch performs the following changes:
* Rename columnwise.load/store to column.major.load/store. This is more
expressive and also more in line with the naming in Clang.
* Changes the stride arguments from i32 to i64. The stride can be
larger than i32 and this makes things more uniform with the way
things are handled in Clang.
* A new boolean argument is added to indicate whether the load/store
is volatile. The lowering respects that when emitting vector
load/store instructions
* MatrixBuilder is updated to require both Alignment and IsVolatile
arguments, which are passed through to the generated intrinsic. The
alignment is set using the `align` attribute.
The changes are grouped together in a single patch, to have a single
commit that breaks the compatibility. We probably should be fine with
updating the intrinsics, as we did not yet officially support them in
the last stable release. If there are any concerns, we can add
auto-upgrade rules for the columnwise intrinsics though.
Reviewers: anemet, Gerolf, hfinkel, andrew.w.kaylor, LuoYuanke, nicolasvasilache, rjmccall, ftynse
Reviewed By: anemet, nicolasvasilache
Differential Revision: https://reviews.llvm.org/D81472
Summary:
This patch adds optional field into function summary,
implements asm and bitcode serialization. YAML
serialization is omitted and can be added later if
needed.
This patch includes this information into summary only
if module contains at least one sanitize_memtag function.
In a near future MTE is the user of the analysis.
Later if needed we can provede more direct control
on when information is included into summary.
Reviewers: eugenis
Subscribers: hiraditya, steven_wu, dexonsmith, arphaman, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D80908
Currently, gc.relocates are defined in terms of indices into the statepoint's operand list. Given the gc args are at the end of a variable length list of operands, this makes interpreting their indices by hand a tad challenging. We can simplify the statepoint sequence and improve readability quite a bit by pulling these new operands into their own named operand bundle.
This patch defines a new operand bundle tag "gc-live". The semantics of the bundle are the same as the existing gc arguments of a statepoint. This patch simply introduces the definition and codegen for the bundle, future patches will migrate RS4GC to emitting the new form.
Interestingly, with this done and the recent migration to using deopt and gc-transition bundles, we really don't have much left in the statepoint itself. It really looks like the existing ID and flags fields are redundant; we have (existing!) attributes for all of them. I think we'll be able to reduce the gc.statepoint signature to simply a wrapped call (e.g. actual target and actual arguments).
Differential Revision: https://reviews.llvm.org/D80937
Currently all code instances within the matrix lowering pass consider
matrix A to be MxN and B to be NxK, producing C which is MxK. Anyone
interacting with this API after reading the docs but without reading the pass
would expect A: MxK, B: KxN, and C: MxN. These changes bring the documentation
in line with the implementation.
One point of concern with this, the original signature as described in the docs
may be better or at least more expected. The interface as it was written
reflected other common matrix multiplication interfaces such as BLAS'[1], where
the matrices are MxK, KxN, MxN respectively. Choosing to honor this requires
changing code and tests instead, but should be mostly just renaming of variables.
Patch by Braedy Kuzma <braedy@ualberta.ca>
[1] http://www.netlib.org/lapack/explore-html/db/dc9/group__single__blas__level3_gafe51bacb54592ff5de056acabd83c260.html#gafe51bacb54592ff5de056acabd83c260
Reviewers: anemet, LuoYuanke, nicolasvasilache, fhahn
Reviewed By: fhahn
Differential Revision: https://reviews.llvm.org/D80663
This is split off from D79100 and:
- adds a intrinsic description/definition for @llvm.get.active.lane.mask(), and
- describe its semantics in LangRef.
As described (in more detail) in its LangRef section, it is semantically
equivalent to an icmp with the vector induction variable and the back-edge
taken count, and generates a mask of active/inactive vector lanes.
It will have several use cases. First, it will be used by the
ExpandVectorPredication pass for the VP intrinsics, to expand VP intrinsics for
scalable vectors on targets that do not support the `%evl` parameter, see
D78203.
Also, this is part of, and essential for our ARM MVE tail-predication story:
- this intrinsic will be emitted by the LoopVectorizer in D79100, when
the scalar epilogue is tail-folded into the vector body. This new intrinsic
will generate the predicate for the masked loads/stores, and it takes the
back-edge taken count as an argument. The back-edge taken count represents the
number of elements processed by the loop, which we need to setup MVE
tail-predication.
- Emitting the intrinsic is controlled by a new TTI hook, see D80597.
- We pick up this new intrinsic in an ARM MVETailPredication backend pass, see
D79175, and convert it to a MVE target specific intrinsic/instruction to
create a tail-predicated loop.
Differential Revision: https://reviews.llvm.org/D80596
The HardwareLoop intrinsics were missing and not described in LangRef. This
adds these descriptions/definitions.
Differential Revision: https://reviews.llvm.org/D80316
Summary:
preallocated and musttail can work together, but we don't want to call
@llvm.call.preallocated.setup() to modify the stack in musttail calls.
So we shouldn't have the "preallocated" operand bundle when a
preallocated call is musttail.
Also disallow use of preallocated on calls without preallocated.
Codegen not yet implemented.
Subscribers: hiraditya, llvm-commits
Tags: #llvm
Differential Revision: https://reviews.llvm.org/D80581
Summary:
A struct argument can be passed-by-value to a callee via a pointer to a
temporary stack copy. Add support for emitting an entry value DBG_VALUE
when an indirect parameter DBG_VALUE becomes unavailable. This is done
by omitting DW_OP_stack_value from the entry value expression, to make
the expression describe the location of an object.
rdar://63373691
Reviewers: djtodoro, aprantl, dstenb
Subscribers: hiraditya, lldb-commits, llvm-commits
Tags: #lldb, #llvm
Differential Revision: https://reviews.llvm.org/D80345
This intrinsic implements IEEE-754 operation roundToIntegralTiesToEven,
and performs rounding to the nearest integer value, rounding halfway
cases to even. The intrinsic represents the missed case of IEEE-754
rounding operations and now llvm provides full support of the rounding
operations defined by the standard.
Differential Revision: https://reviews.llvm.org/D75670
This change makes minor correction to the implementation of intrinsic
`llvm.flt.rounds`:
- Added documentation entry in LangRef,
- Attributes of the intrinsic changed to be in line with other functions
dependent of floating-point environment.
Differential Revision: https://reviews.llvm.org/D79322
Summary: 'A' constraint requires an immediate int or fp constant that can be inlined in an instruction encoding.
Reviewers: arsenm, rampitec
Differential Revision: https://reviews.llvm.org/D78494
The "null-pointer-is-valid" attribute needs to be checked by many
pointer-related combines. To make the check more efficient, convert
it from a string into an enum attribute.
In the future, this attribute may be replaced with data layout
properties.
Differential Revision: https://reviews.llvm.org/D78862
Sanjay Patel