These are identified by misc-include-cleaner. I've filtered out those
that break builds. Also, I'm staying away from llvm-config.h,
config.h, and Compiler.h, which likely cause platform- or
compiler-specific build failures.
These are identified by misc-include-cleaner. I've filtered out those
that break builds. Also, I'm staying away from llvm-config.h,
config.h, and Compiler.h, which likely cause platform- or
compiler-specific build failures.
This revision adds DeviceMaskingAttrInterface and extends
DeviceMappingArrayAttr to accept a union of DeviceMappingAttrInterface
and DeviceMaskingAttrInterface.
Support is added to GPUTransformOps to take advantage of this
information and lower to block/warpgroup/warp/thread specialization when
mapped to linear ids.
The revision also connects to scf::ForallOp and uses the new attribute
to implement warp specialization.
The implementation is in the form of a GPUMappingMaskAttr, which can be
additionally passed to the scf.forall.mapping attribute to specify a
mask on compute resources that should be active.
In the first implementation the masking is a bitfield that specifies for
each processing unit whether it is active or not.
In the future, we may want to implement this as a symbol to refer to
dynamically defined values.
Extending op semantics with an operand is deemed too intrusive at this
time.
---------
Co-authored-by: Oleksandr "Alex" Zinenko <git@ozinenko.com>
This revision adds a new attribute for mapping `scf.forall` to linear
lane ids.
Example:
```
// %arg2 and %arg3 map to lanes [0, 6) and are turned into epxressions
// involving threadIdx.x/y by the map_nested_forall_to_threads
// transformation. This results in a if (linear_thread_id < 6) conditional.
scf.forall (%arg2, %arg3) in (2, 3) {
...
} {mapping = [#gpu.lane<linear_dim_0>, #gpu.lane<linear_dim_1>]}
```
---------
Co-authored-by: Oleksandr "Alex" Zinenko <git@ozinenko.com>
Note that PointerUnion::{is,get} have been soft deprecated in
PointerUnion.h:
// FIXME: Replace the uses of is(), get() and dyn_cast() with
// isa<T>, cast<T> and the llvm::dyn_cast<T>
I'm not touching PointerUnion::dyn_cast for now because it's a bit
complicated; we could blindly migrate it to dyn_cast_if_present, but
we should probably use dyn_cast when the operand is known to be
non-null.
Transform interfaces are implemented, direction or via extensions, in
libraries belonging to multiple other dialects. Those dialects don't
need to depend on the non-interface part of the transform dialect, which
includes the growing number of ops and transitive dependency footprint.
Split out the interfaces into a separate library. This in turn requires
flipping the dependency from the interface on the dialect that has crept
in because both co-existed in one library. The interface shouldn't
depend on the transform dialect either.
As a consequence of splitting, the capability of the interpreter to
automatically walk the payload IR to identify payload ops of a certain
kind based on the type used for the entry point symbol argument is
disabled. This is a good move by itself as it simplifies the interpreter
logic. This functionality can be trivially replaced by a
`transform.structured.match` operation.
This revision significantly simplifies the specification and implementation of mapping loops to GPU ids.
Each type of mapping (block, warpgroup, warp, thread) now comes with 2 mapping modes:
1. a 3-D "grid-like" mode, subject to alignment considerations on threadIdx.x, on which predication
may occur on a per-dimension 3-D sub-rectangle basis.
2. a n-D linearized mode, on which predication may only occur on a linear basis.
In the process, better size and alignment requirement inference are introduced along with improved runtime verification messages.
The `warp_dims` attribute was deemed confusing and is removed from the transform in favor of better size inference.
Differential Revision: https://reviews.llvm.org/D155941
