Add intrinsic for launch-sized workgroup memory on GPUs
Workgroup memory is a memory region that is shared between all
threads in a workgroup on GPUs. Workgroup memory can be allocated
statically or after compilation, when launching a gpu-kernel.
The intrinsic added here returns the pointer to the memory that is
allocated at launch-time.
# Interface
With this change, workgroup memory can be accessed in Rust by
calling the new `gpu_launch_sized_workgroup_mem<T>() -> *mut T`
intrinsic.
It returns the pointer to workgroup memory guaranteeing that it is
aligned to at least the alignment of `T`.
The pointer is dereferencable for the size specified when launching the
current gpu-kernel (which may be the size of `T` but can also be larger
or smaller or zero).
All calls to this intrinsic return a pointer to the same address.
See the intrinsic documentation for more details.
## Alternative Interfaces
It was also considered to expose dynamic workgroup memory as extern
static variables in Rust, like they are represented in LLVM IR.
However, due to the pointer not being guaranteed to be dereferencable
(that depends on the allocated size at runtime), such a global must be
zero-sized, which makes global variables a bad fit.
# Implementation Details
Workgroup memory in amdgpu and nvptx lives in address space 3.
Workgroup memory from a launch is implemented by creating an
external global variable in address space 3. The global is declared with
size 0, as the actual size is only known at runtime. It is defined
behavior in LLVM to access an external global outside the defined size.
There is no similar way to get the allocated size of launch-sized
workgroup memory on amdgpu an nvptx, so users have to pass this
out-of-band or rely on target specific ways for now.
Tracking issue: rust-lang/rust#135516
Workgroup memory is a memory region that is shared between all
threads in a workgroup on GPUs. Workgroup memory can be allocated
statically or after compilation, when launching a gpu-kernel.
The intrinsic added here returns the pointer to the memory that is
allocated at launch-time.
# Interface
With this change, workgroup memory can be accessed in Rust by
calling the new `gpu_launch_sized_workgroup_mem<T>() -> *mut T`
intrinsic.
It returns the pointer to workgroup memory guaranteeing that it is
aligned to at least the alignment of `T`.
The pointer is dereferencable for the size specified when launching the
current gpu-kernel (which may be the size of `T` but can also be larger
or smaller or zero).
All calls to this intrinsic return a pointer to the same address.
See the intrinsic documentation for more details.
## Alternative Interfaces
It was also considered to expose dynamic workgroup memory as extern
static variables in Rust, like they are represented in LLVM IR.
However, due to the pointer not being guaranteed to be dereferencable
(that depends on the allocated size at runtime), such a global must be
zero-sized, which makes global variables a bad fit.
# Implementation Details
Workgroup memory in amdgpu and nvptx lives in address space 3.
Workgroup memory from a launch is implemented by creating an
external global variable in address space 3. The global is declared with
size 0, as the actual size is only known at runtime. It is defined
behavior in LLVM to access an external global outside the defined size.
There is no similar way to get the allocated size of launch-sized
workgroup memory on amdgpu an nvptx, so users have to pass this
out-of-band or rely on target specific ways for now.
Abstract over the existing `simd_cast` intrinsic to implement a new
`sve_cast` intrinsic - this is better than allowing scalable vectors to
be used with all of the generic `simd_*` intrinsics.
Clang changed to representing tuples of scalable vectors as
structs rather than as wide vectors (that is, scalable vector types
where the `N` part of the `<vscale x N x ty>` type was multiplied by
the number of vectors). rustc mirrored this in the initial implementation
of scalable vectors.
Earlier versions of our patches used the wide vector representation and
our intrinsic patches used the legacy
`llvm.aarch64.sve.tuple.{create,get,set}{2,3,4}` intrinsics for creating
these tuples/getting/setting the vectors, which were only supported
due to LLVM's `AutoUpgrade` pass converting these intrinsics into
`llvm.vector.insert`. `AutoUpgrade` only supports these legacy intrinsics
with the wide vector representation.
With the current struct representation, Clang has special handling in
codegen for generating `insertvalue`/`extractvalue` instructions for
these operations, which must be replicated by rustc's codegen for our
intrinsics to use. This patch implements new intrinsics in
`core::intrinsics::scalable` (mirroring the structure of
`core::intrinsics::simd`) which rustc lowers to the appropriate
`insertvalue`/`extractvalue` instructions.
simd_fmin/fmax: make semantics and name consistent with scalar intrinsics
This is the SIMD version of https://github.com/rust-lang/rust/pull/153343: change the documented semantics of the SIMD float min/max intrinsics to that of the scalar intrinsics, and also make the name consistent. The overall semantic change this amounts to is that we restrict the non-determinism: the old semantics effectively mean "when one input is an SNaN, the result non-deterministically is a NaN or the other input"; the new semantics say that in this case the other input must be returned. For all other cases, old and new semantics are equivalent. This means all users of these intrinsics that were correct with the old semantics are still correct: the overall set of possible behaviors has become smaller, no new possible behaviors are being added.
In terms of providers of this API:
- Miri, GCC, and cranelift already implement the new semantics, so no changes are needed.
- LLVM is adjusted to use `minimumnum nsz` instead of `minnum`, thus giving us the new semantics.
In terms of consumers of this API:
- Portable SIMD almost certainly wants to match the scalar behavior, so this is strictly a bugfix here.
- Stdarch mostly stopped using the intrinsic, except on nvptx, where arguably the new semantics are closer to what we actually want than the old semantics (https://github.com/rust-lang/stdarch/issues/2056).
Q: Should there be an `f` in the intrinsic name to indicate that it is for floats? E.g., `simd_fminimum_number_nsz`?
Also see https://github.com/rust-lang/rust/issues/153395.
add `simd_splat` intrinsic
Add `simd_splat` which lowers to the LLVM canonical splat sequence.
```llvm
insertelement <N x elem> poison, elem %x, i32 0
shufflevector <N x elem> v0, <N x elem> poison, <N x i32> zeroinitializer
```
Right now we try to fake it using one of
```rust
fn splat(x: u32) -> u32x8 {
u32x8::from_array([x; 8])
}
```
or (in `stdarch`)
```rust
fn splat(value: $elem_type) -> $name {
#[derive(Copy, Clone)]
#[repr(simd)]
struct JustOne([$elem_type; 1]);
let one = JustOne([value]);
// SAFETY: 0 is always in-bounds because we're shuffling
// a simd type with exactly one element.
unsafe { simd_shuffle!(one, one, [0; $len]) }
}
```
Both of these can confuse the LLVM optimizer, producing sub-par code. Some examples:
- https://github.com/rust-lang/rust/issues/60637
- https://github.com/rust-lang/rust/issues/137407
- https://github.com/rust-lang/rust/issues/122623
- https://github.com/rust-lang/rust/issues/97804
---
As far as I can tell there is no way to provide a fallback implementation for this intrinsic, because there is no `const` way of evaluating the number of elements (there might be issues beyond that, too). So, I added implementations for all 4 backends.
Both GCC and const-eval appear to have some issues with simd vectors containing pointers. I have a workaround for GCC, but haven't yet been able to make const-eval work. See the comments below.
Currently this just adds the intrinsic, it does not actually use it anywhere yet.
Add amdgpu_dispatch_ptr intrinsic
There is an ongoing discussion in rust-lang/rust#150452 about using address spaces from the Rust language in some way.
As that discussion will likely not conclude soon, this PR adds one rustc_intrinsic with an addrspacecast to unblock getting basic information like launch and workgroup size and make it possible to implement something like `core::gpu`.
Add a rustc intrinsic `amdgpu_dispatch_ptr` to access the kernel dispatch packet on amdgpu.
The HSA kernel dispatch packet contains important information like the launch size and workgroup size.
The Rust intrinsic lowers to the `llvm.amdgcn.dispatch.ptr` LLVM intrinsic, which returns a `ptr addrspace(4)`, plus an addrspacecast to `addrspace(0)`, so it can be returned as a Rust reference.
The returned pointer/reference is valid for the whole program lifetime, and is therefore `'static`.
The return type of the intrinsic (`&'static ()`) does not mention the struct so that rustc does not need to know the exact struct type. An alternative would be to define the struct as lang item or add a generic argument to the function.
Is this ok or is there a better way (also, should it return a pointer instead of a reference)?
Short version:
```rust
#[cfg(target_arch = "amdgpu")]
pub fn amdgpu_dispatch_ptr() -> *const ();
```
Tracking issue: rust-lang/rust#135024
Add a rustc intrinsic `amdgpu_dispatch_ptr` to access the kernel
dispatch packet on amdgpu.
The HSA kernel dispatch packet contains important information like the
launch size and workgroup size.
The Rust intrinsic lowers to the `llvm.amdgcn.dispatch.ptr` LLVM
intrinsic, which returns a `ptr addrspace(4)`, plus an addrspacecast to
`addrspace(0)`, so it can be returned as a Rust reference.
The returned pointer/reference is valid for the whole program lifetime,
and is therefore `'static`.
The return type of the intrinsic (`*const ()`) does not mention the
struct so that rustc does not need to know the exact struct type.
An alternative would be to define the struct as lang item or add a
generic argument to the function.
Short version:
```rust
#[cfg(target_arch = "amdgpu")]
pub fn amdgpu_dispatch_ptr() -> *const ();
```
Add `overflow_checks` intrinsic
This adds an intrinsic which allows code in a pre-built library to inherit the overflow checks option from a crate depending on it. This enables code in the standard library to explicitly change behavior based on whether `overflow_checks` are enabled, regardless of the setting used when standard library was compiled.
This is very similar to the `ub_checks` intrinsic, and refactors the two to use a common mechanism.
The primary use case for this is to allow the new `RangeFrom` iterator to yield the maximum element before overflowing, as requested [here](https://github.com/rust-lang/rust/issues/125687#issuecomment-2151118208). This PR includes a working `IterRangeFrom` implementation based on this new intrinsic that exhibits the desired behavior.
[Prior discussion on Zulip](https://rust-lang.zulipchat.com/#narrow/stream/219381-t-libs/topic/Ability.20to.20select.20code.20based.20on.20.60overflow_checks.60.3F)
The contract_checks compiler flag is now used to determine
if runtime contract checks should be enabled, as opposed
to the compiler intrinsic as previously.
Refactor contract HIR lowering to ensure no contract code is
executed when contract-checks are disabled.
The call to contract_checks is moved to inside the lowered fn
body, and contract closures are built conditionally, ensuring
no side-effects present in contracts occur when those are disabled.
Jonathan Brouwer