| commit | 77f256070f985029a53da5e76ecd85ca7686a7ea | [log] [tgz] |
|---|---|---|
| author | Fabian Mora <fmora.dev@gmail.com> | Mon Sep 01 07:16:17 2025 -0400 |
| committer | GitHub <noreply@github.com> | Mon Sep 01 11:16:17 2025 +0000 |
| tree | 593a20c0fa6e3140bb52af0f7515e32e5d783cbc | |
| parent | a3dfedf09e017b00cb3cf5235c24cc57b96eb5b6 [diff] |
[mlir][ptr] Add load and store ops. (#156093)
This patch adds the load and store operations to the ptr dialect. It's
future work to implement SROA and Mem2Reg interfaces, as well as
conversion to LLVM, and add alias information.
This patch also fixes a bug in `OptionalProp` that was causing the
bytecode writer to exit early of writing the Op props if an optional
prop had the default value.
Example:
```mlir
func.func @load_ops(%arg0: !ptr.ptr<#ptr.generic_space>) -> (f32, f32, f32, f32, f32, i64, i32) {
%0 = ptr.load %arg0 : !ptr.ptr<#ptr.generic_space> -> f32
%1 = ptr.load volatile %arg0 : !ptr.ptr<#ptr.generic_space> -> f32
%2 = ptr.load %arg0 nontemporal : !ptr.ptr<#ptr.generic_space> -> f32
%3 = ptr.load %arg0 invariant : !ptr.ptr<#ptr.generic_space> -> f32
%4 = ptr.load %arg0 invariant_group : !ptr.ptr<#ptr.generic_space> -> f32
%5 = ptr.load %arg0 atomic monotonic alignment = 8 : !ptr.ptr<#ptr.generic_space> -> i64
%6 = ptr.load volatile %arg0 atomic syncscope("workgroup") acquire nontemporal alignment = 4 : !ptr.ptr<#ptr.generic_space> -> i32
return %0, %1, %2, %3, %4, %5, %6 : f32, f32, f32, f32, f32, i64, i32
}
func.func @store_ops(%arg0: !ptr.ptr<#ptr.generic_space>, %arg1: f32, %arg2: i64, %arg3: i32) {
ptr.store %arg1, %arg0 : f32, !ptr.ptr<#ptr.generic_space>
ptr.store volatile %arg1, %arg0 : f32, !ptr.ptr<#ptr.generic_space>
ptr.store %arg1, %arg0 nontemporal : f32, !ptr.ptr<#ptr.generic_space>
ptr.store %arg1, %arg0 invariant_group : f32, !ptr.ptr<#ptr.generic_space>
ptr.store %arg2, %arg0 atomic monotonic alignment = 8 : i64, !ptr.ptr<#ptr.generic_space>
ptr.store volatile %arg3, %arg0 atomic syncscope("workgroup") release nontemporal alignment = 4 : i32, !ptr.ptr<#ptr.generic_space>
return
}
```
Finally, this patch allows testing more advanced features of ptr memory
spaces, for example:
```mlir
// mlir-opt -verify-diagnostics
func.func @store_const(%arg0: !ptr.ptr<#test.const_memory_space>, %arg1: i64) {
// expected-error@+1 {{memory space is read-only}}
ptr.store %arg1, %arg0 atomic monotonic alignment = 8 : i64, !ptr.ptr<#test.const_memory_space>
return
}
```
Welcome to the LLVM project!
This repository contains the source code for LLVM, a toolkit for the construction of highly optimized compilers, optimizers, and run-time environments.
The LLVM project has multiple components. The core of the project is itself called “LLVM”. This contains all of the tools, libraries, and header files needed to process intermediate representations and convert them into object files. Tools include an assembler, disassembler, bitcode analyzer, and bitcode optimizer.
C-like languages use the Clang frontend. This component compiles C, C++, Objective-C, and Objective-C++ code into LLVM bitcode -- and from there into object files, using LLVM.
Other components include: the libc++ C++ standard library, the LLD linker, and more.
Consult the Getting Started with LLVM page for information on building and running LLVM.
For information on how to contribute to the LLVM project, please take a look at the Contributing to LLVM guide.
Join the LLVM Discourse forums, Discord chat, LLVM Office Hours or Regular sync-ups.
The LLVM project has adopted a code of conduct for participants to all modes of communication within the project.