[Downstream Change] sincos() vectorization
#87
Comments
This was referenced Feb 11, 2025
|
Can you also provide links to the llvm-project main branch patches? Have these landed or are they in review? |
Added the upstream PR links (above). The upstream patches have not landed. |
dcandler
pushed a commit
that referenced
this issue
Feb 24, 2025
…izing literal struct return values (#82) This patch adds initial support for vectorizing literal struct return values. Currently, this is limited to the case where the struct is homogeneous (all elements have the same type) and not packed. The users of the call also must all be `extractvalue` instructions. The intended use case for this is vectorizing intrinsics such as: ``` declare { float, float } @llvm.sincos.f32(float %x) ``` Mapping them to structure-returning library calls such as: ``` declare { <4 x float>, <4 x i32> } @Sleef_sincosf4_u10advsimd(<4 x float>) ``` Or their widened form (such as `@llvm.sincos.v4f32` in this case). Implementing this required two main changes: 1. Supporting widening `extractvalue` 2. Adding support for vectorized struct types in LV * This is mostly limited to parts of the cost model and scalarization Since the supported use case is narrow, the required changes are relatively small. --- Downstream issue: #87
dcandler
pushed a commit
that referenced
this issue
Feb 24, 2025
…l] calls to llvm.sincos.* when -fno-math-errno is set (#83) This will allow vectorizing these calls (after a few more patches). This should not change the codegen for targets that enable the use of AA during the codegen (in `TargetSubtargetInfo::useAA()`). This includes targets such as AArch64. This notably does not include x86 but can be worked around by passing `-mllvm -combiner-global-alias-analysis=true` to clang. --- Downstream issue: #87
|
Update: All patches have landed upstream (so should be available in the 21.x release without any downstream changes). |
kiranchandramohan
pushed a commit
that referenced
this issue
Feb 28, 2025
…nd vectorize llvm.sincos intrinsics (#84) This teaches the loop vectorizer that `llvm.sincos` is trivially vectorizable. Additionally, this patch updates the cost model to cost intrinsics that return multiple values correctly. Previously, the cost model only thought intrinsics that return `VectorType` need scalarizing, which meant it cost intrinsics that return multiple vectors (that need scalarizing) way too cheap (giving it the cost of a single function call). The `llvm.sincos` intrinsic also has a custom cost when a vector function library is available, as certain VFs can be expanded (later in code-gen) to a vector function, reducing the cost to a single call (+ the possible loads from the vector function returns values via output pointers). --- Downstream issue: #87
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
This issue tracks the downstream changes needed to enable
sincos()vectorization in the Arm toolchain. These changes are only for the release 20.x branch (https://github.com/arm/arm-toolchain/tree/release/arm-software/20.x) as we intend to complete up-streaming the changes for LLVM 21.llvm.sincosllvm/llvm-project#123210sincos()becomes vectorizable with-fveclib=ArmPL/SLEEFThe patches were not possible to upstream for LLVM 20 due to long review times.
The text was updated successfully, but these errors were encountered: