projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 88de8b2) | patch
LLVM dialect: introduce fmuladd intrinsic as operation
authorAlex Zinenko <zinenko@google.com>
Tue, 13 Aug 2019 10:40:20 +0000 (03:40 -0700)
committerA. Unique TensorFlower <gardener@tensorflow.org>
Tue, 13 Aug 2019 10:40:51 +0000 (03:40 -0700)
commit5f0a843144289050844109f16693c86168135dac
treefea50d7cf774933a8d32246fe2713c60f2c96db4tree | snapshot
parent88de8b2a2bb09e4e3ebb829949e0256c3e86c609commit | diff
LLVM dialect: introduce fmuladd intrinsic as operation

This operation is important to achieve decent performance in computational
kernels.  In LLVM, it is implemented as an intrinsic (through function
declaration and function call).  Thanks to MLIR's extendable set of operations,
it does not have to differentiate between built-ins and intrinsics, so fmuladd
is introduced as a general type-polymorphic operation.  Custom printing and
parsing will be added later.

PiperOrigin-RevId: 263106305
mlir/include/mlir/LLVMIR/LLVMOps.td diff | blob | history
mlir/test/Target/llvmir-intrinsics.mlir [new file with mode: 0644] blob
Domain: System / Toolchain;