Switch from shfl.bfly to shfl.down.
authorChristian Sigg <csigg@google.com>
Thu, 12 Dec 2019 09:27:27 +0000 (01:27 -0800)
committerA. Unique TensorFlower <gardener@tensorflow.org>
Thu, 12 Dec 2019 09:28:01 +0000 (01:28 -0800)
commitf68ac464d818629e0fe10c23b44ac782d64a12d2
treecbfb66a4d41d92a13554d8df1505e392810bf004
parent851a8516d3883088b0c02601f9c1bb269e85051c
Switch from shfl.bfly to shfl.down.

Both work for the current use case, but the latter allows implementing
prefix sums and is a little easier to understand for partial warps.

PiperOrigin-RevId: 285145287
mlir/include/mlir/Dialect/LLVMIR/NVVMOps.td
mlir/lib/Conversion/GPUToNVVM/LowerGpuOpsToNVVMOps.cpp
mlir/lib/Dialect/LLVMIR/IR/NVVMDialect.cpp
mlir/lib/Target/LLVMIR/ConvertToNVVMIR.cpp
mlir/test/Conversion/GPUToNVVM/gpu-to-nvvm.mlir
mlir/test/Dialect/LLVMIR/invalid.mlir
mlir/test/Dialect/LLVMIR/nvvm.mlir
mlir/test/Target/nvvmir.mlir