Skip to content

ToeplitzPSF — fwd (cuda, torch.complex64, inner, default)

← Back to compiled benchmarks

Result

Field Value
Operator ToeplitzPSF
Category linop
Method fwd
Device cuda
Dtype torch.complex64
Compile level inner
Compile config default
Status ok
Class compiles_slower
Graph breaks 0
Recompiles —
Eager time 16.24 us
Compiled time 32.78 us
Speedup 0.50x
Max abs diff 0.0
Max rel diff 0.0
Degraded False

Tensors

Tensor Shape Dtype Device Numel Bytes View Contiguous Stride
input [2, 20, 20] torch.complex64 cuda:0 800 6.25 KB False True [400, 20, 1]
output [2, 20, 20] torch.complex64 cuda:0 800 6.25 KB False True [400, 20, 1]

Notes

—