Assemble circuits and transpiler peakmem benchmarks#641
Conversation
kdk
force-pushed
the
assemble_circuits-and-transpiler-peakmem-benchmarks
branch
from
4a70a89 to
8e65e26
Compare
kdk
force-pushed
the
assemble_circuits-and-transpiler-peakmem-benchmarks
branch
from
8e65e26 to
b14d240
Compare
mtreinish
left a comment
mtreinish
left a comment
There was a problem hiding this comment.
LGTM, a couple inline questions but nothing blocking
|
|
||
|
|
||
| class AssemblerBenchmarks: | ||
| params = ([1, 2, 5, 8], |
There was a problem hiding this comment.
Do we want to expand the number of qubits here at all. Just to get a feel for how this scales as we approach the largest sizes?
There was a problem hiding this comment.
Yes and no. There are performance issues in assemble for larger n_qubits (e.g. we do some List.indexs that should be dictionary look ups) but right now, far and away, performance scales with the number of gates. I limited n_qubits in the interest of keeping the benchmark runtime reasonable, but we can always expand it.
(As an aside, since we generate random circuits by depth and not by gate count, increasing n_qubits does dramatically increase the benchmark runtime but only because the test circuits contain more gates, so there's not a lot of interesting information there, at least right now. Maybe we should update random_circuit to accept either depth or gate_count.)
There was a problem hiding this comment.
That's a good point, I figured it was for run time. Yeah most of the time is spent looping over the instructions in experiments.
If we wanted to add a gate_count parameter it'd be simple to do (my first implementation of random circuit which we didn't use in terra actually did it this way). We should do this before we adopt any benchmarks with it though because we'd have to regenerate any data.
Co-Authored-By: Matthew Treinish <[email protected]>
|
Ci Passed, not that we actually run benchmarks in CI (still waiting on the new asv release with the |
…tapackage#641) Adds a peakmem tracking for transpiler_level benchmarks, and a benchmark for assemble_circuits. * Add peakmem versions of transpiler level benchmarks. * Add assemble_circuit benchmark. * Import assemble from qiskit.compile.
…tapackage#641) Adds a peakmem tracking for transpiler_level benchmarks, and a benchmark for assemble_circuits. * Add peakmem versions of transpiler level benchmarks. * Add assemble_circuit benchmark. * Import assemble from qiskit.compile.
2 participants
Adds a peakmem tracking for transpiler_level benchmarks, and a benchmark for assemble_circuits.