Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Greedy topological sort of the binst graph to minimize qubit allocations / deallocations #1099

Merged
merged 9 commits into from
Jul 6, 2024

Conversation

tanujkhattar
Copy link
Collaborator

This PR adds a new stable greedy topological sorting method and changes the default topological sort to use this method when iterating over nodes of a bloq instance graph.

This is useful to

  • Improve qubit cost analysis
  • Improve cirq-interop and minimize qubit allocations when specifying a GreedyQubitManager.

Fixes the issue mentioned in #963 (comment), would be useful for #1097 and can be improved by addressing #1098

if TYPE_CHECKING:
from qualtran import BloqInstance

_INFINITY: int = int(1e18)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

float('inf')?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Eventually, I want to set _INFINITY + binst.i as the priority instead of _INFINITY; assuming we can make binst.i preserve insertion order following #1098

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In that case, a name other than "infinity" should be used if we're going to literally do "infinity + 1" as a meaningful distinction :p

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

renamed

goal to minimize qubit allocations and deallocations by pushing allocations to the
right and de-allocations to the left.
"""
yield from nx.lexicographical_topological_sort(binst_graph, key=_priority)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this function has slimmed down considerably! Thanks @anurudhp

@@ -65,20 +65,20 @@ def test_qpic_data_for_reflect_using_prepare():
reg_2 / \textrm{\scalebox{0.5}{QAny(2)}}
reg_3:on G:width=25:shape=box \textrm{\scalebox{0.8}{alloc}}
reg_3 / \textrm{\scalebox{0.5}{QAny(1)}}
selection G:width=121:shape=box \textrm{\scalebox{0.8}{StatePreparationAliasSampling}} reg G:width=37:shape=box \textrm{\scalebox{0.8}{sigma\_mu}} reg_1 G:width=17:shape=box \textrm{\scalebox{0.8}{alt}} reg_2 G:width=21:shape=box \textrm{\scalebox{0.8}{keep}} reg_3 G:width=65:shape=box \textrm{\scalebox{0.8}{less\_than\_equal}}
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

should we scrap this test? I guess it sortof forces you to look at the circuit so you can paste in the new data but it's not very DAMP https://testing.googleblog.com/2019/12/testing-on-toilet-tests-too-dry-make.html

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

removed

Copy link
Collaborator

@mpharrigan mpharrigan left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

please consider changing the documentation & name for "infinity" if we're going to be doing meaningful arithmetic on it

@tanujkhattar tanujkhattar enabled auto-merge (squash) July 6, 2024 00:53
@tanujkhattar tanujkhattar merged commit 19928a3 into quantumlib:main Jul 6, 2024
7 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants