QRAM template #8368

isaacdevlugt · 2025-10-01T18:42:00Z

Context: QRAM implementation based on https://arxiv.org/pdf/2502.06767

Description of the Change: Adds QRAMtemplate, tests, and documentation

Benefits:

Possible Drawbacks:

Related GitHub Issues:

github-actions · 2025-10-01T18:42:20Z

Hello. You may have forgotten to update the changelog!
Please edit doc/releases/changelog-dev.md with:

A one-to-two sentence description of the change. You may include a small working example for new features.
A link back to this PR.
Your name (or GitHub username) in the contributors section.

isaacdevlugt · 2025-10-01T18:52:00Z

doc/introduction/templates.rst


+.. gallery-item::
+    :description: :doc:`QRAM <../code/api/pennylane.QRAM>`
+    :figure: _static/templates/TODO


Curious how this image will be provided? Where will it come from?

I don't think it needs to be more complicated than this^ 🤔

Any reason we say QROM here rather than QRAM?

Just using this image for now...

Ok I have permission from Josh and Tarik to take a shot at creating the QRAM template in inkscape. I will make one that closely matches the above.

isaacdevlugt · 2025-10-01T18:56:32Z

pennylane/templates/subroutines/qrom.py

        work_wires (Sequence[int]): the auxiliary wires used for the computation
        clean (bool): if True, the work wires are not altered by operator, default is ``True``

+    ..see-also:: :class:`~.QRAM`, :class:`~.QROMStatePreparation`


TODO: update name of QRAM class if needed

comp-phys-marc

Some really great work here. My initial thoughts / questions!

comp-phys-marc · 2025-11-14T18:49:28Z

pennylane/templates/subroutines/qramv1.py

+                for p in range(1 << k):
+                    #change to  in_wire later
+                    parent = _node_index(k - 1, p >> 1)
+                    if p % 2 == 0:
+                        ops.append(qml.SWAP(wires =[self.portL_wires[parent], self._router(k, p)]))
+                    else:
+                        ops.append(qml.SWAP(wires =[self.portR_wires[parent], self._router(k, p)]))


Is there a bit of duplication of logic between here and _route_bus_down_first_k_levels?

comp-phys-marc · 2025-11-14T18:50:31Z

pennylane/templates/subroutines/qramv1.py

+                op0 = qml.CSWAP(wires=[d, in_w, R])
+                ops.append(op0)
+                # dir==0 ⇒ SWAP(in, L)
+                op = qml.SWAP(wires=[in_w, L])
+                ops.append(qml.ctrl(op, control=[d], control_values=[0]))


I am not sure why we sometimes use CSWAP and sometimes use ctrl(SWAP, ...)?

If there's truly no difference we can easily replace with whatever is most efficient :)

There will be no material difference. Just a matter of picking something and sticking with it.

I think the reason we use ctrl here is because it takes control_values, right?

comp-phys-marc · 2025-11-14T18:51:13Z

pennylane/templates/subroutines/qramv1.py

+    # def _route_bus_down(self) -> list:
+    #     """Route the bus from root to leaf across all levels using dir-controlled CSWAPs."""
+    #     ops = []
+    #     for k in range(self.n_k):
+    #         for p in range(1 << k):
+    #             in_w = self._node_in_wire(k, p)
+    #             L = self._portL(k, p)
+    #             R = self._portR(k, p)
+    #             d = self._router(k, p)
+    #             if k == 0:
+    #                 upper_ctrls, upper_vals = [], []
+    #             else:
+    #                 upper_ctrls = [self._router(j, p >> (k - j)) for j in range(k)]
+    #                 upper_vals = [(p >> (k - 1 - j)) & 1 for j in range(k)]
+    #             op0 = qml.SWAP(wires=[in_w, L])
+    #             op1 = qml.SWAP(wires=[in_w, R])
+    #             ops.append(
+    #                 qml.ctrl(op0, control=[d] + upper_ctrls, control_values=[0] + upper_vals)
+    #                 if upper_ctrls
+    #                 else qml.ctrl(op0, control=[d], control_values=[0])
+    #             )
+    #             ops.append(
+    #                 qml.ctrl(op1, control=[d] + upper_ctrls, control_values=[1] + upper_vals)
+    #                 if upper_ctrls
+    #                 else qml.ctrl(op1, control=[d], control_values=[1])
+    #             )
+    #     return ops
+
+    # def _route_bus_up(self) -> list:
+    #     """Inverse of `_route_bus_down`."""
+    #     return list(reversed(self._route_bus_down()))
+
+    # # ---------- Select controls----------
+    # def _select_ctrls(self, s: int):
+    #     if self.k == 0:
+    #         return [], []
+    #     ctrls = list(self.select_wires)
+    #     vals = [(s >> (self.k - 1 - j)) & 1 for j in range(self.k)]
+    #     return ctrls, vals


What is the story / plan with all this commented out code?

comp-phys-marc · 2025-11-14T18:52:44Z

pennylane/templates/subroutines/qramv1.py

+class SelectBucketBrigadeBusQRAM(qml.operation.Operation):
+    r"""Bucket-brigade QRAM with **explicit bus routing** using 3 qubits per node,
+    and an optional **select (MSB) prefix**, plus **hybrid** support.


The one part of Bucket Brigade QRAM I am still unclear about is how once we have traversed the tree and have an active path from the root to a leaf, we actually translate this to a loading of the corresponding data into the data register?

The BB QRAM undergoes three steps to query the data, including address loading, data retrieval, and address unloading. Address loading can be understood as a unitary operation that implements a transformation between binary encoding and unary encoding. Data retrieval, where the loading of the corresponding data happens, directly applies the data for each entry in the unary encoding to the corresponding location. Since unary encoding every entry in the address has seperated qubits, such loading won't affect each other. Finally the unloading of the address ``compress'' the encoding back into binary and finish the query.

pennylane/templates/subroutines/qramv1.py

comp-phys-marc · 2025-11-18T14:38:12Z

pennylane/templates/subroutines/qramv1.py

+        portL_wires: Sequence[int],
+        portR_wires: Sequence[int],
+        *,
+        mode: str = "quantum",


What is the benefit of hybrid mode? Is it merely a trade-off of qubit count versus speedup?

What is the advantage of being able to do an access in superposition?

Basically yes, it is trade off between qubit and time

comp-phys-marc · 2025-11-18T14:50:20Z

pennylane/templates/subroutines/qramv1.py

+    )
+
+# -----------------------------
+# TODOs / Extensions


Are these TODOs something we can help with? How would we like to plan out and distribute this work?

comp-phys-marc · 2025-11-18T15:06:08Z

I also just want to invite @albi3ro to share any of the thoughts she had on refactoring / structuring this code. :)

comp-phys-marc · 2025-11-18T15:44:03Z

pennylane/templates/subroutines/qramv1.py

+            else:
+                target = self._portR(self.n_k - 1, p >> 1)
+            bit = self.bitstrings[p][j] 
+            if bit == "1":


How will the leaves' ops need to be handled once we actually have this template running on a quantum backend? Will there need to be a change?

comp-phys-marc · 2025-11-18T15:58:00Z

Is there value in making this capture compatible?

- rename link to template - use QROM image thumbnail

comp-phys-marc · 2025-11-19T21:39:07Z

pennylane/templates/subroutines/qramv1.py

+            ops.append(SWAP(wires=[tw, self.bus_wire[0]]))
+        return ops
+
+    # not work yet


I am curious what about this decomposition particularly doesn't work yet?

comp-phys-marc · 2025-11-19T21:39:30Z

pennylane/templates/subroutines/qramv1.py

+        ops += self._unmark_routers_via_bus()
+        return ops
+
+    # Not work yet


Curious also about what in particular doesn't work about this decomposition yet.

open PR

8c67a4d

templates pageg

4ca4032

isaacdevlugt commented Oct 1, 2025

View reviewed changes

references

9607fbd

isaacdevlugt changed the title ~~open PR~~ QRAM template Oct 1, 2025

isaacdevlugt commented Oct 1, 2025

View reviewed changes

ShifanXu and others added 3 commits October 21, 2025 21:03

Add Bucket-Brigade QRAM implementation and test case

3c2c17d

[WIP] dev comments

2a09fcd

[WIP] dev comment

6f3ce3f

comp-phys-marc reviewed Nov 14, 2025

View reviewed changes

comp-phys-marc reviewed Nov 18, 2025

View reviewed changes

comp-phys-marc added 6 commits November 19, 2025 08:34

Merge remote-tracking branch 'origin' into qram-feature-yale

31c5e9f

Merge branch 'master' into qram-feature-yale

570b9e8

pylint

2f04ddc

Imports

976d9f6

TODOs

d555d7e

- rename link to template - use QROM image thumbnail

Thumbnail

6210141

comp-phys-marc reviewed Nov 19, 2025

View reviewed changes

Merge branch 'master' into qram-feature-yale

b45553e

QRAM template #8368

Are you sure you want to change the base?

QRAM template #8368

Uh oh!

Conversation

isaacdevlugt commented Oct 1, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Oct 1, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

comp-phys-marc Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

isaacdevlugt Oct 1, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

comp-phys-marc left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

comp-phys-marc commented Nov 18, 2025

Uh oh!

comp-phys-marc Nov 18, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

comp-phys-marc commented Nov 18, 2025

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

isaacdevlugt commented Oct 1, 2025 •

edited

Loading

comp-phys-marc Nov 18, 2025 •

edited

Loading

isaacdevlugt Oct 1, 2025 •

edited

Loading

comp-phys-marc Nov 18, 2025 •

edited

Loading