Introducing any_qpu type to hold both qpus based on the QPU base class and trait based qpus. #3685

Renaud-K · 2025-12-16T00:42:20Z

In preparation for the runtime refactor and the transition of QPU based classes (qpu.h) to trait based classes. I propose the use of the any_qpu type in the quantum platform, which call hold both.

Introducing a new directory where the new files are added. The idea is to make it easier to remove the old files once the migration is done.
The quantum_platforms holds the any_qpu type, instead of the QPU type. It turns out, that this was the only place where the change has to happen.
A non LLVM based type-erased registry is introduced to hold both types. The LLVM based registry only supports type which have a common base, however trait based class have no common base.
- The registry is made to work with more than QPUs, so it is split in 2 files: type_erased_registry.h and any_qpu.h
- The interface to the registry mimics that of the current LLVM based registry.
In the case of overloaded members, all overloads need to be present in a class for the look-up mechanism to work.
For this registry to work across executables, shared libs and dynamically loaded libs, the trick it to explicitly instantiate only once the get function of the registry, in this aggregate, to insure that all binaries call a unique one. This is also the reason why there is LLVM_INSTANTIATE_REGISTRY.
There is an emerging task for the refactor to allow the use of the logger in the header files in a way that the fmtlib package does not bleed into user code. For now, I have split it into Logger.h and logger_lite.h, so basic logging can be done. This was a major pain point to debug the code.

copy-pr-bot · 2025-12-16T00:42:26Z

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

python/runtime/utils/PyFermioniqRESTQPU.cpp

runtime/version2/cudaq/qpus/any_qpu.h

sacpis · 2025-12-16T01:28:33Z

runtime/cudaq/version2/cudaq/qpus/any_qpu.h

+    launchVQE = [](void *i, const std::string &name, const void *kernelArgs,
+                   gradient *grad, const spin_op &H, optimizer &opt,
+                   const int n_params, const std::size_t shots) {
+      static_cast<T *>(i)->launchVQE(name, kernelArgs, grad, H, opt, n_params,


I think VQE is going to be moved to CUDA-QX. @bmhowe23 would you please confirm?

runtime/version2/cudaq/qpus/any_qpu.h

runtime/version2/cudaq/utils/type_erased_registry.h

lmondada

I like this proposal a lot!

The only drawback is that the any_qpu definition is very verbose: each method appears three times:

as attribute in the qpu_dispatch_table,
in the build method
as a method on any_qpu

My C++ hackery skills are at a limit here. I don't see how this redundancy could be resolved with a macro. Any chance tablegen could help here? I am not familiar enough with what it can/cannot do.

Anyways, I think this is a good approach and I am in favour of merging this in!

Renaud-K · 2025-12-18T01:56:44Z

Thank you for the support. I have not had a chance to give too much thoughts about the drawbacks mentioned here. I am also not able to reply there. My focus on the past two days has been on making it work across executables, shared libs and dynamically loaded libs. The trick it to explicitly instantiate only once the get function of the registry, this aggregate to insure that all binaries call the same. This is the reason why there is LLVM_INSTANTIATE_REGISTRY.

Reflections would probably help best with this issue. Macros could help a little too I think. I am not sure I want to get too sophisticated this early in the process with tablegen.

Reducing the size of the interface will help too which should be where we are headed.

Renaud-K · 2025-12-18T19:51:00Z

/ok to test 32911ed

Command Bot: Processing...

Renaud-K · 2025-12-18T22:16:15Z

/ok to test cc96114

Command Bot: Processing...

Renaud-K · 2025-12-18T23:49:14Z

/ok to test f2785df

Command Bot: Processing...

Renaud-K · 2025-12-19T00:30:36Z

/ok to test d007f5f

Command Bot: Processing...

Renaud-K · 2025-12-19T17:15:39Z

/ok to test bb314c6

Command Bot: Processing...

Renaud-K · 2026-01-06T17:14:18Z

/ok to test 47771b1

Command Bot: Processing...

Signed-off-by: Renaud Kauffmann <[email protected]>

… singleton accross dynamically loaded libs Signed-off-by: Renaud Kauffmann <[email protected]>

Signed-off-by: Renaud Kauffmann <[email protected]>

…. Removed the extra registry instantiation Signed-off-by: Renaud Kauffmann <[email protected]>

DCO Remediation Commit for Renaud Kauffmann <[email protected]> I, Renaud Kauffmann <[email protected]>, hereby add my Signed-off-by to this commit: e16a515 Signed-off-by: Renaud Kauffmann <[email protected]>

DCO Remediation Commit for Renaud Kauffmann <[email protected]> I, Renaud Kauffmann <[email protected]>, hereby add my Signed-off-by to this commit: 5c8d9ff Signed-off-by: Renaud Kauffmann <[email protected]>

Signed-off-by: Renaud Kauffmann <[email protected]>

…cannot be controlled. Signed-off-by: Renaud Kauffmann <[email protected]>

Signed-off-by: Renaud Kauffmann <[email protected]>

Renaud-K · 2026-01-07T02:11:23Z

/ok to test 8ca28c5

Command Bot: Processing...

copy-pr-bot · 2026-01-07T02:11:27Z

/ok to test 8ca28c5

@Renaud-K, there was an error processing your request: E2

See the following link for more information: https://docs.gha-runners.nvidia.com/cpr/e/2/

Renaud-K · 2026-01-07T02:13:44Z

/ok to test 21ed030

Command Bot: Processing...

github-actions · 2026-01-07T04:06:39Z

CUDA Quantum Docs Bot: A preview of the documentation can be found here.

I, Renaud Kauffmann <[email protected]>, hereby add my Signed-off-by to this commit: 8ca28c5 Signed-off-by: Renaud Kauffmann <[email protected]>

Renaud-K · 2026-01-07T19:51:10Z

/ok to test a8f4f3e

Command Bot: Processing...

github-actions · 2026-01-07T21:44:54Z

CUDA Quantum Docs Bot: A preview of the documentation can be found here.

Renaud-K requested review from boschmitt, khalatepradnya, lmondada and sacpis December 16, 2025 00:42

sacpis reviewed Dec 16, 2025

View reviewed changes

python/runtime/utils/PyFermioniqRESTQPU.cpp Outdated Show resolved Hide resolved

sacpis reviewed Dec 16, 2025

View reviewed changes

runtime/version2/cudaq/qpus/any_qpu.h Outdated Show resolved Hide resolved

sacpis reviewed Dec 16, 2025

View reviewed changes

runtime/version2/cudaq/qpus/any_qpu.h Outdated Show resolved Hide resolved

sacpis reviewed Dec 16, 2025

View reviewed changes

runtime/version2/cudaq/qpus/any_qpu.h Outdated Show resolved Hide resolved

sacpis reviewed Dec 16, 2025

View reviewed changes

runtime/version2/cudaq/qpus/any_qpu.h Outdated Show resolved Hide resolved

sacpis reviewed Dec 16, 2025

View reviewed changes

runtime/version2/cudaq/utils/type_erased_registry.h Outdated Show resolved Hide resolved

Renaud-K force-pushed the any_qpu branch from 7b3cff8 to 9106e4d Compare December 16, 2025 20:23

Renaud-K requested review from 1tnguyen, bettinaheim, bmhowe23 and schweitzpgi as code owners December 17, 2025 00:14

lmondada reviewed Dec 17, 2025

View reviewed changes

Renaud-K force-pushed the any_qpu branch from 69c9dd6 to 6363b46 Compare December 18, 2025 04:33

Renaud-K changed the title ~~WIP: any_qpu type~~ Introducinf any_qpu type to hold both qpus based on the QPU base class and trait based qpus. Dec 18, 2025

Renaud-K changed the title ~~Introducinf any_qpu type to hold both qpus based on the QPU base class and trait based qpus.~~ Introducing any_qpu type to hold both qpus based on the QPU base class and trait based qpus. Dec 18, 2025

Renaud-K force-pushed the any_qpu branch 2 times, most recently from 17d1b0c to cc96114 Compare December 18, 2025 22:02

Renaud-K force-pushed the any_qpu branch from 4608100 to f2785df Compare December 18, 2025 23:48

Renaud-K added 18 commits January 6, 2026 18:09

Rebased branch

8ca28c5

Formatting

d40059e

Signed-off-by: Renaud Kauffmann <[email protected]>

Reapplying formatting

4b47146

Signed-off-by: Renaud Kauffmann <[email protected]>

Moving version 2 under cudaq

b1f1c62

Signed-off-by: Renaud Kauffmann <[email protected]>

Fixed spelling

9384bc2

Signed-off-by: Renaud Kauffmann <[email protected]>

Introduce CUDAQ_INSTANTIATE_TYPE_ERASED_REGISTRY to guaranty a unique…

f7d1511

… singleton accross dynamically loaded libs Signed-off-by: Renaud Kauffmann <[email protected]>

Formatting changes + Warning issued

454a3ee

Signed-off-by: Renaud Kauffmann <[email protected]>

Introduced the logger_lite, to still be able to log from header files…

947d0e9

…. Removed the extra registry instantiation Signed-off-by: Renaud Kauffmann <[email protected]>

Corrected spelling

b2a1bb9

DCO Remediation Commit for Renaud Kauffmann <[email protected]> I, Renaud Kauffmann <[email protected]>, hereby add my Signed-off-by to this commit: e16a515 Signed-off-by: Renaud Kauffmann <[email protected]>

Adding lit test

45b6466

DCO Remediation Commit for Renaud Kauffmann <[email protected]> I, Renaud Kauffmann <[email protected]>, hereby add my Signed-off-by to this commit: 5c8d9ff Signed-off-by: Renaud Kauffmann <[email protected]>

Fixed copyright headers

5934aa5

Signed-off-by: Renaud Kauffmann <[email protected]>

Fixing and simplify lit test

0fc0af1

Signed-off-by: Renaud Kauffmann <[email protected]>

Fixing lit test - Cannot predict order of destruction

2d5c08d

Signed-off-by: Renaud Kauffmann <[email protected]>

Fixing doxygen error

b99f53a

Signed-off-by: Renaud Kauffmann <[email protected]>

There are 3 qpus in the array, all of which the order of destruction …

902b949

…cannot be controlled. Signed-off-by: Renaud Kauffmann <[email protected]>

Correcting include

eede208

Signed-off-by: Renaud Kauffmann <[email protected]>

Clean-ups after conflict resolution

130bbf2

Signed-off-by: Renaud Kauffmann <[email protected]>

Formatting

bd5c2eb

Signed-off-by: Renaud Kauffmann <[email protected]>

Renaud-K force-pushed the any_qpu branch from 47771b1 to 21ed030 Compare January 7, 2026 02:10

github-actions bot pushed a commit that referenced this pull request Jan 7, 2026

Docs preview for PR #3685.

79703bf

DCO Remediation Commit for Renaud Kauffmann <[email protected]>

a8f4f3e

I, Renaud Kauffmann <[email protected]>, hereby add my Signed-off-by to this commit: 8ca28c5 Signed-off-by: Renaud Kauffmann <[email protected]>

Renaud-K force-pushed the any_qpu branch from 21ed030 to a8f4f3e Compare January 7, 2026 17:19

github-actions bot pushed a commit that referenced this pull request Jan 7, 2026

Docs preview for PR #3685.

2f502b0

Introducing any_qpu type to hold both qpus based on the QPU base class and trait based qpus. #3685

Are you sure you want to change the base?

Introducing any_qpu type to hold both qpus based on the QPU base class and trait based qpus. #3685

Uh oh!

Conversation

Renaud-K commented Dec 16, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

copy-pr-bot bot commented Dec 16, 2025

Uh oh!

Uh oh!

Uh oh!

sacpis Dec 16, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

lmondada left a comment

Choose a reason for hiding this comment

Uh oh!

Renaud-K commented Dec 18, 2025

Uh oh!

Renaud-K commented Dec 18, 2025 • edited by github-actions bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Renaud-K commented Dec 18, 2025 • edited by github-actions bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Renaud-K commented Dec 18, 2025 • edited by github-actions bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Renaud-K commented Dec 19, 2025 • edited by github-actions bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Renaud-K commented Dec 19, 2025 • edited by github-actions bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Renaud-K commented Jan 6, 2026 • edited by github-actions bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Renaud-K commented Jan 7, 2026 • edited by github-actions bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

copy-pr-bot bot commented Jan 7, 2026

Uh oh!

Renaud-K commented Jan 7, 2026 • edited by github-actions bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Jan 7, 2026

Uh oh!

Renaud-K commented Jan 7, 2026 • edited by github-actions bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Jan 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Renaud-K commented Dec 16, 2025 •

edited

Loading

Renaud-K commented Dec 18, 2025 •

edited by github-actions bot

Loading

Renaud-K commented Dec 18, 2025 •

edited by github-actions bot

Loading

Renaud-K commented Dec 18, 2025 •

edited by github-actions bot

Loading

Renaud-K commented Dec 19, 2025 •

edited by github-actions bot

Loading

Renaud-K commented Dec 19, 2025 •

edited by github-actions bot

Loading

Renaud-K commented Jan 6, 2026 •

edited by github-actions bot

Loading

Renaud-K commented Jan 7, 2026 •

edited by github-actions bot

Loading

Renaud-K commented Jan 7, 2026 •

edited by github-actions bot

Loading

Renaud-K commented Jan 7, 2026 •

edited by github-actions bot

Loading