Add QIR Program Format Support to the DDSIM QDMI Device by rturrado · Pull Request #1766 · munich-quantum-toolkit/core

rturrado · 2026-06-05T21:25:55Z

Description

First contribution to UnitaryHACK for me! 🎉 And man, I've enjoyed working on this one! 😀

Note

The current PR still does not cover all the functionality proposed in the #1695 issue. In particular, the support for QIR adaptive base modules and strings is not yet implemented. But the current solution is ready for review.

There are two main pieces of work:

The introduction of an MQT Core QIR JIT library (MQT::CoreQIRJIT, src/qir/jit). This library contains almost all of the previous Runner.cpp code: LLVM JIT initialization, QIR runtime symbols registration, and program execution. Compared to that previous Runner code, it also lets you parse an IR directly from a string, and not only from a file. The main purpose of the QIR JIT library is to encapsulate the LLVM JIT code so that it can also be used from the QDMI device. The interface is a qir::jit::Session type that can be constructed from an input file or from a byte string (text for LLVM assembly or binary for LLVM bitcode), and then run with or without arguments (e.g., the Runner CLI can run a JIT program with arguments, and the QDMI device without arguments).
The job submission in Device.cpp. By using the MQT Core QIR JIT library, the device can just run a JIT program, get the results, and process them (filter zero and one addresses, sort the addresses, and transform the measurements into a bit string).

The new functionality is tested via two tests in results_sampling_test.cpp, one that exercises the QIR base module path, and the other that exercises the QIR base string path.

For this PR I have used Claude Opus 4.7, mainly from a CLion terminal, and for different purposes: from understanding existing code, to drafting code snippets, catching bugs and typos, running clang-format and clang-tidy and proposing fixes for clang-tidy warnings, reviewing wording in commit messages and docs...

Fixes #1695

Checklist

The pull request only contains commits that are focused and relevant to this change.
I have added appropriate tests that cover the new/changed functionality.
I have updated the documentation to reflect these changes.
I have added entries to the changelog for any noteworthy additions, changes, fixes, or removals.
I have added migration instructions to the upgrade guide (if needed).
The changes follow the project's style guidelines and introduce no new warnings.
The changes are fully tested and pass the CI checks.
I have reviewed my own code changes.

If PR contains AI-assisted content:

I have disclosed the use of AI tools in the PR description as per our AI Usage Guidelines.
AI-assisted commits include an Assisted-by: [Model Name] via [Tool Name] footer.
I confirm that I have personally reviewed and understood all AI-generated content, and accept full responsibility for it.

Add qir::jit::Session to encapsulate the LLJIT initialization, QIR runtime symbols registration, and program execution. The idea is to avoid future duplication of JIT setup both in the Runner and in the Device. Update Runner to use qir::jit::Session.

circuits.hpp, error_handling_test.cpp: add QIR_BELL_PAIR_STATIC with an LLVM assembly string of a Bell pair program, and rename MALFORMED_PROGRAM to QASM3_MALFORMED (to be consistent with the other QASM3_ variables). src/qdmi/devices/dd/CMakeLists.txt: add MQT::CoreQIRJIT library dependency. src/qir/jit/CMakeLists.txt: link libraries privately. test/qdmi/devices/dd/CMakeLists.txt: add LLVM native libs dependencies. Device.hpp: add submitQASMProgram and submitQIRProgram APIs. Device.cpp: handle job submissions of a QASM program and of a QIR base module/string separately. Implement submitQIRProgram: for numShots_, reset the runtime, run the program, process results into a bit string, update measurement counts. Device.cpp, test_utils.cpp: the device stores the program into a std::string of a given size, and submitQIRProgram reads exactly that size. No passing +1 bytes or sometimes retrieving -1 bytes is needed in any case. job_parameters_test.cpp: set QDMI_PROGRAM_FORMAT_QIRBASEMODULE and QDMI_PROGRAM_FORMAT_QIRBASESTRING as supported formats. results_sampling_test.cpp: change HistogramKeysAndValuesSumToShots into a test class, and implement two new test fixtures for QIR base modules and strings. These two new tests exercise the submission of a QIR program for a DD device job. Runtime: add getResults API. Session: make run methods const.

This option enables the QIR format support for QDMI. It is defined in the top level CMakeLists.txt, and it is disabled by default. The QDMI DDSim Device library only links to QIR JIT and QIR Runtime libraries if the option is ON. The same applies to the corresponding test library.

Add a 'QIR Support in the QDMI Device' entry. Assisted-by: Claude Opus 4.7 via Claude Code

Leave just one function where both arguments have default values.

codecov · 2026-06-05T21:29:34Z

Codecov Report

❌ Patch coverage is 89.72222% with 37 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/qir/jit/Session.cpp	84.5%	30 Missing ⚠️
src/qdmi/devices/dd/Device.cpp	93.5%	6 Missing ⚠️
src/qir/jit/IRRewriter.cpp	95.2%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

When BUILD_MQT_CORE_MLIR=OFF , find_package(MLIR) is not called, so LLVM CMake macros like llvm_map_components_to_libnames are undefined. The new qir/jit subdirectory used that macro unconditionally and failed to configure. Make BUILD_MQT_CORE_QDMI_WITH_QIR a cmake_dependent_option on BUILD_MQT_CORE_MLIR (forced OFF when MLIR is off). Add the qir/jit subdirectory only when at least one consumer (Runner CLI or QDMI-with-QIR) is enabled. Assisted-by: Claude Opus 4.7 via Claude Code

The HistogramKeysAndValuesSumToShots fixture class in results_sampling_test.cpp was at file scope, triggering clang-tidy's misc-use-anonymous-namespace check in CI (clang-tidy 22). Wrap it in an anonymous namespace to enforce internal linkage, matching the convention used by the ErrorHandling fixture in error_handling_test.cpp. Assisted-by: Claude Opus 4.7 via Claude Code

Signed-off-by: Roberto Turrado Camblor <rturrado@gmail.com>

Remove merge conflict strings.

ystade

Hi @rturrado,
Your contribution already looks great in this state. Thanks for taking part in the unitaryHACK. I am a big fan of the restructuring, i.e., having the JIT interpreter as its own library. This makes the code a lot better structured. It is actually also cool to see that this refactoring comprises the most part of this PR, and the remaining changes are quite local. The biggest action item that remains in my mind is the documentation of the Session class. In particular, I reviewed all files in detail except the session class because the documentation eases the review here a lot. So, in summary, my biggest request are comprehensive docstrings for the Session class. Overall, this contribution looks already very good and is very valuable.

burgholzer · 2026-06-06T19:04:37Z

@rturrado Just came to say this is a great initial draft! Keep it up 💪🏻

Also move comment from Runner's main into Session::run.

Move filtering and sorting of Result pointers to getResults. Add qir::toBitString to convert a map<Result*, bool> to a bit string. Assisted-by: Claude Opus 4.7 via Claude Code

Add an injectable std::ostream pointer member to Runtime, defaulting to std::cout, with getOstream/setOstream/resetOstream accessors. Update __quantum__rt__result_record_output (QIR.cpp) to write through runtime.getOstream() rather than directly to std::cout. Migrate the test fixtures in test_qir_runtime.cpp (QIRRuntimeTest) and results_sampling_test.cpp (HistogramKeysAndValuesSumToShots). Inject a std::ostringstream sink via setOstream in SetUp and restore the default via resetOstream in TearDown. QIRRuntimeTest's previous std::cout.rdbuf swap is gone. The QDMI sampling tests no longer spew thousands of result-record lines to stdout per run. Link MQT::CoreQIRRuntime to the dd-device test target when BUILD_MQT_CORE_QDMI_WITH_QIR is ON, since results_sampling_test.cpp now includes Runtime.hpp directly. Assisted-by: Claude Opus 4.7 via Claude Code

rturrado · 2026-06-06T21:59:47Z

Hi @rturrado, Your contribution already looks great in this state. Thanks for taking part in the unitaryHACK. I am a big fan of the restructuring, i.e., having the JIT interpreter as its own library. This makes the code a lot better structured. It is actually also cool to see that this refactoring comprises the most part of this PR, and the remaining changes are quite local. The biggest action item that remains in my mind is the documentation of the Session class. In particular, I reviewed all files in detail except the session class because the documentation eases the review here a lot. So, in summary, my biggest request are comprehensive docstrings for the Session class. Overall, this contribution looks already very good and is very valuable.

Hi @ystade ! Many thanks for the comments and for the thorough review. I'll go over each of the items in the following days.

There are still a few things I want to tackle as part of this PR, on top of whatever you consider necessary:

Runner is not try/catching the JIT session. ✔
submitQIRProgram calls Runtime's getResults but then does a lot of post processing (filter zero and one addresses, order addresses, and convert to bit string). I would prefer part of that functionality to be done by the Runtime itself (filter and, optionally, order) and by a utility function (toBitString). ✔
Outputting to an ostream, not just to standard output. ✔
Dynamic modules. This is not tested in the current implementation. ✔
Adaptive program formats. Same. Currently untested. ✔
Output label at result_record_output depending on the output schema. ❓
Code coverage errors.

Assisted-by: Claude Opus 4.7 via Claude Code

Add Runtime::recordOutput(Result*) and Runtime::getRecordedOutputs(), backed by a std::string field. __quantum__rt__result_record_output now records the value into recordedOutputs before emitting the QIR-spec "label: 0|1" line, so dynamic QIR programs that release their Results via reference-count decrement no longer lose data before the QDMI device reads it. submitQIRProgram in the dd device consumes runtime.getRecordedOutputs() directly as the histogram key. Remove the now-unused Runtime::getResults() and the qir::toBitString helper. Add QIR_BELL_PAIR_DYNAMIC to circuits.hpp and the matching QIRBaseModuleDynamic and QIRBaseStringDynamic sampling tests. Flip dynamic_qubit_management and dynamic_result_management to i1 true in BellPairDynamic.ll and the inline dynamic string, for spec-compliant module metadata. Assisted-by: Claude Opus 4.7 via Claude Code

Move BellPairStatic.ll, BellPairDynamic.ll, and GHZ4Dynamic.ll from test/qir to test/circuits, so the same files can serve both the qir-{runner,runtime} tests and the QDMI device tests. Update QIR_FILES_DIR in test/qir/{runner,runtime}/CMakeLists.txt to point at the new location. Add a QIR_FILES_DIR compile definition to the QDMI device test target. Switch the QIR tests in results_sampling_test.cpp to load BellPairStatic.ll and BellPairDynamic.ll from that path. Drop the QIR_BELL_PAIR_STATIC and QIR_BELL_PAIR_DYNAMIC inline strings from circuits.hpp. Assisted-by: Claude Opus 4.7 via Claude Code

Route `QDMI_PROGRAM_FORMAT_QIRADAPTIVE{MODULE,STRING}` to `submitQIRProgram` and accept them in the program-format gate. The existing JIT runtime already handles `__quantum__rt__read_result + br i1`, so a `H; measure; conditional-X; measure` Bell pair via classical correction runs end-to-end with no further runtime change. `bindings/fomac/fomac.cpp` already exposed `QIR_ADAPTIVE_{MODULE,STRING}`, so the Python surface is unchanged. New `test/circuits/BellPairAdaptive.ll`: produces the same `{"00","11"}` correlation as the entangled Bell pair, but via classical control flow on a measured result, so the same assertion validates all three encodings (Static, Dynamic, Adaptive). Refactor `results_sampling_test.cpp` into a shared `HistogramTest` fixture plus two QIR subclasses. `checkHistogram` now asserts both `sum == SHOTS` and the histogram keys are exactly `{"00","11"}`. The new key-shape check is what proves the adaptive branch actually fires for every shot rather than collapsing to a single value. Move the adaptive format constants from the unsupported loop to the supported loop in `job_parameters_test.cpp`. Assisted-by: Claude Opus 4.7 via Claude Code

rturrado · 2026-06-08T19:18:11Z

A question for you, guys!

The last commits add support for QIR adaptive formats. These formats allow the client to record and output values other than Result*, e.g., bool, int, float... At the moment, the Runtime is outputting all these values, but recording only bit values coming from measurement results (i.e., the Boolean ResultStruct.r). It just keeps those results as a bit string because the only consumer of recorded outputs is the QDMI DDsim device, via submitQIRprogram, which gets the bit string and updates the counts_ map with it.

The question would be if we also want to record the other possible values, and if we want to do it as part of this PR. A possible solution would be to change that std::string recordedBitResults data structure into something like a llvm::SmallVector<std::variant<bit, bool, int64_t, double, TupleHeader, ArrayHeader>> with the required accessors. I don't think that the implementation would be very difficult; I just don't know if it would be necessary.

burgholzer · 2026-06-08T19:25:29Z

A question for you, guys!

The last commits add support for QIR adaptive formats. These formats allow the client to record and output values other than Result*, e.g., bool, int, float... At the moment, the Runtime is outputting all these values, but recording only bit values coming from measurement results (i.e., the Boolean ResultStruct.r). It just keeps those results as a bit string because the only consumer of recorded outputs is the QDMI DDsim device, via submitQIRprogram, which gets the bit string and updates the counts_ map with it.

The question would be if we also want to record the other possible values, and if we want to do it as part of this PR. A possible solution would be to change that std::string recordedBitResults data structure into something like a llvm::SmallVector<std::variant<bit, bool, int64_t, double, TupleHeader, ArrayHeader>> with the required accessors. I don't think that the implementation would be very difficult; I just don't know if it would be necessary.

Hm, very good question.
I believe the limiting factor for this at the moment is actually QDMI because it does not really expose any meaningful result types that would make sense except for measurement results.
The reasonable things are SHOTS and COUNTS.
I don't recall seeing many programs that output something else. Have you maybe come across some that seemed reasonable?
How does the official QIR runner handle output recording/results collection?

I generally believe that for the sake of this PR, it is fine to focus on the things that can reasonably be piped through QDMI as results.
Technically, one could use a CUSTOM result type to expose the full output recording (which is probably what the QIR runner does). That might be a reasonable compromise here.

Does that make sense?

rturrado · 2026-06-08T23:15:35Z

Thanks for the answer, @burgholzer.

I don't recall seeing many programs that output something else. Have you maybe come across some that seemed reasonable?

No, not really. I asked Claude today to think about a realistic scenario for using a QIR program in adaptive format, and it came up with a Hamming weight, measuring 3 qubits, and calculating an int weight and a float mean, which I found nice.

But, back to my original question, I think that I was understanding wrongly the QIR concept of recording outputs of different types. I believe now that the spec only refers to logging labels and values somewhere (e.g., an output stream). And that, if we decide to use a data structure for keeping the measurements bit string, that is something completely apart. My confusion came from the idea that we were requested by the QIR spec to store bools, and ints, and records, and so on, into that data structure.

How does the official QIR runner handle output recording/results collection?

The current QIR runner can only record Result*, and doesn't support output schemas. That means that it can only print measured qubits, and in a custom way, e.g., r0: 0. This PR is adding support both for adaptive profiles and output schemas (both to a certain extent), so it can record more data types and in a spec-compliant way, let's say.

ystade · 2026-06-09T07:42:57Z

@rturrado, it is cool to see when the software components that we have, e.g., the runner, become more spec-compliant. In my opinion, to keep the PR focused and avoid overloading it, it should be fine to restrict it to recording results and leave other data types for future Issues/PRs.

ystade · 2026-06-12T06:25:29Z

@rturrado let me know, when I should review this PR again. I'll try to get to it asap then.

rturrado · 2026-06-12T07:29:14Z

@rturrado let me know, when I should review this PR again. I'll try to get to it asap then.

@ystade Thanks! I think that the only pending points are the coverage errors, so if you want to review the last commits, and maybe check if some unresolved convesations can already be closed, you can do it already.

ystade · 2026-06-12T07:53:55Z

@rturrado Thanks for the update, I am happy to wait until the coverage is also satisfied.

burgholzer · 2026-06-12T09:51:11Z

BTW: Do not worry about the Windows CI errors. They are unrelated to your changes and a consequence of some GitHub runner updates for Windows systems not interacting well with our pre-built LLVM distributions. They will be fixed separately.

Lift `src/qir/jit/Session.cpp` diff coverage from 76.8% toward the Codecov 90% target by exercising paths the existing tests skipped: - `LoadModuleFromMemory`: build a `JitSession` from an in-memory IR buffer (the contents of `BellPairStatic.ll`) and run it, covering `loadModuleFromMemory` and the memory-based constructor. - `MalformedIRThrows`: pass unparseable IR and expect `std::runtime_error`, covering `getThreadSafeModuleOrError`'s null-module branch and the throw in `initialize` when the loader hands back an error. - `MissingMainThrows`: pass valid IR with no `main` and expect `std::runtime_error`, covering the `lookup("main")` failure branch. While here, rename the fixture from `JitSessionExecutionMode` to `JitSessionTest` and factor out a `getProgram` helper that mirrors the one in `results_sampling_test.cpp`. The remaining uncovered lines in `Session.cpp` are Cygwin/MinGW paths, the GDB JIT listener, and throw arms behind LLVM `Expected<T>` failures that cannot be portably forced from a test. Files: - test/qir/jit/test_jit_session.cpp: new tests, renamed fixture, and `getProgram` helper. Assisted-by: Claude Opus 4.7 via Claude Code

The `JitSessionErrors.MissingMainThrows` test added in the previous commit failed on macOS CI. The host process's `main` symbol is exported through `DynamicLibrarySearchGenerator::GetForCurrentProcess` on macOS but not on Linux. So the JIT's `lookup("main")` finds the gtest runner's `main` on macOS and the constructor returns without throwing. On Linux the lookup fails and the throw fires, which is what the test expected. We do not control which symbols the process generator exposes from the test side, so the test cannot be made portable. The lost line (the `lookup("main")` failure branch at line 381 of `Session.cpp`) is not worth a platform-specific carve-out. Files: - test/qir/jit/test_jit_session.cpp: remove the `MissingMainThrows` test. Assisted-by: Claude Opus 4.7 via Claude Code

Add `QIRRuntimeTest.PackageResizeWhenEnlargingState`: apply H to qubit address 32, which makes `translateAddresses` call `enlargeState(32)` and pushes `qState.numQubits` past the `dd::Package` default capacity of 32, triggering `qState.dd->resize`. Files: - test/qir/runtime/test_qir_runtime.cpp: new test driving H on qubit index 32. Assisted-by: Claude Opus 4.7 via Claude Code

…into 1695

rturrado · 2026-06-12T12:25:17Z

@ystade @burgholzer PR should be ready now for review. Functionality is there. Code coverage is happy. And tests are passing (except those 2 Windows ones that you mentioned we shouldn't worry about for this PR).

ystade

@rturrado Thanks for this very cool PR extending the QDMI DDSIM device to support QIR in the base and adaptive profile. Your code is very well structured and I like also the few refactoring of the old code base. This is a very valauble contribution.

I have left some minor comments inline. Other then these, you get a green light from my end. I purposefully do not approve the PR yet and leave this to @burgholzer because I do not know whether he also wants to have an eye on the PR before it gets merged.

burgholzer · 2026-06-15T16:57:03Z

I'll give this a proper review tomorrow. Didn't quite get to it today unfortunately.

Apply review suggestions across CMake, docs, and Doxygen comments, plus one parameter rename. - CMakeLists.txt: default `BUILD_MQT_CORE_QDMI_DDSIM_WITH_QIR` to `ON` instead of `OFF`, so QIR support is enabled whenever MLIR is. - test/qir/CMakeLists.txt: drop the outer `if(BUILD_MQT_CORE_QIR_RUNNER)` and `if(BUILD_MQT_CORE_QIR_RUNNER OR BUILD_MQT_CORE_QDMI_DDSIM_WITH_QIR)` guards, since the inner `if(TARGET ...)` checks in each child already subsume them. Assisted-by: Claude Opus 4.7 via Claude Code

Replace three identical inline `getProgram` definitions across the QIR JIT and DDSIM QDMI test files with a single shared utility. - test/qir/helpers/test_utils.{h,c}pp: new `qir_test::getProgram(std::string_view)` that reads a QIR source file from `QIR_FILES_DIR` and returns its contents as a string. - test/qir/jit/test_jit_session.cpp, test/qdmi/devices/dd/{results_sampling_test.cpp, results_statevector_test.cpp}: drop the local `getProgram` and call `qir_test::getProgram` directly. - test/CMakeLists.txt: reorder so `qir` is added before `qdmi`, so the qdmi tests can link against `mqt-core-qir-test-utils`. - test/qir/CMakeLists.txt: add the new `helpers` subdirectory first. - test/qir/helpers/CMakeLists.txt: new `mqt-core-qir-test-utils` STATIC library, with `test/` on its public include path so consumers include via `#include "qir/helpers/test_utils.hpp"`. - test/qir/jit/CMakeLists.txt, test/qdmi/devices/dd/CMakeLists.txt: link the helper. Assisted-by: Claude Opus 4.7 via Claude Code

- include/mqt-core/qdmi/devices/dd/Device.hpp: change `program_` to `std::variant<std::string, std::vector<std::byte>>`. Text formats (QASM2/3, QIR Base/Adaptive String) use the string alternative, binary formats (QIR Base/Adaptive Module) use the byte-vector alternative. - src/qdmi/devices/dd/Device.cpp: write site dispatches on `isTextProgramFormat(format_)`. QIR JIT paths use `std::visit` with a generic lambda so the same lambda extracts `.data()` and `.size()` from either alternative and builds an `llvm::StringRef`. Assisted-by: Claude Opus 4.7 via Claude Code

burgholzer

This looks really clean. Great contribution @rturrado 🙌🏼
I have one small round of feedback that I'd like to see addressed. Afterwards, we can happily merge this 🚀

- Device.cpp: Fix bug at `queryProperty`. The previous `std::get_if<std::string>` guard silently dropped binary payloads. Use ADD_STRING_PROPERTY for the `std::string` alternative and ADD_LIST_PROPERTY for the `std::vector<std::byte>` alternative. - ProgramFormat.hpp: deleted; the predicate is inlined at its two call sites. - index.md: cover Adaptive Profile. - CHANGELOG.md, CMakeLists.txt, index.md: normalize "DDSIM QDMI Device" term. Assisted-by: Claude Opus 4.7 via Claude Code

ystade

@rturrado This PR was truly a very pleasant interaction. This is exactly how we imagine a unitaryHACK contribution to be. Thanks for your contribution!
If you are interested, you are very welcome to contribute to our projects again in the future, either as part of the next unitaryHACK or independently.

rturrado · 2026-06-16T14:45:52Z

@rturrado This PR was truly a very pleasant interaction. This is exactly how we imagine a unitaryHACK contribution to be. Thanks for your contribution!

@ystade Thanks so much! I really appreciate the words. I can also say it was very pleasant for me. You were very quick to answer my questions and provide reviews, and I found the comments very constructive. Not to mention that for me, being able to work in (and learn from) a C++ codebase in this quantum/LLVM/IR world is kind of an ideal job.

If you are interested, you are very welcome to contribute to our projects again in the future, either as part of the next unitaryHACK or independently.

Definitely. I have at least a couple of pending follow-ups: Add QIR Output Schemas Support, and Add QIR Adaptive Profile Support. Do you prefer me to open the PRs directly, or should I create an issue first?

burgholzer

Wonderful 🙌🏼

Before we merge this, could you please add a comment to the original issue here so that we can assign it to you and you get your bounty? 💸

As for tackling further issues: Typically, it's not a bad idea to create tracking issues. Not an absolute must-have, but it kind of gives us the opportunity to discuss some design decisions before any implementation is done.

rturrado · 2026-06-16T16:44:29Z

Wonderful 🙌🏼

🥳

Before we merge this, could you please add a comment to the original issue here so that we can assign it to you and you get your bounty? 💸

Roger that.

As for tackling further issues: Typically, it's not a bad idea to create tracking issues. Not an absolute must-have, but it kind of gives us the opportunity to discuss some design decisions before any implementation is done.

Fair enough!

rturrado added 6 commits June 5, 2026 14:05

Fix typo in comment in src/qir/runtime/CMakeLists.txt

4034994

Update docs/qir/index.md

9422a38

Add a 'QIR Support in the QDMI Device' entry. Assisted-by: Claude Opus 4.7 via Claude Code

Clean up Session::run

ac958b7

Leave just one function where both arguments have default values.

rturrado added 3 commits June 5, 2026 23:43

Update CHANGELOG.md

a20336b

mergify Bot added the conflict label Jun 5, 2026

Merge branch 'main' into 1695

6d763e8

Signed-off-by: Roberto Turrado Camblor <rturrado@gmail.com>

mergify Bot removed the conflict label Jun 5, 2026

rturrado and others added 2 commits June 6, 2026 00:17

Fix CHANGELOG.md

596b6b3

Remove merge conflict strings.

Merge branch 'main' into 1695

e44eba5

ystade requested changes Jun 6, 2026

View reviewed changes

rturrado added 4 commits June 6, 2026 22:16

Try/catch JIT session code in the Runner

71ecf0d

Also move comment from Runner's main into Session::run.

Refactor submitQIRProgram and Runtime::getResults

05e5ee5

Move filtering and sorting of Result pointers to getResults. Add qir::toBitString to convert a map<Result*, bool> to a bit string. Assisted-by: Claude Opus 4.7 via Claude Code

Merge remote-tracking branch 'origin/1695' into 1695

507b457

rturrado added 4 commits June 7, 2026 00:31

Fix missing <exception> header

f920621

Assisted-by: Claude Opus 4.7 via Claude Code

rturrado added 4 commits June 12, 2026 13:13

Merge branch 'main' of https://github.com/munich-quantum-toolkit/core …

4e3a0d0

…into 1695

rturrado requested a review from ystade June 12, 2026 12:25

ystade reviewed Jun 13, 2026

View reviewed changes

rturrado added 4 commits June 15, 2026 20:20

Merge branch 'main' into 1695

1e59bde

rturrado requested a review from ystade June 15, 2026 21:10

burgholzer requested changes Jun 16, 2026

View reviewed changes

Comment thread include/mqt-core/qdmi/devices/dd/ProgramFormat.hpp Outdated

Comment thread docs/qir/index.md Outdated

Comment thread docs/qir/index.md Outdated

Comment thread src/qdmi/devices/dd/Device.cpp Outdated

Comment thread test/qdmi/devices/dd/results_sampling_test.cpp

rturrado requested a review from burgholzer June 16, 2026 12:18

rturrado changed the title ~~Add QIR Program Format Support to the QDMI DDSim Device~~ Add QIR Program Format Support to the DDSIM QDMI Device Jun 16, 2026

ystade approved these changes Jun 16, 2026

View reviewed changes

burgholzer approved these changes Jun 16, 2026

View reviewed changes

rturrado mentioned this pull request Jun 16, 2026

✨ Add QIR Program Format Support to the QDMI DDSim Device #1695

Closed

burgholzer merged commit 12cf546 into munich-quantum-toolkit:main Jun 16, 2026
33 checks passed

rturrado deleted the 1695 branch June 17, 2026 10:10

This was referenced Jun 20, 2026

✨ Add QIR Adaptive Profile non-bit output recording to the QIR runtime #1797

Open

✨ Add non-bit record output functions to the QIR runtime #1799

Open

Uh oh!

Conversation

rturrado commented Jun 5, 2026

Description

Checklist

Uh oh!

codecov Bot commented Jun 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

ystade left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

burgholzer commented Jun 6, 2026

Uh oh!

rturrado commented Jun 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rturrado commented Jun 8, 2026

Uh oh!

burgholzer commented Jun 8, 2026

Uh oh!

rturrado commented Jun 8, 2026

Uh oh!

ystade commented Jun 9, 2026

Uh oh!

ystade commented Jun 12, 2026

Uh oh!

rturrado commented Jun 12, 2026

Uh oh!

ystade commented Jun 12, 2026

Uh oh!

burgholzer commented Jun 12, 2026

Uh oh!

rturrado commented Jun 12, 2026

Uh oh!

ystade left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

burgholzer commented Jun 15, 2026

Uh oh!

burgholzer left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ystade left a comment

Choose a reason for hiding this comment

Uh oh!

rturrado commented Jun 16, 2026

Uh oh!

burgholzer left a comment

Choose a reason for hiding this comment

Uh oh!

rturrado commented Jun 16, 2026

Uh oh!

Uh oh!

codecov Bot commented Jun 5, 2026 •

edited

Loading

rturrado commented Jun 6, 2026 •

edited

Loading