Support using complex numbers and lists of complex numbers inside kernel functions #1518

annagrin · 2024-04-15T22:36:44Z

Make sure we can use complex numbers inside kernel function definitions, for example:

  c = [.70710678 + 0j, 0., 0., 0.70710678]
  
  # Pass a list of complex numbers as a parameter
  @cudaq.kernel
  def test_complex_vec_param(vec : list[complex]):
      c1 = vec

  counts = cudaq.sample(test_complex_vec_param, c)

  # Capture a list of complex numbers
  @cudaq.kernel
  def test_complex_vec_capture():
      c1 = c

   counts = cudaq.sample(test_complex_vec_capture)

  # Define a list of complex numbers inside a kernel
  @cudaq.kernel
  def test_complex_vec_definition():
      c1 = [1.0 + 0j, 0., 0., 1.]

  counts = cudaq.sample(test_complex_vec_definition)

Towards: #1086
Closes: #1454

…tate support

into annagrin/qalloc-state-init

copy-pr-bot · 2024-04-15T22:36:47Z

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.

github-actions · 2024-04-15T22:37:00Z

CLA Assistant Lite bot All Contributors have signed the CLA.

annagrin · 2024-04-15T22:37:54Z

I have read the Contributor License Agreement and I hereby accept the Terms.

annagrin · 2024-04-15T22:39:50Z

recheck

into annagrin/kernel-complex-support

amccaskey · 2024-04-16T22:21:13Z

python/cudaq/kernel/ast_bridge.py

        """
        ty = self.getFloatType()
        return arith.ConstantOp(ty, self.getFloatAttr(ty, value)).result

+    def getComplexType(self):


It would be good to parameterize this with the precision or bitwidth in MLIR terms (32 or 64 is what we support)

amccaskey · 2024-04-16T22:21:31Z

python/cudaq/kernel/ast_bridge.py

+        """
+        return ComplexType.get(self.getFloatType())
+
+    def getConstantComplex(self, value):


Same here (parameterize with 32 or 64)

amccaskey · 2024-04-16T22:23:16Z

python/cudaq/kernel/ast_bridge.py

@@ -2726,6 +2730,22 @@ def visit_Continue(self, node):
        else:
            cc.ContinueOp([])

+    def promote_operand_type(self, ty, operand):


Can we move this up with the other utility methods on this visitor class?

amccaskey · 2024-04-16T22:25:05Z

python/cudaq/kernel/ast_bridge.py

@@ -2896,16 +2906,21 @@ def visit_Name(self, node):

        if node.id in self.capturedVars:
            # Only support a small subset of types here
+            complexType = type(1j)


I think you can just use complex instead of getting the type of 1j

We should also probably check for np.complex64

I think you can just use complex instead of getting the type of 1j

We use complex dialect and it seems to override the complex type. Maybe worth decoupling in a separate PR...

We should also probably check for np.complex64

Done.

amccaskey · 2024-04-16T22:28:44Z

python/tests/kernel/test_kernel_float_complex.py

+    def test_float_vec_param(vec : list[float]):
+        f1 = vec
+
+    counts = cudaq.sample(test_float_vec_param, f)


A better test here would be to define a complex return type on the kernel, and assert test_float_vec_param(f) == that value

something like

@cudaq.kernel def test_float_vec_param(vec : list[float], i : int) -> complex: f1 = vec return vec[i] for i in range(len(f)): assert np.isclose(f[i], test_float_vec_param(f, i), atol=1e-6)

Opps, looks like there is a bug there - function above always returns the value of f[0]. Will investigate...

Changed the tests for now to work around this issue (will update the tests after the issue is fixed in the experimental branch as well)

amccaskey · 2024-04-16T22:30:25Z

python/tests/kernel/test_kernel_float_complex.py

+    counts = cudaq.sample(test_float_np_use)
+    assert len(counts) == 0
+
+def test_complex():


I would suggest also adding support for c = np.array([...], dtype=np.complex64) and demonstrate with the nvidia target. Also it would be good to test that we raise an exception for the use of complex with nvidia and np.complex64 with FP64 backends (all the other ones).

I created a separate issue for supporting np arrays: #1523, will work in it in the next PR

schweitzpgi · 2024-04-17T17:11:00Z

python/cudaq/kernel/ast_bridge.py

+        """
+        Return an MLIR complex type (double precision).
+        """
+        return ComplexType.get(self.getFloatType())


Q: Is there a way to request either F32 or F64 in the python bindings? (This getFloatType() feels more ambiguous.)

schweitzpgi · 2024-04-17T17:14:20Z

python/cudaq/kernel/ast_bridge.py

+
+        if F64Type.isinstance(ty):
+            if IntegerType.isinstance(operand.type):
+                operand = arith.SIToFPOp(ty, operand).result


These conversions make me wonder if we don't want to automatically add FPTrunc and FPExt if the input data is not at the precision expected by the simulator.

I'd like to postpone this to the next PR(s) when I have examples that are running with different data types in np.array and assigning the state to a qvector (to be able to tests my changes properly) - does it is sound reasonable?

annagrin · 2024-04-19T23:17:29Z

Replaced by #1550

annagrin added 5 commits April 12, 2024 13:25

Fix complex arithmetic and capture, add qvector initialization from s…

aced552

…tate support

Merge branch 'features/qalloc' of https://github.com/NVIDIA/cuda-quantum

ab57231

into annagrin/qalloc-state-init

Add more examples

fbbf0ff

Merge with features/qalloc

e4101a6

Support using complex numbers and lists inside kernel functions

4d895a6

cuda-quantum-bot added a commit that referenced this pull request Apr 15, 2024

@annagrin has signed the CLA from Pull Request #1518

3b58777

annagrin added 4 commits April 16, 2024 13:55

Run format and spell checker tools

7aacc28

Support complex numbers in numpy calls

d27477e

Run format and spell checker tools

d264015

Merge branch 'features/qalloc' of https://github.com/NVIDIA/cuda-quantum

a78724a

into annagrin/kernel-complex-support

annagrin requested review from amccaskey and khalatepradnya April 16, 2024 21:37

amccaskey reviewed Apr 16, 2024

View reviewed changes

annagrin requested a review from schweitzpgi April 16, 2024 22:40

Start addressing CR comments

f65842f

schweitzpgi reviewed Apr 17, 2024

View reviewed changes

annagrin added 2 commits April 19, 2024 15:26

Add handing of np.complex128 and np.complex64

b026031

format

6b6e491

annagrin mentioned this pull request Apr 19, 2024

Support using complex numbers and lists in kernel functions #1549

Closed

5 tasks

annagrin closed this Apr 19, 2024

github-actions bot locked and limited conversation to collaborators Apr 19, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Support using complex numbers and lists of complex numbers inside kernel functions #1518

Support using complex numbers and lists of complex numbers inside kernel functions #1518

annagrin commented Apr 15, 2024 •

edited

Loading

copy-pr-bot bot commented Apr 15, 2024

github-actions bot commented Apr 15, 2024 •

edited

Loading

annagrin commented Apr 15, 2024

annagrin commented Apr 15, 2024

amccaskey Apr 16, 2024

annagrin Apr 19, 2024

amccaskey Apr 16, 2024

annagrin Apr 19, 2024

amccaskey Apr 16, 2024

annagrin Apr 19, 2024

amccaskey Apr 16, 2024

amccaskey Apr 16, 2024

annagrin Apr 19, 2024

amccaskey Apr 16, 2024

annagrin Apr 17, 2024

annagrin Apr 19, 2024

amccaskey Apr 16, 2024

annagrin Apr 17, 2024

schweitzpgi Apr 17, 2024

schweitzpgi Apr 17, 2024

annagrin Apr 19, 2024

annagrin commented Apr 19, 2024

Support using complex numbers and lists of complex numbers inside kernel functions #1518

Support using complex numbers and lists of complex numbers inside kernel functions #1518

Conversation

annagrin commented Apr 15, 2024 • edited Loading

copy-pr-bot bot commented Apr 15, 2024

github-actions bot commented Apr 15, 2024 • edited Loading

annagrin commented Apr 15, 2024

annagrin commented Apr 15, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

annagrin commented Apr 19, 2024

annagrin commented Apr 15, 2024 •

edited

Loading

github-actions bot commented Apr 15, 2024 •

edited

Loading