[Builder] Add F64Type support and SinOp for math operations

[zzzDavid]

Summary

• Add F64Type to the type check for scalar math operations in builder.py, so float64 arguments are recognized alongside float32 and integer types
• Add math.SinOp to the supported math operation map, enabling allo.sin() for all float types
• Add dedicated test cases for sin (float32 and float64) and float64 math/trig operations

Split out from #287 per review feedback.

Test plan

• test_sin_float32 — verifies allo.sin() works with float32 arrays
• test_sin_float64 — verifies allo.sin() works with float64 arrays
• test_float64_math_ops — verifies allo.exp(), allo.log(), allo.sqrt() with float64
• test_float64_trig_ops — verifies allo.sin(), allo.cos(), allo.tan() with float64

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR extends Allo’s IR builder to recognize float64 in scalar math op lowering and adds sin support via the MLIR math dialect, with new tests covering sin and other math/trig operations on float64.

Changes:

• Add F64Type to the scalar math-op type check in allo/ir/builder.py.
• Add "sin": math_d.SinOp to the scalar math operation map.
• Add tests for allo.sin() on float32/float64, plus additional float64 math/trig coverage.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
allo/ir/builder.py	Expands scalar math-op lowering to accept f64 and adds sin lowering.
tests/test_types.py	Adds new regression tests for sin and other math/trig ops on float64.

Comments suppressed due to low confidence (1)

allo/ir/builder.py:3126

• opcls can be None when fn_name isn’t in the mapping, but this branch still unconditionally calls opcls(...), which will raise a TypeError instead of the intended unsupported-function error. Add an explicit check for opcls is None and fall through to the later handling / raise a clear RuntimeError for unsupported scalar math ops.

                opcls = {
                    "exp": math_d.ExpOp,
                    "log": math_d.LogOp,
                    "log2": math_d.Log2Op,
                    "log10": math_d.Log10Op,
                    "sqrt": math_d.SqrtOp,
                    "sin": math_d.SinOp,
                    "cos": math_d.CosOp,
                    "tan": math_d.TanOp,
                    "tanh": math_d.TanhOp,
                    "power": math_d.PowFOp,
                    "abs": math_d.AbsIOp,
                }.get(fn_name)
                return opcls(
                    *[ASTTransformer.get_mlir_op_result(ctx, x) for x in new_args],
                    ip=ctx.get_ip(),
                )

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[chhzh123]

LGTM, thanks!