use runmat_accelerate_api::GpuTensorHandle;
use runmat_builtins::{CharArray, ComplexTensor, Tensor, Value};
use runmat_macros::runtime_builtin;
use crate::builtins::common::spec::{
BroadcastSemantics, BuiltinFusionSpec, BuiltinGpuSpec, ConstantStrategy, FusionError,
FusionExprContext, FusionKernelTemplate, GpuOpKind, ProviderHook, ReductionNaN,
ResidencyPolicy, ScalarType, ShapeRequirements,
};
use crate::builtins::common::{gpu_helpers, map_control_flow_with_builtin, tensor};
use crate::builtins::math::type_resolvers::numeric_unary_type;
use crate::{build_runtime_error, BuiltinResult, RuntimeError};
#[runmat_macros::register_gpu_spec(builtin_path = "crate::builtins::math::elementwise::angle")]
pub const GPU_SPEC: BuiltinGpuSpec = BuiltinGpuSpec {
name: "angle",
op_kind: GpuOpKind::Elementwise,
supported_precisions: &[ScalarType::F32, ScalarType::F64],
broadcast: BroadcastSemantics::Matlab,
provider_hooks: &[ProviderHook::Unary { name: "unary_angle" }],
constant_strategy: ConstantStrategy::InlineLiteral,
residency: ResidencyPolicy::NewHandle,
nan_mode: ReductionNaN::Include,
two_pass_threshold: None,
workgroup_size: None,
accepts_nan_mode: false,
notes: "Providers implement unary_angle to evaluate atan2(imag(x), real(x)) on device; the runtime gathers to host whenever the hook is unavailable or when complex tensors need host conversion.",
};
#[runmat_macros::register_fusion_spec(builtin_path = "crate::builtins::math::elementwise::angle")]
pub const FUSION_SPEC: BuiltinFusionSpec = BuiltinFusionSpec {
name: "angle",
shape: ShapeRequirements::BroadcastCompatible,
constant_strategy: ConstantStrategy::InlineLiteral,
elementwise: Some(FusionKernelTemplate {
scalar_precisions: &[ScalarType::F32, ScalarType::F64],
wgsl_body: |ctx: &FusionExprContext| {
let input = ctx
.inputs
.first()
.ok_or(FusionError::MissingInput(0))?;
let zero = match ctx.scalar_ty {
ScalarType::F32 => "0.0".to_string(),
ScalarType::F64 => "f64(0.0)".to_string(),
other => return Err(FusionError::UnsupportedPrecision(other)),
};
Ok(format!("atan2({zero}, {input})"))
},
}),
reduction: None,
emits_nan: false,
notes: "Fusion assumes real-valued inputs (imaginary part zero). Complex tensors are gathered to the host until GPU complex storage lands.",
};
const BUILTIN_NAME: &str = "angle";
fn builtin_error(message: impl Into<String>) -> RuntimeError {
build_runtime_error(message)
.with_builtin(BUILTIN_NAME)
.build()
}
#[runtime_builtin(
name = "angle",
category = "math/elementwise",
summary = "Phase angle of scalars, vectors, matrices, or N-D tensors.",
keywords = "angle,phase,argument,complex,gpu",
accel = "unary",
type_resolver(numeric_unary_type),
builtin_path = "crate::builtins::math::elementwise::angle"
)]
async fn angle_builtin(value: Value) -> BuiltinResult<Value> {
match value {
Value::GpuTensor(handle) => angle_gpu(handle).await,
Value::Complex(re, im) => Ok(Value::Num(angle_scalar(re, im))),
Value::ComplexTensor(ct) => angle_complex_tensor(ct),
Value::CharArray(ca) => angle_char_array(ca),
Value::String(_) | Value::StringArray(_) => {
Err(builtin_error("angle: expected numeric input"))
}
other => angle_real(other),
}
}
async fn angle_gpu(handle: GpuTensorHandle) -> BuiltinResult<Value> {
if let Some(provider) = runmat_accelerate_api::provider_for_handle(&handle) {
if let Ok(device_result) = provider.unary_angle(&handle).await {
return Ok(Value::GpuTensor(device_result));
}
}
let tensor = gpu_helpers::gather_tensor_async(&handle)
.await
.map_err(|flow| map_control_flow_with_builtin(flow, BUILTIN_NAME))?;
Ok(tensor::tensor_into_value(angle_tensor(tensor)?))
}
fn angle_real(value: Value) -> BuiltinResult<Value> {
let tensor = tensor::value_into_tensor_for("angle", value)
.map_err(|e| builtin_error(format!("angle: {e}")))?;
Ok(tensor::tensor_into_value(angle_tensor(tensor)?))
}
fn angle_tensor(tensor: Tensor) -> BuiltinResult<Tensor> {
let Tensor { data, shape, .. } = tensor;
let mapped: Vec<f64> = data.into_iter().map(|re| angle_scalar(re, 0.0)).collect();
Tensor::new(mapped, shape).map_err(|e| builtin_error(format!("angle: {e}")))
}
fn angle_complex_tensor(ct: ComplexTensor) -> BuiltinResult<Value> {
let ComplexTensor { data, shape, .. } = ct;
let mapped: Vec<f64> = data
.into_iter()
.map(|(re, im)| angle_scalar(re, im))
.collect();
let tensor = Tensor::new(mapped, shape).map_err(|e| builtin_error(format!("angle: {e}")))?;
Ok(tensor::tensor_into_value(tensor))
}
fn angle_char_array(ca: CharArray) -> BuiltinResult<Value> {
let CharArray { data, rows, cols } = ca;
let mapped: Vec<f64> = data
.into_iter()
.map(|ch| angle_scalar(ch as u32 as f64, 0.0))
.collect();
let tensor =
Tensor::new(mapped, vec![rows, cols]).map_err(|e| builtin_error(format!("angle: {e}")))?;
Ok(tensor::tensor_into_value(tensor))
}
#[inline]
fn angle_scalar(re: f64, im: f64) -> f64 {
im.atan2(re)
}
#[cfg(test)]
pub(crate) mod tests {
use super::*;
use crate::builtins::common::test_support;
use futures::executor::block_on;
use runmat_builtins::{IntValue, LogicalArray, ResolveContext, StringArray, Type};
use std::f64::consts::PI;
fn angle_builtin(value: Value) -> BuiltinResult<Value> {
block_on(super::angle_builtin(value))
}
#[test]
fn angle_type_preserves_tensor_shape() {
let out = numeric_unary_type(
&[Type::Tensor {
shape: Some(vec![Some(2), Some(3)]),
}],
&ResolveContext::new(Vec::new()),
);
assert_eq!(
out,
Type::Tensor {
shape: Some(vec![Some(2), Some(3)])
}
);
}
#[test]
fn angle_type_scalar_tensor_returns_num() {
let out = numeric_unary_type(
&[Type::Tensor {
shape: Some(vec![Some(1), Some(1)]),
}],
&ResolveContext::new(Vec::new()),
);
assert_eq!(out, Type::Num);
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn angle_real_positive_negative() {
let pos = angle_builtin(Value::Num(5.0)).expect("angle");
assert_eq!(pos, Value::Num(0.0));
let neg = angle_builtin(Value::Num(-3.0)).expect("angle");
if let Value::Num(val) = neg {
assert!((val - PI).abs() < 1e-12);
} else {
panic!("expected numeric result, got {neg:?}");
}
let zero = angle_builtin(Value::Num(0.0)).expect("angle");
assert_eq!(zero, Value::Num(0.0));
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn angle_complex_scalar_matches_atan2() {
let value = Value::Complex(3.0, -4.0);
let result = angle_builtin(value).expect("angle");
if let Value::Num(angle) = result {
assert!((angle - (-4.0f64).atan2(3.0)).abs() < 1e-12);
} else {
panic!("expected numeric result");
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn angle_tensor_values() {
let tensor = Tensor::new(vec![1.0, -1.0, 0.0, 2.0], vec![2, 2]).unwrap();
let result = angle_builtin(Value::Tensor(tensor)).expect("angle");
match result {
Value::Tensor(out) => {
assert_eq!(out.shape, vec![2, 2]);
assert!((out.data[0] - 0.0).abs() < 1e-12);
assert!((out.data[1] - PI).abs() < 1e-12);
assert_eq!(out.data[2], 0.0);
assert_eq!(out.data[3], 0.0);
}
other => panic!("expected tensor result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn angle_logical_and_char_inputs() {
let logical = LogicalArray::new(vec![0, 1, 0, 1], vec![2, 2]).unwrap();
let logical_value = Value::LogicalArray(logical);
let logical_result = angle_builtin(logical_value).expect("angle");
match logical_result {
Value::Tensor(out) => assert!(out.data.iter().all(|&v| v == 0.0)),
other => panic!("expected tensor result, got {other:?}"),
}
let chars = CharArray::new("AB".chars().collect(), 1, 2).unwrap();
let char_result = angle_builtin(Value::CharArray(chars)).expect("angle");
match char_result {
Value::Tensor(out) => assert!(out.data.iter().all(|&v| v == 0.0)),
other => panic!("expected tensor result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn angle_complex_tensor() {
let data = vec![(1.0, 1.0), (-1.0, 1.0), (-1.0, -1.0), (1.0, -1.0)];
let tensor = ComplexTensor::new(data, vec![2, 2]).unwrap();
let result = angle_builtin(Value::ComplexTensor(tensor)).expect("angle");
match result {
Value::Tensor(out) => {
let expected = [
(1.0f64).atan2(1.0),
(1.0f64).atan2(-1.0),
(-1.0f64).atan2(-1.0),
(-1.0f64).atan2(1.0),
];
for (actual, target) in out.data.iter().zip(expected.iter()) {
assert!((actual - target).abs() < 1e-12);
}
}
other => panic!("expected tensor result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn angle_gpu_provider_roundtrip() {
test_support::with_test_provider(|provider| {
let tensor = Tensor::new(vec![1.0, -1.0, 0.5, -0.5], vec![2, 2]).unwrap();
let view = runmat_accelerate_api::HostTensorView {
data: &tensor.data,
shape: &tensor.shape,
};
let handle = provider.upload(&view).expect("upload");
let result = angle_builtin(Value::GpuTensor(handle)).expect("angle");
let gathered = test_support::gather(result).expect("gather");
let expected: Vec<f64> = tensor.data.iter().map(|&v| angle_scalar(v, 0.0)).collect();
assert_eq!(gathered.shape, vec![2, 2]);
for (actual, target) in gathered.data.iter().zip(expected.iter()) {
assert!((actual - target).abs() < 1e-12);
}
});
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn angle_dimensionless_int_input() {
let result = angle_builtin(Value::Int(IntValue::I32(-10))).expect("angle");
if let Value::Num(val) = result {
assert!((val - PI).abs() < 1e-12);
} else {
panic!("expected numeric result");
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn angle_nan_propagates() {
let result = angle_builtin(Value::Num(f64::NAN)).expect("angle");
match result {
Value::Num(v) => assert!(v.is_nan()),
other => panic!("expected numeric result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn angle_rejects_strings() {
let err = angle_builtin(Value::from("hello")).unwrap_err();
assert!(err.message().contains("angle: expected numeric input"));
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn angle_rejects_string_arrays() {
let array = StringArray::new(vec!["a".to_string(), "b".to_string()], vec![1, 2]).unwrap();
let err = angle_builtin(Value::StringArray(array)).unwrap_err();
assert!(err.message().contains("angle: expected numeric input"));
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
#[cfg(feature = "wgpu")]
fn angle_wgpu_matches_cpu() {
let _ = runmat_accelerate::backend::wgpu::provider::register_wgpu_provider(
runmat_accelerate::backend::wgpu::provider::WgpuProviderOptions::default(),
);
let tensor = Tensor::new(vec![1.0, -1.0, 0.5, -0.5], vec![2, 2]).unwrap();
let cpu = angle_tensor(tensor.clone()).unwrap();
let view = runmat_accelerate_api::HostTensorView {
data: &tensor.data,
shape: &tensor.shape,
};
let handle = runmat_accelerate_api::provider()
.unwrap()
.upload(&view)
.unwrap();
let gpu = block_on(angle_gpu(handle)).unwrap();
let gathered = test_support::gather(gpu).expect("gather");
match (Value::Tensor(cpu), gathered) {
(Value::Tensor(ct), gt) => {
assert_eq!(gt.shape, ct.shape);
let tol = match runmat_accelerate_api::provider().unwrap().precision() {
runmat_accelerate_api::ProviderPrecision::F64 => 1e-12,
runmat_accelerate_api::ProviderPrecision::F32 => 1e-5,
};
for (a, b) in gt.data.iter().zip(ct.data.iter()) {
assert!((a - b).abs() < tol);
}
}
_ => panic!("unexpected shapes"),
}
}
}