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, 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::abs")]
pub const GPU_SPEC: BuiltinGpuSpec = BuiltinGpuSpec {
name: "abs",
op_kind: GpuOpKind::Elementwise,
supported_precisions: &[ScalarType::F32, ScalarType::F64],
broadcast: BroadcastSemantics::Matlab,
provider_hooks: &[ProviderHook::Unary { name: "unary_abs" }],
constant_strategy: ConstantStrategy::InlineLiteral,
residency: ResidencyPolicy::NewHandle,
nan_mode: ReductionNaN::Include,
two_pass_threshold: None,
workgroup_size: None,
accepts_nan_mode: false,
notes: "Providers may execute abs in-place; the runtime gathers to host when unary_abs is unavailable or when complex magnitudes are required.",
};
#[runmat_macros::register_fusion_spec(builtin_path = "crate::builtins::math::elementwise::abs")]
pub const FUSION_SPEC: BuiltinFusionSpec = BuiltinFusionSpec {
name: "abs",
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))?;
Ok(format!("abs({input})"))
},
}),
reduction: None,
emits_nan: false,
notes: "Fusion planner emits WGSL abs; providers can swap in specialised kernels.",
};
const BUILTIN_NAME: &str = "abs";
fn builtin_error(message: impl Into<String>) -> RuntimeError {
build_runtime_error(message)
.with_builtin(BUILTIN_NAME)
.build()
}
#[runtime_builtin(
name = "abs",
category = "math/elementwise",
summary = "Absolute value or magnitude of scalars, vectors, matrices, or N-D tensors.",
keywords = "abs,absolute value,magnitude,complex,gpu",
accel = "unary",
type_resolver(numeric_unary_type),
builtin_path = "crate::builtins::math::elementwise::abs"
)]
async fn abs_builtin(value: Value) -> BuiltinResult<Value> {
match value {
Value::GpuTensor(handle) => abs_gpu(handle).await,
Value::Complex(re, im) => Ok(Value::Num(complex_magnitude(re, im))),
Value::ComplexTensor(ct) => abs_complex_tensor(ct),
Value::CharArray(ca) => abs_char_array(ca),
Value::String(_) | Value::StringArray(_) => {
Err(builtin_error("abs: expected numeric input"))
}
other => abs_real(other),
}
}
async fn abs_gpu(handle: GpuTensorHandle) -> BuiltinResult<Value> {
if let Some(provider) = runmat_accelerate_api::provider_for_handle(&handle) {
if let Ok(out) = provider.unary_abs(&handle).await {
return Ok(Value::GpuTensor(out));
}
}
let tensor = gpu_helpers::gather_tensor_async(&handle).await?;
Ok(tensor::tensor_into_value(abs_tensor(tensor)?))
}
fn abs_real(value: Value) -> BuiltinResult<Value> {
let tensor = tensor::value_into_tensor_for("abs", value)?;
Ok(tensor::tensor_into_value(abs_tensor(tensor)?))
}
fn abs_tensor(tensor: Tensor) -> BuiltinResult<Tensor> {
let data = tensor.data.iter().map(|&v| v.abs()).collect::<Vec<_>>();
Tensor::new(data, tensor.shape.clone()).map_err(|e| builtin_error(format!("abs: {e}")))
}
fn abs_complex_tensor(ct: ComplexTensor) -> BuiltinResult<Value> {
let data = ct
.data
.iter()
.map(|&(re, im)| complex_magnitude(re, im))
.collect::<Vec<_>>();
let tensor =
Tensor::new(data, ct.shape.clone()).map_err(|e| builtin_error(format!("abs: {e}")))?;
Ok(tensor::tensor_into_value(tensor))
}
fn abs_char_array(ca: CharArray) -> BuiltinResult<Value> {
let data = ca
.data
.iter()
.map(|&ch| ch as u32 as f64)
.collect::<Vec<_>>();
let tensor = Tensor::new(data, vec![ca.rows, ca.cols])
.map_err(|e| builtin_error(format!("abs: {e}")))?;
Ok(tensor::tensor_into_value(tensor))
}
#[inline]
fn complex_magnitude(re: f64, im: f64) -> f64 {
re.hypot(im)
}
#[cfg(test)]
pub(crate) mod tests {
use super::*;
use crate::builtins::common::test_support;
use futures::executor::block_on;
use runmat_builtins::{IntValue, ResolveContext, Tensor, Type};
fn abs_builtin(value: Value) -> BuiltinResult<Value> {
block_on(super::abs_builtin(value))
}
#[test]
fn abs_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 abs_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 abs_scalar_negative() {
let result = abs_builtin(Value::Num(-3.5)).expect("abs");
match result {
Value::Num(n) => assert!((n - 3.5).abs() < 1e-12),
other => panic!("expected scalar result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn abs_int_promotes() {
let result = abs_builtin(Value::Int(IntValue::I32(-8))).expect("abs");
match result {
Value::Num(n) => assert!((n - 8.0).abs() < 1e-12),
other => panic!("expected scalar result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn abs_tensor_elements() {
let tensor = Tensor::new(vec![-1.0, -2.0, 3.0, -4.0], vec![2, 2]).unwrap();
let result = abs_builtin(Value::Tensor(tensor)).expect("abs");
match result {
Value::Tensor(t) => {
assert_eq!(t.shape, vec![2, 2]);
assert_eq!(t.data, vec![1.0, 2.0, 3.0, 4.0]);
}
other => panic!("expected tensor result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn abs_complex_scalar() {
let result = abs_builtin(Value::Complex(3.0, 4.0)).expect("abs");
match result {
Value::Num(n) => assert!((n - 5.0).abs() < 1e-12),
other => panic!("expected scalar result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn abs_complex_tensor_to_real_tensor() {
let complex = ComplexTensor::new(vec![(3.0, 4.0), (1.0, -1.0)], vec![2, 1]).unwrap();
let result = abs_builtin(Value::ComplexTensor(complex)).expect("abs");
match result {
Value::Tensor(t) => {
assert_eq!(t.shape, vec![2, 1]);
assert!((t.data[0] - 5.0).abs() < 1e-12);
assert!((t.data[1] - (2f64).sqrt()).abs() < 1e-12);
}
other => panic!("expected tensor result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn abs_char_array_codes() {
let char_array = CharArray::new("Az".chars().collect(), 1, 2).unwrap();
let result = abs_builtin(Value::CharArray(char_array)).expect("abs");
match result {
Value::Tensor(t) => {
assert_eq!(t.shape, vec![1, 2]);
assert_eq!(t.data, vec![65.0, 122.0]);
}
other => panic!("expected tensor result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn abs_string_rejected() {
let err = abs_builtin(Value::from("hello")).expect_err("should error");
assert!(err.message().contains("expected numeric"));
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn abs_gpu_provider_roundtrip() {
test_support::with_test_provider(|provider| {
let tensor = Tensor::new(vec![-2.0, -1.0, 0.0, 3.0], vec![4, 1]).unwrap();
let view = runmat_accelerate_api::HostTensorView {
data: &tensor.data,
shape: &tensor.shape,
};
let handle = provider.upload(&view).expect("upload");
let result = abs_builtin(Value::GpuTensor(handle)).expect("abs");
let gathered = test_support::gather(result).expect("gather");
assert_eq!(gathered.shape, vec![4, 1]);
assert_eq!(gathered.data, vec![2.0, 1.0, 0.0, 3.0]);
});
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
#[cfg(feature = "wgpu")]
fn abs_wgpu_matches_cpu_elementwise() {
let _ = runmat_accelerate::backend::wgpu::provider::register_wgpu_provider(
runmat_accelerate::backend::wgpu::provider::WgpuProviderOptions::default(),
);
let tensor = Tensor::new(vec![-3.0, -1.0, 0.5, -0.25], vec![4, 1]).unwrap();
let cpu = abs_real(Value::Tensor(tensor.clone())).unwrap();
let view = runmat_accelerate_api::HostTensorView {
data: &tensor.data,
shape: &tensor.shape,
};
let h = runmat_accelerate_api::provider()
.unwrap()
.upload(&view)
.unwrap();
let gpu = block_on(abs_gpu(h)).unwrap();
let gathered = test_support::gather(gpu).expect("gather");
match (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, "|{} - {}| >= {}", a, b, tol);
}
}
_ => panic!("unexpected result shape"),
}
}
}