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::rounding::fix")]
pub const GPU_SPEC: BuiltinGpuSpec = BuiltinGpuSpec {
name: "fix",
op_kind: GpuOpKind::Elementwise,
supported_precisions: &[ScalarType::F32, ScalarType::F64],
broadcast: BroadcastSemantics::Matlab,
provider_hooks: &[ProviderHook::Unary { name: "unary_fix" }],
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 implement unary_fix to keep fix on device; otherwise the runtime gathers to host and applies CPU truncation.",
};
#[runmat_macros::register_fusion_spec(builtin_path = "crate::builtins::math::rounding::fix")]
pub const FUSION_SPEC: BuiltinFusionSpec = BuiltinFusionSpec {
name: "fix",
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)),
};
let truncated = format!("trunc({input})");
Ok(format!("select({0}, {1}, {0} == {1})", truncated, zero))
},
}),
reduction: None,
emits_nan: false,
notes: "Fusion planner emits WGSL truncation; providers can substitute custom kernels when unary_fix is available.",
};
const BUILTIN_NAME: &str = "fix";
fn builtin_error(message: impl Into<String>) -> RuntimeError {
build_runtime_error(message)
.with_builtin(BUILTIN_NAME)
.build()
}
#[runtime_builtin(
name = "fix",
category = "math/rounding",
summary = "Round scalars, vectors, matrices, or N-D tensors toward zero.",
keywords = "fix,truncate,rounding,toward zero,gpu",
accel = "unary",
type_resolver(numeric_unary_type),
builtin_path = "crate::builtins::math::rounding::fix"
)]
async fn fix_builtin(value: Value) -> BuiltinResult<Value> {
match value {
Value::GpuTensor(handle) => fix_gpu(handle).await,
Value::Complex(re, im) => Ok(Value::Complex(fix_scalar(re), fix_scalar(im))),
Value::ComplexTensor(ct) => fix_complex_tensor(ct),
Value::CharArray(ca) => fix_char_array(ca),
Value::LogicalArray(logical) => {
let tensor = tensor::logical_to_tensor(&logical).map_err(|err| builtin_error(err))?;
fix_tensor(tensor).map(tensor::tensor_into_value)
}
Value::String(_) | Value::StringArray(_) => {
Err(builtin_error("fix: expected numeric or logical input"))
}
other => fix_numeric(other),
}
}
async fn fix_gpu(handle: GpuTensorHandle) -> BuiltinResult<Value> {
if let Some(provider) = runmat_accelerate_api::provider_for_handle(&handle) {
if let Ok(out) = provider.unary_fix(&handle).await {
return Ok(gpu_helpers::resident_gpu_value(out));
}
}
let tensor = gpu_helpers::gather_tensor_async(&handle).await?;
fix_tensor(tensor).map(tensor::tensor_into_value)
}
fn fix_numeric(value: Value) -> BuiltinResult<Value> {
match value {
Value::Num(n) => Ok(Value::Num(fix_scalar(n))),
Value::Int(i) => Ok(Value::Num(fix_scalar(i.to_f64()))),
Value::Bool(b) => Ok(Value::Num(fix_scalar(if b { 1.0 } else { 0.0 }))),
Value::Tensor(t) => fix_tensor(t).map(tensor::tensor_into_value),
other => {
let tensor =
tensor::value_into_tensor_for("fix", other).map_err(|err| builtin_error(err))?;
Ok(fix_tensor(tensor).map(tensor::tensor_into_value)?)
}
}
}
fn fix_tensor(mut tensor: Tensor) -> BuiltinResult<Tensor> {
for value in &mut tensor.data {
*value = fix_scalar(*value);
}
Ok(tensor)
}
fn fix_complex_tensor(ct: ComplexTensor) -> BuiltinResult<Value> {
let data = ct
.data
.iter()
.map(|&(re, im)| (fix_scalar(re), fix_scalar(im)))
.collect::<Vec<_>>();
let tensor = ComplexTensor::new(data, ct.shape.clone())
.map_err(|e| builtin_error(format!("fix: {e}")))?;
Ok(Value::ComplexTensor(tensor))
}
fn fix_char_array(ca: CharArray) -> BuiltinResult<Value> {
let data = ca
.data
.iter()
.map(|&ch| fix_scalar(ch as u32 as f64))
.collect::<Vec<_>>();
let tensor = Tensor::new(data, vec![ca.rows, ca.cols])
.map_err(|e| builtin_error(format!("fix: {e}")))?;
Ok(Value::Tensor(tensor))
}
fn fix_scalar(value: f64) -> f64 {
if !value.is_finite() {
return value;
}
let truncated = value.trunc();
if truncated == 0.0 {
0.0
} else {
truncated
}
}
#[cfg(test)]
pub(crate) mod tests {
use super::*;
use crate::builtins::common::test_support;
use crate::RuntimeError;
use futures::executor::block_on;
use runmat_builtins::{ComplexTensor, IntValue, LogicalArray, ResolveContext, Type};
fn fix_builtin(value: Value) -> BuiltinResult<Value> {
block_on(super::fix_builtin(value))
}
fn assert_error_contains(error: RuntimeError, needle: &str) {
assert!(
error.message().contains(needle),
"unexpected error: {}",
error.message()
);
}
#[test]
fn fix_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 fix_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 fix_scalar_positive_and_negative() {
let input = Value::Tensor(
Tensor::new(vec![-3.7, -2.4, -0.6, 0.0, 0.6, 2.4, 3.7], vec![7, 1]).unwrap(),
);
let result = fix_builtin(input).expect("fix");
match result {
Value::Tensor(t) => {
assert_eq!(t.data, vec![-3.0, -2.0, 0.0, 0.0, 0.0, 2.0, 3.0]);
}
other => panic!("expected tensor result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn fix_tensor_matrix() {
let tensor = Tensor::new(vec![1.9, 4.1, -2.8, 0.5], vec![2, 2]).unwrap();
let result = fix_builtin(Value::Tensor(tensor)).expect("fix");
match result {
Value::Tensor(t) => {
assert_eq!(t.shape, vec![2, 2]);
assert_eq!(t.data, vec![1.0, 4.0, -2.0, 0.0]);
}
other => panic!("expected tensor result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn fix_complex_number() {
let result = fix_builtin(Value::Complex(1.9, -2.2)).expect("fix");
match result {
Value::Complex(re, im) => {
assert_eq!(re, 1.0);
assert_eq!(im, -2.0);
}
other => panic!("expected complex result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn fix_char_array_returns_numeric_tensor() {
let chars = CharArray::new("ABC".chars().collect(), 1, 3).unwrap();
let result = fix_builtin(Value::CharArray(chars)).expect("fix");
match result {
Value::Tensor(t) => {
assert_eq!(t.shape, vec![1, 3]);
assert_eq!(t.data, vec![65.0, 66.0, 67.0]);
}
other => panic!("expected tensor result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn fix_logical_array() {
let logical = LogicalArray::new(vec![1, 0, 1, 1], vec![2, 2]).unwrap();
let result = fix_builtin(Value::LogicalArray(logical)).expect("fix");
match result {
Value::Tensor(t) => assert_eq!(t.data, vec![1.0, 0.0, 1.0, 1.0]),
other => panic!("expected tensor result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn fix_bool_promotes_to_numeric() {
let result = fix_builtin(Value::Bool(true)).expect("fix");
match result {
Value::Num(v) => assert_eq!(v, 1.0),
other => panic!("expected scalar result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn fix_int_value_promotes() {
let value = Value::Int(IntValue::I32(-42));
let result = fix_builtin(value).expect("fix");
match result {
Value::Num(v) => assert_eq!(v, -42.0),
other => panic!("expected scalar result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn fix_string_errors() {
let err = fix_builtin(Value::from("abc")).unwrap_err();
assert_error_contains(err, "expected numeric");
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn fix_preserves_special_values_and_canonicalizes_negative_zero() {
let tensor = Tensor::new(
vec![f64::NAN, f64::INFINITY, f64::NEG_INFINITY, -0.0],
vec![4, 1],
)
.unwrap();
let result = fix_builtin(Value::Tensor(tensor)).expect("fix");
let Value::Tensor(out) = result else {
panic!("expected tensor result");
};
assert!(out.data[0].is_nan(), "NaN should propagate");
assert_eq!(out.data[1], f64::INFINITY);
assert_eq!(out.data[2], f64::NEG_INFINITY);
assert_eq!(out.data[3], 0.0);
assert!(
out.data[3].is_sign_positive(),
"negative zero should canonicalize to +0"
);
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn fix_complex_tensor_rounds_components() {
let tensor = ComplexTensor::new(vec![(1.9, -2.6), (-3.4, 0.2)], vec![2, 1]).unwrap();
let result = fix_builtin(Value::ComplexTensor(tensor)).expect("fix");
let Value::ComplexTensor(out) = result else {
panic!("expected complex tensor result");
};
assert_eq!(out.shape, vec![2, 1]);
assert_eq!(out.data, vec![(1.0, -2.0), (-3.0, 0.0)]);
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn fix_gpu_provider_roundtrip() {
test_support::with_test_provider(|provider| {
let tensor = Tensor::new(vec![-1.9, -0.1, 0.1, 2.6], 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 = fix_builtin(Value::GpuTensor(handle)).expect("fix");
let gathered = test_support::gather(result).expect("gather");
assert_eq!(gathered.data, vec![-1.0, 0.0, 0.0, 2.0]);
});
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
#[cfg(feature = "wgpu")]
fn fix_wgpu_matches_cpu() {
let _ = runmat_accelerate::backend::wgpu::provider::register_wgpu_provider(
runmat_accelerate::backend::wgpu::provider::WgpuProviderOptions::default(),
);
let tensor = Tensor::new(vec![-3.7, -0.4, 0.4, 3.7], vec![4, 1]).unwrap();
let cpu = fix_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(fix_gpu(handle)).unwrap();
let gathered = test_support::gather(gpu).expect("gather");
assert_eq!(gathered.shape, cpu.shape);
assert_eq!(gathered.data, cpu.data);
}
}