use runmat_accelerate_api::{GpuTensorHandle, HostTensorView};
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::floor")]
pub const GPU_SPEC: BuiltinGpuSpec = BuiltinGpuSpec {
name: "floor",
op_kind: GpuOpKind::Elementwise,
supported_precisions: &[ScalarType::F32, ScalarType::F64],
broadcast: BroadcastSemantics::Matlab,
provider_hooks: &[ProviderHook::Unary { name: "unary_floor" }],
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 floor directly on the device; the runtime gathers to the host when unary_floor is unavailable.",
};
#[runmat_macros::register_fusion_spec(builtin_path = "crate::builtins::math::rounding::floor")]
pub const FUSION_SPEC: BuiltinFusionSpec = BuiltinFusionSpec {
name: "floor",
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!("floor({input})"))
},
}),
reduction: None,
emits_nan: false,
notes: "Fusion planner emits WGSL `floor` calls; providers can substitute custom kernels when available.",
};
const BUILTIN_NAME: &str = "floor";
fn builtin_error(message: impl Into<String>) -> RuntimeError {
build_runtime_error(message)
.with_builtin(BUILTIN_NAME)
.build()
}
#[runtime_builtin(
name = "floor",
category = "math/rounding",
summary = "Round values toward negative infinity.",
keywords = "floor,rounding,integers,gpu",
accel = "unary",
type_resolver(numeric_unary_type),
builtin_path = "crate::builtins::math::rounding::floor"
)]
async fn floor_builtin(value: Value, rest: Vec<Value>) -> BuiltinResult<Value> {
let args = parse_arguments(&rest)?;
let base = match value {
Value::GpuTensor(handle) => floor_gpu(handle, &args).await?,
Value::Complex(re, im) => Value::Complex(
apply_floor_scalar(re, args.strategy),
apply_floor_scalar(im, args.strategy),
),
Value::ComplexTensor(ct) => floor_complex_tensor(ct, args.strategy)?,
Value::CharArray(ca) => floor_char_array(ca, args.strategy)?,
Value::LogicalArray(logical) => {
let tensor = tensor::logical_to_tensor(&logical).map_err(|err| builtin_error(err))?;
let floored = floor_tensor(tensor, args.strategy)?;
tensor::tensor_into_value(floored)
}
Value::String(_) | Value::StringArray(_) => {
return Err(builtin_error("floor: expected numeric or logical input"));
}
other => floor_numeric(other, args.strategy)?,
};
apply_output_template(base, &args.output).await
}
fn floor_numeric(value: Value, strategy: FloorStrategy) -> BuiltinResult<Value> {
let tensor = tensor::value_into_tensor_for("floor", value).map_err(|err| builtin_error(err))?;
let floored = floor_tensor(tensor, strategy)?;
Ok(tensor::tensor_into_value(floored))
}
fn floor_tensor(mut tensor: Tensor, strategy: FloorStrategy) -> BuiltinResult<Tensor> {
for value in &mut tensor.data {
*value = apply_floor_scalar(*value, strategy);
}
Ok(tensor)
}
fn floor_complex_tensor(ct: ComplexTensor, strategy: FloorStrategy) -> BuiltinResult<Value> {
let data: Vec<(f64, f64)> = ct
.data
.iter()
.map(|&(re, im)| {
(
apply_floor_scalar(re, strategy),
apply_floor_scalar(im, strategy),
)
})
.collect();
let tensor = ComplexTensor::new(data, ct.shape.clone())
.map_err(|e| builtin_error(format!("floor: {e}")))?;
Ok(Value::ComplexTensor(tensor))
}
fn floor_char_array(ca: CharArray, strategy: FloorStrategy) -> BuiltinResult<Value> {
let mut data = Vec::with_capacity(ca.data.len());
for ch in ca.data {
data.push(apply_floor_scalar(ch as u32 as f64, strategy));
}
let tensor = Tensor::new(data, vec![ca.rows, ca.cols])
.map_err(|e| builtin_error(format!("floor: {e}")))?;
Ok(Value::Tensor(tensor))
}
async fn floor_gpu(handle: GpuTensorHandle, args: &FloorArgs) -> BuiltinResult<Value> {
if matches!(args.strategy, FloorStrategy::Integer) {
if let Some(provider) = runmat_accelerate_api::provider_for_handle(&handle) {
if let Ok(out) = provider.unary_floor(&handle).await {
return Ok(Value::GpuTensor(out));
}
}
}
let tensor = gpu_helpers::gather_tensor_async(&handle).await?;
let floored = floor_tensor(tensor, args.strategy)?;
Ok(tensor::tensor_into_value(floored))
}
#[derive(Clone, Copy, Debug, PartialEq, Eq)]
enum FloorStrategy {
Integer,
Decimals(i32),
Significant(i32),
}
#[derive(Clone, Debug)]
struct FloorArgs {
strategy: FloorStrategy,
output: OutputTemplate,
}
#[derive(Clone, Debug)]
enum OutputTemplate {
Default,
Like(Value),
}
fn parse_arguments(args: &[Value]) -> BuiltinResult<FloorArgs> {
let (strategy_len, output) = parse_output_template(args)?;
let strategy = match strategy_len {
0 => FloorStrategy::Integer,
1 => FloorStrategy::Decimals(parse_digits(&args[0])?),
2 => {
let digits = parse_digits(&args[0])?;
let mode = parse_mode(&args[1])?;
match mode {
FloorMode::Decimals => FloorStrategy::Decimals(digits),
FloorMode::Significant => {
if digits <= 0 {
return Err(builtin_error(
"floor: N must be a positive integer for 'significant' rounding",
));
}
FloorStrategy::Significant(digits)
}
}
}
_ => return Err(builtin_error("floor: too many input arguments")),
};
Ok(FloorArgs { strategy, output })
}
fn parse_output_template(args: &[Value]) -> BuiltinResult<(usize, OutputTemplate)> {
if !args.is_empty() && is_keyword(&args[args.len() - 1], "like") {
return Err(builtin_error("floor: expected prototype after 'like'"));
}
if args.len() >= 2 && is_keyword(&args[args.len() - 2], "like") {
let proto = &args[args.len() - 1];
if matches!(
proto,
Value::String(_) | Value::StringArray(_) | Value::CharArray(_)
) {
return Err(builtin_error("floor: unsupported prototype for 'like'"));
}
return Ok((args.len() - 2, OutputTemplate::Like(proto.clone())));
}
Ok((args.len(), OutputTemplate::Default))
}
fn parse_digits(value: &Value) -> BuiltinResult<i32> {
let err = || builtin_error("floor: N must be an integer scalar");
let raw = match value {
Value::Int(i) => i.to_i64(),
Value::Num(n) => {
if !n.is_finite() {
return Err(err());
}
let rounded = n.round();
if (rounded - n).abs() > f64::EPSILON {
return Err(err());
}
rounded as i64
}
Value::Bool(b) => {
if *b {
1
} else {
0
}
}
other => {
return Err(builtin_error(format!(
"floor: N must be numeric, got {:?}",
other
)))
}
};
if raw > i32::MAX as i64 || raw < i32::MIN as i64 {
return Err(builtin_error("floor: integer overflow in N"));
}
Ok(raw as i32)
}
#[derive(Clone, Copy, Debug, PartialEq, Eq)]
enum FloorMode {
Decimals,
Significant,
}
fn parse_mode(value: &Value) -> BuiltinResult<FloorMode> {
let Some(text) = tensor::value_to_string(value) else {
return Err(builtin_error(
"floor: mode must be a character vector or string scalar",
));
};
let lowered = text.trim().to_ascii_lowercase();
match lowered.as_str() {
"significant" => Ok(FloorMode::Significant),
"decimal" | "decimals" => Ok(FloorMode::Decimals),
other => Err(builtin_error(format!(
"floor: unknown rounding mode '{other}'"
))),
}
}
fn is_keyword(value: &Value, target: &str) -> bool {
tensor::value_to_string(value)
.map(|s| s.trim().eq_ignore_ascii_case(target))
.unwrap_or(false)
}
fn apply_floor_scalar(value: f64, strategy: FloorStrategy) -> f64 {
if !value.is_finite() {
return value;
}
match strategy {
FloorStrategy::Integer => value.floor(),
FloorStrategy::Decimals(digits) => floor_with_decimals(value, digits),
FloorStrategy::Significant(digits) => floor_with_significant(value, digits),
}
}
fn floor_with_decimals(value: f64, digits: i32) -> f64 {
if digits == 0 {
return value.floor();
}
let factor = 10f64.powi(digits);
if !factor.is_finite() || factor == 0.0 {
return value;
}
(value * factor).floor() / factor
}
fn floor_with_significant(value: f64, digits: i32) -> f64 {
if value == 0.0 {
return 0.0;
}
let abs_val = value.abs();
let order = abs_val.log10().floor();
let scale_power = digits - 1 - order as i32;
let scale = 10f64.powi(scale_power);
if !scale.is_finite() || scale == 0.0 {
return value;
}
(value * scale).floor() / scale
}
async fn apply_output_template(value: Value, output: &OutputTemplate) -> BuiltinResult<Value> {
match output {
OutputTemplate::Default => Ok(value),
OutputTemplate::Like(proto) => match proto {
Value::GpuTensor(_) => convert_to_gpu(value),
Value::Tensor(_)
| Value::Num(_)
| Value::Int(_)
| Value::Bool(_)
| Value::LogicalArray(_)
| Value::Complex(_, _)
| Value::ComplexTensor(_) => convert_to_host_like(value).await,
_ => Err(builtin_error(
"floor: unsupported prototype for 'like'; provide a numeric or gpuArray prototype",
)),
},
}
}
fn convert_to_gpu(value: Value) -> BuiltinResult<Value> {
let provider = runmat_accelerate_api::provider().ok_or_else(|| {
builtin_error(
"floor: GPU output requested via 'like' but no acceleration provider is active",
)
})?;
match value {
Value::GpuTensor(handle) => Ok(Value::GpuTensor(handle)),
Value::Tensor(tensor) => {
let view = HostTensorView {
data: &tensor.data,
shape: &tensor.shape,
};
let handle = provider
.upload(&view)
.map_err(|e| builtin_error(format!("floor: {e}")))?;
Ok(Value::GpuTensor(handle))
}
Value::Num(n) => {
let tensor = Tensor::new(vec![n], vec![1, 1])
.map_err(|e| builtin_error(format!("floor: {e}")))?;
convert_to_gpu(Value::Tensor(tensor))
}
Value::LogicalArray(logical) => {
let tensor = tensor::logical_to_tensor(&logical).map_err(|err| builtin_error(err))?;
convert_to_gpu(Value::Tensor(tensor))
}
other => Err(builtin_error(format!(
"floor: 'like' GPU prototypes are only supported for real numeric outputs (got {other:?})"
))),
}
}
async fn convert_to_host_like(value: Value) -> BuiltinResult<Value> {
match value {
Value::GpuTensor(handle) => {
let proxy = Value::GpuTensor(handle);
gpu_helpers::gather_value_async(&proxy).await
}
other => Ok(other),
}
}
#[cfg(test)]
pub(crate) mod tests {
use super::*;
use crate::builtins::common::test_support;
use crate::RuntimeError;
use futures::executor::block_on;
use runmat_accelerate_api::HostTensorView;
use runmat_builtins::{IntValue, LogicalArray, ResolveContext, Tensor, Type, Value};
fn floor_builtin(value: Value, rest: Vec<Value>) -> BuiltinResult<Value> {
block_on(super::floor_builtin(value, rest))
}
fn assert_error_contains(error: RuntimeError, needle: &str) {
assert!(
error.message().contains(needle),
"unexpected error: {}",
error.message()
);
}
#[test]
fn floor_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 floor_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 floor_scalar_positive_and_negative() {
let value = Value::Num(-2.7);
let result = floor_builtin(value, Vec::new()).expect("floor");
match result {
Value::Num(v) => assert_eq!(v, -3.0),
other => panic!("expected scalar result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn floor_integer_tensor() {
let tensor = Tensor::new(vec![1.2, 4.7, -3.4, 5.0], vec![2, 2]).unwrap();
let result = floor_builtin(Value::Tensor(tensor), Vec::new()).expect("floor");
match result {
Value::Tensor(t) => {
assert_eq!(t.shape, vec![2, 2]);
assert_eq!(t.data, vec![1.0, 4.0, -4.0, 5.0]);
}
other => panic!("expected tensor result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn floor_complex_value() {
let result = floor_builtin(Value::Complex(1.7, -2.3), Vec::new()).expect("floor");
match result {
Value::Complex(re, im) => {
assert_eq!(re, 1.0);
assert_eq!(im, -3.0);
}
other => panic!("expected complex result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn floor_char_array_to_tensor() {
let chars = CharArray::new("AB".chars().collect(), 1, 2).unwrap();
let result = floor_builtin(Value::CharArray(chars), Vec::new()).expect("floor");
match result {
Value::Tensor(t) => {
assert_eq!(t.shape, vec![1, 2]);
assert_eq!(t.data, vec![65.0, 66.0]);
}
other => panic!("expected tensor result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn floor_logical_array_remains_same() {
let logical = LogicalArray::new(vec![1, 0, 1, 1], vec![2, 2]).unwrap();
let result = floor_builtin(Value::LogicalArray(logical), Vec::new()).expect("floor");
match result {
Value::Tensor(t) => {
assert_eq!(t.shape, vec![2, 2]);
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 floor_int_value_passthrough() {
let result = floor_builtin(Value::Int(IntValue::I32(-4)), Vec::new()).expect("floor");
match result {
Value::Num(v) => assert_eq!(v, -4.0),
other => panic!("expected scalar result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn floor_gpu_provider_roundtrip() {
test_support::with_test_provider(|provider| {
let tensor = Tensor::new(vec![0.2, 1.9, -0.1, -3.8], vec![2, 2]).unwrap();
let view = HostTensorView {
data: &tensor.data,
shape: &tensor.shape,
};
let handle = provider.upload(&view).expect("upload");
let result = floor_builtin(Value::GpuTensor(handle), Vec::new()).expect("floor");
let gathered = test_support::gather(result).expect("gather");
assert_eq!(gathered.shape, vec![2, 2]);
assert_eq!(gathered.data, vec![0.0, 1.0, -1.0, -4.0]);
});
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn floor_decimal_digits() {
let value = Value::Num(21.456);
let args = vec![Value::Int(IntValue::I32(2))];
let result = floor_builtin(value, args).expect("floor");
match result {
Value::Num(v) => assert!((v - 21.45).abs() < 1e-12),
other => panic!("expected scalar result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn floor_negative_digits() {
let tensor = Tensor::new(vec![123.4, -987.6], vec![2, 1]).unwrap();
let args = vec![Value::Int(IntValue::I32(-2))];
let result = floor_builtin(Value::Tensor(tensor), args).expect("floor");
match result {
Value::Tensor(t) => assert_eq!(t.data, vec![100.0, -1000.0]),
other => panic!("expected tensor result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn floor_significant_digits() {
let value = Value::Num(98765.4321);
let args = vec![Value::Int(IntValue::I32(3)), Value::from("significant")];
let result = floor_builtin(value, args).expect("floor");
match result {
Value::Num(v) => assert_eq!(v, 98700.0),
other => panic!("expected scalar result, got {other:?}"),
}
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn floor_significant_requires_positive_digits() {
let args = vec![Value::Int(IntValue::I32(0)), Value::from("significant")];
let err = floor_builtin(Value::Num(1.23), args).unwrap_err();
assert_error_contains(err, "positive integer");
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn floor_string_input_errors() {
let err = floor_builtin(Value::from("hello"), Vec::new()).unwrap_err();
assert_error_contains(err, "numeric");
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn floor_like_invalid_prototype_errors() {
let args = vec![Value::from("like"), Value::from("prototype")];
let err = floor_builtin(Value::Num(1.0), args).unwrap_err();
assert_error_contains(err, "unsupported prototype");
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn floor_like_gpu_output() {
test_support::with_test_provider(|provider| {
let tensor = Tensor::new(vec![0.9, -1.2, 2.7, -3.4], vec![2, 2]).unwrap();
let like_proto = {
let proto = Tensor::new(vec![0.0], vec![1, 1]).unwrap();
let view = HostTensorView {
data: &proto.data,
shape: &proto.shape,
};
provider.upload(&view).expect("upload proto")
};
let args = vec![Value::from("like"), Value::GpuTensor(like_proto)];
let result = floor_builtin(Value::Tensor(tensor), args).expect("floor");
match result {
Value::GpuTensor(handle) => {
let gathered = test_support::gather(Value::GpuTensor(handle)).expect("gather");
assert_eq!(gathered.shape, vec![2, 2]);
assert_eq!(gathered.data, vec![0.0, -2.0, 2.0, -4.0]);
}
other => panic!("expected GPU tensor, got {other:?}"),
}
});
}
#[cfg_attr(target_arch = "wasm32", wasm_bindgen_test::wasm_bindgen_test)]
#[test]
fn floor_bool_value() {
let result = floor_builtin(Value::Bool(true), Vec::new()).expect("floor");
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]
#[cfg(feature = "wgpu")]
fn floor_wgpu_matches_cpu() {
let _ = runmat_accelerate::backend::wgpu::provider::register_wgpu_provider(
runmat_accelerate::backend::wgpu::provider::WgpuProviderOptions::default(),
);
let t = Tensor::new(vec![0.3, 1.1, -0.2, -1.7], vec![2, 2]).unwrap();
let cpu = floor_numeric(Value::Tensor(t.clone()), FloorStrategy::Integer).unwrap();
let view = HostTensorView {
data: &t.data,
shape: &t.shape,
};
let h = runmat_accelerate_api::provider()
.unwrap()
.upload(&view)
.unwrap();
let gpu = block_on(floor_gpu(
h,
&FloorArgs {
strategy: FloorStrategy::Integer,
output: OutputTemplate::Default,
},
))
.unwrap();
let gathered = test_support::gather(gpu).expect("gather");
match (cpu, gathered) {
(Value::Tensor(ct), gt) => {
assert_eq!(gt.shape, ct.shape);
assert_eq!(gt.data, ct.data);
}
(Value::Num(c), gt) => {
assert_eq!(gt.data, vec![c]);
}
other => panic!("unexpected comparison {other:?}"),
}
}
}