Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Parse Floating Point Constants #2195

Open
wants to merge 3 commits into
base: calyx-float
Choose a base branch
from
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
24 changes: 24 additions & 0 deletions calyx-frontend/src/parser.rs
Original file line number Diff line number Diff line change
Expand Up @@ -294,6 +294,29 @@ impl CalyxParser {
u64::from_str_radix(input.as_str(), 2)
.map_err(|_| input.error("Expected binary number"))
}
fn float(input: Node) -> ParseResult<f64> {
let float_str = input.as_str();
let float_val: f64 = float_str
.parse()
.map_err(|_| input.error("Expected valid floating-point number"))?;
Ok(float_val)
}

fn ieee754_const(input: Node) -> ParseResult<BitNum> {
println!("parsing iee754");
let span = Self::get_span(&input);
let val = match_nodes!(
input.clone().into_children();
[float(val)] => val
);
let bit_pattern = val.to_bits();
Ok(BitNum {
width: 64,
num_type: NumType::Hex,
val: bit_pattern,
span,
})
}

fn num_lit(input: Node) -> ParseResult<BitNum> {
let span = Self::get_span(&input);
Expand Down Expand Up @@ -323,6 +346,7 @@ impl CalyxParser {
val,
span
},
[ieee754_const(val)] => val,

);

Expand Down
9 changes: 7 additions & 2 deletions calyx-frontend/src/syntax.pest
Original file line number Diff line number Diff line change
Expand Up @@ -16,17 +16,22 @@ binary = @{ ASCII_HEX_DIGIT+ }
decimal = @{ ASCII_HEX_DIGIT+ }
octal = @{ ASCII_HEX_DIGIT+ }
hex = @{ ASCII_HEX_DIGIT+ }
float = @{ ASCII_DIGIT+ ~ "." ~ ASCII_DIGIT+ }

ieee754_const = { "ieee754_const(" ~ float ~ ")" }
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

separate out iee754_const( into "ieee754" ~ "("


// `$` creates a compound rule which ignores whitespace while allowing for
// inner rules (`@` makes inner rules silent).
// See: https://pest.rs/book/print.html#atomic
num_lit = ${
bitwidth
(bitwidth
~ "'"
~ ( "d" ~ decimal
| "b" ~ binary
| "x" ~ hex
| "o" ~ octal )
| "o" ~ octal)
)
| ieee754_const
}

char = { !"\"" ~ ANY }
Expand Down
21 changes: 21 additions & 0 deletions calyx-py/calyx/numeric_types.py
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,7 @@
from decimal import Decimal, getcontext
import math
import logging as log
import struct


class InvalidNumericType(Exception):
Expand Down Expand Up @@ -335,3 +336,23 @@ def bitnum_to_fixed(bitnum: Bitnum, int_width: int) -> FixedPoint:
int_width=int_width,
is_signed=bitnum.is_signed,
)


@dataclass
class FloatingPoint(NumericType):
"""Represents a floating point number."""

def __init__(self, value: str, width: int, is_signed: bool):
super().__init__(value, width, is_signed)

if self.bit_string_repr is None and self.hex_string_repr is None:
# The decimal representation was passed in.
packed = struct.pack('!f', float(self.string_repr))
unpacked = struct.unpack('>I', packed)[0]
self.bit_string_repr = f'{unpacked:0{self.width}b}'
self.uint_repr = int(self.bit_string_repr, 2)
self.hex_string_repr = np.base_repr(self.uint_repr, 16)

def to_dec(self, round_place: int):
float_value = struct.unpack('!f', int(self.bit_string_repr, 2).to_bytes(4, byteorder='big'))[0]
return round(float_value, round_place)
28 changes: 18 additions & 10 deletions fud/fud/stages/verilator/json_to_dat.py
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
import simplejson as sjson
import numpy as np
from calyx.numeric_types import FixedPoint, Bitnum, InvalidNumericType
from calyx.numeric_types import FixedPoint, Bitnum, FloatingPoint, InvalidNumericType
from pathlib import Path
from fud.errors import Malformed
import logging as log
Expand All @@ -14,7 +14,7 @@ def float_to_fixed(value: float, N: int) -> float:
return round(value * w) / float(w)


def parse_dat(path, args):
def parse_dat(path, is_bn, args):
"""Parses a number with the given numeric type
arguments from the array at the given `path`.
"""
Expand All @@ -34,12 +34,15 @@ def parse(hex_value: str):
hex_value = f"0x{hex_value}"
if "int_width" in args:
return FixedPoint(hex_value, **args).str_value()
else:
elif is_bn:
bn = Bitnum(hex_value, **args)
if bn.is_undef:
return bn.str_value()
else:
return int(bn.str_value())
else:
fp = FloatingPoint(hex_value, **args)
return fp.to_dec(round_place=2)

with path.open("r") as f:
lines = []
Expand Down Expand Up @@ -90,11 +93,14 @@ def provided(x, y):
)


def convert(x, round: bool, is_signed: bool, width: int, int_width=None):
def convert(x, round: bool, is_signed: bool, width: int, is_bn: bool, int_width=None):
with_prefix = False
# If `int_width` is not defined, then this is a `Bitnum`
if int_width is None:
return Bitnum(x, width, is_signed).hex_string(with_prefix)
if is_bn:
return Bitnum(x, width, is_signed).hex_string(with_prefix)
else:
return FloatingPoint(x, width, is_signed).hex_string(with_prefix)

try:
return FixedPoint(x, width, int_width, is_signed).hex_string(with_prefix)
Expand Down Expand Up @@ -133,14 +139,14 @@ def convert2dat(output_dir, data, extension, round: bool):
numeric_type = format["numeric_type"]
is_signed = format["is_signed"]

if numeric_type not in {"bitnum", "fixed_point"}:
raise InvalidNumericType('Fud only supports "fixed_point" and "bitnum".')
if numeric_type not in {"bitnum", "fixed_point", "floating_point"}:
raise InvalidNumericType('Fud only supports "fixed_point", "bitnum", and "floating_point".')

is_fp = numeric_type == "fixed_point"
if is_fp:
width, int_width = parse_fp_widths(format)
else:
# `Bitnum`s only have a bit width.
# `Bitnum`s and `FloatingPoint`s only have a bit width
width = format["width"]
int_width = None

Expand All @@ -154,7 +160,8 @@ def convert2dat(output_dir, data, extension, round: bool):

with path.open("w") as f:
for v in arr.flatten():
f.write(convert(v, round, is_signed, width, int_width) + "\n")
is_bn=numeric_type == "bitnum"
f.write(convert(v, round, is_signed, width, is_bn, int_width) + "\n")

shape[k]["shape"] = list(arr.shape)
shape[k]["numeric_type"] = numeric_type
Expand Down Expand Up @@ -185,8 +192,9 @@ def convert2json(input_dir, extension):
# for building the FixedPoint or Bitnum classes.
args = form.copy()
del args["shape"]
is_bn = args["numeric_type"] == "bitnum"
del args["numeric_type"]
arr = parse_dat(path, args)
arr = parse_dat(path, is_bn, args)
if form["shape"] == [0]:
raise Malformed(
"Data format shape",
Expand Down
22 changes: 22 additions & 0 deletions tests/correctness/float/float-const.data
Original file line number Diff line number Diff line change
@@ -0,0 +1,22 @@
{
"mem_read": {
"data": [
4.2
],
"format": {
"is_signed": true,
"numeric_type": "floating_point",
"width": 32
}
},
"mem_write": {
"data": [
0.0
],
"format": {
"is_signed": true,
"numeric_type": "floating_point",
"width": 32
}
}
}
22 changes: 22 additions & 0 deletions tests/correctness/float/float-const.expect
Original file line number Diff line number Diff line change
@@ -0,0 +1,22 @@
{
"mem_read": {
"data": [
4.2
],
"format": {
"is_signed": true,
"numeric_type": "floating_point",
"width": 32
}
},
"mem_write": {
"data": [
4.2
Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Oh the second thought is that it only proves that it reads/writes the same value, but doesn't prove that it's not necessarily intepretting 4.2 as the decimal representation..

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you explain what you mean by that? The comment is lacking enough context to make sense to me.

],
"format": {
"is_signed": true,
"numeric_type": "floating_point",
"width": 32
}
}
}
31 changes: 31 additions & 0 deletions tests/correctness/float/float-const.futil
Original file line number Diff line number Diff line change
@@ -0,0 +1,31 @@
import "primitives/compile.futil";
import "primitives/memories/comb.futil";

component main<"toplevel"=1,>(@clk clk: 1, @reset reset: 1, @go go: 1) -> (@done done: 1) {
cells {
reg0 = std_reg(32);
@external mem_read = comb_mem_d1(32, 1, 1);
@external mem_write = comb_mem_d1(32, 1, 1);
}
wires {
group read {
mem_read.addr0 = 1'b0;
reg0.in = ieee754_const(1.00);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@sampsyo thoughts on this representation as opposed to the idea of only having a primitive called ieee754 and using its output port?

reg0.write_en = 1'b1;
read[done] = reg0.done;
}

group write {
mem_write.addr0 = 1'b0;
mem_write.write_en = 1'b1;
mem_write.write_data = reg0.out;
write[done] = mem_write.done;
}
}
control {
seq {
read;
write;
}
}
}