Taking the sqrt of an integer shouldn't require an explicit cast.

[dlangBugzillaToGithub]

dsimcha (@dsimcha) reported this on 2010-07-13T13:56:00Z

Transfered from https://issues.dlang.org/show_bug.cgi?id=4455

CC List

• bearophile_hugs
• dlang (@Vild)

Description

This one's been around for ages, but has just now started bothering me enough to file a bug report:

import std.math;

void main() {
    uint num = 1;
    auto ans = sqrt(num);
}

Error:
test.d(5): Error: function std.math.sqrt called with argument types:
	((uint))
matches both:
	std.math.sqrt(float x)
and:
	std.math.sqrt(real x)

I guess the solution is to make long and ulong overloads that forward to std.math.sqrt(real x).

[dlangBugzillaToGithub]

bearophile_hugs commented on 2010-07-13T14:25:18Z

Partially unrelated: an efficient D compiler can desire to implement the sqrt with SSE instructions like SQRTSS RSQRTSS SQRTPS and RSQRTPS, that have floats or doubles in input or output. So I think a double sqrt(double) too can be useful, to avoid the useless argument passing of 10-12-16 bytes (necessary for an argument of type real) for the computation of sqrt on a double.

[dlangBugzillaToGithub]

dsimcha commented on 2010-08-11T19:45:25Z

Fixed SVN.

[dlangBugzillaToGithub]

dlang (@Vild) commented on 2019-09-12T07:46:28Z

Still there, or crept in again.