www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - How to detect overflow

reply Namal <sotis22 mail.ru> writes:
Is there a way to detect overflow for example for:

	int i = 2_000_000_000;
	
	int a = i*i*i;
	
	writeln(a);

-> 1073741824
Nov 03 2015
parent reply BBasile <bb.temp gmx.com> writes:
On Wednesday, 4 November 2015 at 03:55:13 UTC, Namal wrote:

 Is there a way to detect overflow for example for:

 	int i = 2_000_000_000;
 	
 	int a = i*i*i;
 	
 	writeln(a);

 -> 1073741824


You can use core.checkedint [1]

---

http://dlang.org/phobos/core_checkedint.html
Nov 03 2015
next sibling parent reply Namal <sotis22 mail.ru> writes:
On Wednesday, 4 November 2015 at 04:22:03 UTC, BBasile wrote:

 On Wednesday, 4 November 2015 at 03:55:13 UTC, Namal wrote:

 Is there a way to detect overflow for example for:

 	int i = 2_000_000_000;
 	
 	int a = i*i*i;
 	
 	writeln(a);

 -> 1073741824


 You can use core.checkedint [1]

 ---

 http://dlang.org/phobos/core_checkedint.html


Is it just an error in the documentation that the return value is 
stated as sum for the multiplication functions?
Nov 03 2015
parent reply =?UTF-8?Q?Ali_=c3=87ehreli?= <acehreli yahoo.com> writes:
On 11/03/2015 10:34 PM, Namal wrote:


 http://dlang.org/phobos/core_checkedint.html


 Is it just an error in the documentation that the return value is stated
 as sum for the multiplication functions?


Yeah, looks like classic copy-paste errors. :)

Ali
Nov 03 2015
parent reply BBasile <bb.temp gmx.com> writes:
On Wednesday, 4 November 2015 at 07:19:09 UTC, Ali Çehreli wrote:

 On 11/03/2015 10:34 PM, Namal wrote:


 http://dlang.org/phobos/core_checkedint.html


 Is it just an error in the documentation that the return value 
 is stated
 as sum for the multiplication functions?


 Yeah, looks like classic copy-paste errors. :)

 Ali


I take the token for this ddoc fix:

https://github.com/D-Programming-Language/druntime/pull/1429
Nov 03 2015
parent reply =?UTF-8?Q?Ali_=c3=87ehreli?= <acehreli yahoo.com> writes:
On 11/03/2015 11:34 PM, BBasile wrote:

 On Wednesday, 4 November 2015 at 07:19:09 UTC, Ali Çehreli wrote:

 On 11/03/2015 10:34 PM, Namal wrote:


 http://dlang.org/phobos/core_checkedint.html


 Is it just an error in the documentation that the return value is stated
 as sum for the multiplication functions?


 Yeah, looks like classic copy-paste errors. :)

 Ali


 I take the token for this ddoc fix:

 https://github.com/D-Programming-Language/druntime/pull/1429


Thanks. I've noticed that the parameter of the subtraction functions 
should be named 'underflow', no?

Ali
Nov 04 2015
parent reply Ola Fosheim =?UTF-8?B?R3LDuHN0YWQ=?= writes:
On Wednesday, 4 November 2015 at 08:18:00 UTC, Ali Çehreli wrote:

 Thanks. I've noticed that the parameter of the subtraction 
 functions should be named 'underflow', no?


Integer math cannot underflow, unless you define division to be 
equivalent to division over reals.

overflow => higher/lower than max/min

underflow => wrongly rounded to zero  (smaller than epsilon)

Underflow can lead to unexpected division by zero errors in float 
expressions.
Nov 04 2015
parent reply =?UTF-8?Q?Ali_=c3=87ehreli?= <acehreli yahoo.com> writes:
On 11/04/2015 02:01 AM, Ola Fosheim Grøstad wrote:

 On Wednesday, 4 November 2015 at 08:18:00 UTC, Ali Çehreli wrote:

 Thanks. I've noticed that the parameter of the subtraction functions
 should be named 'underflow', no?


 Integer math cannot underflow, unless you define division to be
 equivalent to division over reals.

 overflow => higher/lower than max/min

 underflow => wrongly rounded to zero  (smaller than epsilon)

 Underflow can lead to unexpected division by zero errors in float
 expressions.


Thanks. It looks like I've been making stuff up on this page: :(

   http://ddili.org/ders/d.en/arithmetic.html

Ali
Nov 04 2015
parent Ola Fosheim =?UTF-8?B?R3LDuHN0YWQ=?= writes:
On Wednesday, 4 November 2015 at 10:06:47 UTC, Ali Çehreli wrote:

 Thanks. It looks like I've been making stuff up on this page: :(

   http://ddili.org/ders/d.en/arithmetic.html


It's a common source for confusion, the word "underflow" is a bit 
misleading. Maybe better to use the term "zero-flushed".
Nov 04 2015
prev sibling parent reply Namal <sotis22 mail.ru> writes:
On Wednesday, 4 November 2015 at 04:22:03 UTC, BBasile wrote:

 On Wednesday, 4 November 2015 at 03:55:13 UTC, Namal wrote:

 Is there a way to detect overflow for example for:

 	int i = 2_000_000_000;
 	
 	int a = i*i*i;
 	
 	writeln(a);

 -> 1073741824


 You can use core.checkedint [1]

 ---

 http://dlang.org/phobos/core_checkedint.html


It says:


"The overflow is sticky, meaning a sequence of operations can be 
done and overflow need only be checked at the end."

But how can I make multiple operations? I can only put 2 values 
in the function.
Nov 03 2015
parent reply =?UTF-8?Q?Ali_=c3=87ehreli?= <acehreli yahoo.com> writes:
On 11/03/2015 11:52 PM, Namal wrote:


 http://dlang.org/phobos/core_checkedint.html


 It says:


 "The overflow is sticky, meaning a sequence of operations can be done
 and overflow need only be checked at the end."

 But how can I make multiple operations? I can only put 2 values in the
 function.


import core.checkedint;

void main() {
     bool overflowed;
     auto result = adds(int.max, 1, overflowed); // this overflows
     adds(1, 2, overflowed);     // this does not reset the flag

     assert(overflowed);
}

Ali
Nov 03 2015
parent reply Namal <sotis22 mail.ru> writes:
On Wednesday, 4 November 2015 at 07:59:44 UTC, Ali Çehreli wrote:

 On 11/03/2015 11:52 PM, Namal wrote:


 http://dlang.org/phobos/core_checkedint.html


 It says:


 "The overflow is sticky, meaning a sequence of operations can 
 be done
 and overflow need only be checked at the end."

 But how can I make multiple operations? I can only put 2 
 values in the
 function.


 import core.checkedint;

 void main() {
     bool overflowed;
     auto result = adds(int.max, 1, overflowed); // this 
 overflows
     adds(1, 2, overflowed);     // this does not reset the flag

     assert(overflowed);
 }

 Ali


wow, this I don't understand at all, how do those two operations 
connected to each other? By the bool value?
Nov 04 2015
parent =?UTF-8?Q?Ali_=c3=87ehreli?= <acehreli yahoo.com> writes:
On 11/04/2015 12:11 AM, Namal wrote:


 import core.checkedint;

 void main() {
     bool overflowed;
     auto result = adds(int.max, 1, overflowed); // this overflows
     adds(1, 2, overflowed);     // this does not reset the flag

     assert(overflowed);
 }

 Ali


 wow, this I don't understand at all, how do those two operations
 connected to each other? By the bool value?


The 'overflow' parameter is a 'ref' parameter, meaning that these 
functions modify the actual 'overflowed' variable above. You can imagine 
that they do not set it to 'false' ever:

Imagine that adds() is implemented like the following:

pure nothrow  nogc  safe
int adds(int x, int y, ref bool overflow) {
     // ...
     if (over_flow_detected) {
         overflow = true;
     }
}

So, the caller's variable can only go from 'false' to 'true', never the 
other way around. (All of this is just a guess. :) )

Ali
Nov 04 2015