• =?UTF-8?B?Ik5vcmRsw7Z3Ig==?= (13/13) Jan 21 2015 As a follow up to
• deadalnix (4/17) Jan 21 2015 That is reasonable to expect a compiler to remove such a check
• bearophile (5/8) Jan 21 2015 I think this thread is about an enhancement request for a

=?UTF-8?B?Ik5vcmRsw7Z3Ig==?= <per.nordlow gmail.com> writes:

As a follow up to

https://github.com/D-Programming-Language/dmd/pull/4293

would it be possible to also avoid range-checking in expressions 
such as

     x[0 .. $/n]

and

     x[$/n .. $]

where n is an integer compile-time constant >= 2.

I'm askingbecause this is a reoccurring pattern in D-style binary 
divide-and-conquer algorithms and would be a good show-case for D 
from a performance standpoint.

Is DMD prepared for (easily) adding this to e2ir.c and 
expression.c?

"deadalnix" <deadalnix gmail.com> writes:

On Wednesday, 21 January 2015 at 23:50:44 UTC, Nordlöw wrote:
 As a follow up to

 https://github.com/D-Programming-Language/dmd/pull/4293

 would it be possible to also avoid range-checking in 
 expressions such as

     x[0 .. $/n]

 and

     x[$/n .. $]

 where n is an integer compile-time constant >= 2.

 I'm askingbecause this is a reoccurring pattern in D-style 
 binary divide-and-conquer algorithms and would be a good 
 show-case for D from a performance standpoint.

 Is DMD prepared for (easily) adding this to e2ir.c and 
 expression.c?

That is reasonable to expect a compiler to remove such a check 
with optimization enabled already. I'm not sure about DMD, but 
I'd be willing to bet that LDC and GDC already do it.

"bearophile" <bearophileHUGS lycos.com> writes:

deadalnix:

 That is reasonable to expect a compiler to remove such a check 
 with optimization enabled already. I'm not sure about DMD, but 
 I'd be willing to bet that LDC and GDC already do it.

I think this thread is about an enhancement request for a 
guaranteed front-end optimization.

Bye,
bearophile