Re: [fpc-devel] Discussion on a particular optimisation development (WARNING: Technical!)

2021-02-03 Thread Florian Klämpfl via fpc-devel
Am 03.02.21 um 22:36 schrieb J. Gareth Moreton via fpc-devel: On 03/02/2021 21:18, Florian Klämpfl via fpc-devel wrote: Am 03.02.21 um 22:14 schrieb J. Gareth Moreton via fpc-devel: Rats, I might have messed up with some of the arithmetic, as well as the dangers of crossing bitwise with logical

Re: [fpc-devel] Discussion on a particular optimisation development (WARNING: Technical!)

2021-02-03 Thread J. Gareth Moreton via fpc-devel
On 03/02/2021 21:18, Florian Klämpfl via fpc-devel wrote: Am 03.02.21 um 22:14 schrieb J. Gareth Moreton via fpc-devel: Rats, I might have messed up with some of the arithmetic, as well as the dangers of crossing bitwise with logical Boolean operations, although some combinations still work - i

Re: [fpc-devel] Discussion on a particular optimisation development (WARNING: Technical!)

2021-02-03 Thread Florian Klämpfl via fpc-devel
Am 03.02.21 um 22:14 schrieb J. Gareth Moreton via fpc-devel: Rats, I might have messed up with some of the arithmetic, as well as the dangers of crossing bitwise with logical Boolean operations, although some combinations still work - if the conditions are "x = 0" rather than "x <> 0" in the e

Re: [fpc-devel] Discussion on a particular optimisation development (WARNING: Technical!)

2021-02-03 Thread J. Gareth Moreton via fpc-devel
Rats, I might have messed up with some of the arithmetic, as well as the dangers of crossing bitwise with logical Boolean operations, although some combinations still work - if the conditions are "x = 0" rather than "x <> 0" in the example: testq    %rbx,%rbx seteb    %al testq 

Re: [fpc-devel] Discussion on a particular optimisation development (WARNING: Technical!)

2021-02-03 Thread Florian Klämpfl via fpc-devel
Am 03.02.21 um 21:23 schrieb Sven Barth via fpc-devel: Am 03.02.2021 um 20:25 schrieb Florian Klämpfl via fpc-devel: Am 02.02.21 um 22:06 schrieb J. Gareth Moreton via fpc-devel: Hi everyone, I've found a potential optimisation for conditions of the form "(x <> 0) and (y <> 0)", which are ver

Re: [fpc-devel] Discussion on a particular optimisation development (WARNING: Technical!)

2021-02-03 Thread Sven Barth via fpc-devel
Am 03.02.2021 um 20:25 schrieb Florian Klämpfl via fpc-devel: Am 02.02.21 um 22:06 schrieb J. Gareth Moreton via fpc-devel: Hi everyone, I've found a potential optimisation for conditions of the form "(x <> 0) and (y <> 0)", which are very common because this is semantically equivalent to "As

Re: [fpc-devel] Discussion on a particular optimisation development (WARNING: Technical!)

2021-02-03 Thread Florian Klämpfl via fpc-devel
Am 02.02.21 um 22:06 schrieb J. Gareth Moreton via fpc-devel: Hi everyone, I've found a potential optimisation for conditions of the form "(x <> 0) and (y <> 0)", which are very common because this is semantically equivalent to "Assigned(x) and Assigned(y)", for example, and such a construct

Re: [fpc-devel] Discussion on a particular optimisation development (WARNING: Technical!)

2021-02-02 Thread Nikolay Nikolov via fpc-devel
On 2/2/21 11:06 PM, J. Gareth Moreton via fpc-devel wrote: Hi everyone, I've found a potential optimisation for conditions of the form "(x <> 0) and (y <> 0)", which are very common because this is semantically equivalent to "Assigned(x) and Assigned(y)", for example, and such a construct is

[fpc-devel] Discussion on a particular optimisation development (WARNING: Technical!)

2021-02-02 Thread J. Gareth Moreton via fpc-devel
Hi everyone, I've found a potential optimisation for conditions of the form "(x <> 0) and (y <> 0)", which are very common because this is semantically equivalent to "Assigned(x) and Assigned(y)", for example, and such a construct is generated implicity in TObject.Destroy, for example.  The n