This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Might I say something?
- From: "Geek Moses" <elgersmad at vfemail dot net>
- To: <fortran at gcc dot gnu dot org>
- Date: Sun, 14 Sep 2008 11:40:37 -0700
- Subject: Might I say something?
- Reply-to: "Geek Moses" <elgersmad at vfemail dot net>
Hello,
As far as your loop tests are concerned, did you ever sit down and come
up with a Convetion, in so much as to say, we will use this register to
store a value, or place it on the stack for the next piece of code? Most of
the problems revolving around macros seems to be the lack of convention.
When I finish working out COS of a given number, where do I store the
Result. When I produce a loop, do I store the result in an Array, or do I
keep the last itteration? There has to be a Unified Set of Choices
Decisions made of what to do with Variables in order to maintain order.
It seems more like we should be passing around a map of the CPU and
Registers in x86 and 64 Bit.
I personally want to get away from Variable Names, and work on Memory
Locations for variables, to shorten and speed up code. Memory Locations are
Node Logic, and figuring out how the compiler would produce a list of
variable names, place them in the header, enumerate them and use the
enumerations in the program. If it's a command line program, those
variables show up in the header as available switches in the text that is
ignored by the processor. That way the access is there for any
user/programmer, and they use the header as a look-up table, and it doesn't
take away from the speed of execution.