30 Oct
2018
30 Oct
'18
8:02 a.m.
“entry” was a reserved word in K&R Ed.1,
my personal favourite C trivia. I have never seen it used outside fortran on mainframes
though.
-Steve
30 Oct
30 Oct
8:51 a.m.
Steve Simon wrote:
“entry” was a reserved word in K&R Ed.1,
my personal favourite C trivia. I have never seen it used outside
fortran on mainframes though.
It's there in Snyder's C compiler, but not used. From the grammar:
statement:
...
| ENTRY identifier ':' statement {val=aentry(#2,#4);}
But the supporting code in the compiler:
aentry (idn, stmt)
{return (stmt);} /* not implemented */
1:32 p.m.
On 10/30/18, Angelo Papenhoff <aap(a)papnet.eu> wrote:
On 30/10/18, Steve Simon wrote:
PL/1 inherited the concept of routines with multiple entry points from Fortran.
Did the early K&R C compilers actually implement multiple entry point
routines, or was the keyword simply reserved for future use?
-Paul W.
“entry” was a reserved word in K&R Ed.1,
my personal favourite C trivia. I have never seen it used outside fortran
on mainframes though.
I think PL/1 on Multics uses it, which is probably how it "got into" C.
1:56 p.m.
I'd imagine you'd use it like so (in more modern C):
void *memcpy(void *src, void *dst, size_t len)
{
char *rv = dst;
top:
while (len--) *dst++ =*src; /* not always a trivial body */
return(rv);
entry bcopy:
rv = dst; /* Swap args and jump to memcpy */
dst = src;
src = rv;
goto top:
}
Variations on this theme are often done in assembler for these routines,
though the example is an attempt to cope with the diversity of interfaces
that grew up after v6 unix was released in the world, so it's not likely
this particular example inspired it...
Warner
On Tue, Oct 30, 2018 at 4:00 AM Angelo Papenhoff <aap(a)papnet.eu> wrote:
On 30/10/18, Steve Simon wrote:
“entry” was a reserved word in K&R Ed.1,
my personal favourite C trivia. I have never seen it used outside
fortran on
mainframes though.
I think PL/1 on Multics uses it, which is probably how it "got into" C.
aap
5:53 p.m.
In Fortran secondary entry points served two main purposes. It
allowed for sharing of code for similar routines (as in Angelo
Papenhoff's C example), which was important when your program had to
fit in only a few K of memory. Secondly, it provided partial relief
for Fortran's very restrictive variable-scoping rules.
It's a pain in the butt for compiler optimizers, although certain
modern interprocedural optimizations emit code that is the moral
equivalent of a routine with multiple entry points.
-Paul W.
2014
days inactive
2014
days old
5 comments
5 participants
participants (5)
-
Angelo Papenhoff -
Lars Brinkhoff -
Paul Winalski -
Steve Simon -
Warner Losh