Michael,
On Wed, Jun 11, 2014 at 02:05:35PM +0200, Michael Haggerty wrote:
> On 06/10/2014 12:19 AM, Jeremiah Mahler wrote:
> > Currently, the data in a strbuf is modified using add operations. To
> > set the buffer to some data a reset must be performed before an add.
> >
...
> > diff --git a/strbuf.c b/strbuf.c
> > index ac62982..9d64b00 100644
> > --- a/strbuf.c
> > +++ b/strbuf.c
> > @@ -189,6 +189,27 @@ void strbuf_splice(struct strbuf *sb, size_t pos, size_t len,
> > strbuf_setlen(sb, sb->len + dlen - len);
> > }
> >
> > +void strbuf_set(struct strbuf *sb, const void *data, size_t len)
> > +{
> > + strbuf_reset(sb);
> > + strbuf_add(sb, data, len);
> > +}
> > +
>
> I never know how much intelligence to attribute to modern compilers, but
> it seems like even after optimization this function will be doing more
> work than necessary:
>
> strbuf_reset(sb)
> -> strbuf_setlen(sb, 0)
> -> sb->len = 0
> -> sb->buf[len] = '\0'
> -> strbuf_add(sb, data, len)
> -> strbuf_grow(sb, len)
> -> ...lots of stuff...
> -> memcpy(sb->buf + sb->len, data, len)
> -> strbuf_setlen(sb, sb->len + len)
> -> sb->len = len
> -> sb->buf[len] = '\0'
>
> If there were a function like strbuf_grow_to(sb, len):
>
> void strbuf_grow_to(struct strbuf *sb, size_t len)
> {
> int new_buf = !sb->alloc;
> if (unsigned_add_overflows(len, 1))
> die("you want to use way too much memory");
> if (new_buf)
> sb->buf = NULL;
> ALLOC_GROW(sb->buf, len + 1, sb->alloc);
> if (new_buf)
> sb->buf[0] = '\0';
> }
>
grow_to() which could reduce in size, interesting.
> (strbuf_grow() could call it:
>
> static inline void strbuf_grow(struct strbuf *sb, size_t extra)
> {
> if (unsigned_add_overflows(sb->len, extra))
> die("you want to use way too much memory");
> strbuf_grow_to(sb, sb->len + extra);
> }
>
> ), then your function could be minimized to
>
> void strbuf_set(struct strbuf *sb, const void *data, size_t len)
> {
> strbuf_grow_to(sb, len);
> memcpy(sb->buf, data, len);
> strbuf_setlen(sb, len);
> }
>
> I think strbuf_grow_to() would be useful in other situations too.
>
> This is just an idea; I'm not saying that you have to make this change.
>
I like your idea. I am leaning towards doing the un-optimized
strbuf_set operations first, then optimizing in a later patch.
> > [...]
>
> Michael
>
> --
> Michael Haggerty
> mhagger@xxxxxxxxxxxx
> http://softwareswirl.blogspot.com/
--
Jeremiah Mahler
jmmahler@xxxxxxxxx
http://github.com/jmahler
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html