On Tue, 2022-07-12 at 16:46 +0300, Dan Carpenter wrote:
> On Sun, Jul 10, 2022 at 02:04:18PM -0700, Binyi Han wrote:
> > Fix indentation issue to adhere to Linux kernel coding style,
> > Issue found by checkpatch. Change the long for loop into 3 lines. And
> > optimize by avoiding the multiplication.
>
> There is no possible way this optimization helps benchmarks. Better to
> focus on readability.
I think removing the multiply _improves_ readability.
> > diff --git a/drivers/staging/qlge/qlge_main.c b/drivers/staging/qlge/qlge_main.c
[]
> > @@ -3007,10 +3007,12 @@ static int qlge_start_rx_ring(struct qlge_adapter *qdev, struct rx_ring *rx_ring
> > tmp = (u64)rx_ring->lbq.base_dma;
> > base_indirect_ptr = rx_ring->lbq.base_indirect;
> >
> > - for (page_entries = 0; page_entries <
> > - MAX_DB_PAGES_PER_BQ(QLGE_BQ_LEN); page_entries++)
> > - base_indirect_ptr[page_entries] =
> > - cpu_to_le64(tmp + (page_entries * DB_PAGE_SIZE));
> > + for (page_entries = 0;
> > + page_entries < MAX_DB_PAGES_PER_BQ(QLGE_BQ_LEN);
> > + page_entries++) {
> > + base_indirect_ptr[page_entries] = cpu_to_le64(tmp);
> > + tmp += DB_PAGE_SIZE;
>
> I've previously said that using "int i;" is clearer here. You would
> kind of expect "page_entries" to be the number of entries, so it's kind
> of misleading. In other words, it's not just harmless wordiness and
> needless exposition, it's actively bad.
Likely true.
> I would probably just put it on one line:
>
> for (i = 0; i MAX_DB_PAGES_PER_BQ(QLGE_BQ_LEN); i++)
> base_indirect_ptr[i] = cpu_to_le64(tmp + (i * DB_PAGE_SIZE));
>
> But if you want to break it up you could do:
>
> for (i = 0; i MAX_DB_PAGES_PER_BQ(QLGE_BQ_LEN); i++)
> base_indirect_ptr[i] = cpu_to_le64(tmp +
> (i * DB_PAGE_SIZE));
>
> "tmp" is kind of a bad name. Also "base_indirect_ptr" would be better
> as "base_indirect".
tmp is a poor name here. Maybe dma would be better.
MAX_DB_PAGES_PER_BQ(QLGE_BQ_LEN) is also a poorly named macro
where all the existing uses are QLGE_BQ_LEN.
And there's base_indirect_ptr and base_indirect_dma in the struct
so just base_indirect might not be the best.
tmp = (u64)rx_ring->lbq.base_dma;
base_indirect_ptr = rx_ring->lbq.base_indirect;
And clarity is good.
Though here, clarity to value for effort though is dubious.
btw: this code got moved to staging 3 years ago.
Maybe it's getting closer to removal time.
