On Mon, Jul 17, 2023 at 5:03 PM Andy Shevchenko
[off-list ref] wrote:

On Mon, Jul 17, 2023 at 04:53:51PM +0200, Alexander Potapenko wrote:

...

quoted

quoted

quoted

I remember that this construction may bring horrible code on some architectures
with some version(s) of the compiler (*).

Wow, even the trunk Clang and GCC seem to generate better code for
your version of this line: https://godbolt.org/z/36Kqxhe6j

Oh wait.
First, my Godbolt reproducer is incorrect, it is using sizeof(unsigned
long) instead of 8 * sizeof(unsigned long) for BITS_PER_LONG.

Still slightly better. And note, that the same GENMASK() is used at the
beginning of the function. Compiler actually might cache it.

I think that the compiler might consider this transformation invalid
because "GENMASK(nbits - 1, 0) << offset" and "GENMASK(nbits + offset
- 1, offset)" have different values in the case "nbits + offset"
exceed 64.
Restricting the domain of bitmap_set_value() might result in better
code, but it is easier to stick to your version, which prevents the
overflow.

quoted

quoted

quoted

To fix that I found an easy refactoring:

                map[index] &= ~(GENMASK(nbits, 0) << offset));

Second, the line above should probably be:
  map[index] &= ~(GENMASK(nbits - 1, 0) << offset));

, right?

Yes.

This "nbits vs. nbits - 1" was not detected by test_set_get_value(),
so I'll add an extra test case for it.

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel