On 2017/01/24 04:13PM, David Laight wrote:

From: 'Naveen N. Rao'

quoted

Sent: 23 January 2017 19:22
On 2017/01/15 09:00AM, Benjamin Herrenschmidt wrote:

quoted

On Fri, 2017-01-13 at 23:22 +0530, 'Naveen N. Rao' wrote:

quoted

quoted

That rather depends on whether the processor has a store to load forwarder
that will satisfy the read from the store buffer.
I don't know about ppc, but at least some x86 will do that.

Interesting - good to know that.

However, I don't think powerpc does that and in-register swap is likely
faster regardless. Note also that gcc prefers this form at higher
optimization levels.

Of course powerpc has a load-store forwarder these days, however, I
wouldn't be surprised if the in-register form was still faster on some
implementations, but this needs to be tested.

Thanks for clarifying! To test this, I wrote a simple (perhaps naive)
test that just issues a whole lot of endian swaps and in _that_ test, it
does look like the load-store forwarder is doing pretty well.

...

quoted

This is all in a POWER8 vm. On POWER7, the in-register variant is around
4 times faster than the ldbrx variant.

...

I wonder which is faster on the little 1GHz embedded ppc we use here.

Worth a test, for sure.
FWIW, this patch won't matter since eBPF JIT is for ppc64.

Thanks,
Naveen