Re: [patch 2/3] x86/fpu/xsave: Prepare for optimized compaction

From: Thomas Gleixner <tglx@linutronix.de>
To: Dave Hansen <dave.hansen@intel.com>, LKML <linux-kernel@vger.kernel.org>
Cc: x86@kernel.org, Andrew Cooper <andrew.cooper3@citrix.com>,
	"Edgecombe, Rick P" <rick.p.edgecombe@intel.com>
Subject: Re: [patch 2/3] x86/fpu/xsave: Prepare for optimized compaction
Date: Tue, 19 Apr 2022 15:33:31 +0200	[thread overview]
Message-ID: <87h76p9p04.ffs@tglx> (raw)
In-Reply-To: <87k0bl9rhz.ffs@tglx>

On Tue, Apr 19 2022 at 14:39, Thomas Gleixner wrote:
> On Thu, Apr 14 2022 at 08:46, Dave Hansen wrote:
>>> +	for (i = fls64(xtmp) - 1; i >= FIRST_EXTENDED_XFEATURE;
>>> +	     i = fls64(xtmp) - 1) {
>>> +		unsigned int to, from;
>>
>> Is it worth a check here like:
>>
>> 		/* Do not move features in their init state: */
>> 		if (!(xcur & BIT_ULL(i))) {
>> 			xtmp &= ~BIT_ULL(i);
>> 			continue;
>> 		}
>
> That would also require to clear the bit in xall, but we can't do that
> in the loop as that affects offsets. Let me think about that.

Duh, misread it. Yes it's possible to do that. OTOH, with the optimized
compaction there won't be a bit set in xbv, but not in xfeatures, except
for the initial state when a task starts.

Thanks,

        tglx