[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Z9hThFNFrrbXjkjc@google.com>
Date: Mon, 17 Mar 2025 09:53:24 -0700
From: Sean Christopherson <seanjc@...gle.com>
To: Thomas Gleixner <tglx@...utronix.de>
Cc: Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
Dave Hansen <dave.hansen@...ux.intel.com>, x86@...nel.org,
Paolo Bonzini <pbonzini@...hat.com>, linux-kernel@...r.kernel.org, kvm@...r.kernel.org,
Jacob Pan <jacob.jun.pan@...ux.intel.com>, Jim Mattson <jmattson@...gle.com>
Subject: Re: [PATCH 2/8] x86/irq: Track if IRQ was found in PIR during initial
loop (to load PIR vals)
On Mon, Mar 17, 2025, Thomas Gleixner wrote:
> On Fri, Mar 14 2025 at 20:06, Sean Christopherson wrote:
> > @@ -409,25 +409,28 @@ static __always_inline bool handle_pending_pir(u64 *pir, struct pt_regs *regs)
> > {
> > int i, vec = FIRST_EXTERNAL_VECTOR;
> > unsigned long pir_copy[4];
> > - bool handled = false;
> > + bool found_irq = false;
> >
> > - for (i = 0; i < 4; i++)
> > + for (i = 0; i < 4; i++) {
> > pir_copy[i] = READ_ONCE(pir[i]);
> > + if (pir_copy[i])
> > + found_irq = true;
> > + }
>
> That's four extra conditional branches. You can avoid them completely. See
> delta patch below.
Huh. gcc elides the conditional branches when computing found_irq regardless of
the approach; the JEs in the changelog are from skipping the XCHG.
But clang-14 does not. I'll slot this in.
Thanks!
Powered by blists - more mailing lists