Date: Mon, 18 Feb 2002 21:48:18 +0100 From: Poul-Henning Kamp <phk@critter.freebsd.dk> To: Matthew Dillon <dillon@apollo.backplane.com> Cc: Mike Smith <msmith@hub.freebsd.org>, John Polstra <jdp@polstra.com>, hackers@FreeBSD.ORG Subject: Re: ACPI timecounter tests original fast version vs masked version Message-ID: <45878.1014065298@critter.freebsd.dk> In-Reply-To: Your message of "Mon, 18 Feb 2002 12:40:59 PST." <200202182040.g1IKex936264@apollo.backplane.com>
next in thread | previous in thread | raw e-mail | index | archive | help
In message <200202182040.g1IKex936264@apollo.backplane.com>, Matthew Dillon wri tes: > In anycase, I think this is reasonable especially if we explicitly > use the fast code for those chipsets known to be good. I would like > to commit it. It can handle *ANY* sort of ripple or fast-carry breakage, > really any type of breakage since it looks for two identical samples > after masking rather then using an inequality. My only worry here is laptops. If my laptop is connected to the wall-power when it boots it comes up at full CPU speed during the probe and mask determination. Then I unplug it and the CPU speed drops by a large factor and suddenly the mask is too tight for the CPU to get two samples. I think we need an approach where we loosen the mask at runtime if some significant number of iterations fail to get a count may be needed: #define N 25 static unsigned mask = 0xfffffffe; for (;;) { for (i = 0; i < N; i++) { c = READ_TIMER & mask; d = READ_TIMER & mask; if (c == d) return(d); } printf("Adjusting mask to %08x\n", mask); mask <<= 1; } -- Poul-Henning Kamp | UNIX since Zilog Zeus 3.20 phk@FreeBSD.ORG | TCP/IP since RFC 956 FreeBSD committer | BSD since 4.3-tahoe Never attribute to malice what can adequately be explained by incompetence. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?45878.1014065298>