| To: | "Maciej W. Rozycki" <macro@ds2.pg.gda.pl> |
|---|---|
| Subject: | R4[04]00SC |
| From: | Ralf Baechle <ralf@linux-mips.org> |
| Date: | Fri, 6 Feb 2004 15:22:01 +0100 |
| Cc: | linux-mips@linux-mips.org |
| Original-recipient: | rfc822;linux-mips@linux-mips.org |
| Sender: | linux-mips-bounce@linux-mips.org |
| User-agent: | Mutt/1.4.1i |
We still have this stuff in flush_cache_mm():
/*
* Kludge alert. For obscure reasons R4000SC and R4400SC go nuts if we
* only flush the primary caches but R10000 and R12000 behave sane ...
*/
if (current_cpu_data.cputype == CPU_R4000SC ||
current_cpu_data.cputype == CPU_R4000MC ||
current_cpu_data.cputype == CPU_R4400SC ||
current_cpu_data.cputype == CPU_R4400MC)
r4k_blast_scache();
You have any idea what might make this necessary? This slows down
SC systems quite badly but makes the compiler from eleminating the
call to r4k_blast_scache() on systems that don't have one of these
processors.
Could be kludged a bit by also testing cpu_has_subset_pcaches() but
that'd be a hack.
Ralf
|
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | Re: Indy R4000PC problems, Markus Dahms |
|---|---|
| Next by Date: | Re: [PATCH][2.4] Removed no-used files for vr41xx, Ralf Baechle |
| Previous by Thread: | [PATCH][2.4] Merged vr41xx_pci_ops into vr41xx.h, Yoichi Yuasa |
| Next by Thread: | [PATCH][2.4] Changed NEC Eagle's IRQ routines, Yoichi Yuasa |
| Indexes: | [Date] [Thread] [Top] [All Lists] |