summaryrefslogtreecommitdiff
path: root/sys/arch
AgeCommit message (Collapse)Author
1995-11-13varargs.h needs ... in va_dcl for gcc >= 2; from arnej@imf.unit.no; netbsd ↵Theo de Raadt
pr#1695
1995-11-13do not use -fno-builtin for kernels; the reasons why are manifoldTheo de Raadt
1995-11-13wrap isaphysmem in NISADMA; from andrew@wipux2.wifo.uni-mannheim.de; netbsd ↵Theo de Raadt
pr#1735
1995-11-13from netbsd:Theo de Raadt
Prevent endless loop (-> kernel crash) when a signal is sent to a process waiting for a vt to become active. Move one ifdef XSERVER to the right place.
1995-11-13can have multiple cg2Theo de Raadt
1995-11-13accept ufs or ffsTheo de Raadt
1995-11-13from netbsdTheo de Raadt
1995-11-13from netbsd:Need to define KERNTEXTOFF for kvm_mkdbTheo de Raadt
1995-11-13require UNIONTheo de Raadt
1995-11-13SETUIDSCRIPTS are always onTheo de Raadt
1995-11-13update to netbsdTheo de Raadt
1995-11-13correct some copyrightsTheo de Raadt
1995-11-13allow ufs or ffsTheo de Raadt
1995-11-13comment a bit betterTheo de Raadt
1995-11-10fix DVMA problem on sun4 systems with writeback cache. you needchuck
to flush the cache after map in so that the info gets written into main memory. you really only need to do this when writing data (e.g. disk write), but there is currently no way to tell if you are writing so we do it for all cases (XXX this is stupid, fix later). This causes 4/200's to be able to write disk files on SMD disks without data corruption.
1995-11-09correct sun4 calls to fb_attach()Theo de Raadt
1995-11-09don't panic for multiple /dev/fb declarersTheo de Raadt
1995-11-09$NetBSD$ -> $Id$Theo de Raadt
1995-11-09no installTheo de Raadt
1995-11-07../m68k is a subdirTheo de Raadt
1995-11-07fpsp is a subdirTheo de Raadt
1995-11-07optimized in_cksum from Zubin Dittia <zubin@dworkin.wustl.edu>chuck
Zubin says: The checksum computation code here is significantly faster than its vanilla C counterpart (by significantly, I mean 2-3 times faster if the data is in cache, and 1.5-2 times faster if the data is not in cache). We optimize on three fronts: 1. By using the add-with-carry (addxcc) instruction, we can use 32-bit operations instead of 16-bit operations. 2. By unrolling the main loop to reduce branch overheads. 3. By doing a sequence of load,load,add,add,load,load,add,add, we can avoid the extra stall cycle which is incurred if the instruction immediately following a load tries to use the target register of the load. Another possible optimization is to replace a pair of 32-bit loads with a single 64-bit load (ldd) instruction, but I found that although this improves performance somewhat on Sun4c machines, it actually reduces performance considerably on Sun4m machines (because of their superscaler architecture). So I chose to leave it out.
1995-11-07$Id$ throughoutTheo de Raadt
update many copyrights
1995-11-07$Id$ linesTheo de Raadt
MVME167: repair job
1995-11-07fix type-o by pk. a sun4 cpumodel looks like "SUN-4" rather thanchuck
"SUN/4" ... sun4's definitely do *not* want the a.out header installed.
1995-11-06clean two more .o filesTheo de Raadt
1995-11-06from NetBSDTheo de Raadt
1995-11-06avoid bus_tmp() because mapiodev() may reuse that mapping, which would be a ↵Theo de Raadt
catastrophe
1995-11-03can map pixels in VME32 spaceTheo de Raadt
1995-11-03no obio/vme on sun4cTheo de Raadt
1995-11-01from netbsdTheo de Raadt
1995-11-01tls rearranges GENERIC/INSTALLTheo de Raadt
1995-11-01gwr says:Theo de Raadt
New SCSI driver for the NCR5380, by David Jones. Does DMA with interrupts. Much faster than our old driver which did only PIO transfers. (Thanks David!) Could be used on the amiga, and probably others...
1995-11-01xy and xd drivers from chuckTheo de Raadt
1995-11-01LANCE_REVC_BUG fixesTheo de Raadt
1995-11-01new xy & 5380 driversTheo de Raadt
1995-11-01from NetBSDTheo de Raadt
1995-11-01vtrace updateTheo de Raadt
1995-11-01Add the function isbad(), needed by xd/xy driversTheo de Raadt
1995-11-01Fix previous change that broke kd consoleTheo de Raadt
1995-11-01Copyright stuffTheo de Raadt
1995-10-30pms is part of pcconsTheo de Raadt
1995-10-30pms & pc were mergedTheo de Raadt
1995-10-30NLOG of 8 is more usefulTheo de Raadt
1995-10-30locore.s has mips3 code in it (for speed)Theo de Raadt
1995-10-30mislabeled END()Theo de Raadt
1995-10-29correct cursor movementTheo de Raadt
1995-10-29fix camdump(), and use wbflush() to serialize chip acccesses in sngetaddr()Theo de Raadt
1995-10-29fixes from daleTheo de Raadt
1995-10-29line feed support is in putcTheo de Raadt