Port-sgimips archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: CPUFLAGS for IP32



-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

On Aug 15, 2008, at 11:40 AM, Izumi Tsutsui wrote:

In article <080815225618.M0105311%mirage.ceres.dti.ne.jp@localhost>
I wrote:

with revised one:
---
# ttcp -ts -n 10000 192.168.20.1
 :
ttcp-t: 81920000 bytes in 8.03 real seconds = 9957.48 KB/sec +++
ttcp-t: 10000 I/O calls, msec/call = 0.82, calls/sec = 1244.69
ttcp-t: 0.0user 6.7sys 0:08real 84% 0i+0d 0maxrss 0+2pf -1+49csw

Note these kernels have the following CPUFLAGS:
---
makeoptions     CPUFLAGS="-march=mips3 -mabi=32 -mtune=vr5000"

I've found something odd - I used to build my kernels with - march=mips3 -mabi=32 -mtune=mips3 -pipe, until a few weeks ago that worked just fine but then I got odd panics from gzip when doing tar xzf <whatever>:
trap: TLB miss ( load or instr. fetch )
status=0xff03 cause=0x08 vaddr=0x30
uvm_load
pipe_write
dowritefile
sys_write
syscall_plain
mips3_SystemCall

On a hunch I removed -march and -mtune parameters, rebuilt the kernel and it works fine. Now I'm wondering if there was some messed up .o file or if any of the CPU specific options triggered some compiler bug.

have fun
Michael

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iQEVAwUBSKXwvcpnzkX8Yg2nAQITXAgAllNG+cWYb14oRCvc3YvxJ0g9iZ3Qzwb1
bJlmuf5oQ7rw1+UXurqsnXOcDpafZ6Os6FId8HYiy+IJ9dRlSskykF4dX954Maw+
yqnG0wrppcqTdyQcXo5Dp5CkuT33Nv4UyjGeHE3hvEgE5l2Ww/s1J2biOTaeTnFM
ML/TJoV6KlLf4cZThVZ6Jr3fECD/ih4oYpBnMHeUS5I7tFxnjWnas7jMCHVs5vw7
uVQT4Dmh9o1Ftqvv5X5n+9R4xVFnPBWD2AYkkoMVmRWGmAdp550+RRfRbj+EWDFT
N0ko/CNgUAfk0GfgSGcp3ZzOZedB/kL0Cv+/4ulsB69Eg9xbD2z6WQ==
=xE/d
-----END PGP SIGNATURE-----


Home | Main Index | Thread Index | Old Index