| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 93.1 | solar flare? cometary visitors? | MSBCS::SCHNEIDER | Say it with ASCII | Thu Apr 10 1997 11:29 | 8 |
| Hi, Robert,
Knowing your sense of humor, I'm not sure if you really want an
answer, but if so I reckon you'd have to ask the board designers
what's connected to the associated pins on the Alpha chip.
Cheers,
Chuck
|
| 93.2 | re. .1 | HYDRA::SHEN | | Thu Apr 10 1997 16:28 | 12 |
| re. .1
Hi, Chuck,
Very nice to hear from you.
I was really not trying to be funny. My ISV partner in Colorado
definitely would ask me that question as they are keen on making the
best out of dpw's board level cache for their application.
Robert
|
| 93.3 | Its counting FILLs | ROCK::BANNON | | Tue Apr 29 1997 13:42 | 12 |
| Heres the code from the EV5 model:
/*
* Each fill idicates that a miss was generated and seviced by memory
*/
c_bsq->pm_bc_miss_a_h = EXTW(c_bsq_dtb->bsq_state_10b_h, C_BSQ__FILL);
The counter counts fills because there are cases where the a read
that misses will be restarted. We didn't want to count these cases,
so we counted the number of fills.
Pete
|
| 93.4 | re. .3 | HYDRA::SHEN | | Thu May 01 1997 18:31 | 27 |
| Hi, Pete,
Thanks for your reply of .3.
Does it mean that, on a machine without board level cache, the
dcache_miss rate is the same as bcache_miss rate and iprobe simply
reported
so ? If it is the case, then I have peace of mind and my curiosity is
cured.
I would like to report a somewhat different, though related problem
with iprobe "Nt40IprobeT23Ev5"
(MD5 hash value: d0 52 bd da 60 ba c8 6f 23 d4 3f 0b cd 58
c9 7b)
the following iprobe command would crash the system and give you a
blue screen STOP:0x0000001E on WNT4.0/ALPHA:
C:> iprobe -method sample -duration 30 dcache_access bcache_miss
Please take a look to see if you can reproduce it on other
WNT/Alpha
varieties. Thanks
Robert
|