Something wrong with recent 2.4.21-rc* kernels?

[ghugh Song]

I am comparing

2.4.21-pre3-ac4
2.4.21-rc5-ac1
2.4.21-rc6-ac1

In the last two kernels, something is definitely not right.
LaTeXing a big file takes about twice the time compared to
the same job under 2.4.21-pre3-ac4.
However, I am not totally satisfied with the 2.4.21-pre3-ac4, either.
In the beginning just after a reboot, the condition is really nice.
However, after a while, "time"ing a job reveals that something bad is
happening in the first column of "real" time, which means the wall-clock
timing.  One might say that it is still better with 2.4.21-pre3-ac4
than the situation with 2.4.21-rc5-ac1 or with 2.4.21-rc6-ac1, because
it still does not contaminate "user" time in the middle column.

With 2.4.21-pre3-ac4, as everyone comlained, there seems to be a
problem with the cursor movement while kernel-compilation in the back.
The mouse movement seemed to have improved in 2.4.21-rc6-ac1.  But
overall performance hit is just not acceptable.

BTW, the .config file for 2.4.21-rc6-ac1 has been taken from
2.4.21-pre3-ac4 during compilation.

The motherboard is an Asus with the Nvidia2 chipset.
Main system is running off SCSI hard disks under Adaptec U2-SCSI PCI
controller.  No ATAPI hard disk is running at the moment.
I never used the installed ATAPI CDrom during the test.

In the emaillist, I have been reading about random halts in recent
kernels, i.e., 2.4.21-rc*.  Well. I did not experience such a
drastic things. maybe because I did not wait long enough.

Regards,

G. Hugh Song