looking at a high resolution clock for each of the following events:
function call, function return, interrupt entry, interrupt exit,
and interesting branches. The differences between the times of
these events are added at appropriate places in a ordinary histogram
(as if very fast statistical profiling sampled the pc at those
places) so that ordinary gprof can be used to analyze the times.
gmon.h:
Histogram counters need to be 4 bytes for microsecond resolutions.
They will need to be larger for the 586 clock.
The comments were vax-centric and wrong even on vaxes. Does anyone
disagree?
gprof4.c:
The standard gprof should support counters of all integral sizes
and the size of the counter should be in the gmon header. This
hack will do until then. (Use gprof4 -u to examine the results
of non-statistical profiling.)
config/*:
Non-statistical profiling is configured with `config -pp'.
`config -p' still gives ordinary profiling.
kgmon/*:
Non-statistical profiling is enabled with `kgmon -B'. `kgmon -b'
still enables ordinary profiling (and distables non-statistical
profiling) if non-statistical profiling is configured.
Note that this code is dormant unless the options files exist.
Also, parsing of quoted options in the config files is improved.
What this allows, is all the options in LINT to be specified to be
configured as #defines in a file rather than on the CC command line at
kernel build time. This means that 'make depend' will catch dependencies
on actual *options*, meaning that you can run 'config' and 'make depend'
in complete safety WITHOUT removing the compile directory each time.
Unfortunately, this requires a pass over the source to get the individual
files to #include the new .h files that would be generated by config.
This has a small compile time penalty (appears up to about 2% slower)
from a "fresh" build. Of course, you should not be needing to do complete
rebuilds very often once this was completed, so it would be an overall
win for most people.
Since this code is dormant and we've got a lot of other things happening
on the kernel tree at the moment (prototypes, devfs, static declarations
etc) I am not planning on doing any changes to activate this feature just
yet.
/*
* filename [ standard | optional ] [ config-dependent ]
* [ dev* | profiling-routine ] [ device-driver] [ no-obj ]
* [ compile-with "compile rule" [no-implicit-rule] ]
* [ dependancy "dependancy-list"]
*/
I added
no-obj - This entry does not create anything linkable to the kernel.
dependancy - Add additional dependancy rules to a target.
no-implicit-rule - Don't assume .c -> .o type rules. Config is really
dumb in this area and assumes that everything is a .c file
irregarless of extention. This was the best choice really
since there may even be .c file that you don't want to follow
the standard rules.
This was all done so that the building to the aic7770 assembler and using
the aic7770 assembler in the building of the aic7770 driver could be config
dependant. I can now have an entry like this for the driver:
aic7770 optional ahc device-driver \
compile-with "${CC} $> -o $@" \
dependancy "$S/gnu/misc/aic7770/aic7770.c" \
no-obj no-implicit-rule
aic7770_seq.h optional ahc device-driver \
compile-with "${.CURDIR}/aic7770 -o $@ $S/gnu/misc/aic7770/aic7770.seq"\
dependancy "$S/gnu/misc/aic7770/aic7770.seq aic7770" \
no-obj no-implicit-rule
i386/isa/aic7770.c optional ahc device-driver \
dependancy "aic7770_seq.h"
I also added '\' escaping to newlines so that this doesn't look as gross as
it could have.
Reviewed by: jkh
/usr/src/usr.sbin.config:
o -DSTATCLOCK gives kludges to support the rtc non-device as well as
old kludges to support the clk non-device.
/usr/src/usr.sbin/config.8:
o Document the trivialness of the new vector.h.
/usr/src/usr.sbin/mkglue.c:
o Only print DEVICE_NAMES and NR_DEVICES in vector.h. These are
only required to support vmstat. The vmstat interface will need
to be improved for dynamic loading.
/usr/src/usr.sbin/mkioconf.c:
o Print device ids to be used as indexes into DEVICE_NAMES.
o Print secondary interrupt handler entry points (xxxintr()) instead
of primary ones (VdevU()). Primary ones are now XintrI() and
XfastintrI() and are independent of the config so they are not
handled here.
o Minor cleanups.
Submitted by: Bruce Evans
