I did double check the source code of btt/output.c. However, still it
does not make sense to have a value like 164% for G2I. In other words
it seems to me that the value of G2I should always be smaller than
Q2C, since G2I is included in Q2C. Note that I have no IO merge in my
block IO workload.
I will appreciate if anyone can help me to understand how the average
value of G2I can be larger than average Q2C ?
262 void __output_dip_prep_ohead(struct d_info *dip, void *arg)
263 {
264 if (dip->avgs.q2c.n > 0 && dip->avgs.q2c.total > 0) {
265 char dev_info[15];
266 struct ohead_data *odp = arg;
267 double q2c_total = (double)(dip->avgs.q2c.total);
268
269 fprintf(odp->ofp,
270 "%10s | %8.4lf%% %8.4lf%% %8.4lf%%
%8.4lf%% %8.4lf%%\n",
271 make_dev_hdr(dev_info, 15, dip, 1),
272 100.0 * (double)(dip->avgs.q2g.total) / q2c_total,
273 100.0 * (double)(dip->avgs.g2i.total) / q2c_total,
274 100.0 * (double)(dip->avgs.q2m.total) / q2c_total,
275 100.0 * (double)(dip->avgs.i2d.total) / q2c_total,
276 100.0 * (double)(dip->avgs.d2c.total) / q2c_total);
277
278 __update_odp(odp, dip, q2g);
279 __update_odp(odp, dip, g2i);
280 __update_odp(odp, dip, q2m);
281 __update_odp(odp, dip, i2d);
282 __update_odp(odp, dip, d2c);
283 __update_odp(odp, dip, q2c);
284 }
285 }
On Wed, Dec 5, 2012 at 10:01 PM, Alireza Haghdoost <haghdoost@xxxxxxxxx> wrote:
> Hi
>
> I changed device driver queue depth and run an IO workload on the
> device while capturing block trace. In the final results btt report
> value of 164% in Q2I timing segment. I was wondering how this
> percentage value calculated and how it does make sense to see more
> than 100% value ?
>
> Thanks
> Alireza
--
To unsubscribe from this list: send the line "unsubscribe linux-btrace" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
