> From: Fan Yu <fan.yu9@xxxxxxxxxx>
>
> The original delaytop tool always displayed detailed memory
> subsystem breakdown, which could be overwhelming for users
> who only need high-level overview.
>
> Add flexible display control allowing users to choose their
> preferred information granularity.
>
> The new flexibility provides:
> 1) For quick monitoring: use normal mode to reduce visual clutter
> 2) For deep analysis: use verbose mode to see all memory subsystem details
>
> Signed-off-by: Fan Yu <fan.yu9@xxxxxxxxxx>
This feature is very useful to analyze specific delay sources due to memory operations.
But these shown datas are basically average values rather than min/max, it's not enough
for debugging delay jitters. Will you add showing min/max values to delaytop in future?
> ---
> tools/accounting/delaytop.c | 111 ++++++++++++++++++++++++++++--------
> 1 file changed, 87 insertions(+), 24 deletions(-)
>
> /* PSI statistics structure */
> @@ -163,13 +166,14 @@ static void usage(void)
> {
> printf("Usage: delaytop [Options]\n"
> "Options:\n"
> - " -h, --help Show this help message and exit\n"
> - " -d, --delay=SECONDS Set refresh interval (default: 2 seconds, min: 1)\n"
> - " -n, --iterations=COUNT Set number of updates (default: 0 = infinite)\n"
> - " -P, --processes=NUMBER Set maximum number of processes to show (default: 20, max: 1000)\n"
> - " -o, --once Display once and exit\n"
> - " -p, --pid=PID Monitor only the specified PID\n"
> - " -C, --container=PATH Monitor the container at specified cgroup path\n");
> + " -h, --help Show this help message and exit\n"
> + " -d, --delay=SECONDS Set refresh interval (default: 2 seconds, min: 1)\n"
> + " -n, --iterations=COUNT Set number of updates (default: 0 = infinite)\n"
> + " -P, --processes=NUMBER Set maximum number of processes to show (default: 20, max: 1000)\n"
> + " -o, --once Display once and exit\n"
> + " -p, --pid=PID Monitor only the specified PID\n"
> + " -C, --container=PATH Monitor the container at specified cgroup path\n"
> + " -M, --memverbose Display memory detailed information\n");
> exit(0);
> }
>
..
>
> + suc &= BOOL_FPRINT(out, "%8s %8s %-17s", "PID", "TGID", "COMMAND");
> +
> + if (!cfg.mem_verbose_mode) {
> + suc &= BOOL_FPRINT(out, "%8s %8s %8s %8s\n",
> + "CPU(ms)", "IO(ms)", "IRQ(ms)", "MEM(ms)");
> + suc &= BOOL_FPRINT(out, "-----------------------");
> + suc &= BOOL_FPRINT(out, "-----------------------");
> + suc &= BOOL_FPRINT(out, "--------------------------\n");
> + } else {
> + suc &= BOOL_FPRINT(out, "%8s %8s %8s %8s %8s %8s %8s %8s %8s\n",
> + "CPU(ms)", "IO(ms)", "IRQ(ms)", "MEM(ms)",
> + "SWAP(ms)", "RCL(ms)", "THR(ms)", "CMP(ms)", "WP(ms)");
> + suc &= BOOL_FPRINT(out, "-----------------------");
> + suc &= BOOL_FPRINT(out, "-----------------------");
> + suc &= BOOL_FPRINT(out, "-----------------------");
> + suc &= BOOL_FPRINT(out, "-----------------------");
> + suc &= BOOL_FPRINT(out, "-------------------------\n");
> + }
> +
>
> - suc &= BOOL_FPRINT(out, "-----------------------------------------------");
> - suc &= BOOL_FPRINT(out, "----------------------------------------------\n");
> count = task_count < cfg.max_processes ? task_count : cfg.max_processes;
>
> for (i = 0; i < count; i++) {
> - suc &= BOOL_FPRINT(out, "%5d %5d %-15s",
> + suc &= BOOL_FPRINT(out, "%8d %8d %-15s",
> tasks[i].pid, tasks[i].tgid, tasks[i].command);
> - suc &= BOOL_FPRINT(out, "%7.2f %7.2f %7.2f %7.2f %7.2f %7.2f %7.2f %7.2f\n",
> - average_ms(tasks[i].cpu_delay_total, tasks[i].cpu_count),
> - average_ms(tasks[i].blkio_delay_total, tasks[i].blkio_count),
> - average_ms(tasks[i].swapin_delay_total, tasks[i].swapin_count),
> - average_ms(tasks[i].thrashing_delay_total, tasks[i].thrashing_count),
> - average_ms(tasks[i].compact_delay_total, tasks[i].compact_count),
> - average_ms(tasks[i].wpcopy_delay_total, tasks[i].wpcopy_count),
> - average_ms(tasks[i].irq_delay_total, tasks[i].irq_count));
> + if (!cfg.mem_verbose_mode) {
> + suc &= BOOL_FPRINT(out, FMT_NORMAL,
> + average_ms(tasks[i].cpu_delay_total,
> + tasks[i].cpu_count),
> + average_ms(tasks[i].blkio_delay_total,
> + tasks[i].blkio_count),
> + average_ms(tasks[i].irq_delay_total,
> + tasks[i].irq_count),
> + average_ms(task_total_mem_delay(&tasks[i]),
> + task_total_mem_count(&tasks[i])));
> + } else {
> + suc &= BOOL_FPRINT(out, FMT_MEMVERBOSE,
> + average_ms(tasks[i].cpu_delay_total,
> + tasks[i].cpu_count),
> + average_ms(tasks[i].blkio_delay_total,
> + tasks[i].blkio_count),
> + average_ms(tasks[i].irq_delay_total,
> + tasks[i].irq_count),
> + average_ms(task_total_mem_delay(&tasks[i]),
> + task_total_mem_count(&tasks[i])),
> + average_ms(tasks[i].swapin_delay_total,
> + tasks[i].swapin_count),
> + average_ms(tasks[i].freepages_delay_total,
> + tasks[i].freepages_count),
> + average_ms(tasks[i].thrashing_delay_total,
> + tasks[i].thrashing_count),
> + average_ms(tasks[i].compact_delay_total,
> + tasks[i].compact_count),
> + average_ms(tasks[i].wpcopy_delay_total,
> + tasks[i].wpcopy_count));
> + }
> }
I think if users press 'M', delaytop doesn't need to shown CPU/IRQ/IO delays, just show the detailed memory delays.
>
> suc &= BOOL_FPRINT(out, "\n");
> --
> 2.25.1
