Skip to contentSkip to conversation

perf bench numa: make possible testing on uneven and/or overlapping CPU ranges

PATCH Applied Alexander Gordeev 4 participants 15 messages Aug 10, 2020 – Aug 13, 2020 perf events

lore
Share & tools
lore
#1 lore raw

PATCH v2 1/3 perf bench numa: use numa_node_to_cpus() to bind tasks to nodes

In reply to Alexander Gordeev

It is currently assumed that each node contains at most nr_cpus/nr_nodes CPUs and node CPU ranges do not overlap. That assumption is generally incorrect as there are archs where a CPU number does not depend on to its node number.

This update removes the described assumption by simply calling numa_node_to_cpus() interface and using the returned mask for binding CPUs to nodes. It also tightens a cpumask allocation failure check a bit.

Cc Satheesh Rajendran <[email protected]>
Cc Srikar Dronamraju <[email protected]>
Cc Naveen N. Rao <[email protected]>
Cc Balamuruhan S <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>
---
 tools/perf/bench/numa.c | 27 +++++++++++++--------------
 1 file changed, 13 insertions(+), 14 deletions(-)
tools/perf/bench/numa.c +13 -14
index 5797253..23e224e 100644
@@ -247,12 +247,13 @@ static int is_node_present(int node)
  */
 static bool node_has_cpus(int node)
 {
-	struct bitmask *cpu = numa_allocate_cpumask();
+	struct bitmask *cpumask = numa_allocate_cpumask();
 	unsigned int i;
 
-	if (cpu && !numa_node_to_cpus(node, cpu)) {
-		for (i = 0; i < cpu->size; i++) {
-			if (numa_bitmask_isbitset(cpu, i))
+	BUG_ON(!cpumask);
+	if (!numa_node_to_cpus(node, cpumask)) {
+		for (i = 0; i < cpumask->size; i++) {
+			if (numa_bitmask_isbitset(cpumask, i))
 				return true;
 		}
 	}
@@ -288,14 +289,10 @@ static cpu_set_t bind_to_cpu(int target_cpu)
 
 static cpu_set_t bind_to_node(int target_node)
 {
-	int cpus_per_node = g->p.nr_cpus / nr_numa_nodes();
 	cpu_set_t orig_mask, mask;
 	int cpu;
 	int ret;
 
-	BUG_ON(cpus_per_node * nr_numa_nodes() != g->p.nr_cpus);
-	BUG_ON(!cpus_per_node);
-
 	ret = sched_getaffinity(0, sizeof(orig_mask), &orig_mask);
 	BUG_ON(ret);
 
@@ -305,13 +302,15 @@ static cpu_set_t bind_to_node(int target_node)
 		for (cpu = 0; cpu < g->p.nr_cpus; cpu++)
 			CPU_SET(cpu, &mask);
 	} else {
-		int cpu_start = (target_node + 0) * cpus_per_node;
-		int cpu_stop  = (target_node + 1) * cpus_per_node;
-
-		BUG_ON(cpu_stop > g->p.nr_cpus);
+		struct bitmask *cpumask = numa_allocate_cpumask();
 
-		for (cpu = cpu_start; cpu < cpu_stop; cpu++)
-			CPU_SET(cpu, &mask);
+		BUG_ON(!cpumask);
+		if (!numa_node_to_cpus(target_node, cpumask)) {
+			for (cpu = 0; cpu < (int)cpumask->size; cpu++) {
+				if (numa_bitmask_isbitset(cpumask, cpu))
+					CPU_SET(cpu, &mask);
+			}
+		}
 	}
 
 	ret = sched_setaffinity(0, sizeof(mask), &mask);
Signature · 1 line
1.8.3.1
#0 lore raw

v2 cover perf bench numa: make possible testing on uneven and/or overlapping CPU ranges

This series allows running the tool on some configurations that do not conform to an assumption each node contains nr_cpus/nr_nodes CPUs at most. Instead, the actual node-to- CPU mapping is acquired dynamically.

patch 1 fixes the described issue
patches 2,3 are follow-up fixes

Changes since v1:
- numa01* and numa02* test names left intact;
- "2x3-convergence" fix moved out to separate patch

Cc Peter Zijlstra <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>

Alexander Gordeev (3):
perf bench numa: use numa_node_to_cpus() to bind tasks to nodes perf bench numa: fix number of processes in "2x3-convergence" test perf bench numa: fix benchmark names

 tools/perf/bench/numa.c | 64 ++++++++++++++++++++++++-------------------------
 1 file changed, 31 insertions(+), 33 deletions(-)
Signature · 1 line
1.8.3.1
#10 lore raw

PATCH v2 2/3 perf bench numa: fix number of processes in "2x3-convergence" test

Cc Peter Zijlstra <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>
---
 tools/perf/bench/numa.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/bench/numa.c +1 -1
index 23e224e..90639c9 100644
@@ -1754,7 +1754,7 @@ static int run_bench_numa(const char *name, const char **argv)
    { " 1x3-convergence,", "mem",  "-p",  "1", "-t",  "3", "-P",  "512", OPT_CONV },
    { " 1x4-convergence,", "mem",  "-p",  "1", "-t",  "4", "-P",  "512", OPT_CONV },
    { " 1x6-convergence,", "mem",  "-p",  "1", "-t",  "6", "-P", "1020", OPT_CONV },
-   { " 2x3-convergence,", "mem",  "-p",  "3", "-t",  "3", "-P", "1020", OPT_CONV },
+   { " 2x3-convergence,", "mem",  "-p",  "2", "-t",  "3", "-P", "1020", OPT_CONV },
    { " 3x3-convergence,", "mem",  "-p",  "3", "-t",  "3", "-P", "1020", OPT_CONV },
    { " 4x4-convergence,", "mem",  "-p",  "4", "-t",  "4", "-P",  "512", OPT_CONV },
    { " 4x4-convergence-NOTHP,",
Signature · 1 line
1.8.3.1
#13 lore raw

PATCH v2 3/3 perf bench numa: fix benchmark names

In reply to Alexander Gordeev

Standard benchmark names let users know the tests specifics. For example "2x1-bw-process" name tells that two processes one thread each are run and the RAM bandwidth is measured.

Several benchmarks names do not correspond to their actual running configuration. Fix that and also some whitespace and comment inconsistencies.

Cc Peter Zijlstra <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>
---
 tools/perf/bench/numa.c | 35 +++++++++++++++++------------------
 1 file changed, 17 insertions(+), 18 deletions(-)
tools/perf/bench/numa.c +17 -18
index 90639c9..3b4b63f 100644
@@ -812,12 +812,12 @@ static u64 do_work(u8 *__data, long bytes, int nr, int nr_max, int loop, u64 val
 			}
 		}
 	} else if (!g->p.data_backwards || (nr + loop) & 1) {
+		/* Process data forwards: */
 
 		d0 = data + off;
 		d  = data + off + 1;
 		d1 = data + words;
 
-		/* Process data forwards: */
 		for (;;) {
 			if (unlikely(d >= d1))
 				d = data;
@@ -835,7 +835,6 @@ static u64 do_work(u8 *__data, long bytes, int nr, int nr_max, int loop, u64 val
 		d  = data + off - 1;
 		d1 = data + words;
 
-		/* Process data forwards: */
 		for (;;) {
 			if (unlikely(d < data))
 				d = data + words-1;
@@ -1732,12 +1731,12 @@ static int run_bench_numa(const char *name, const char **argv)
  */
 static const char *tests[][MAX_ARGS] = {
    /* Basic single-stream NUMA bandwidth measurements: */
-   { "RAM-bw-local,",	  "mem",  "-p",  "1",  "-t",  "1", "-P", "1024",
+   { "RAM-bw-local,",     "mem",  "-p",  "1",  "-t",  "1", "-P", "1024",
 			  "-C" ,   "0", "-M",   "0", OPT_BW_RAM },
    { "RAM-bw-local-NOTHP,",
 			  "mem",  "-p",  "1",  "-t",  "1", "-P", "1024",
 			  "-C" ,   "0", "-M",   "0", OPT_BW_RAM_NOTHP },
-   { "RAM-bw-remote,",	  "mem",  "-p",  "1",  "-t",  "1", "-P", "1024",
+   { "RAM-bw-remote,",    "mem",  "-p",  "1",  "-t",  "1", "-P", "1024",
 			  "-C" ,   "0", "-M",   "1", OPT_BW_RAM },
 
    /* 2-stream NUMA bandwidth measurements: */
@@ -1779,24 +1778,24 @@ static int run_bench_numa(const char *name, const char **argv)
 			  "mem",  "-p",  "8", "-t",  "1", "-P", " 512", OPT_BW_NOTHP },
    { "16x1-bw-process,",  "mem",  "-p", "16", "-t",  "1", "-P",  "256", OPT_BW },
 
-   { " 4x1-bw-thread,",	  "mem",  "-p",  "1", "-t",  "4", "-T",  "256", OPT_BW },
-   { " 8x1-bw-thread,",	  "mem",  "-p",  "1", "-t",  "8", "-T",  "256", OPT_BW },
-   { "16x1-bw-thread,",   "mem",  "-p",  "1", "-t", "16", "-T",  "128", OPT_BW },
-   { "32x1-bw-thread,",   "mem",  "-p",  "1", "-t", "32", "-T",   "64", OPT_BW },
+   { " 1x4-bw-thread,",   "mem",  "-p",  "1", "-t",  "4", "-T",  "256", OPT_BW },
+   { " 1x8-bw-thread,",   "mem",  "-p",  "1", "-t",  "8", "-T",  "256", OPT_BW },
+   { "1x16-bw-thread,",   "mem",  "-p",  "1", "-t", "16", "-T",  "128", OPT_BW },
+   { "1x32-bw-thread,",   "mem",  "-p",  "1", "-t", "32", "-T",   "64", OPT_BW },
 
-   { " 2x3-bw-thread,",	  "mem",  "-p",  "2", "-t",  "3", "-P",  "512", OPT_BW },
-   { " 4x4-bw-thread,",	  "mem",  "-p",  "4", "-t",  "4", "-P",  "512", OPT_BW },
-   { " 4x6-bw-thread,",	  "mem",  "-p",  "4", "-t",  "6", "-P",  "512", OPT_BW },
-   { " 4x8-bw-thread,",	  "mem",  "-p",  "4", "-t",  "8", "-P",  "512", OPT_BW },
-   { " 4x8-bw-thread-NOTHP,",
+   { " 2x3-bw-process,",  "mem",  "-p",  "2", "-t",  "3", "-P",  "512", OPT_BW },
+   { " 4x4-bw-process,",  "mem",  "-p",  "4", "-t",  "4", "-P",  "512", OPT_BW },
+   { " 4x6-bw-process,",  "mem",  "-p",  "4", "-t",  "6", "-P",  "512", OPT_BW },
+   { " 4x8-bw-process,",  "mem",  "-p",  "4", "-t",  "8", "-P",  "512", OPT_BW },
+   { " 4x8-bw-process-NOTHP,",
 			  "mem",  "-p",  "4", "-t",  "8", "-P",  "512", OPT_BW_NOTHP },
-   { " 3x3-bw-thread,",	  "mem",  "-p",  "3", "-t",  "3", "-P",  "512", OPT_BW },
-   { " 5x5-bw-thread,",	  "mem",  "-p",  "5", "-t",  "5", "-P",  "512", OPT_BW },
+   { " 3x3-bw-process,",  "mem",  "-p",  "3", "-t",  "3", "-P",  "512", OPT_BW },
+   { " 5x5-bw-process,",  "mem",  "-p",  "5", "-t",  "5", "-P",  "512", OPT_BW },
 
-   { "2x16-bw-thread,",   "mem",  "-p",  "2", "-t", "16", "-P",  "512", OPT_BW },
-   { "1x32-bw-thread,",   "mem",  "-p",  "1", "-t", "32", "-P", "2048", OPT_BW },
+   { "2x16-bw-process,",  "mem",  "-p",  "2", "-t", "16", "-P",  "512", OPT_BW },
+   { "1x32-bw-process,",  "mem",  "-p",  "1", "-t", "32", "-P", "2048", OPT_BW },
 
-   { "numa02-bw,",	  "mem",  "-p",  "1", "-t", "32", "-T",   "32", OPT_BW },
+   { "numa02-bw,",        "mem",  "-p",  "1", "-t", "32", "-T",   "32", OPT_BW },
    { "numa02-bw-NOTHP,",  "mem",  "-p",  "1", "-t", "32", "-T",   "32", OPT_BW_NOTHP },
    { "numa01-bw-thread,", "mem",  "-p",  "2", "-t", "16", "-T",  "192", OPT_BW },
    { "numa01-bw-thread-NOTHP,",
Signature · 1 line
1.8.3.1
maintainer
#2 lore raw

Re: [PATCH v2 1/3] perf bench numa: use numa_node_to_cpus() to bind tasks to nodes

In reply to Alexander Gordeev

Hello,

Quoted 73 lines · Alexander Gordeev

It is currently assumed that each node contains at most nr_cpus/nr_nodes CPUs and node CPU ranges do not overlap. That assumption is generally incorrect as there are archs where a CPU number does not depend on to its node number.

This update removes the described assumption by simply calling numa_node_to_cpus() interface and using the returned mask for binding CPUs to nodes. It also tightens a cpumask allocation failure check a bit.

Cc Satheesh Rajendran <[email protected]>
Cc Srikar Dronamraju <[email protected]>
Cc Naveen N. Rao <[email protected]>
Cc Balamuruhan S <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>
---
 tools/perf/bench/numa.c | 27 +++++++++++++--------------
 1 file changed, 13 insertions(+), 14 deletions(-)
tools/perf/bench/numa.c +6 -12
index 5797253..23e224e 100644
@@ -247,12 +247,13 @@ static int is_node_present(int node)
  */
 static bool node_has_cpus(int node)
 {
-       struct bitmask *cpu = numa_allocate_cpumask();
+       struct bitmask *cpumask = numa_allocate_cpumask();
        unsigned int i;
 
-       if (cpu && !numa_node_to_cpus(node, cpu)) {
-               for (i = 0; i < cpu->size; i++) {
-                       if (numa_bitmask_isbitset(cpu, i))
+       BUG_ON(!cpumask);
+       if (!numa_node_to_cpus(node, cpumask)) {
+               for (i = 0; i < cpumask->size; i++) {
+                       if (numa_bitmask_isbitset(cpumask, i))
                                return true;
                }
        }
@@ -288,14 +289,10 @@ static cpu_set_t bind_to_cpu(int target_cpu)
 
 static cpu_set_t bind_to_node(int target_node)
 {
-       int cpus_per_node = g->p.nr_cpus / nr_numa_nodes();
        cpu_set_t orig_mask, mask;
        int cpu;
        int ret;
 
-       BUG_ON(cpus_per_node * nr_numa_nodes() != g->p.nr_cpus);
-       BUG_ON(!cpus_per_node);
-
        ret = sched_getaffinity(0, sizeof(orig_mask), &orig_mask);
        BUG_ON(ret);
 
@@ -305,13 +302,15 @@ static cpu_set_t bind_to_node(int target_node)
                for (cpu = 0; cpu < g->p.nr_cpus; cpu++)
                        CPU_SET(cpu, &mask);
        } else {
-               int cpu_start = (target_node + 0) * cpus_per_node;
-               int cpu_stop  = (target_node + 1) * cpus_per_node;
-
-               BUG_ON(cpu_stop > g->p.nr_cpus);
+               struct bitmask *cpumask = numa_allocate_cpumask();

tools/perf/bench/numa.c static cpu_set_t bind_to_node(int target_node) · line 313

Quoted 10 lines
tools/perf/bench/numa.c +7 -2
 
-               for (cpu = cpu_start; cpu < cpu_stop; cpu++)
-                       CPU_SET(cpu, &mask);
+               BUG_ON(!cpumask);
+               if (!numa_node_to_cpus(target_node, cpumask)) {
+                       for (cpu = 0; cpu < (int)cpumask->size; cpu++) {
+                               if (numa_bitmask_isbitset(cpumask, cpu))
+                                       CPU_SET(cpu, &mask);
+                       }
+               }

It seems you need to call numa_free_cpumask() for both functions.

Thanks
Namhyung

Quoted 6 lines
tools/perf/bench/numa.c +0 -1
        }
 
        ret = sched_setaffinity(0, sizeof(mask), &mask);
--

1.8.3.1

maintainer Acked-by on v2 2/3
#11 lore raw

Re: [PATCH v2 2/3] perf bench numa: fix number of processes in "2x3-convergence" test

In reply to Alexander Gordeev
Quoted 9 lines · Alexander Gordeev
Cc Peter Zijlstra <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>

Thanks
Namhyung

Quoted 20 lines
---
 tools/perf/bench/numa.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/bench/numa.c +1 -2
index 23e224e..90639c9 100644
@@ -1754,7 +1754,7 @@ static int run_bench_numa(const char *name, const char **argv)
    { " 1x3-convergence,", "mem",  "-p",  "1", "-t",  "3", "-P",  "512", OPT_CONV },
    { " 1x4-convergence,", "mem",  "-p",  "1", "-t",  "4", "-P",  "512", OPT_CONV },
    { " 1x6-convergence,", "mem",  "-p",  "1", "-t",  "6", "-P", "1020", OPT_CONV },
-   { " 2x3-convergence,", "mem",  "-p",  "3", "-t",  "3", "-P", "1020", OPT_CONV },
+   { " 2x3-convergence,", "mem",  "-p",  "2", "-t",  "3", "-P", "1020", OPT_CONV },
    { " 3x3-convergence,", "mem",  "-p",  "3", "-t",  "3", "-P", "1020", OPT_CONV },
    { " 4x4-convergence,", "mem",  "-p",  "4", "-t",  "4", "-P",  "512", OPT_CONV },
    { " 4x4-convergence-NOTHP,",
--

1.8.3.1

#3 lore raw

Re: [PATCH v2 1/3] perf bench numa: use numa_node_to_cpus() to bind tasks to nodes

In reply to Namhyung Kim
Quoted 78 lines · Namhyung Kim

Hello,

Quoted 74 lines · Alexander Gordeev

It is currently assumed that each node contains at most nr_cpus/nr_nodes CPUs and node CPU ranges do not overlap. That assumption is generally incorrect as there are archs where a CPU number does not depend on to its node number.

This update removes the described assumption by simply calling numa_node_to_cpus() interface and using the returned mask for binding CPUs to nodes. It also tightens a cpumask allocation failure check a bit.

Cc Satheesh Rajendran <[email protected]>
Cc Srikar Dronamraju <[email protected]>
Cc Naveen N. Rao <[email protected]>
Cc Balamuruhan S <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>
---
 tools/perf/bench/numa.c | 27 +++++++++++++--------------
 1 file changed, 13 insertions(+), 14 deletions(-)
tools/perf/bench/numa.c +6 -12
index 5797253..23e224e 100644
@@ -247,12 +247,13 @@ static int is_node_present(int node)
  */
 static bool node_has_cpus(int node)
 {
-       struct bitmask *cpu = numa_allocate_cpumask();
+       struct bitmask *cpumask = numa_allocate_cpumask();
        unsigned int i;
 
-       if (cpu && !numa_node_to_cpus(node, cpu)) {
-               for (i = 0; i < cpu->size; i++) {
-                       if (numa_bitmask_isbitset(cpu, i))
+       BUG_ON(!cpumask);
+       if (!numa_node_to_cpus(node, cpumask)) {
+               for (i = 0; i < cpumask->size; i++) {
+                       if (numa_bitmask_isbitset(cpumask, i))
                                return true;
                }
        }
@@ -288,14 +289,10 @@ static cpu_set_t bind_to_cpu(int target_cpu)
 
 static cpu_set_t bind_to_node(int target_node)
 {
-       int cpus_per_node = g->p.nr_cpus / nr_numa_nodes();
        cpu_set_t orig_mask, mask;
        int cpu;
        int ret;
 
-       BUG_ON(cpus_per_node * nr_numa_nodes() != g->p.nr_cpus);
-       BUG_ON(!cpus_per_node);
-
        ret = sched_getaffinity(0, sizeof(orig_mask), &orig_mask);
        BUG_ON(ret);
 
@@ -305,13 +302,15 @@ static cpu_set_t bind_to_node(int target_node)
                for (cpu = 0; cpu < g->p.nr_cpus; cpu++)
                        CPU_SET(cpu, &mask);
        } else {
-               int cpu_start = (target_node + 0) * cpus_per_node;
-               int cpu_stop  = (target_node + 1) * cpus_per_node;
-
-               BUG_ON(cpu_stop > g->p.nr_cpus);
+               struct bitmask *cpumask = numa_allocate_cpumask();
 
Quoted 11 lines

tools/perf/bench/numa.c static cpu_set_t bind_to_node(int target_node) · line 313

Quoted 9 lines
tools/perf/bench/numa.c +7 -2
-               for (cpu = cpu_start; cpu < cpu_stop; cpu++)
-                       CPU_SET(cpu, &mask);
+               BUG_ON(!cpumask);
+               if (!numa_node_to_cpus(target_node, cpumask)) {
+                       for (cpu = 0; cpu < (int)cpumask->size; cpu++) {
+                               if (numa_bitmask_isbitset(cpumask, cpu))
+                                       CPU_SET(cpu, &mask);
+                       }
+               }

It seems you need to call numa_free_cpumask() for both functions.

Well, the whole approach to memory allocation is rather relaxed troughout the code. I.e cpumasks do not get deallocated (*), strdup() return values are not checked etc.

If it worth fixing all that then it would be a separate effort, as far as I am concerned.

Quoted 9 lines

Thanks
Namhyung

Quoted 6 lines
tools/perf/bench/numa.c +0 -1
        }
 
        ret = sched_setaffinity(0, sizeof(mask), &mask);
--

1.8.3.1

maintainer Applied
#12 lore raw

Re: [PATCH v2 2/3] perf bench numa: fix number of processes in "2x3-convergence" test

In reply to Namhyung Kim

Em Tue, Aug 11, 2020 at 04:27:35PM +0900, Namhyung Kim escreveu:

Quoted 12 lines
Quoted 9 lines · Alexander Gordeev
Cc Peter Zijlstra <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>
Quoted 1 line

Thanks, applied,

- Arnaldo

Quoted 23 lines

Thanks
Namhyung

Quoted 20 lines
---
 tools/perf/bench/numa.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/bench/numa.c +1 -2
index 23e224e..90639c9 100644
@@ -1754,7 +1754,7 @@ static int run_bench_numa(const char *name, const char **argv)
    { " 1x3-convergence,", "mem",  "-p",  "1", "-t",  "3", "-P",  "512", OPT_CONV },
    { " 1x4-convergence,", "mem",  "-p",  "1", "-t",  "4", "-P",  "512", OPT_CONV },
    { " 1x6-convergence,", "mem",  "-p",  "1", "-t",  "6", "-P", "1020", OPT_CONV },
-   { " 2x3-convergence,", "mem",  "-p",  "3", "-t",  "3", "-P", "1020", OPT_CONV },
+   { " 2x3-convergence,", "mem",  "-p",  "2", "-t",  "3", "-P", "1020", OPT_CONV },
    { " 3x3-convergence,", "mem",  "-p",  "3", "-t",  "3", "-P", "1020", OPT_CONV },
    { " 4x4-convergence,", "mem",  "-p",  "4", "-t",  "4", "-P",  "512", OPT_CONV },
    { " 4x4-convergence-NOTHP,",
--

1.8.3.1

Signature · 1 line
- Arnaldo
#14 lore raw

Re: [PATCH v2 3/3] perf bench numa: fix benchmark names

In reply to Alexander Gordeev

Em Mon, Aug 10, 2020 at 08:22:00AM +0200, Alexander Gordeev escreveu:

Quoted 7 lines

Standard benchmark names let users know the tests specifics. For example "2x1-bw-process" name tells that two processes one thread each are run and the RAM bandwidth is measured.

Several benchmarks names do not correspond to their actual running configuration. Fix that and also some whitespace and comment inconsistencies.

Looks, ok, applied.

- Arnaldo

Quoted 95 lines
Cc Peter Zijlstra <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>
---
 tools/perf/bench/numa.c | 35 +++++++++++++++++------------------
 1 file changed, 17 insertions(+), 18 deletions(-)
tools/perf/bench/numa.c +17 -18
index 90639c9..3b4b63f 100644
@@ -812,12 +812,12 @@ static u64 do_work(u8 *__data, long bytes, int nr, int nr_max, int loop, u64 val
 			}
 		}
 	} else if (!g->p.data_backwards || (nr + loop) & 1) {
+		/* Process data forwards: */
 
 		d0 = data + off;
 		d  = data + off + 1;
 		d1 = data + words;
 
-		/* Process data forwards: */
 		for (;;) {
 			if (unlikely(d >= d1))
 				d = data;
@@ -835,7 +835,6 @@ static u64 do_work(u8 *__data, long bytes, int nr, int nr_max, int loop, u64 val
 		d  = data + off - 1;
 		d1 = data + words;
 
-		/* Process data forwards: */
 		for (;;) {
 			if (unlikely(d < data))
 				d = data + words-1;
@@ -1732,12 +1731,12 @@ static int run_bench_numa(const char *name, const char **argv)
  */
 static const char *tests[][MAX_ARGS] = {
    /* Basic single-stream NUMA bandwidth measurements: */
-   { "RAM-bw-local,",	  "mem",  "-p",  "1",  "-t",  "1", "-P", "1024",
+   { "RAM-bw-local,",     "mem",  "-p",  "1",  "-t",  "1", "-P", "1024",
 			  "-C" ,   "0", "-M",   "0", OPT_BW_RAM },
    { "RAM-bw-local-NOTHP,",
 			  "mem",  "-p",  "1",  "-t",  "1", "-P", "1024",
 			  "-C" ,   "0", "-M",   "0", OPT_BW_RAM_NOTHP },
-   { "RAM-bw-remote,",	  "mem",  "-p",  "1",  "-t",  "1", "-P", "1024",
+   { "RAM-bw-remote,",    "mem",  "-p",  "1",  "-t",  "1", "-P", "1024",
 			  "-C" ,   "0", "-M",   "1", OPT_BW_RAM },
 
    /* 2-stream NUMA bandwidth measurements: */
@@ -1779,24 +1778,24 @@ static int run_bench_numa(const char *name, const char **argv)
 			  "mem",  "-p",  "8", "-t",  "1", "-P", " 512", OPT_BW_NOTHP },
    { "16x1-bw-process,",  "mem",  "-p", "16", "-t",  "1", "-P",  "256", OPT_BW },
 
-   { " 4x1-bw-thread,",	  "mem",  "-p",  "1", "-t",  "4", "-T",  "256", OPT_BW },
-   { " 8x1-bw-thread,",	  "mem",  "-p",  "1", "-t",  "8", "-T",  "256", OPT_BW },
-   { "16x1-bw-thread,",   "mem",  "-p",  "1", "-t", "16", "-T",  "128", OPT_BW },
-   { "32x1-bw-thread,",   "mem",  "-p",  "1", "-t", "32", "-T",   "64", OPT_BW },
+   { " 1x4-bw-thread,",   "mem",  "-p",  "1", "-t",  "4", "-T",  "256", OPT_BW },
+   { " 1x8-bw-thread,",   "mem",  "-p",  "1", "-t",  "8", "-T",  "256", OPT_BW },
+   { "1x16-bw-thread,",   "mem",  "-p",  "1", "-t", "16", "-T",  "128", OPT_BW },
+   { "1x32-bw-thread,",   "mem",  "-p",  "1", "-t", "32", "-T",   "64", OPT_BW },
 
-   { " 2x3-bw-thread,",	  "mem",  "-p",  "2", "-t",  "3", "-P",  "512", OPT_BW },
-   { " 4x4-bw-thread,",	  "mem",  "-p",  "4", "-t",  "4", "-P",  "512", OPT_BW },
-   { " 4x6-bw-thread,",	  "mem",  "-p",  "4", "-t",  "6", "-P",  "512", OPT_BW },
-   { " 4x8-bw-thread,",	  "mem",  "-p",  "4", "-t",  "8", "-P",  "512", OPT_BW },
-   { " 4x8-bw-thread-NOTHP,",
+   { " 2x3-bw-process,",  "mem",  "-p",  "2", "-t",  "3", "-P",  "512", OPT_BW },
+   { " 4x4-bw-process,",  "mem",  "-p",  "4", "-t",  "4", "-P",  "512", OPT_BW },
+   { " 4x6-bw-process,",  "mem",  "-p",  "4", "-t",  "6", "-P",  "512", OPT_BW },
+   { " 4x8-bw-process,",  "mem",  "-p",  "4", "-t",  "8", "-P",  "512", OPT_BW },
+   { " 4x8-bw-process-NOTHP,",
 			  "mem",  "-p",  "4", "-t",  "8", "-P",  "512", OPT_BW_NOTHP },
-   { " 3x3-bw-thread,",	  "mem",  "-p",  "3", "-t",  "3", "-P",  "512", OPT_BW },
-   { " 5x5-bw-thread,",	  "mem",  "-p",  "5", "-t",  "5", "-P",  "512", OPT_BW },
+   { " 3x3-bw-process,",  "mem",  "-p",  "3", "-t",  "3", "-P",  "512", OPT_BW },
+   { " 5x5-bw-process,",  "mem",  "-p",  "5", "-t",  "5", "-P",  "512", OPT_BW },
 
-   { "2x16-bw-thread,",   "mem",  "-p",  "2", "-t", "16", "-P",  "512", OPT_BW },
-   { "1x32-bw-thread,",   "mem",  "-p",  "1", "-t", "32", "-P", "2048", OPT_BW },
+   { "2x16-bw-process,",  "mem",  "-p",  "2", "-t", "16", "-P",  "512", OPT_BW },
+   { "1x32-bw-process,",  "mem",  "-p",  "1", "-t", "32", "-P", "2048", OPT_BW },
 
-   { "numa02-bw,",	  "mem",  "-p",  "1", "-t", "32", "-T",   "32", OPT_BW },
+   { "numa02-bw,",        "mem",  "-p",  "1", "-t", "32", "-T",   "32", OPT_BW },
    { "numa02-bw-NOTHP,",  "mem",  "-p",  "1", "-t", "32", "-T",   "32", OPT_BW_NOTHP },
    { "numa01-bw-thread,", "mem",  "-p",  "2", "-t", "16", "-T",  "192", OPT_BW },
    { "numa01-bw-thread-NOTHP,",
Signature · 1 line
1.8.3.1
Signature · 1 line
- Arnaldo
#4 lore raw

PATCH v3 1/2 perf bench numa: fix cpumask memory leak in node_has_cpus()

In reply to Namhyung Kim

Couple numa_allocate_cpumask() and numa_free_cpumask() functions

---
 tools/perf/bench/numa.c | 13 +++++++++----
 1 file changed, 9 insertions(+), 4 deletions(-)
tools/perf/bench/numa.c +9 -4
index 31e2601..9066511 100644
@@ -248,16 +248,21 @@ static int is_node_present(int node)
 static bool node_has_cpus(int node)
 {
 	struct bitmask *cpu = numa_allocate_cpumask();
+	bool ret = false; /* fall back to nocpus */
 	unsigned int i;
 
-	if (cpu && !numa_node_to_cpus(node, cpu)) {
+	BUG_ON(!cpu);
+	if (!numa_node_to_cpus(node, cpu)) {
 		for (i = 0; i < cpu->size; i++) {
-			if (numa_bitmask_isbitset(cpu, i))
-				return true;
+			if (numa_bitmask_isbitset(cpu, i)) {
+				ret = true;
+				break;
+			}
 		}
 	}
+	numa_free_cpumask(cpu);
 
-	return false; /* lets fall back to nocpus safely */
+	return ret;
 }
 
 static cpu_set_t bind_to_cpu(int target_cpu)
Signature · 1 line
1.8.3.1
#7 lore raw

PATCH v3 1/2 perf bench numa: use numa_node_to_cpus() to bind tasks to nodes

In reply to Namhyung Kim

It is currently assumed that each node contains at most nr_cpus/nr_nodes CPUs and nodes' CPU ranges do not overlap. That assumption is generally incorrect as there are archs where a CPU number does not depend on to its node number.

This update removes the described assumption by simply calling numa_node_to_cpus() interface and using the returned mask for binding CPUs to nodes.

Also, variable types and names made consistent in functions using cpumask.

Cc Satheesh Rajendran <[email protected]>
Cc Srikar Dronamraju <[email protected]>
Cc Naveen N. Rao <[email protected]>
Cc Balamuruhan S <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>
---
 tools/perf/bench/numa.c | 33 ++++++++++++++++-----------------
 1 file changed, 16 insertions(+), 17 deletions(-)
tools/perf/bench/numa.c +16 -17
index 9066511..6d5c890 100644
@@ -247,20 +247,20 @@ static int is_node_present(int node)
  */
 static bool node_has_cpus(int node)
 {
-	struct bitmask *cpu = numa_allocate_cpumask();
+	struct bitmask *cpumask = numa_allocate_cpumask();
 	bool ret = false; /* fall back to nocpus */
-	unsigned int i;
+	int cpu;
 
-	BUG_ON(!cpu);
-	if (!numa_node_to_cpus(node, cpu)) {
-		for (i = 0; i < cpu->size; i++) {
-			if (numa_bitmask_isbitset(cpu, i)) {
+	BUG_ON(!cpumask);
+	if (!numa_node_to_cpus(node, cpumask)) {
+		for (cpu = 0; cpu < (int)cpumask->size; cpu++) {
+			if (numa_bitmask_isbitset(cpumask, cpu)) {
 				ret = true;
 				break;
 			}
 		}
 	}
-	numa_free_cpumask(cpu);
+	numa_free_cpumask(cpumask);
 
 	return ret;
 }
@@ -293,14 +293,10 @@ static cpu_set_t bind_to_cpu(int target_cpu)
 
 static cpu_set_t bind_to_node(int target_node)
 {
-	int cpus_per_node = g->p.nr_cpus / nr_numa_nodes();
 	cpu_set_t orig_mask, mask;
 	int cpu;
 	int ret;
 
-	BUG_ON(cpus_per_node * nr_numa_nodes() != g->p.nr_cpus);
-	BUG_ON(!cpus_per_node);
-
 	ret = sched_getaffinity(0, sizeof(orig_mask), &orig_mask);
 	BUG_ON(ret);
 
@@ -310,13 +306,16 @@ static cpu_set_t bind_to_node(int target_node)
 		for (cpu = 0; cpu < g->p.nr_cpus; cpu++)
 			CPU_SET(cpu, &mask);
 	} else {
-		int cpu_start = (target_node + 0) * cpus_per_node;
-		int cpu_stop  = (target_node + 1) * cpus_per_node;
-
-		BUG_ON(cpu_stop > g->p.nr_cpus);
+		struct bitmask *cpumask = numa_allocate_cpumask();
 
-		for (cpu = cpu_start; cpu < cpu_stop; cpu++)
-			CPU_SET(cpu, &mask);
+		BUG_ON(!cpumask);
+		if (!numa_node_to_cpus(target_node, cpumask)) {
+			for (cpu = 0; cpu < (int)cpumask->size; cpu++) {
+				if (numa_bitmask_isbitset(cpumask, cpu))
+					CPU_SET(cpu, &mask);
+			}
+		}
+		numa_free_cpumask(cpumask);
 	}
 
 	ret = sched_setaffinity(0, sizeof(mask), &mask);
Signature · 1 line
1.8.3.1
Reviewed-by on v3 1/2
#5 lore raw

Re: [PATCH v3 1/2] perf bench numa: fix cpumask memory leak in node_has_cpus()

In reply to Alexander Gordeev
Quoted 40 lines · Alexander Gordeev

Couple numa_allocate_cpumask() and numa_free_cpumask() functions

---
 tools/perf/bench/numa.c | 13 +++++++++----
 1 file changed, 9 insertions(+), 4 deletions(-)
tools/perf/bench/numa.c +9 -4
index 31e2601..9066511 100644
@@ -248,16 +248,21 @@ static int is_node_present(int node)
 static bool node_has_cpus(int node)
 {
 	struct bitmask *cpu = numa_allocate_cpumask();
+	bool ret = false; /* fall back to nocpus */
 	unsigned int i;
 
-	if (cpu && !numa_node_to_cpus(node, cpu)) {
+	BUG_ON(!cpu);
+	if (!numa_node_to_cpus(node, cpu)) {
 		for (i = 0; i < cpu->size; i++) {
-			if (numa_bitmask_isbitset(cpu, i))
-				return true;
+			if (numa_bitmask_isbitset(cpu, i)) {
+				ret = true;
+				break;
+			}
 		}
 	}
+	numa_free_cpumask(cpu);
 
-	return false; /* lets fall back to nocpus safely */
+	return ret;
 }
 
 static cpu_set_t bind_to_cpu(int target_cpu)
Signature · 1 line
1.8.3.1

Looks good to me.

Signature · 2 lines
Thanks and Regards
Srikar Dronamraju
Reviewed-by on v3 1/2
#8 lore raw

Re: [PATCH v3 1/2] perf bench numa: use numa_node_to_cpus() to bind tasks to nodes

In reply to Alexander Gordeev
Quoted 51 lines · Alexander Gordeev

It is currently assumed that each node contains at most nr_cpus/nr_nodes CPUs and nodes' CPU ranges do not overlap. That assumption is generally incorrect as there are archs where a CPU number does not depend on to its node number.

This update removes the described assumption by simply calling numa_node_to_cpus() interface and using the returned mask for binding CPUs to nodes.

Also, variable types and names made consistent in functions using cpumask.

Cc Satheesh Rajendran <[email protected]>
Cc Srikar Dronamraju <[email protected]>
Cc Naveen N. Rao <[email protected]>
Cc Balamuruhan S <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Alexander Shishkin <[email protected]>
Cc Jiri Olsa <[email protected]>
Cc Namhyung Kim <[email protected]>

---

tools/perf/bench/numa.c +9 -6
@@ -310,13 +306,16 @@ static cpu_set_t bind_to_node(int target_node)
 		for (cpu = 0; cpu < g->p.nr_cpus; cpu++)
 			CPU_SET(cpu, &mask);
 	} else {
-		int cpu_start = (target_node + 0) * cpus_per_node;
-		int cpu_stop  = (target_node + 1) * cpus_per_node;
-
-		BUG_ON(cpu_stop > g->p.nr_cpus);
+		struct bitmask *cpumask = numa_allocate_cpumask();
 
-		for (cpu = cpu_start; cpu < cpu_stop; cpu++)
-			CPU_SET(cpu, &mask);
+		BUG_ON(!cpumask);
+		if (!numa_node_to_cpus(target_node, cpumask)) {
+			for (cpu = 0; cpu < (int)cpumask->size; cpu++) {
+				if (numa_bitmask_isbitset(cpumask, cpu))
+					CPU_SET(cpu, &mask);
+			}
+		}
+		numa_free_cpumask(cpumask);
 	}
 
 	ret = sched_setaffinity(0, sizeof(mask), &mask);
Signature · 1 line
1.8.3.1

Looks good to me.

Signature · 2 lines
Thanks and Regards
Srikar Dronamraju