Skip to contentSkip to conversation

perf env: Find correct branch counter info on hybrid

PATCH In review [email protected] 3 participants 7 messages Sep 9, 2024 – Sep 11, 2024 perf events

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 perf env: Find correct branch counter info on hybrid

From: Kan Liang <[email protected]>

No event is printed in the "Branch Counter" column on hybrid machines.

For example,
$perf record -e "{cpu_core/branch-instructions/pp,cpu_core/branches/}:S" -j any,counter
$perf report --total-cycles

# Branch counter abbr list:

 # cpu_core/branch-instructions/pp = A
 # cpu_core/branches/ = B
 # '-' No event occurs
 # '+' Event occurrences may be lost due to branch counter saturated
 #
 # Sampled Cycles%  Sampled Cycles  Avg Cycles%  Avg Cycles  Branch Counter
 # ...............  ..............  ...........  ..........  ..............
           44.54%          727.1K        0.00%           1   |+   |+   |
           36.31%          592.7K        0.00%           2   |+   |+   |
           17.83%          291.1K        0.00%           1   |+   |+   |

The branch counter information (br_cntr_width and br_cntr_nr) in the perf_env is retrieved from the CPU_PMU_CAPS. However, the CPU_PMU_CAPS is not available on hybrid machines. Without the width information, the number of occurrences of an event cannot be calculated.

For a hybrid machine, the caps information should be retrieved from the PMU_CAPS, and stored in the perf_env->pmu_caps.

Add a perf_env__find_br_cntr_info() to return the correct branch counter information from the corresponding fields.

Fixes 6f9d8d1de2c6 ("perf script: Add branch counters")
---
 tools/perf/builtin-script.c |  5 +++--
 tools/perf/util/annotate.c  |  5 +++--
 tools/perf/util/env.c       | 15 +++++++++++++++
 tools/perf/util/env.h       |  3 +++
 tools/perf/util/session.c   |  6 ++++--
 5 files changed, 28 insertions(+), 6 deletions(-)
tools/perf/builtin-script.c +3 -2
index dbe792b52c5c..a644787fa9e1 100644
@@ -1241,10 +1241,11 @@ static int ip__fprintf_jump(uint64_t ip, struct branch_entry *en,
 	}
 
 	if (PRINT_FIELD(BRCNTR)) {
-		unsigned int width = evsel__env(evsel)->br_cntr_width;
-		unsigned int i = 0, j, num, mask = (1L << width) - 1;
 		struct evsel *pos = evsel__leader(evsel);
+		unsigned int i = 0, j, num, mask, width;
 
+		perf_env__find_br_cntr_info(evsel__env(evsel), NULL, &width);
+		mask = (1L << width) - 1;
 		printed += fprintf(fp, "br_cntr: ");
 		evlist__for_each_entry_from(evsel->evlist, pos) {
 			if (!(pos->core.attr.branch_sample_type & PERF_SAMPLE_BRANCH_COUNTERS))
tools/perf/util/annotate.c +3 -2
index 4990c70b1794..c6ebde9d40a8 100644
@@ -333,14 +333,15 @@ static int symbol__account_br_cntr(struct annotated_branch *branch,
 {
 	unsigned int br_cntr_nr = evsel__leader(evsel)->br_cntr_nr;
 	unsigned int base = evsel__leader(evsel)->br_cntr_idx;
-	unsigned int width = evsel__env(evsel)->br_cntr_width;
 	unsigned int off = offset * evsel->evlist->nr_br_cntr;
-	unsigned int i, mask = (1L << width) - 1;
 	u64 *branch_br_cntr = branch->br_cntr;
+	unsigned int i, mask, width;
 
 	if (!br_cntr || !branch_br_cntr)
 		return 0;
 
+	perf_env__find_br_cntr_info(evsel__env(evsel), NULL, &width);
+	mask = (1L << width) - 1;
 	for (i = 0; i < br_cntr_nr; i++) {
 		u64 cntr = (br_cntr >> i * width) & mask;
 
tools/perf/util/env.c +15 -0
index a459374d0a1a..1edbccfc3281 100644
@@ -624,3 +624,18 @@ char *perf_env__find_pmu_cap(struct perf_env *env, const char *pmu_name,
 	free(cap_eq);
 	return NULL;
 }
+
+void perf_env__find_br_cntr_info(struct perf_env *env,
+				 unsigned int *nr,
+				 unsigned int *width)
+{
+	if (nr) {
+		*nr = env->cpu_pmu_caps ? env->br_cntr_nr :
+					  env->pmu_caps->br_cntr_nr;
+	}
+
+	if (width) {
+		*width = env->cpu_pmu_caps ? env->br_cntr_width :
+					     env->pmu_caps->br_cntr_width;
+	}
+}
tools/perf/util/env.h +3 -0
index 2a2c37cc40b7..51b36c36019b 100644
@@ -192,4 +192,7 @@ char *perf_env__find_pmu_cap(struct perf_env *env, const char *pmu_name,
 			     const char *cap);
 
 bool perf_env__has_pmu_mapping(struct perf_env *env, const char *pmu_name);
+void perf_env__find_br_cntr_info(struct perf_env *env,
+				 unsigned int *nr,
+				 unsigned int *width);
 #endif /* __PERF_ENV_H */
tools/perf/util/session.c +4 -2
index b492300ec959..dbaf07bf6c5f 100644
@@ -856,7 +856,6 @@ static void branch_stack__printf(struct perf_sample *sample,
 	struct branch_entry *entries = perf_sample__branch_entries(sample);
 	bool callstack = evsel__has_branch_callstack(evsel);
 	u64 *branch_stack_cntr = sample->branch_stack_cntr;
-	struct perf_env *env = evsel__env(evsel);
 	uint64_t i;
 
 	if (!callstack) {
@@ -900,8 +899,11 @@ static void branch_stack__printf(struct perf_sample *sample,
 	}
 
 	if (branch_stack_cntr) {
+		unsigned int br_cntr_width, br_cntr_nr;
+
+		perf_env__find_br_cntr_info(evsel__env(evsel), &br_cntr_nr, &br_cntr_width);
 		printf("... branch stack counters: nr:%" PRIu64 " (counter width: %u max counter nr:%u)\n",
-			sample->branch_stack->nr, env->br_cntr_width, env->br_cntr_nr);
+			sample->branch_stack->nr, br_cntr_width, br_cntr_nr);
 		for (i = 0; i < sample->branch_stack->nr; i++)
 			printf("..... %2"PRIu64": %016" PRIx64 "\n", i, branch_stack_cntr[i]);
 	}
Signature · 1 line
2.38.1
reviewer Reviewed-by on v1 1/1
#1 lore raw
Quoted 27 lines · On Mon, Sep 9, 2024 at 11:41 AM <[email protected]> wrote:

From: Kan Liang <[email protected]>

No event is printed in the "Branch Counter" column on hybrid machines.

For example,
$perf record -e "{cpu_core/branch-instructions/pp,cpu_core/branches/}:S" -j any,counter
$perf report --total-cycles

# Branch counter abbr list:

 # cpu_core/branch-instructions/pp = A
 # cpu_core/branches/ = B
 # '-' No event occurs
 # '+' Event occurrences may be lost due to branch counter saturated
 #
 # Sampled Cycles%  Sampled Cycles  Avg Cycles%  Avg Cycles  Branch Counter
 # ...............  ..............  ...........  ..........  ..............
           44.54%          727.1K        0.00%           1   |+   |+   |
           36.31%          592.7K        0.00%           2   |+   |+   |
           17.83%          291.1K        0.00%           1   |+   |+   |

The branch counter information (br_cntr_width and br_cntr_nr) in the perf_env is retrieved from the CPU_PMU_CAPS. However, the CPU_PMU_CAPS is not available on hybrid machines. Without the width information, the number of occurrences of an event cannot be calculated.

Quoted 8 lines

For a hybrid machine, the caps information should be retrieved from the PMU_CAPS, and stored in the perf_env->pmu_caps.

Add a perf_env__find_br_cntr_info() to return the correct branch counter information from the corresponding fields.

Fixes 6f9d8d1de2c6 ("perf script: Add branch counters")

Thanks,
Ian

Quoted 111 lines
---
 tools/perf/builtin-script.c |  5 +++--
 tools/perf/util/annotate.c  |  5 +++--
 tools/perf/util/env.c       | 15 +++++++++++++++
 tools/perf/util/env.h       |  3 +++
 tools/perf/util/session.c   |  6 ++++--
 5 files changed, 28 insertions(+), 6 deletions(-)
tools/perf/builtin-script.c +3 -2
index dbe792b52c5c..a644787fa9e1 100644
@@ -1241,10 +1241,11 @@ static int ip__fprintf_jump(uint64_t ip, struct branch_entry *en,
        }
 
        if (PRINT_FIELD(BRCNTR)) {
-               unsigned int width = evsel__env(evsel)->br_cntr_width;
-               unsigned int i = 0, j, num, mask = (1L << width) - 1;
                struct evsel *pos = evsel__leader(evsel);
+               unsigned int i = 0, j, num, mask, width;
 
+               perf_env__find_br_cntr_info(evsel__env(evsel), NULL, &width);
+               mask = (1L << width) - 1;
                printed += fprintf(fp, "br_cntr: ");
                evlist__for_each_entry_from(evsel->evlist, pos) {
                        if (!(pos->core.attr.branch_sample_type & PERF_SAMPLE_BRANCH_COUNTERS))
tools/perf/util/annotate.c +3 -2
index 4990c70b1794..c6ebde9d40a8 100644
@@ -333,14 +333,15 @@ static int symbol__account_br_cntr(struct annotated_branch *branch,
 {
        unsigned int br_cntr_nr = evsel__leader(evsel)->br_cntr_nr;
        unsigned int base = evsel__leader(evsel)->br_cntr_idx;
-       unsigned int width = evsel__env(evsel)->br_cntr_width;
        unsigned int off = offset * evsel->evlist->nr_br_cntr;
-       unsigned int i, mask = (1L << width) - 1;
        u64 *branch_br_cntr = branch->br_cntr;
+       unsigned int i, mask, width;
 
        if (!br_cntr || !branch_br_cntr)
                return 0;
 
+       perf_env__find_br_cntr_info(evsel__env(evsel), NULL, &width);
+       mask = (1L << width) - 1;
        for (i = 0; i < br_cntr_nr; i++) {
                u64 cntr = (br_cntr >> i * width) & mask;
 
tools/perf/util/env.c +15 -0
index a459374d0a1a..1edbccfc3281 100644
@@ -624,3 +624,18 @@ char *perf_env__find_pmu_cap(struct perf_env *env, const char *pmu_name,
        free(cap_eq);
        return NULL;
 }
+
+void perf_env__find_br_cntr_info(struct perf_env *env,
+                                unsigned int *nr,
+                                unsigned int *width)
+{
+       if (nr) {
+               *nr = env->cpu_pmu_caps ? env->br_cntr_nr :
+                                         env->pmu_caps->br_cntr_nr;
+       }
+
+       if (width) {
+               *width = env->cpu_pmu_caps ? env->br_cntr_width :
+                                            env->pmu_caps->br_cntr_width;
+       }
+}
tools/perf/util/env.h +3 -0
index 2a2c37cc40b7..51b36c36019b 100644
@@ -192,4 +192,7 @@ char *perf_env__find_pmu_cap(struct perf_env *env, const char *pmu_name,
                             const char *cap);
 
 bool perf_env__has_pmu_mapping(struct perf_env *env, const char *pmu_name);
+void perf_env__find_br_cntr_info(struct perf_env *env,
+                                unsigned int *nr,
+                                unsigned int *width);
 #endif /* __PERF_ENV_H */
tools/perf/util/session.c +4 -3
index b492300ec959..dbaf07bf6c5f 100644
@@ -856,7 +856,6 @@ static void branch_stack__printf(struct perf_sample *sample,
        struct branch_entry *entries = perf_sample__branch_entries(sample);
        bool callstack = evsel__has_branch_callstack(evsel);
        u64 *branch_stack_cntr = sample->branch_stack_cntr;
-       struct perf_env *env = evsel__env(evsel);
        uint64_t i;
 
        if (!callstack) {
@@ -900,8 +899,11 @@ static void branch_stack__printf(struct perf_sample *sample,
        }
 
        if (branch_stack_cntr) {
+               unsigned int br_cntr_width, br_cntr_nr;
+
+               perf_env__find_br_cntr_info(evsel__env(evsel), &br_cntr_nr, &br_cntr_width);
                printf("... branch stack counters: nr:%" PRIu64 " (counter width: %u max counter nr:%u)\n",
-                       sample->branch_stack->nr, env->br_cntr_width, env->br_cntr_nr);
+                       sample->branch_stack->nr, br_cntr_width, br_cntr_nr);
                for (i = 0; i < sample->branch_stack->nr; i++)
                        printf("..... %2"PRIu64": %016" PRIx64 "\n", i, branch_stack_cntr[i]);
        }
--

2.38.1

#2 lore raw
In reply to [email protected]
Quoted 8 lines · On Mon, Sep 09, 2024 at 11:42:00AM -0700, [email protected] wrote:

From: Kan Liang <[email protected]>

No event is printed in the "Branch Counter" column on hybrid machines.

For example,
$perf record -e "{cpu_core/branch-instructions/pp,cpu_core/branches/}:S" -j any,counter
$perf report --total-cycles

So I tried this on an Intel Hybrid system, I start the TUI, then try 'B' and, after applying your patch:

  0.66%  67.8K  0.00%   75                    ┌─Branch counter abbr list────────────┐ibc_write+18]          libc.so.6
  0.63%  65.4K  0.00%  106                  [p│                                     │r_active+159]          [kernel.kallsyms]
  0.61%  63.2K  0.01%  175  [native_queued_spi│ The branch counter is not available.│in_lock_slowpath+528]  [kernel.kallsyms]
  0.57%  58.4K  0.00%   39         [syscall_ex│Press any key...                     │ser_mode+109]          [kernel.kallsyms]
  0.54%  55.6K  0.00%    1  [_IO_file_xsputn@@└─────────────────────────────────────┘BC_2.2.5+316]          libc.so.6

The perf.data was obtained from:

root@number:~# perf record -e "{cpu_core/branch-instructions/pp,cpu_core/branches/}:S" -j any,counter find /

And:

root@number:~# perf evlist --group 
{cpu_core/branch-instructions/pp,cpu_core/branches/}
dummy:u
root@number:~# 

Only mishap above we discussed already, the missing :S, and then:

root@number:~# perf evlist -v
cpu_core/branch-instructions/pp: type: 4 (cpu_core), size: 136, config: 0xc4 (branch-instructions), { sample_period, sample_freq }: 4000, sample_type: IP|TID|TIME|READ|PERIOD|BRANCH_STACK|IDENTIFIER, read_format: ID|GROUP|LOST, disabled: 1, freq: 1, enable_on_exec: 1, precise_ip: 2, sample_id_all: 1, exclude_guest: 1, branch_sample_type: ANY
cpu_core/branches/: type: 0 (PERF_TYPE_HARDWARE), size: 136, config: 0x400000004, sample_type: IP|TID|TIME|READ|PERIOD|BRANCH_STACK|IDENTIFIER, read_format: ID|GROUP|LOST, sample_id_all: 1, exclude_guest: 1, branch_sample_type: ANY
dummy:u: type: 1 (software), size: 136, config: 0x9 (PERF_COUNT_SW_DUMMY), { sample_period, sample_freq }: 1, sample_type: IP|TID|TIME|IDENTIFIER, read_format: ID|LOST, inherit: 1, exclude_kernel: 1, exclude_hv: 1, mmap: 1, comm: 1, task: 1, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1, ksymbol: 1, bpf_event: 1
root@number:~#

Version: Intel(R) Core(TM) i7-14700K
Voltage: 1.0 V
External Clock: 100 MHz
Max Speed: 8500 MHz
Current Speed: 3400 MHz
Status: Populated, Enabled
Upgrade: Socket LGA1700

What am I doing wrong not to get to see that "branch counter abbr list" info to then check before/after your patch?

- Arnaldo

Quoted 135 lines

# Branch counter abbr list:

 # cpu_core/branch-instructions/pp = A
 # cpu_core/branches/ = B
 # '-' No event occurs
 # '+' Event occurrences may be lost due to branch counter saturated
 #
 # Sampled Cycles%  Sampled Cycles  Avg Cycles%  Avg Cycles  Branch Counter
 # ...............  ..............  ...........  ..........  ..............
           44.54%          727.1K        0.00%           1   |+   |+   |
           36.31%          592.7K        0.00%           2   |+   |+   |
           17.83%          291.1K        0.00%           1   |+   |+   |

The branch counter information (br_cntr_width and br_cntr_nr) in the perf_env is retrieved from the CPU_PMU_CAPS. However, the CPU_PMU_CAPS is not available on hybrid machines. Without the width information, the number of occurrences of an event cannot be calculated.

For a hybrid machine, the caps information should be retrieved from the PMU_CAPS, and stored in the perf_env->pmu_caps.

Add a perf_env__find_br_cntr_info() to return the correct branch counter information from the corresponding fields.

Fixes 6f9d8d1de2c6 ("perf script: Add branch counters")
---
 tools/perf/builtin-script.c |  5 +++--
 tools/perf/util/annotate.c  |  5 +++--
 tools/perf/util/env.c       | 15 +++++++++++++++
 tools/perf/util/env.h       |  3 +++
 tools/perf/util/session.c   |  6 ++++--
 5 files changed, 28 insertions(+), 6 deletions(-)
tools/perf/builtin-script.c +3 -2
index dbe792b52c5c..a644787fa9e1 100644
@@ -1241,10 +1241,11 @@ static int ip__fprintf_jump(uint64_t ip, struct branch_entry *en,
 	}
 
 	if (PRINT_FIELD(BRCNTR)) {
-		unsigned int width = evsel__env(evsel)->br_cntr_width;
-		unsigned int i = 0, j, num, mask = (1L << width) - 1;
 		struct evsel *pos = evsel__leader(evsel);
+		unsigned int i = 0, j, num, mask, width;
 
+		perf_env__find_br_cntr_info(evsel__env(evsel), NULL, &width);
+		mask = (1L << width) - 1;
 		printed += fprintf(fp, "br_cntr: ");
 		evlist__for_each_entry_from(evsel->evlist, pos) {
 			if (!(pos->core.attr.branch_sample_type & PERF_SAMPLE_BRANCH_COUNTERS))
tools/perf/util/annotate.c +3 -2
index 4990c70b1794..c6ebde9d40a8 100644
@@ -333,14 +333,15 @@ static int symbol__account_br_cntr(struct annotated_branch *branch,
 {
 	unsigned int br_cntr_nr = evsel__leader(evsel)->br_cntr_nr;
 	unsigned int base = evsel__leader(evsel)->br_cntr_idx;
-	unsigned int width = evsel__env(evsel)->br_cntr_width;
 	unsigned int off = offset * evsel->evlist->nr_br_cntr;
-	unsigned int i, mask = (1L << width) - 1;
 	u64 *branch_br_cntr = branch->br_cntr;
+	unsigned int i, mask, width;
 
 	if (!br_cntr || !branch_br_cntr)
 		return 0;
 
+	perf_env__find_br_cntr_info(evsel__env(evsel), NULL, &width);
+	mask = (1L << width) - 1;
 	for (i = 0; i < br_cntr_nr; i++) {
 		u64 cntr = (br_cntr >> i * width) & mask;
 
tools/perf/util/env.c +15 -0
index a459374d0a1a..1edbccfc3281 100644
@@ -624,3 +624,18 @@ char *perf_env__find_pmu_cap(struct perf_env *env, const char *pmu_name,
 	free(cap_eq);
 	return NULL;
 }
+
+void perf_env__find_br_cntr_info(struct perf_env *env,
+				 unsigned int *nr,
+				 unsigned int *width)
+{
+	if (nr) {
+		*nr = env->cpu_pmu_caps ? env->br_cntr_nr :
+					  env->pmu_caps->br_cntr_nr;
+	}
+
+	if (width) {
+		*width = env->cpu_pmu_caps ? env->br_cntr_width :
+					     env->pmu_caps->br_cntr_width;
+	}
+}
tools/perf/util/env.h +3 -0
index 2a2c37cc40b7..51b36c36019b 100644
@@ -192,4 +192,7 @@ char *perf_env__find_pmu_cap(struct perf_env *env, const char *pmu_name,
 			     const char *cap);
 
 bool perf_env__has_pmu_mapping(struct perf_env *env, const char *pmu_name);
+void perf_env__find_br_cntr_info(struct perf_env *env,
+				 unsigned int *nr,
+				 unsigned int *width);
 #endif /* __PERF_ENV_H */
tools/perf/util/session.c +4 -2
index b492300ec959..dbaf07bf6c5f 100644
@@ -856,7 +856,6 @@ static void branch_stack__printf(struct perf_sample *sample,
 	struct branch_entry *entries = perf_sample__branch_entries(sample);
 	bool callstack = evsel__has_branch_callstack(evsel);
 	u64 *branch_stack_cntr = sample->branch_stack_cntr;
-	struct perf_env *env = evsel__env(evsel);
 	uint64_t i;
 
 	if (!callstack) {
@@ -900,8 +899,11 @@ static void branch_stack__printf(struct perf_sample *sample,
 	}
 
 	if (branch_stack_cntr) {
+		unsigned int br_cntr_width, br_cntr_nr;
+
+		perf_env__find_br_cntr_info(evsel__env(evsel), &br_cntr_nr, &br_cntr_width);
 		printf("... branch stack counters: nr:%" PRIu64 " (counter width: %u max counter nr:%u)\n",
-			sample->branch_stack->nr, env->br_cntr_width, env->br_cntr_nr);
+			sample->branch_stack->nr, br_cntr_width, br_cntr_nr);
 		for (i = 0; i < sample->branch_stack->nr; i++)
 			printf("..... %2"PRIu64": %016" PRIx64 "\n", i, branch_stack_cntr[i]);
 	}
Signature · 1 line
2.38.1
author
#3 lore raw
Quoted 30 lines · Arnaldo Carvalho de Melo
Quoted 8 lines · On Mon, Sep 09, 2024 at 11:42:00AM -0700, [email protected] wrote:

From: Kan Liang <[email protected]>

No event is printed in the "Branch Counter" column on hybrid machines.

For example,
$perf record -e "{cpu_core/branch-instructions/pp,cpu_core/branches/}:S" -j any,counter
$perf report --total-cycles

So I tried this on an Intel Hybrid system, I start the TUI, then try 'B' and, after applying your patch:

  0.66%  67.8K  0.00%   75                    ┌─Branch counter abbr list────────────┐ibc_write+18]          libc.so.6
  0.63%  65.4K  0.00%  106                  [p│                                     │r_active+159]          [kernel.kallsyms]
  0.61%  63.2K  0.01%  175  [native_queued_spi│ The branch counter is not available.│in_lock_slowpath+528]  [kernel.kallsyms]
  0.57%  58.4K  0.00%   39         [syscall_ex│Press any key...                     │ser_mode+109]          [kernel.kallsyms]
  0.54%  55.6K  0.00%    1  [_IO_file_xsputn@@└─────────────────────────────────────┘BC_2.2.5+316]          libc.so.6

The perf.data was obtained from:

root@number:~# perf record -e "{cpu_core/branch-instructions/pp,cpu_core/branches/}:S" -j any,counter find /

And:

root@number:~# perf evlist --group 
{cpu_core/branch-instructions/pp,cpu_core/branches/}
dummy:u
root@number:~# 
Quoted 4 lines

Only mishap above we discussed already, the missing :S, and then:

root@number:~# perf evlist -v
cpu_core/branch-instructions/pp: type: 4 (cpu_core), size: 136, config: 0xc4 (branch-instructions), { sample_period, sample_freq }: 4000, sample_type: IP|TID|TIME|READ|PERIOD|BRANCH_STACK|IDENTIFIER, read_format: ID|GROUP|LOST, disabled: 1, freq: 1, enable_on_exec: 1, precise_ip: 2, sample_id_all: 1, exclude_guest: 1, branch_sample_type: ANY

For a hybrid client, the "Branch Counter" feature is only supported starting from the just released Lunar Lake. Perf falls back to only "ANY" on your Raptor Lake.

The "The branch counter is not available" message is expected.

Here is the perf evlist result from my Lunar Lake machine,

#perf evlist -v
cpu_core/branch-instructions/pp: type: 4 (cpu_core), size: 136, config:
0xc4 (branch-instructions), { sample_period, sample_freq }: 4000,
sample_type: IP|TID|TIME|READ|PERIOD|BRANCH_STACK|IDENTIFIER,
read_format: ID|GROUP|LOST, disabled: 1, freq: 1, enable_on_exec: 1,
precise_ip: 2, sample_id_all: 1, exclude_guest: 1, branch_sample_type:
ANY|COUNTERS

Thanks,
Kan

Quoted 153 lines

cpu_core/branches/: type: 0 (PERF_TYPE_HARDWARE), size: 136, config: 0x400000004, sample_type: IP|TID|TIME|READ|PERIOD|BRANCH_STACK|IDENTIFIER, read_format: ID|GROUP|LOST, sample_id_all: 1, exclude_guest: 1, branch_sample_type: ANY
dummy:u: type: 1 (software), size: 136, config: 0x9 (PERF_COUNT_SW_DUMMY), { sample_period, sample_freq }: 1, sample_type: IP|TID|TIME|IDENTIFIER, read_format: ID|LOST, inherit: 1, exclude_kernel: 1, exclude_hv: 1, mmap: 1, comm: 1, task: 1, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1, ksymbol: 1, bpf_event: 1 root@number:~#

Version: Intel(R) Core(TM) i7-14700K
Voltage: 1.0 V
External Clock: 100 MHz
Max Speed: 8500 MHz
Current Speed: 3400 MHz
Status: Populated, Enabled
Upgrade: Socket LGA1700

What am I doing wrong not to get to see that "branch counter abbr list" info to then check before/after your patch?

- Arnaldo

Quoted 135 lines

# Branch counter abbr list:

 # cpu_core/branch-instructions/pp = A
 # cpu_core/branches/ = B
 # '-' No event occurs
 # '+' Event occurrences may be lost due to branch counter saturated
 #
 # Sampled Cycles%  Sampled Cycles  Avg Cycles%  Avg Cycles  Branch Counter
 # ...............  ..............  ...........  ..........  ..............
           44.54%          727.1K        0.00%           1   |+   |+   |
           36.31%          592.7K        0.00%           2   |+   |+   |
           17.83%          291.1K        0.00%           1   |+   |+   |

The branch counter information (br_cntr_width and br_cntr_nr) in the perf_env is retrieved from the CPU_PMU_CAPS. However, the CPU_PMU_CAPS is not available on hybrid machines. Without the width information, the number of occurrences of an event cannot be calculated.

For a hybrid machine, the caps information should be retrieved from the PMU_CAPS, and stored in the perf_env->pmu_caps.

Add a perf_env__find_br_cntr_info() to return the correct branch counter information from the corresponding fields.

Fixes 6f9d8d1de2c6 ("perf script: Add branch counters")
---
 tools/perf/builtin-script.c |  5 +++--
 tools/perf/util/annotate.c  |  5 +++--
 tools/perf/util/env.c       | 15 +++++++++++++++
 tools/perf/util/env.h       |  3 +++
 tools/perf/util/session.c   |  6 ++++--
 5 files changed, 28 insertions(+), 6 deletions(-)
tools/perf/builtin-script.c +3 -2
index dbe792b52c5c..a644787fa9e1 100644
@@ -1241,10 +1241,11 @@ static int ip__fprintf_jump(uint64_t ip, struct branch_entry *en,
 	}
 
 	if (PRINT_FIELD(BRCNTR)) {
-		unsigned int width = evsel__env(evsel)->br_cntr_width;
-		unsigned int i = 0, j, num, mask = (1L << width) - 1;
 		struct evsel *pos = evsel__leader(evsel);
+		unsigned int i = 0, j, num, mask, width;
 
+		perf_env__find_br_cntr_info(evsel__env(evsel), NULL, &width);
+		mask = (1L << width) - 1;
 		printed += fprintf(fp, "br_cntr: ");
 		evlist__for_each_entry_from(evsel->evlist, pos) {
 			if (!(pos->core.attr.branch_sample_type & PERF_SAMPLE_BRANCH_COUNTERS))
tools/perf/util/annotate.c +3 -2
index 4990c70b1794..c6ebde9d40a8 100644
@@ -333,14 +333,15 @@ static int symbol__account_br_cntr(struct annotated_branch *branch,
 {
 	unsigned int br_cntr_nr = evsel__leader(evsel)->br_cntr_nr;
 	unsigned int base = evsel__leader(evsel)->br_cntr_idx;
-	unsigned int width = evsel__env(evsel)->br_cntr_width;
 	unsigned int off = offset * evsel->evlist->nr_br_cntr;
-	unsigned int i, mask = (1L << width) - 1;
 	u64 *branch_br_cntr = branch->br_cntr;
+	unsigned int i, mask, width;
 
 	if (!br_cntr || !branch_br_cntr)
 		return 0;
 
+	perf_env__find_br_cntr_info(evsel__env(evsel), NULL, &width);
+	mask = (1L << width) - 1;
 	for (i = 0; i < br_cntr_nr; i++) {
 		u64 cntr = (br_cntr >> i * width) & mask;
 
tools/perf/util/env.c +15 -0
index a459374d0a1a..1edbccfc3281 100644
@@ -624,3 +624,18 @@ char *perf_env__find_pmu_cap(struct perf_env *env, const char *pmu_name,
 	free(cap_eq);
 	return NULL;
 }
+
+void perf_env__find_br_cntr_info(struct perf_env *env,
+				 unsigned int *nr,
+				 unsigned int *width)
+{
+	if (nr) {
+		*nr = env->cpu_pmu_caps ? env->br_cntr_nr :
+					  env->pmu_caps->br_cntr_nr;
+	}
+
+	if (width) {
+		*width = env->cpu_pmu_caps ? env->br_cntr_width :
+					     env->pmu_caps->br_cntr_width;
+	}
+}
tools/perf/util/env.h +3 -0
index 2a2c37cc40b7..51b36c36019b 100644
@@ -192,4 +192,7 @@ char *perf_env__find_pmu_cap(struct perf_env *env, const char *pmu_name,
 			     const char *cap);
 
 bool perf_env__has_pmu_mapping(struct perf_env *env, const char *pmu_name);
+void perf_env__find_br_cntr_info(struct perf_env *env,
+				 unsigned int *nr,
+				 unsigned int *width);
 #endif /* __PERF_ENV_H */
tools/perf/util/session.c +4 -2
index b492300ec959..dbaf07bf6c5f 100644
@@ -856,7 +856,6 @@ static void branch_stack__printf(struct perf_sample *sample,
 	struct branch_entry *entries = perf_sample__branch_entries(sample);
 	bool callstack = evsel__has_branch_callstack(evsel);
 	u64 *branch_stack_cntr = sample->branch_stack_cntr;
-	struct perf_env *env = evsel__env(evsel);
 	uint64_t i;
 
 	if (!callstack) {
@@ -900,8 +899,11 @@ static void branch_stack__printf(struct perf_sample *sample,
 	}
 
 	if (branch_stack_cntr) {
+		unsigned int br_cntr_width, br_cntr_nr;
+
+		perf_env__find_br_cntr_info(evsel__env(evsel), &br_cntr_nr, &br_cntr_width);
 		printf("... branch stack counters: nr:%" PRIu64 " (counter width: %u max counter nr:%u)\n",
-			sample->branch_stack->nr, env->br_cntr_width, env->br_cntr_nr);
+			sample->branch_stack->nr, br_cntr_width, br_cntr_nr);
 		for (i = 0; i < sample->branch_stack->nr; i++)
 			printf("..... %2"PRIu64": %016" PRIx64 "\n", i, branch_stack_cntr[i]);
 	}
Signature · 1 line
2.38.1
#4 lore raw
Quoted 46 lines · Kan
Quoted 30 lines · Arnaldo Carvalho de Melo
Quoted 8 lines · On Mon, Sep 09, 2024 at 11:42:00AM -0700, [email protected] wrote:

From: Kan Liang <[email protected]>

No event is printed in the "Branch Counter" column on hybrid machines.

For example,
$perf record -e "{cpu_core/branch-instructions/pp,cpu_core/branches/}:S" -j any,counter
$perf report --total-cycles

So I tried this on an Intel Hybrid system, I start the TUI, then try 'B' and, after applying your patch:

  0.66%  67.8K  0.00%   75                    ┌─Branch counter abbr list────────────┐ibc_write+18]          libc.so.6
  0.63%  65.4K  0.00%  106                  [p│                                     │r_active+159]          [kernel.kallsyms]
  0.61%  63.2K  0.01%  175  [native_queued_spi│ The branch counter is not available.│in_lock_slowpath+528]  [kernel.kallsyms]
  0.57%  58.4K  0.00%   39         [syscall_ex│Press any key...                     │ser_mode+109]          [kernel.kallsyms]
  0.54%  55.6K  0.00%    1  [_IO_file_xsputn@@└─────────────────────────────────────┘BC_2.2.5+316]          libc.so.6

The perf.data was obtained from:

root@number:~# perf record -e "{cpu_core/branch-instructions/pp,cpu_core/branches/}:S" -j any,counter find /

And:

root@number:~# perf evlist --group 
{cpu_core/branch-instructions/pp,cpu_core/branches/}
dummy:u
root@number:~# 
Quoted 4 lines

Only mishap above we discussed already, the missing :S, and then:

root@number:~# perf evlist -v
cpu_core/branch-instructions/pp: type: 4 (cpu_core), size: 136, config: 0xc4 (branch-instructions), { sample_period, sample_freq }: 4000, sample_type: IP|TID|TIME|READ|PERIOD|BRANCH_STACK|IDENTIFIER, read_format: ID|GROUP|LOST, disabled: 1, freq: 1, enable_on_exec: 1, precise_ip: 2, sample_id_all: 1, exclude_guest: 1, branch_sample_type: ANY

For a hybrid client, the "Branch Counter" feature is only supported starting from the just released Lunar Lake. Perf falls back to only "ANY" on your Raptor Lake.

The "The branch counter is not available" message is expected.

Here is the perf evlist result from my Lunar Lake machine,

Quoted 7 lines

#perf evlist -v
cpu_core/branch-instructions/pp: type: 4 (cpu_core), size: 136, config:
0xc4 (branch-instructions), { sample_period, sample_freq }: 4000,
sample_type: IP|TID|TIME|READ|PERIOD|BRANCH_STACK|IDENTIFIER,
read_format: ID|GROUP|LOST, disabled: 1, freq: 1, enable_on_exec: 1,
precise_ip: 2, sample_id_all: 1, exclude_guest: 1, branch_sample_type:
ANY|COUNTERS

So I'll just take the patch and add these explanations there, ok?

- Arnaldo

Quoted 156 lines

Thanks,
Kan

Quoted 153 lines

cpu_core/branches/: type: 0 (PERF_TYPE_HARDWARE), size: 136, config: 0x400000004, sample_type: IP|TID|TIME|READ|PERIOD|BRANCH_STACK|IDENTIFIER, read_format: ID|GROUP|LOST, sample_id_all: 1, exclude_guest: 1, branch_sample_type: ANY
dummy:u: type: 1 (software), size: 136, config: 0x9 (PERF_COUNT_SW_DUMMY), { sample_period, sample_freq }: 1, sample_type: IP|TID|TIME|IDENTIFIER, read_format: ID|LOST, inherit: 1, exclude_kernel: 1, exclude_hv: 1, mmap: 1, comm: 1, task: 1, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1, ksymbol: 1, bpf_event: 1 root@number:~#

Version: Intel(R) Core(TM) i7-14700K
Voltage: 1.0 V
External Clock: 100 MHz
Max Speed: 8500 MHz
Current Speed: 3400 MHz
Status: Populated, Enabled
Upgrade: Socket LGA1700

What am I doing wrong not to get to see that "branch counter abbr list" info to then check before/after your patch?

- Arnaldo

Quoted 135 lines

# Branch counter abbr list:

 # cpu_core/branch-instructions/pp = A
 # cpu_core/branches/ = B
 # '-' No event occurs
 # '+' Event occurrences may be lost due to branch counter saturated
 #
 # Sampled Cycles%  Sampled Cycles  Avg Cycles%  Avg Cycles  Branch Counter
 # ...............  ..............  ...........  ..........  ..............
           44.54%          727.1K        0.00%           1   |+   |+   |
           36.31%          592.7K        0.00%           2   |+   |+   |
           17.83%          291.1K        0.00%           1   |+   |+   |

The branch counter information (br_cntr_width and br_cntr_nr) in the perf_env is retrieved from the CPU_PMU_CAPS. However, the CPU_PMU_CAPS is not available on hybrid machines. Without the width information, the number of occurrences of an event cannot be calculated.

For a hybrid machine, the caps information should be retrieved from the PMU_CAPS, and stored in the perf_env->pmu_caps.

Add a perf_env__find_br_cntr_info() to return the correct branch counter information from the corresponding fields.

Fixes 6f9d8d1de2c6 ("perf script: Add branch counters")
---
 tools/perf/builtin-script.c |  5 +++--
 tools/perf/util/annotate.c  |  5 +++--
 tools/perf/util/env.c       | 15 +++++++++++++++
 tools/perf/util/env.h       |  3 +++
 tools/perf/util/session.c   |  6 ++++--
 5 files changed, 28 insertions(+), 6 deletions(-)
tools/perf/builtin-script.c +3 -2
index dbe792b52c5c..a644787fa9e1 100644
@@ -1241,10 +1241,11 @@ static int ip__fprintf_jump(uint64_t ip, struct branch_entry *en,
 	}
 
 	if (PRINT_FIELD(BRCNTR)) {
-		unsigned int width = evsel__env(evsel)->br_cntr_width;
-		unsigned int i = 0, j, num, mask = (1L << width) - 1;
 		struct evsel *pos = evsel__leader(evsel);
+		unsigned int i = 0, j, num, mask, width;
 
+		perf_env__find_br_cntr_info(evsel__env(evsel), NULL, &width);
+		mask = (1L << width) - 1;
 		printed += fprintf(fp, "br_cntr: ");
 		evlist__for_each_entry_from(evsel->evlist, pos) {
 			if (!(pos->core.attr.branch_sample_type & PERF_SAMPLE_BRANCH_COUNTERS))
tools/perf/util/annotate.c +3 -2
index 4990c70b1794..c6ebde9d40a8 100644
@@ -333,14 +333,15 @@ static int symbol__account_br_cntr(struct annotated_branch *branch,
 {
 	unsigned int br_cntr_nr = evsel__leader(evsel)->br_cntr_nr;
 	unsigned int base = evsel__leader(evsel)->br_cntr_idx;
-	unsigned int width = evsel__env(evsel)->br_cntr_width;
 	unsigned int off = offset * evsel->evlist->nr_br_cntr;
-	unsigned int i, mask = (1L << width) - 1;
 	u64 *branch_br_cntr = branch->br_cntr;
+	unsigned int i, mask, width;
 
 	if (!br_cntr || !branch_br_cntr)
 		return 0;
 
+	perf_env__find_br_cntr_info(evsel__env(evsel), NULL, &width);
+	mask = (1L << width) - 1;
 	for (i = 0; i < br_cntr_nr; i++) {
 		u64 cntr = (br_cntr >> i * width) & mask;
 
tools/perf/util/env.c +15 -0
index a459374d0a1a..1edbccfc3281 100644
@@ -624,3 +624,18 @@ char *perf_env__find_pmu_cap(struct perf_env *env, const char *pmu_name,
 	free(cap_eq);
 	return NULL;
 }
+
+void perf_env__find_br_cntr_info(struct perf_env *env,
+				 unsigned int *nr,
+				 unsigned int *width)
+{
+	if (nr) {
+		*nr = env->cpu_pmu_caps ? env->br_cntr_nr :
+					  env->pmu_caps->br_cntr_nr;
+	}
+
+	if (width) {
+		*width = env->cpu_pmu_caps ? env->br_cntr_width :
+					     env->pmu_caps->br_cntr_width;
+	}
+}
tools/perf/util/env.h +3 -0
index 2a2c37cc40b7..51b36c36019b 100644
@@ -192,4 +192,7 @@ char *perf_env__find_pmu_cap(struct perf_env *env, const char *pmu_name,
 			     const char *cap);
 
 bool perf_env__has_pmu_mapping(struct perf_env *env, const char *pmu_name);
+void perf_env__find_br_cntr_info(struct perf_env *env,
+				 unsigned int *nr,
+				 unsigned int *width);
 #endif /* __PERF_ENV_H */
tools/perf/util/session.c +4 -2
index b492300ec959..dbaf07bf6c5f 100644
@@ -856,7 +856,6 @@ static void branch_stack__printf(struct perf_sample *sample,
 	struct branch_entry *entries = perf_sample__branch_entries(sample);
 	bool callstack = evsel__has_branch_callstack(evsel);
 	u64 *branch_stack_cntr = sample->branch_stack_cntr;
-	struct perf_env *env = evsel__env(evsel);
 	uint64_t i;
 
 	if (!callstack) {
@@ -900,8 +899,11 @@ static void branch_stack__printf(struct perf_sample *sample,
 	}
 
 	if (branch_stack_cntr) {
+		unsigned int br_cntr_width, br_cntr_nr;
+
+		perf_env__find_br_cntr_info(evsel__env(evsel), &br_cntr_nr, &br_cntr_width);
 		printf("... branch stack counters: nr:%" PRIu64 " (counter width: %u max counter nr:%u)\n",
-			sample->branch_stack->nr, env->br_cntr_width, env->br_cntr_nr);
+			sample->branch_stack->nr, br_cntr_width, br_cntr_nr);
 		for (i = 0; i < sample->branch_stack->nr; i++)
 			printf("..... %2"PRIu64": %016" PRIx64 "\n", i, branch_stack_cntr[i]);
 	}
Signature · 1 line
2.38.1
author
#5 lore raw
Quoted 55 lines · Arnaldo Carvalho de Melo
Quoted 46 lines · Kan
Quoted 30 lines · Arnaldo Carvalho de Melo
Quoted 8 lines · On Mon, Sep 09, 2024 at 11:42:00AM -0700, [email protected] wrote:

From: Kan Liang <[email protected]>

No event is printed in the "Branch Counter" column on hybrid machines.

For example,
$perf record -e "{cpu_core/branch-instructions/pp,cpu_core/branches/}:S" -j any,counter
$perf report --total-cycles

So I tried this on an Intel Hybrid system, I start the TUI, then try 'B' and, after applying your patch:

  0.66%  67.8K  0.00%   75                    ┌─Branch counter abbr list────────────┐ibc_write+18]          libc.so.6
  0.63%  65.4K  0.00%  106                  [p│                                     │r_active+159]          [kernel.kallsyms]
  0.61%  63.2K  0.01%  175  [native_queued_spi│ The branch counter is not available.│in_lock_slowpath+528]  [kernel.kallsyms]
  0.57%  58.4K  0.00%   39         [syscall_ex│Press any key...                     │ser_mode+109]          [kernel.kallsyms]
  0.54%  55.6K  0.00%    1  [_IO_file_xsputn@@└─────────────────────────────────────┘BC_2.2.5+316]          libc.so.6

The perf.data was obtained from:

root@number:~# perf record -e "{cpu_core/branch-instructions/pp,cpu_core/branches/}:S" -j any,counter find /

And:

root@number:~# perf evlist --group 
{cpu_core/branch-instructions/pp,cpu_core/branches/}
dummy:u
root@number:~# 
Quoted 4 lines

Only mishap above we discussed already, the missing :S, and then:

root@number:~# perf evlist -v
cpu_core/branch-instructions/pp: type: 4 (cpu_core), size: 136, config: 0xc4 (branch-instructions), { sample_period, sample_freq }: 4000, sample_type: IP|TID|TIME|READ|PERIOD|BRANCH_STACK|IDENTIFIER, read_format: ID|GROUP|LOST, disabled: 1, freq: 1, enable_on_exec: 1, precise_ip: 2, sample_id_all: 1, exclude_guest: 1, branch_sample_type: ANY

For a hybrid client, the "Branch Counter" feature is only supported starting from the just released Lunar Lake. Perf falls back to only "ANY" on your Raptor Lake.

The "The branch counter is not available" message is expected.

Here is the perf evlist result from my Lunar Lake machine,

Quoted 7 lines

#perf evlist -v
cpu_core/branch-instructions/pp: type: 4 (cpu_core), size: 136, config:
0xc4 (branch-instructions), { sample_period, sample_freq }: 4000,
sample_type: IP|TID|TIME|READ|PERIOD|BRANCH_STACK|IDENTIFIER,
read_format: ID|GROUP|LOST, disabled: 1, freq: 1, enable_on_exec: 1,
precise_ip: 2, sample_id_all: 1, exclude_guest: 1, branch_sample_type:
ANY|COUNTERS

Quoted 2 lines

So I'll just take the patch and add these explanations there, ok?

Sounds good. Thanks!

Kan

Quoted 158 lines

- Arnaldo

Quoted 156 lines

Thanks,
Kan

Quoted 153 lines

cpu_core/branches/: type: 0 (PERF_TYPE_HARDWARE), size: 136, config: 0x400000004, sample_type: IP|TID|TIME|READ|PERIOD|BRANCH_STACK|IDENTIFIER, read_format: ID|GROUP|LOST, sample_id_all: 1, exclude_guest: 1, branch_sample_type: ANY
dummy:u: type: 1 (software), size: 136, config: 0x9 (PERF_COUNT_SW_DUMMY), { sample_period, sample_freq }: 1, sample_type: IP|TID|TIME|IDENTIFIER, read_format: ID|LOST, inherit: 1, exclude_kernel: 1, exclude_hv: 1, mmap: 1, comm: 1, task: 1, sample_id_all: 1, exclude_guest: 1, mmap2: 1, comm_exec: 1, ksymbol: 1, bpf_event: 1 root@number:~#

Version: Intel(R) Core(TM) i7-14700K
Voltage: 1.0 V
External Clock: 100 MHz
Max Speed: 8500 MHz
Current Speed: 3400 MHz
Status: Populated, Enabled
Upgrade: Socket LGA1700

What am I doing wrong not to get to see that "branch counter abbr list" info to then check before/after your patch?

- Arnaldo

Quoted 135 lines

# Branch counter abbr list:

 # cpu_core/branch-instructions/pp = A
 # cpu_core/branches/ = B
 # '-' No event occurs
 # '+' Event occurrences may be lost due to branch counter saturated
 #
 # Sampled Cycles%  Sampled Cycles  Avg Cycles%  Avg Cycles  Branch Counter
 # ...............  ..............  ...........  ..........  ..............
           44.54%          727.1K        0.00%           1   |+   |+   |
           36.31%          592.7K        0.00%           2   |+   |+   |
           17.83%          291.1K        0.00%           1   |+   |+   |

The branch counter information (br_cntr_width and br_cntr_nr) in the perf_env is retrieved from the CPU_PMU_CAPS. However, the CPU_PMU_CAPS is not available on hybrid machines. Without the width information, the number of occurrences of an event cannot be calculated.

For a hybrid machine, the caps information should be retrieved from the PMU_CAPS, and stored in the perf_env->pmu_caps.

Add a perf_env__find_br_cntr_info() to return the correct branch counter information from the corresponding fields.

Fixes 6f9d8d1de2c6 ("perf script: Add branch counters")
---
 tools/perf/builtin-script.c |  5 +++--
 tools/perf/util/annotate.c  |  5 +++--
 tools/perf/util/env.c       | 15 +++++++++++++++
 tools/perf/util/env.h       |  3 +++
 tools/perf/util/session.c   |  6 ++++--
 5 files changed, 28 insertions(+), 6 deletions(-)
tools/perf/builtin-script.c +3 -2
index dbe792b52c5c..a644787fa9e1 100644
@@ -1241,10 +1241,11 @@ static int ip__fprintf_jump(uint64_t ip, struct branch_entry *en,
 	}
 
 	if (PRINT_FIELD(BRCNTR)) {
-		unsigned int width = evsel__env(evsel)->br_cntr_width;
-		unsigned int i = 0, j, num, mask = (1L << width) - 1;
 		struct evsel *pos = evsel__leader(evsel);
+		unsigned int i = 0, j, num, mask, width;
 
+		perf_env__find_br_cntr_info(evsel__env(evsel), NULL, &width);
+		mask = (1L << width) - 1;
 		printed += fprintf(fp, "br_cntr: ");
 		evlist__for_each_entry_from(evsel->evlist, pos) {
 			if (!(pos->core.attr.branch_sample_type & PERF_SAMPLE_BRANCH_COUNTERS))
tools/perf/util/annotate.c +3 -2
index 4990c70b1794..c6ebde9d40a8 100644
@@ -333,14 +333,15 @@ static int symbol__account_br_cntr(struct annotated_branch *branch,
 {
 	unsigned int br_cntr_nr = evsel__leader(evsel)->br_cntr_nr;
 	unsigned int base = evsel__leader(evsel)->br_cntr_idx;
-	unsigned int width = evsel__env(evsel)->br_cntr_width;
 	unsigned int off = offset * evsel->evlist->nr_br_cntr;
-	unsigned int i, mask = (1L << width) - 1;
 	u64 *branch_br_cntr = branch->br_cntr;
+	unsigned int i, mask, width;
 
 	if (!br_cntr || !branch_br_cntr)
 		return 0;
 
+	perf_env__find_br_cntr_info(evsel__env(evsel), NULL, &width);
+	mask = (1L << width) - 1;
 	for (i = 0; i < br_cntr_nr; i++) {
 		u64 cntr = (br_cntr >> i * width) & mask;
 
tools/perf/util/env.c +15 -0
index a459374d0a1a..1edbccfc3281 100644
@@ -624,3 +624,18 @@ char *perf_env__find_pmu_cap(struct perf_env *env, const char *pmu_name,
 	free(cap_eq);
 	return NULL;
 }
+
+void perf_env__find_br_cntr_info(struct perf_env *env,
+				 unsigned int *nr,
+				 unsigned int *width)
+{
+	if (nr) {
+		*nr = env->cpu_pmu_caps ? env->br_cntr_nr :
+					  env->pmu_caps->br_cntr_nr;
+	}
+
+	if (width) {
+		*width = env->cpu_pmu_caps ? env->br_cntr_width :
+					     env->pmu_caps->br_cntr_width;
+	}
+}
tools/perf/util/env.h +3 -0
index 2a2c37cc40b7..51b36c36019b 100644
@@ -192,4 +192,7 @@ char *perf_env__find_pmu_cap(struct perf_env *env, const char *pmu_name,
 			     const char *cap);
 
 bool perf_env__has_pmu_mapping(struct perf_env *env, const char *pmu_name);
+void perf_env__find_br_cntr_info(struct perf_env *env,
+				 unsigned int *nr,
+				 unsigned int *width);
 #endif /* __PERF_ENV_H */
tools/perf/util/session.c +4 -2
index b492300ec959..dbaf07bf6c5f 100644
@@ -856,7 +856,6 @@ static void branch_stack__printf(struct perf_sample *sample,
 	struct branch_entry *entries = perf_sample__branch_entries(sample);
 	bool callstack = evsel__has_branch_callstack(evsel);
 	u64 *branch_stack_cntr = sample->branch_stack_cntr;
-	struct perf_env *env = evsel__env(evsel);
 	uint64_t i;
 
 	if (!callstack) {
@@ -900,8 +899,11 @@ static void branch_stack__printf(struct perf_sample *sample,
 	}
 
 	if (branch_stack_cntr) {
+		unsigned int br_cntr_width, br_cntr_nr;
+
+		perf_env__find_br_cntr_info(evsel__env(evsel), &br_cntr_nr, &br_cntr_width);
 		printf("... branch stack counters: nr:%" PRIu64 " (counter width: %u max counter nr:%u)\n",
-			sample->branch_stack->nr, env->br_cntr_width, env->br_cntr_nr);
+			sample->branch_stack->nr, br_cntr_width, br_cntr_nr);
 		for (i = 0; i < sample->branch_stack->nr; i++)
 			printf("..... %2"PRIu64": %016" PRIx64 "\n", i, branch_stack_cntr[i]);
 	}
Signature · 1 line
2.38.1
#6 lore raw
In reply to Arnaldo Carvalho de Melo
Quoted 53 lines · Arnaldo Carvalho de Melo
Quoted 44 lines · Kan
Quoted 30 lines · Arnaldo Carvalho de Melo
Quoted 8 lines · On Mon, Sep 09, 2024 at 11:42:00AM -0700, [email protected] wrote:

From: Kan Liang <[email protected]>

No event is printed in the "Branch Counter" column on hybrid machines.

For example,
$perf record -e "{cpu_core/branch-instructions/pp,cpu_core/branches/}:S" -j any,counter
$perf report --total-cycles

So I tried this on an Intel Hybrid system, I start the TUI, then try 'B' and, after applying your patch:

  0.66%  67.8K  0.00%   75                    ┌─Branch counter abbr list────────────┐ibc_write+18]          libc.so.6
  0.63%  65.4K  0.00%  106                  [p│                                     │r_active+159]          [kernel.kallsyms]
  0.61%  63.2K  0.01%  175  [native_queued_spi│ The branch counter is not available.│in_lock_slowpath+528]  [kernel.kallsyms]
  0.57%  58.4K  0.00%   39         [syscall_ex│Press any key...                     │ser_mode+109]          [kernel.kallsyms]
  0.54%  55.6K  0.00%    1  [_IO_file_xsputn@@└─────────────────────────────────────┘BC_2.2.5+316]          libc.so.6

The perf.data was obtained from:

root@number:~# perf record -e "{cpu_core/branch-instructions/pp,cpu_core/branches/}:S" -j any,counter find /

And:

root@number:~# perf evlist --group 
{cpu_core/branch-instructions/pp,cpu_core/branches/}
dummy:u
root@number:~# 
Quoted 4 lines

Only mishap above we discussed already, the missing :S, and then:

root@number:~# perf evlist -v
cpu_core/branch-instructions/pp: type: 4 (cpu_core), size: 136, config: 0xc4 (branch-instructions), { sample_period, sample_freq }: 4000, sample_type: IP|TID|TIME|READ|PERIOD|BRANCH_STACK|IDENTIFIER, read_format: ID|GROUP|LOST, disabled: 1, freq: 1, enable_on_exec: 1, precise_ip: 2, sample_id_all: 1, exclude_guest: 1, branch_sample_type: ANY

For a hybrid client, the "Branch Counter" feature is only supported starting from the just released Lunar Lake. Perf falls back to only "ANY" on your Raptor Lake.

The "The branch counter is not available" message is expected.

Here is the perf evlist result from my Lunar Lake machine,

Quoted 7 lines

#perf evlist -v
cpu_core/branch-instructions/pp: type: 4 (cpu_core), size: 136, config:
0xc4 (branch-instructions), { sample_period, sample_freq }: 4000,
sample_type: IP|TID|TIME|READ|PERIOD|BRANCH_STACK|IDENTIFIER,
read_format: ID|GROUP|LOST, disabled: 1, freq: 1, enable_on_exec: 1,
precise_ip: 2, sample_id_all: 1, exclude_guest: 1, branch_sample_type:
ANY|COUNTERS

Quoted 1 line

So I'll just take the patch and add these explanations there, ok?

Added this:

Committer notes:

While testing I couldn't s ee those "Branch counter" columns enabled by pressing 'B' on the TUI, after reporting it to the list Kan explained the situation:

<quote Kan Liang>
For a hybrid client, the "Branch Counter" feature is only supported starting from the just released Lunar Lake. Perf falls back to only "ANY" on your Raptor Lake.

The "The branch counter is not available" message is expected.

Here is the 'perf evlist' result from my Lunar Lake machine,

  # perf evlist -v
  cpu_core/branch-instructions/pp: type: 4 (cpu_core), size: 136, config: 0xc4 (branch-instructions), { sample_period, sample_freq }: 4000, sample_type: IP|TID|TIME|READ|PERIOD|BRANCH_STACK|IDENTIFIER, read_format: ID|GROUP|LOST, disabled: 1, freq: 1, enable_on_exec: 1, precise_ip: 2, sample_id_all: 1, exclude_guest: 1, branch_sample_type: ANY|COUNTERS
  #
</quote>