Skip to contentSkip to conversation

perf metric fixes and test

PATCH Superseded Ian Rogers 6 participants 36 messages Apr 22, 2020 – May 6, 2020 perf events

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Ian Rogers

author reviewer
#0 lore raw

v2 cover perf metric fixes and test

Add a test that all perf metrics (for your architecture) are parsable. Fix bugs in the expr parser and in x86 metrics. Untested on architectures other than x86.

v2 adds Fixes tags to commit messages for when broken metrics were
first added. Adds a debug warning for division by zero in expr, and adds a workaround for id values in the expr test necessary for powerpc. It also fixes broken power8 and power9 metrics.

Ian Rogers (11):
perf expr: unlimited escaped characters in a symbol perf metrics: fix parse errors in cascade lake metrics perf metrics: fix parse errors in skylake metrics perf expr: allow ',' to be an other token perf expr: increase max other
perf expr: parse numbers as doubles
perf expr: debug lex if debugging yacc
perf metrics: fix parse errors in power8 metrics perf metrics: fix parse errors in power9 metrics perf expr: print a debug message for division by zero perf test: add expr test for pmu metrics

 .../arch/powerpc/power8/metrics.json          |  4 +-
 .../arch/powerpc/power9/metrics.json          |  2 +-
 .../arch/x86/cascadelakex/clx-metrics.json    | 10 +-
 .../arch/x86/skylakex/skx-metrics.json        |  4 +-
 tools/perf/tests/builtin-test.c               |  5 +
 tools/perf/tests/expr.c                       | 96 ++++++++++++++++++-
 tools/perf/tests/tests.h                      |  2 +
 tools/perf/util/expr.c                        |  1 +
 tools/perf/util/expr.h                        |  2 +-
 tools/perf/util/expr.l                        | 16 ++--
 tools/perf/util/expr.y                        | 16 +++-
 11 files changed, 135 insertions(+), 23 deletions(-)
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
author reviewer
#1 lore raw

PATCH v2 1/11 perf expr: unlimited escaped characters in a symbol

Current expression allows 2 escaped '-,=' characters. However, some metrics require more, for example Haswell DRAM_BW_Use.

Fixes 26226a97724d (perf expr: Move expr lexer to flex)
---
 tools/perf/util/expr.l | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/expr.l +1 -1
index 2582c2464938..95bcf3629edf 100644
@@ -73,7 +73,7 @@ number		[0-9]+
 sch		[-,=]
 spec		\\{sch}
 sym		[0-9a-zA-Z_\.:@]+
-symbol		{spec}*{sym}*{spec}*{sym}*
+symbol		({spec}|{sym})+
 
 %%
 	struct expr_scanner_ctx *sctx = expr_get_extra(yyscanner);
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
#2 lore raw

Re: [PATCH v2 01/11] perf expr: unlimited escaped characters in a symbol

Quoted 14 lines · Ian Rogers

Current expression allows 2 escaped '-,=' characters. However, some metrics require more, for example Haswell DRAM_BW_Use.

Fixes 26226a97724d (perf expr: Move expr lexer to flex)
---
 tools/perf/util/expr.l | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/expr.l +0 -0
index 2582c2464938..95bcf3629edf 100644

tools/perf/util/expr.l number [0-9]+ · line 76

Quoted 5 lines
tools/perf/util/expr.l +1 -1
@@ -73,7 +73,7 @@ number		[0-9]+
 sch		[-,=]
 spec		\\{sch}
 sym		[0-9a-zA-Z_\.:@]+
-symbol		{spec}*{sym}*{spec}*{sym}*
+symbol		({spec}|{sym})+

yep, much better ;-)

thanks,
jirka

Quoted 6 lines
tools/perf/util/expr.l +0 -0
 %%
 	struct expr_scanner_ctx *sctx = expr_get_extra(yyscanner);
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
#3 lore raw

Re: [PATCH v2 01/11] perf expr: unlimited escaped characters in a symbol

Quoted 10 lines · Jiri Olsa
Quoted 9 lines · Ian Rogers

Current expression allows 2 escaped '-,=' characters. However, some metrics require more, for example Haswell DRAM_BW_Use.

Fixes 26226a97724d (perf expr: Move expr lexer to flex)
---
 tools/perf/util/expr.l | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
Quoted 10 lines

tools/perf/util/expr.l number [0-9]+ · line 76

Quoted 10 lines
tools/perf/util/expr.l +1 -1
index 2582c2464938..95bcf3629edf 100644
@@ -73,7 +73,7 @@ number		[0-9]+
 sch		[-,=]
 spec		\\{sch}
 sym		[0-9a-zA-Z_\.:@]+
-symbol		{spec}*{sym}*{spec}*{sym}*
+symbol		({spec}|{sym})+

Tested by: [email protected]

Thanks,
Kajol Jain

Quoted 13 lines

yep, much better ;-)

thanks,
jirka

Quoted 6 lines
tools/perf/util/expr.l +0 -0
 %%
 	struct expr_scanner_ctx *sctx = expr_get_extra(yyscanner);
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
author reviewer
#4 lore raw

PATCH v2 2/11 perf metrics: fix parse errors in cascade lake metrics

In reply to Ian Rogers

Remove over escaping with \\.
Remove extraneous if 1 if 0 == 1 else 0 else 0.

Fixes fd5500989c8f (perf vendor events intel: Update metrics from TMAM 3.5)
---
 .../pmu-events/arch/x86/cascadelakex/clx-metrics.json  | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
tools/perf/pmu-events/arch/x86/cascadelakex/clx-metrics.json +5 -5
index 7fde0d2943cd..d25eebce34c9 100644
@@ -328,31 +328,31 @@
     },
     {
         "BriefDescription": "Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches",
-        "MetricExpr": "1000000000 * ( cha@event\\=0x36\\\\\\,umask\\=0x21@ / cha@event\\=0x35\\\\\\,umask\\=0x21@ ) / ( cha_0@event\\=0x0@ / duration_time )",
+        "MetricExpr": "1000000000 * ( cha@event\\=0x36\\,umask\\=0x21@ / cha@event\\=0x35\\,umask\\=0x21@ ) / ( cha_0@event\\=0x0@ / duration_time )",
         "MetricGroup": "Memory_Lat",
         "MetricName": "DRAM_Read_Latency"
     },
     {
         "BriefDescription": "Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches",
-        "MetricExpr": "cha@event\\=0x36\\\\\\,umask\\=0x21@ / cha@event\\=0x36\\\\\\,umask\\=0x21\\\\\\,thresh\\=1@",
+        "MetricExpr": "cha@event\\=0x36\\,umask\\=0x21@ / cha@event\\=0x36\\,umask\\=0x21\\,thresh\\=1@",
         "MetricGroup": "Memory_BW",
         "MetricName": "DRAM_Parallel_Reads"
     },
     {
         "BriefDescription": "Average latency of data read request to external 3D X-Point memory [in nanoseconds]. Accounts for demand loads and L1/L2 data-read prefetches",
-        "MetricExpr": "( 1000000000 * ( imc@event\\=0xe0\\\\\\,umask\\=0x1@ / imc@event\\=0xe3@ ) / imc_0@event\\=0x0@ ) if 1 if 0 == 1 else 0 else 0",
+        "MetricExpr": "( 1000000000 * ( imc@event\\=0xe0\\,umask\\=0x1@ / imc@event\\=0xe3@ ) / imc_0@event\\=0x0@ )",
         "MetricGroup": "Memory_Lat",
         "MetricName": "MEM_PMM_Read_Latency"
     },
     {
         "BriefDescription": "Average 3DXP Memory Bandwidth Use for reads [GB / sec]",
-        "MetricExpr": "( ( 64 * imc@event\\=0xe3@ / 1000000000 ) / duration_time ) if 1 if 0 == 1 else 0 else 0",
+        "MetricExpr": "( ( 64 * imc@event\\=0xe3@ / 1000000000 ) / duration_time )",
         "MetricGroup": "Memory_BW",
         "MetricName": "PMM_Read_BW"
     },
     {
         "BriefDescription": "Average 3DXP Memory Bandwidth Use for Writes [GB / sec]",
-        "MetricExpr": "( ( 64 * imc@event\\=0xe7@ / 1000000000 ) / duration_time ) if 1 if 0 == 1 else 0 else 0",
+        "MetricExpr": "( ( 64 * imc@event\\=0xe7@ / 1000000000 ) / duration_time )",
         "MetricGroup": "Memory_BW",
         "MetricName": "PMM_Write_BW"
     },
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
author reviewer
#5 lore raw

PATCH v2 3/11 perf metrics: fix parse errors in skylake metrics

In reply to Ian Rogers

Remove over escaping with \\.

Fixes fd5500989c8f (perf vendor events intel: Update metrics from TMAM 3.5)
---
 tools/perf/pmu-events/arch/x86/skylakex/skx-metrics.json | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
tools/perf/pmu-events/arch/x86/skylakex/skx-metrics.json +2 -2
index b4f91137f40c..390bdab1be9d 100644
@@ -328,13 +328,13 @@
     },
     {
         "BriefDescription": "Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches",
-        "MetricExpr": "1000000000 * ( cha@event\\=0x36\\\\\\,umask\\=0x21@ / cha@event\\=0x35\\\\\\,umask\\=0x21@ ) / ( cha_0@event\\=0x0@ / duration_time )",
+        "MetricExpr": "1000000000 * ( cha@event\\=0x36\\,umask\\=0x21@ / cha@event\\=0x35\\,umask\\=0x21@ ) / ( cha_0@event\\=0x0@ / duration_time )",
         "MetricGroup": "Memory_Lat",
         "MetricName": "DRAM_Read_Latency"
     },
     {
         "BriefDescription": "Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches",
-        "MetricExpr": "cha@event\\=0x36\\\\\\,umask\\=0x21@ / cha@event\\=0x36\\\\\\,umask\\=0x21\\\\\\,thresh\\=1@",
+        "MetricExpr": "cha@event\\=0x36\\,umask\\=0x21@ / cha@event\\=0x36\\,umask\\=0x21\\,thresh\\=1@",
         "MetricGroup": "Memory_BW",
         "MetricName": "DRAM_Parallel_Reads"
     },
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
author reviewer
#6 lore raw

PATCH v2 4/11 perf expr: allow ',' to be an other token

In reply to Ian Rogers

Corrects parse errors in expr__find_other of expressions with min.

---
 tools/perf/util/expr.y | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/expr.y +1 -1
index cd17486c1c5d..54260094b947 100644
@@ -80,7 +80,7 @@ other: ID
 	ctx->ids[ctx->num_ids++].name = $1;
 }
 |
-MIN | MAX | IF | ELSE | SMT_ON | NUMBER | '|' | '^' | '&' | '-' | '+' | '*' | '/' | '%' | '(' | ')'
+MIN | MAX | IF | ELSE | SMT_ON | NUMBER | '|' | '^' | '&' | '-' | '+' | '*' | '/' | '%' | '(' | ')' | ','
 
 
 all_expr: if_expr			{ *final_val = $1; }
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
#7 lore raw

Re: [PATCH v2 04/11] perf expr: allow ',' to be an other token

Quoted 12 lines · Ian Rogers

Corrects parse errors in expr__find_other of expressions with min.

---
 tools/perf/util/expr.y | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/expr.y +0 -0
index cd17486c1c5d..54260094b947 100644

tools/perf/util/expr.y other: ID · line 83

Quoted 5 lines
tools/perf/util/expr.y +1 -1
@@ -80,7 +80,7 @@ other: ID
 	ctx->ids[ctx->num_ids++].name = $1;
 }
 |
-MIN | MAX | IF | ELSE | SMT_ON | NUMBER | '|' | '^' | '&' | '-' | '+' | '*' | '/' | '%' | '(' | ')'
+MIN | MAX | IF | ELSE | SMT_ON | NUMBER | '|' | '^' | '&' | '-' | '+' | '*' | '/' | '%' | '(' | ')' | ','

ugh, thanks

jirka

Quoted 6 lines
tools/perf/util/expr.y +0 -0
 all_expr: if_expr			{ *final_val = $1; }
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
author reviewer
#8 lore raw

PATCH v2 5/11 perf expr: increase max other

In reply to Ian Rogers

Large metrics such as Branch_Misprediction_Cost_SMT on x86 broadwell need more space.

---
 tools/perf/util/expr.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/expr.h +1 -1
index 0938ad166ece..4938bfc608b7 100644
@@ -2,7 +2,7 @@
 #ifndef PARSE_CTX_H
 #define PARSE_CTX_H 1
 
-#define EXPR_MAX_OTHER 20
+#define EXPR_MAX_OTHER 64
 #define MAX_PARSE_ID EXPR_MAX_OTHER
 
 struct expr_parse_id {
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
#9 lore raw

Re: [PATCH v2 05/11] perf expr: increase max other

Quoted 24 lines · Ian Rogers

Large metrics such as Branch_Misprediction_Cost_SMT on x86 broadwell need more space.

---
 tools/perf/util/expr.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/expr.h +1 -1
index 0938ad166ece..4938bfc608b7 100644
@@ -2,7 +2,7 @@
 #ifndef PARSE_CTX_H
 #define PARSE_CTX_H 1
 
-#define EXPR_MAX_OTHER 20
+#define EXPR_MAX_OTHER 64
 #define MAX_PARSE_ID EXPR_MAX_OTHER
 
 struct expr_parse_id {
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog

ok, and we should probably start to think about what Andi suggested
in here: https://lore.kernel.org/lkml/[email protected]/

jirka

author reviewer
#10 lore raw

Re: [PATCH v2 05/11] perf expr: increase max other

Quoted 22 lines · Jiri Olsa
Quoted 20 lines · Ian Rogers

Large metrics such as Branch_Misprediction_Cost_SMT on x86 broadwell need more space.

---
 tools/perf/util/expr.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/expr.h +1 -1
index 0938ad166ece..4938bfc608b7 100644
@@ -2,7 +2,7 @@
 #ifndef PARSE_CTX_H
 #define PARSE_CTX_H 1
 
-#define EXPR_MAX_OTHER 20
+#define EXPR_MAX_OTHER 64
 #define MAX_PARSE_ID EXPR_MAX_OTHER
 
Quoted 7 lines
Quoted 4 lines
tools/perf/util/expr.h +0 -1
 struct expr_parse_id {
--

2.26.2.303.gf8c07b1a785-goog

ok, and we should probably start to think about what Andi suggested
in here: https://lore.kernel.org/lkml/[email protected]/

Agreed, a hash table would make sense. This was the smallest value that would let the test on x86 pass.

Thanks,
Ian

Quoted 2 lines

jirka

author reviewer
#11 lore raw

Re: [PATCH v2 05/11] perf expr: increase max other

Quoted 32 lines · Ian Rogers
Quoted 22 lines · Jiri Olsa
Quoted 20 lines · Ian Rogers

Large metrics such as Branch_Misprediction_Cost_SMT on x86 broadwell need more space.

---
 tools/perf/util/expr.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/util/expr.h +1 -1
index 0938ad166ece..4938bfc608b7 100644
@@ -2,7 +2,7 @@
 #ifndef PARSE_CTX_H
 #define PARSE_CTX_H 1
 
-#define EXPR_MAX_OTHER 20
+#define EXPR_MAX_OTHER 64
 #define MAX_PARSE_ID EXPR_MAX_OTHER
 
Quoted 7 lines
Quoted 4 lines
tools/perf/util/expr.h +0 -1
 struct expr_parse_id {
--

2.26.2.303.gf8c07b1a785-goog

ok, and we should probably start to think about what Andi suggested
in here: https://lore.kernel.org/lkml/[email protected]/

Quoted 2 lines

Agreed, a hash table would make sense. This was the smallest value that would let the test on x86 pass.

Fwiw, I have done this based on tools/lib/bpf/hashmap.h in CLs following on from this patch set. I'm holding off sending so I can rebase on acme's perf/next when the CLs acked by Jirka land:
https://lore.kernel.org/lkml/20200503221650.GA1916255@krava/
The libbpf dependency for a hashmap is counter intuitive, so maybe there's something better to do there.

Thanks,
Ian

Quoted 5 lines

Thanks,
Ian

Quoted 2 lines

jirka

author reviewer
#12 lore raw

PATCH v2 6/11 perf expr: parse numbers as doubles

In reply to Ian Rogers

This is expected in expr.y and metrics use floating point values such as x86 broadwell IFetch_Line_Utilization.

Fixes 26226a97724d (perf expr: Move expr lexer to flex)
---
 tools/perf/util/expr.l | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)
tools/perf/util/expr.l +7 -7
index 95bcf3629edf..0efda2ce2766 100644
@@ -10,12 +10,12 @@
 char *expr_get_text(yyscan_t yyscanner);
 YYSTYPE *expr_get_lval(yyscan_t yyscanner);
 
-static int __value(YYSTYPE *yylval, char *str, int base, int token)
+static double __value(YYSTYPE *yylval, char *str, int token)
 {
-	u64 num;
+	double num;
 
 	errno = 0;
-	num = strtoull(str, NULL, base);
+	num = strtod(str, NULL);
 	if (errno)
 		return EXPR_ERROR;
 
@@ -23,12 +23,12 @@ static int __value(YYSTYPE *yylval, char *str, int base, int token)
 	return token;
 }
 
-static int value(yyscan_t scanner, int base)
+static int value(yyscan_t scanner)
 {
 	YYSTYPE *yylval = expr_get_lval(scanner);
 	char *text = expr_get_text(scanner);
 
-	return __value(yylval, text, base, NUMBER);
+	return __value(yylval, text, NUMBER);
 }
 
 /*
@@ -68,7 +68,7 @@ static int str(yyscan_t scanner, int token)
 }
 %}
 
-number		[0-9]+
+number		[0-9]*\.?[0-9]+
 
 sch		[-,=]
 spec		\\{sch}
@@ -92,7 +92,7 @@ min		{ return MIN; }
 if		{ return IF; }
 else		{ return ELSE; }
 #smt_on		{ return SMT_ON; }
-{number}	{ return value(yyscanner, 10); }
+{number}	{ return value(yyscanner); }
 {symbol}	{ return str(yyscanner, ID); }
 "|"		{ return '|'; }
 "^"		{ return '^'; }
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
#13 lore raw

Re: [PATCH v2 06/11] perf expr: parse numbers as doubles

Quoted 2 lines · Ian Rogers

This is expected in expr.y and metrics use floating point values such as x86 broadwell IFetch_Line_Utilization.

ugh, I wonder why we did not get any compiler warning when expr.y expects double.. good catch

jirka

Quoted 63 lines
Fixes 26226a97724d (perf expr: Move expr lexer to flex)
---
 tools/perf/util/expr.l | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)
tools/perf/util/expr.l +7 -7
index 95bcf3629edf..0efda2ce2766 100644
@@ -10,12 +10,12 @@
 char *expr_get_text(yyscan_t yyscanner);
 YYSTYPE *expr_get_lval(yyscan_t yyscanner);
 
-static int __value(YYSTYPE *yylval, char *str, int base, int token)
+static double __value(YYSTYPE *yylval, char *str, int token)
 {
-	u64 num;
+	double num;
 
 	errno = 0;
-	num = strtoull(str, NULL, base);
+	num = strtod(str, NULL);
 	if (errno)
 		return EXPR_ERROR;
 
@@ -23,12 +23,12 @@ static int __value(YYSTYPE *yylval, char *str, int base, int token)
 	return token;
 }
 
-static int value(yyscan_t scanner, int base)
+static int value(yyscan_t scanner)
 {
 	YYSTYPE *yylval = expr_get_lval(scanner);
 	char *text = expr_get_text(scanner);
 
-	return __value(yylval, text, base, NUMBER);
+	return __value(yylval, text, NUMBER);
 }
 
 /*
@@ -68,7 +68,7 @@ static int str(yyscan_t scanner, int token)
 }
 %}
 
-number		[0-9]+
+number		[0-9]*\.?[0-9]+
 
 sch		[-,=]
 spec		\\{sch}
@@ -92,7 +92,7 @@ min		{ return MIN; }
 if		{ return IF; }
 else		{ return ELSE; }
 #smt_on		{ return SMT_ON; }
-{number}	{ return value(yyscanner, 10); }
+{number}	{ return value(yyscanner); }
 {symbol}	{ return str(yyscanner, ID); }
 "|"		{ return '|'; }
 "^"		{ return '^'; }
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
#14 lore raw

Re: [PATCH v2 06/11] perf expr: parse numbers as doubles

In reply to Ian Rogers
Quoted 45 lines · Ian Rogers

This is expected in expr.y and metrics use floating point values such as x86 broadwell IFetch_Line_Utilization.

Fixes 26226a97724d (perf expr: Move expr lexer to flex)
---
 tools/perf/util/expr.l | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)
tools/perf/util/expr.l +5 -5
index 95bcf3629edf..0efda2ce2766 100644
@@ -10,12 +10,12 @@
 char *expr_get_text(yyscan_t yyscanner);
 YYSTYPE *expr_get_lval(yyscan_t yyscanner);
 
-static int __value(YYSTYPE *yylval, char *str, int base, int token)
+static double __value(YYSTYPE *yylval, char *str, int token)
 {
-	u64 num;
+	double num;
 
 	errno = 0;
-	num = strtoull(str, NULL, base);
+	num = strtod(str, NULL);
 	if (errno)
 		return EXPR_ERROR;
 
@@ -23,12 +23,12 @@ static int __value(YYSTYPE *yylval, char *str, int base, int token)
 	return token;
 }
 
-static int value(yyscan_t scanner, int base)
+static int value(yyscan_t scanner)
 {
 	YYSTYPE *yylval = expr_get_lval(scanner);
 	char *text = expr_get_text(scanner);
 
-	return __value(yylval, text, base, NUMBER);
+	return __value(yylval, text, NUMBER);
 }
 
 /*

tools/perf/util/expr.l static int str(yyscan_t scanner, int token) · line 72

Quoted 6 lines
tools/perf/util/expr.l +1 -1
@@ -68,7 +68,7 @@ static int str(yyscan_t scanner, int token)
 }
 %}
 
-number		[0-9]+
+number		[0-9]*\.?[0-9]+
 

Hi Ian,

    In this patch I saw the parsing of expression with '+number		[0-9]*\.?[0-9]+' 

Could you please explain why '?' is introduced here, so that I can be sure that this is not conflicting with my change to add '?'

In this patch : https://lkml.org/lkml/2020/4/1/1427
I have used '?' symbol as part of metric expression in order to replace '?' with runtime parameter.

Thanks,
Kajol Jain

Quoted 12 lines
tools/perf/util/expr.l +1 -1
 sch		[-,=]
 spec		\\{sch}
@@ -92,7 +92,7 @@ min		{ return MIN; }
 if		{ return IF; }
 else		{ return ELSE; }
 #smt_on		{ return SMT_ON; }
-{number}	{ return value(yyscanner, 10); }
+{number}	{ return value(yyscanner); }
 {symbol}	{ return str(yyscanner, ID); }
 "|"		{ return '|'; }
 "^"		{ return '^'; }
author reviewer
#15 lore raw

Re: [PATCH v2 06/11] perf expr: parse numbers as doubles

Quoted 50 lines · kajoljain
Quoted 48 lines · Ian Rogers

This is expected in expr.y and metrics use floating point values such as x86 broadwell IFetch_Line_Utilization.

Fixes 26226a97724d (perf expr: Move expr lexer to flex)
---
 tools/perf/util/expr.l | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)
tools/perf/util/expr.l +5 -5
index 95bcf3629edf..0efda2ce2766 100644
@@ -10,12 +10,12 @@
 char *expr_get_text(yyscan_t yyscanner);
 YYSTYPE *expr_get_lval(yyscan_t yyscanner);
 
-static int __value(YYSTYPE *yylval, char *str, int base, int token)
+static double __value(YYSTYPE *yylval, char *str, int token)
 {
-     u64 num;
+     double num;
 
      errno = 0;
-     num = strtoull(str, NULL, base);
+     num = strtod(str, NULL);
      if (errno)
              return EXPR_ERROR;
 
@@ -23,12 +23,12 @@ static int __value(YYSTYPE *yylval, char *str, int base, int token)
      return token;
 }
 
-static int value(yyscan_t scanner, int base)
+static int value(yyscan_t scanner)
 {
      YYSTYPE *yylval = expr_get_lval(scanner);
      char *text = expr_get_text(scanner);
 
-     return __value(yylval, text, base, NUMBER);
+     return __value(yylval, text, NUMBER);
 }
 
 /*
@@ -68,7 +68,7 @@ static int str(yyscan_t scanner, int token)
 }
 %}
 
Quoted 8 lines

tools/perf/util/expr.l static int str(yyscan_t scanner, int token) · line 72

Quoted 3 lines
tools/perf/util/expr.l +1 -1
-number               [0-9]+
+number               [0-9]*\.?[0-9]+
 

Hi Ian,
In this patch I saw the parsing of expression with '+number [0-9]*\.?[0-9]+'
Could you please explain why '?' is introduced here, so that I can be sure that this is not conflicting with my change to add '?'

Hi Kajol,
the '?' here is part of the regular expression. Basically it is saying that a number is a possible set of integers possibly followed by a '.' and then 1 or more integers. The expression comes from having seen Intel's metrics are of the form '.1234' in some of their topdown spreadsheets, so we need to be able to handle cases like '1.2', '123' and '.1234'. Having looked at your patch I don't believe it interferes.

Quoted 3 lines

In this patch : https://lkml.org/lkml/2020/4/1/1427
I have used '?' symbol as part of metric expression in order to replace '?' with runtime parameter.

Interesting, I'll follow up with comments on that e-mail.

Thanks!
Ian

Quoted 16 lines

Thanks,
Kajol Jain

Quoted 12 lines
tools/perf/util/expr.l +1 -1
 sch          [-,=]
 spec         \\{sch}
@@ -92,7 +92,7 @@ min         { return MIN; }
 if           { return IF; }
 else         { return ELSE; }
 #smt_on              { return SMT_ON; }
-{number}     { return value(yyscanner, 10); }
+{number}     { return value(yyscanner); }
 {symbol}     { return str(yyscanner, ID); }
 "|"          { return '|'; }
 "^"          { return '^'; }
#16 lore raw

Re: [PATCH v2 06/11] perf expr: parse numbers as doubles

In reply to Ian Rogers
Quoted 45 lines · Ian Rogers

This is expected in expr.y and metrics use floating point values such as x86 broadwell IFetch_Line_Utilization.

Fixes 26226a97724d (perf expr: Move expr lexer to flex)
---
 tools/perf/util/expr.l | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)
tools/perf/util/expr.l +5 -5
index 95bcf3629edf..0efda2ce2766 100644
@@ -10,12 +10,12 @@
 char *expr_get_text(yyscan_t yyscanner);
 YYSTYPE *expr_get_lval(yyscan_t yyscanner);
 
-static int __value(YYSTYPE *yylval, char *str, int base, int token)
+static double __value(YYSTYPE *yylval, char *str, int token)
 {
-	u64 num;
+	double num;
 
 	errno = 0;
-	num = strtoull(str, NULL, base);
+	num = strtod(str, NULL);
 	if (errno)
 		return EXPR_ERROR;
 
@@ -23,12 +23,12 @@ static int __value(YYSTYPE *yylval, char *str, int base, int token)
 	return token;
 }
 
-static int value(yyscan_t scanner, int base)
+static int value(yyscan_t scanner)
 {
 	YYSTYPE *yylval = expr_get_lval(scanner);
 	char *text = expr_get_text(scanner);
 
-	return __value(yylval, text, base, NUMBER);
+	return __value(yylval, text, NUMBER);
 }
 
 /*

tools/perf/util/expr.l static int str(yyscan_t scanner, int token) · line 72

Quoted 6 lines
tools/perf/util/expr.l +1 -1
@@ -68,7 +68,7 @@ static int str(yyscan_t scanner, int token)
 }
 %}
 
-number		[0-9]+
+number		[0-9]*\.?[0-9]+
 

Acked By: Kajol Jain <[email protected]>

Thanks,
Kajol Jain

Quoted 12 lines
tools/perf/util/expr.l +1 -1
 sch		[-,=]
 spec		\\{sch}
@@ -92,7 +92,7 @@ min		{ return MIN; }
 if		{ return IF; }
 else		{ return ELSE; }
 #smt_on		{ return SMT_ON; }
-{number}	{ return value(yyscanner, 10); }
+{number}	{ return value(yyscanner); }
 {symbol}	{ return str(yyscanner, ID); }
 "|"		{ return '|'; }
 "^"		{ return '^'; }
author reviewer
#17 lore raw

PATCH v2 7/11 perf expr: debug lex if debugging yacc

In reply to Ian Rogers

Only effects parser debugging (disabled by default). Enables displaying '--accepting rule at line .. ("...").

---
 tools/perf/util/expr.c | 1 +
 1 file changed, 1 insertion(+)
tools/perf/util/expr.c +1 -0
index c3382d58cf40..8694bec86f1a 100644
@@ -44,6 +44,7 @@ __expr__parse(double *val, struct expr_parse_ctx *ctx, const char *expr,
 
 #ifdef PARSER_DEBUG
 	expr_debug = 1;
+	expr_set_debug(1, scanner);
 #endif
 
 	ret = expr_parse(val, ctx, scanner);
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
author reviewer
#18 lore raw

PATCH v2 8/11 perf metrics: fix parse errors in power8 metrics

In reply to Ian Rogers

Mismatched parentheses.

Fixes dd81eafacc52 (perf vendor events power8: Cpi_breakdown & estimated_dcache_miss_cpi metrics)
---
 tools/perf/pmu-events/arch/powerpc/power8/metrics.json | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
tools/perf/pmu-events/arch/powerpc/power8/metrics.json +2 -2
index bffb2d4a6420..ad71486a38e3 100644
@@ -169,7 +169,7 @@
     },
     {
         "BriefDescription": "Cycles GCT empty where dispatch was held",
-        "MetricExpr": "(PM_GCT_NOSLOT_DISP_HELD_MAP + PM_GCT_NOSLOT_DISP_HELD_SRQ + PM_GCT_NOSLOT_DISP_HELD_ISSQ + PM_GCT_NOSLOT_DISP_HELD_OTHER) / PM_RUN_INST_CMPL)",
+        "MetricExpr": "(PM_GCT_NOSLOT_DISP_HELD_MAP + PM_GCT_NOSLOT_DISP_HELD_SRQ + PM_GCT_NOSLOT_DISP_HELD_ISSQ + PM_GCT_NOSLOT_DISP_HELD_OTHER) / PM_RUN_INST_CMPL",
         "MetricGroup": "cpi_breakdown",
         "MetricName": "gct_empty_disp_held_cpi"
     },
@@ -886,7 +886,7 @@
     },
     {
         "BriefDescription": "GCT slot utilization (11 to 14) as a % of cycles this thread had atleast 1 slot valid",
-        "MetricExpr": "PM_GCT_UTIL_11_14_ENTRIES / ( PM_RUN_CYC - PM_GCT_NOSLOT_CYC) * 100",
+        "MetricExpr": "PM_GCT_UTIL_11_14_ENTRIES / ( PM_RUN_CYC - PM_GCT_NOSLOT_CYC ) * 100",
         "MetricGroup": "general",
         "MetricName": "gct_util_11to14_slots_percent"
     },
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
Reviewed-by on v2 8/11
#19 lore raw

Re: [PATCH v2 08/11] perf metrics: fix parse errors in power8 metrics

Quoted 13 lines · Ian Rogers

Mismatched parentheses.

Fixes dd81eafacc52 (perf vendor events power8: Cpi_breakdown & estimated_dcache_miss_cpi metrics)
---
 tools/perf/pmu-events/arch/powerpc/power8/metrics.json | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
tools/perf/pmu-events/arch/powerpc/power8/metrics.json +0 -0
index bffb2d4a6420..ad71486a38e3 100644

tools/perf/pmu-events/arch/powerpc/power8/metrics.json line 172

Quoted 5 lines
tools/perf/pmu-events/arch/powerpc/power8/metrics.json +1 -1
@@ -169,7 +169,7 @@
     },
     {
         "BriefDescription": "Cycles GCT empty where dispatch was held",
-        "MetricExpr": "(PM_GCT_NOSLOT_DISP_HELD_MAP + PM_GCT_NOSLOT_DISP_HELD_SRQ + PM_GCT_NOSLOT_DISP_HELD_ISSQ + PM_GCT_NOSLOT_DISP_HELD_OTHER) / PM_RUN_INST_CMPL)",
+        "MetricExpr": "(PM_GCT_NOSLOT_DISP_HELD_MAP + PM_GCT_NOSLOT_DISP_HELD_SRQ + PM_GCT_NOSLOT_DISP_HELD_ISSQ + PM_GCT_NOSLOT_DISP_HELD_OTHER) / PM_RUN_INST_CMPL",

OK. (Thank you!)

tools/perf/pmu-events/arch/powerpc/power8/metrics.json line 889

Quoted 9 lines
tools/perf/pmu-events/arch/powerpc/power8/metrics.json +1 -1
         "MetricGroup": "cpi_breakdown",
         "MetricName": "gct_empty_disp_held_cpi"
     },
@@ -886,7 +886,7 @@
     },
     {
         "BriefDescription": "GCT slot utilization (11 to 14) as a % of cycles this thread had atleast 1 slot valid",
-        "MetricExpr": "PM_GCT_UTIL_11_14_ENTRIES / ( PM_RUN_CYC - PM_GCT_NOSLOT_CYC) * 100",
+        "MetricExpr": "PM_GCT_UTIL_11_14_ENTRIES / ( PM_RUN_CYC - PM_GCT_NOSLOT_CYC ) * 100",

I think this is just a whitespace change? Is it necessary? Curiosity, more than complaint.

tools/perf/pmu-events/arch/powerpc/power8/metrics.json line 892

Quoted 3 lines
tools/perf/pmu-events/arch/powerpc/power8/metrics.json +0 -0
         "MetricGroup": "general",
         "MetricName": "gct_util_11to14_slots_percent"
     },

PC

author reviewer
#20 lore raw

Re: [PATCH v2 08/11] perf metrics: fix parse errors in power8 metrics

Quoted 23 lines · Paul Clarke
Quoted 13 lines · Ian Rogers

Mismatched parentheses.

Fixes dd81eafacc52 (perf vendor events power8: Cpi_breakdown & estimated_dcache_miss_cpi metrics)
---
 tools/perf/pmu-events/arch/powerpc/power8/metrics.json | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
tools/perf/pmu-events/arch/powerpc/power8/metrics.json +0 -0
index bffb2d4a6420..ad71486a38e3 100644

tools/perf/pmu-events/arch/powerpc/power8/metrics.json line 172

Quoted 5 lines
tools/perf/pmu-events/arch/powerpc/power8/metrics.json +1 -1
@@ -169,7 +169,7 @@
     },
     {
         "BriefDescription": "Cycles GCT empty where dispatch was held",
-        "MetricExpr": "(PM_GCT_NOSLOT_DISP_HELD_MAP + PM_GCT_NOSLOT_DISP_HELD_SRQ + PM_GCT_NOSLOT_DISP_HELD_ISSQ + PM_GCT_NOSLOT_DISP_HELD_OTHER) / PM_RUN_INST_CMPL)",
+        "MetricExpr": "(PM_GCT_NOSLOT_DISP_HELD_MAP + PM_GCT_NOSLOT_DISP_HELD_SRQ + PM_GCT_NOSLOT_DISP_HELD_ISSQ + PM_GCT_NOSLOT_DISP_HELD_OTHER) / PM_RUN_INST_CMPL",

OK. (Thank you!)

Quoted 12 lines

tools/perf/pmu-events/arch/powerpc/power8/metrics.json line 889

Quoted 9 lines
tools/perf/pmu-events/arch/powerpc/power8/metrics.json +1 -1
         "MetricGroup": "cpi_breakdown",
         "MetricName": "gct_empty_disp_held_cpi"
     },
@@ -886,7 +886,7 @@
     },
     {
         "BriefDescription": "GCT slot utilization (11 to 14) as a % of cycles this thread had atleast 1 slot valid",
-        "MetricExpr": "PM_GCT_UTIL_11_14_ENTRIES / ( PM_RUN_CYC - PM_GCT_NOSLOT_CYC) * 100",
+        "MetricExpr": "PM_GCT_UTIL_11_14_ENTRIES / ( PM_RUN_CYC - PM_GCT_NOSLOT_CYC ) * 100",

I think this is just a whitespace change? Is it necessary? Curiosity, more than complaint.

Sorry about that, the space isn't necessary and this doesn't need to change. For the curious, originally the parse test would make all metrics equal to 1.0 and this metric would trigger a divide by zero because of this. This motivated adding a debug print for this case.

Thanks,
Ian

Quoted 7 lines

tools/perf/pmu-events/arch/powerpc/power8/metrics.json line 892

Quoted 3 lines
tools/perf/pmu-events/arch/powerpc/power8/metrics.json +0 -0
         "MetricGroup": "general",
         "MetricName": "gct_util_11to14_slots_percent"
     },

PC

author reviewer
#21 lore raw

PATCH v2 9/11 perf metrics: fix parse errors in power9 metrics

In reply to Ian Rogers

Mismatched parentheses.

Fixes 7f3cf5ac7743 (perf vendor events power9: Cpi_breakdown & estimated_dcache_miss_cpi metrics)
---
 tools/perf/pmu-events/arch/powerpc/power9/metrics.json | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/pmu-events/arch/powerpc/power9/metrics.json +1 -1
index 811c2a8c1c9e..f427436f2c0a 100644
@@ -362,7 +362,7 @@
     },
     {
         "BriefDescription": "Completion stall for other reasons",
-        "MetricExpr": "PM_CMPLU_STALL - PM_CMPLU_STALL_NTC_DISP_FIN - PM_CMPLU_STALL_NTC_FLUSH - PM_CMPLU_STALL_LSU - PM_CMPLU_STALL_EXEC_UNIT - PM_CMPLU_STALL_BRU)/PM_RUN_INST_CMPL",
+        "MetricExpr": "(PM_CMPLU_STALL - PM_CMPLU_STALL_NTC_DISP_FIN - PM_CMPLU_STALL_NTC_FLUSH - PM_CMPLU_STALL_LSU - PM_CMPLU_STALL_EXEC_UNIT - PM_CMPLU_STALL_BRU)/PM_RUN_INST_CMPL",
         "MetricGroup": "cpi_breakdown",
         "MetricName": "other_stall_cpi"
     },
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
Reviewed-by on v2 9/11
#22 lore raw

Re: [PATCH v2 09/11] perf metrics: fix parse errors in power9 metrics

Quoted 13 lines · Ian Rogers

Mismatched parentheses.

Fixes 7f3cf5ac7743 (perf vendor events power9: Cpi_breakdown & estimated_dcache_miss_cpi metrics)
---
 tools/perf/pmu-events/arch/powerpc/power9/metrics.json | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
tools/perf/pmu-events/arch/powerpc/power9/metrics.json +0 -0
index 811c2a8c1c9e..f427436f2c0a 100644

tools/perf/pmu-events/arch/powerpc/power9/metrics.json line 365

Quoted 5 lines
tools/perf/pmu-events/arch/powerpc/power9/metrics.json +1 -1
@@ -362,7 +362,7 @@
     },
     {
         "BriefDescription": "Completion stall for other reasons",
-        "MetricExpr": "PM_CMPLU_STALL - PM_CMPLU_STALL_NTC_DISP_FIN - PM_CMPLU_STALL_NTC_FLUSH - PM_CMPLU_STALL_LSU - PM_CMPLU_STALL_EXEC_UNIT - PM_CMPLU_STALL_BRU)/PM_RUN_INST_CMPL",
+        "MetricExpr": "(PM_CMPLU_STALL - PM_CMPLU_STALL_NTC_DISP_FIN - PM_CMPLU_STALL_NTC_FLUSH - PM_CMPLU_STALL_LSU - PM_CMPLU_STALL_EXEC_UNIT - PM_CMPLU_STALL_BRU)/PM_RUN_INST_CMPL",

OK. (Thank you!!)

tools/perf/pmu-events/arch/powerpc/power9/metrics.json line 368

Quoted 3 lines
tools/perf/pmu-events/arch/powerpc/power9/metrics.json +0 -0
         "MetricGroup": "cpi_breakdown",
         "MetricName": "other_stall_cpi"
     },

(I like how you got the "power8" changes to be 8th in the series and the "power9" changes to be 9th ;-)

PC

author reviewer
#23 lore raw

PATCH v2 10/11 perf expr: print a debug message for division by zero

In reply to Ian Rogers

If an expression yields 0 and is then divided-by/modulus-by then the parsing aborts. Add a debug error message to better enable debugging when this happens.

---
 tools/perf/util/expr.y | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)
tools/perf/util/expr.y +12 -2
index 54260094b947..21e82a1e11a2 100644
@@ -103,8 +103,18 @@ expr:	  NUMBER
 	| expr '+' expr		{ $$ = $1 + $3; }
 	| expr '-' expr		{ $$ = $1 - $3; }
 	| expr '*' expr		{ $$ = $1 * $3; }
-	| expr '/' expr		{ if ($3 == 0) YYABORT; $$ = $1 / $3; }
-	| expr '%' expr		{ if ((long)$3 == 0) YYABORT; $$ = (long)$1 % (long)$3; }
+	| expr '/' expr		{ if ($3 == 0) {
+					pr_debug("division by zero\n");
+					YYABORT;
+				  }
+				  $$ = $1 / $3;
+	                        }
+	| expr '%' expr		{ if ((long)$3 == 0) {
+					pr_debug("division by zero\n");
+					YYABORT;
+				  }
+				  $$ = (long)$1 % (long)$3;
+	                        }
 	| '-' expr %prec NEG	{ $$ = -$2; }
 	| '(' if_expr ')'	{ $$ = $2; }
 	| MIN '(' expr ',' expr ')' { $$ = $3 < $5 ? $3 : $5; }
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
#24 lore raw

Re: [PATCH v2 10/11] perf expr: print a debug message for division by zero

Quoted 16 lines · Ian Rogers

If an expression yields 0 and is then divided-by/modulus-by then the parsing aborts. Add a debug error message to better enable debugging when this happens.

---
 tools/perf/util/expr.y | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)
tools/perf/util/expr.y +0 -0
index 54260094b947..21e82a1e11a2 100644
@@ -103,8 +103,18 @@ expr:	  NUMBER
 	| expr '+' expr		{ $$ = $1 + $3; }
 	| expr '-' expr		{ $$ = $1 - $3; }

tools/perf/util/expr.y expr: NUMBER · line 112

Quoted 10 lines
tools/perf/util/expr.y +7 -2
 	| expr '*' expr		{ $$ = $1 * $3; }
-	| expr '/' expr		{ if ($3 == 0) YYABORT; $$ = $1 / $3; }
-	| expr '%' expr		{ if ((long)$3 == 0) YYABORT; $$ = (long)$1 % (long)$3; }
+	| expr '/' expr		{ if ($3 == 0) {
+					pr_debug("division by zero\n");
+					YYABORT;
+				  }
+				  $$ = $1 / $3;
+	                        }
+	| expr '%' expr		{ if ((long)$3 == 0) {

is that (long) cast necessary? it's missing for the '/' case

jirka

Quoted 11 lines
tools/perf/util/expr.y +5 -0
+					pr_debug("division by zero\n");
+					YYABORT;
+				  }
+				  $$ = (long)$1 % (long)$3;
+	                        }
 	| '-' expr %prec NEG	{ $$ = -$2; }
 	| '(' if_expr ')'	{ $$ = $2; }
 	| MIN '(' expr ',' expr ')' { $$ = $3 < $5 ? $3 : $5; }
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
author reviewer
#25 lore raw

Re: [PATCH v2 10/11] perf expr: print a debug message for division by zero

Quoted 22 lines · Jiri Olsa
Quoted 20 lines · Ian Rogers

If an expression yields 0 and is then divided-by/modulus-by then the parsing aborts. Add a debug error message to better enable debugging when this happens.

---
 tools/perf/util/expr.y | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)
tools/perf/util/expr.y +1 -2
index 54260094b947..21e82a1e11a2 100644
@@ -103,8 +103,18 @@ expr:      NUMBER
      | expr '+' expr         { $$ = $1 + $3; }
      | expr '-' expr         { $$ = $1 - $3; }
      | expr '*' expr         { $$ = $1 * $3; }
-     | expr '/' expr         { if ($3 == 0) YYABORT; $$ = $1 / $3; }
-     | expr '%' expr         { if ((long)$3 == 0) YYABORT; $$ = (long)$1 % (long)$3; }
+     | expr '/' expr         { if ($3 == 0) {
Quoted 8 lines

tools/perf/util/expr.y expr: NUMBER · line 112

Quoted 6 lines
tools/perf/util/expr.y +6 -0
+                                     pr_debug("division by zero\n");
+                                     YYABORT;
+                               }
+                               $$ = $1 / $3;
+                             }
+     | expr '%' expr         { if ((long)$3 == 0) {

is that (long) cast necessary? it's missing for the '/' case

Andi Kleen could say for sure :-) From my observation, the values are typically doubles and definitely need to be in the divide case. There is also code handling them as longs such as |, &, ^:
https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/tree/tools/perf/util/expr.y#n100
There's an experiment to see if all of the long handling code could be removed. This change isn't modifying the existing behavior.

Thanks,
Ian

Quoted 14 lines

jirka

Quoted 11 lines
tools/perf/util/expr.y +5 -1
+                                     pr_debug("division by zero\n");
+                                     YYABORT;
+                               }
+                               $$ = (long)$1 % (long)$3;
+                             }
      | '-' expr %prec NEG    { $$ = -$2; }
      | '(' if_expr ')'       { $$ = $2; }
      | MIN '(' expr ',' expr ')' { $$ = $3 < $5 ? $3 : $5; }
--

2.26.2.303.gf8c07b1a785-goog

author reviewer
#26 lore raw

PATCH v2 11/11 perf test: add expr test for pmu metrics

In reply to Ian Rogers

Add basic floating point number test.
Verify that all pmu metrics, for the current architecture, parse.

---
 tools/perf/tests/builtin-test.c |  5 ++
 tools/perf/tests/expr.c         | 96 ++++++++++++++++++++++++++++++++-
 tools/perf/tests/tests.h        |  2 +
 3 files changed, 102 insertions(+), 1 deletion(-)
tools/perf/tests/builtin-test.c +5 -0
index b6322eb0f423..28d547951f6b 100644
@@ -63,6 +63,11 @@ static struct test generic_tests[] = {
 	{
 		.desc = "Simple expression parser",
 		.func = test__expr,
+		.subtest = {
+			.get_nr		= test__expr_subtest_get_nr,
+			.get_desc	= test__expr_subtest_get_desc,
+		},
+
 	},
 	{
 		.desc = "PERF_RECORD_* events & perf_sample fields",
tools/perf/tests/expr.c +95 -1

120 lines not on this page · raw message

tools/perf/tests/tests.h +2 -0
index 61a1ab032080..315d64ffd14c 100644
@@ -72,6 +72,8 @@ int test__keep_tracking(struct test *test, int subtest);
 int test__parse_no_sample_id_all(struct test *test, int subtest);
 int test__dwarf_unwind(struct test *test, int subtest);
 int test__expr(struct test *test, int subtest);
+const char *test__expr_subtest_get_desc(int subtest);
+int test__expr_subtest_get_nr(void);
 int test__hists_filter(struct test *test, int subtest);
 int test__mmap_thread_lookup(struct test *test, int subtest);
 int test__thread_maps_share(struct test *test, int subtest);
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
#27 lore raw

Re: [PATCH v2 11/11] perf test: add expr test for pmu metrics

On Wed, Apr 22, 2020 at 03:04:30PM -0700, Ian Rogers wrote:

SNIP

Quoted 16 lines
tools/perf/tests/expr.c +16 -0
+
+static int pmu_tests(void)
+{
+	struct pmu_events_map *map;
+	struct pmu_event *pe;
+	int i, j, k;
+	const char **ids;
+	int idnum;
+	int ret = 0;
+	struct expr_parse_ctx ctx;
+	double result;
+
+	i = 0;
+	for (;;) {
+		map = &pmu_events_map[i++];
+		if (!map->table) {

tools/perf/tests/expr.c int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) · line 96

Quoted 10 lines
tools/perf/tests/expr.c +10 -0
+			map = NULL;
+			break;
+		}
+		j = 0;
+		for (;;) {
+			pe = &map->table[j++];
+			if (!pe->name && !pe->metric_group && !pe->metric_name)
+				break;
+			if (!pe->metric_expr)
+				continue;

so we go throught all the metrics for the current cpu and test the parsing on them.. great!

thanks,
jirka

author reviewer
#28 lore raw

Re: [PATCH v2 11/11] perf test: add expr test for pmu metrics

Quoted 26 lines · Jiri Olsa

On Wed, Apr 22, 2020 at 03:04:30PM -0700, Ian Rogers wrote:

SNIP

Quoted 21 lines
tools/perf/tests/expr.c +21 -0
+
+static int pmu_tests(void)
+{
+     struct pmu_events_map *map;
+     struct pmu_event *pe;
+     int i, j, k;
+     const char **ids;
+     int idnum;
+     int ret = 0;
+     struct expr_parse_ctx ctx;
+     double result;
+
+     i = 0;
+     for (;;) {
+             map = &pmu_events_map[i++];
+             if (!map->table) {
+                     map = NULL;
+                     break;
+             }
+             j = 0;
+             for (;;) {
Quoted 8 lines

tools/perf/tests/expr.c int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) · line 96

Quoted 5 lines
tools/perf/tests/expr.c +5 -0
+                     pe = &map->table[j++];
+                     if (!pe->name && !pe->metric_group && !pe->metric_name)
+                             break;
+                     if (!pe->metric_expr)
+                             continue;

so we go throught all the metrics for the current cpu and test the parsing on them.. great!

It's not just the current CPU (such as skylake) it is every map (skylake, cascade lake, etc), but this only works for the architecture that jevents built.
If jevents built all architectures then this could check them as well. Perhaps there should be a jevents test suite, but I think even then this test has value.
A worthy addition to this is checking that the events within the expression parse, but this is good progress and worth landing.

Thanks,
Ian

Quoted 3 lines

thanks,
jirka

#29 lore raw

Re: [PATCH v2 11/11] perf test: add expr test for pmu metrics

Quoted 36 lines · Ian Rogers
Quoted 26 lines · Jiri Olsa

On Wed, Apr 22, 2020 at 03:04:30PM -0700, Ian Rogers wrote:

SNIP

Quoted 21 lines
tools/perf/tests/expr.c +21 -0
+
+static int pmu_tests(void)
+{
+     struct pmu_events_map *map;
+     struct pmu_event *pe;
+     int i, j, k;
+     const char **ids;
+     int idnum;
+     int ret = 0;
+     struct expr_parse_ctx ctx;
+     double result;
+
+     i = 0;
+     for (;;) {
+             map = &pmu_events_map[i++];
+             if (!map->table) {
+                     map = NULL;
+                     break;
+             }
+             j = 0;
+             for (;;) {
Quoted 8 lines

tools/perf/tests/expr.c int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) · line 96

Quoted 5 lines
tools/perf/tests/expr.c +5 -0
+                     pe = &map->table[j++];
+                     if (!pe->name && !pe->metric_group && !pe->metric_name)
+                             break;
+                     if (!pe->metric_expr)
+                             continue;

so we go throught all the metrics for the current cpu and test the parsing on them.. great!

Quoted 6 lines

It's not just the current CPU (such as skylake) it is every map (skylake, cascade lake, etc), but this only works for the architecture that jevents built.
If jevents built all architectures then this could check them as well. Perhaps there should be a jevents test suite, but I think even then this test has value.

note: there is test__pmu_events(), which verifies that some test events generated in pmu-events.c are as expected, and also verifies that we create PMU events aliases as expected (for those test events). Nothing is done for metrics, ATM.

Thanks,
John

author reviewer
#30 lore raw

Re: [PATCH v2 11/11] perf test: add expr test for pmu metrics

Quoted 45 lines · John Garry
Quoted 36 lines · Ian Rogers
Quoted 26 lines · Jiri Olsa

On Wed, Apr 22, 2020 at 03:04:30PM -0700, Ian Rogers wrote:

SNIP

Quoted 21 lines
tools/perf/tests/expr.c +21 -0
+
+static int pmu_tests(void)
+{
+     struct pmu_events_map *map;
+     struct pmu_event *pe;
+     int i, j, k;
+     const char **ids;
+     int idnum;
+     int ret = 0;
+     struct expr_parse_ctx ctx;
+     double result;
+
+     i = 0;
+     for (;;) {
+             map = &pmu_events_map[i++];
+             if (!map->table) {
+                     map = NULL;
+                     break;
+             }
+             j = 0;
+             for (;;) {
Quoted 8 lines

tools/perf/tests/expr.c int test__expr(struct test *t __maybe_unused, int subtest __maybe_unused) · line 96

Quoted 5 lines
tools/perf/tests/expr.c +5 -0
+                     pe = &map->table[j++];
+                     if (!pe->name && !pe->metric_group && !pe->metric_name)
+                             break;
+                     if (!pe->metric_expr)
+                             continue;

so we go throught all the metrics for the current cpu and test the parsing on them.. great!

Quoted 6 lines

It's not just the current CPU (such as skylake) it is every map (skylake, cascade lake, etc), but this only works for the architecture that jevents built.
If jevents built all architectures then this could check them as well. Perhaps there should be a jevents test suite, but I think even then this test has value.

Quoted 4 lines

note: there is test__pmu_events(), which verifies that some test events generated in pmu-events.c are as expected, and also verifies that we create PMU events aliases as expected (for those test events). Nothing is done for metrics, ATM.

Thanks John, that sounds like the right place to start.

Ian

Quoted 2 lines

Thanks,
John

#31 lore raw
In reply to Ian Rogers
Quoted 8 lines · Ian Rogers

Add a test that all perf metrics (for your architecture) are parsable. Fix bugs in the expr parser and in x86 metrics. Untested on architectures other than x86.

v2 adds Fixes tags to commit messages for when broken metrics were
first added. Adds a debug warning for division by zero in expr, and adds a workaround for id values in the expr test necessary for powerpc. It also fixes broken power8 and power9 metrics.

looks good to me

Jin Yao, is there a metric that's not working for you with this patchset applied?

thanks,
jirka

Quoted 30 lines

Ian Rogers (11):
perf expr: unlimited escaped characters in a symbol perf metrics: fix parse errors in cascade lake metrics perf metrics: fix parse errors in skylake metrics perf expr: allow ',' to be an other token perf expr: increase max other
perf expr: parse numbers as doubles
perf expr: debug lex if debugging yacc
perf metrics: fix parse errors in power8 metrics perf metrics: fix parse errors in power9 metrics perf expr: print a debug message for division by zero perf test: add expr test for pmu metrics

 .../arch/powerpc/power8/metrics.json          |  4 +-
 .../arch/powerpc/power9/metrics.json          |  2 +-
 .../arch/x86/cascadelakex/clx-metrics.json    | 10 +-
 .../arch/x86/skylakex/skx-metrics.json        |  4 +-
 tools/perf/tests/builtin-test.c               |  5 +
 tools/perf/tests/expr.c                       | 96 ++++++++++++++++++-
 tools/perf/tests/tests.h                      |  2 +
 tools/perf/util/expr.c                        |  1 +
 tools/perf/util/expr.h                        |  2 +-
 tools/perf/util/expr.l                        | 16 ++--
 tools/perf/util/expr.y                        | 16 +++-
 11 files changed, 135 insertions(+), 23 deletions(-)
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
#32 lore raw

Hi Jiri,

Quoted 10 lines · Jiri Olsa
Quoted 8 lines · Ian Rogers

Add a test that all perf metrics (for your architecture) are parsable. Fix bugs in the expr parser and in x86 metrics. Untested on architectures other than x86.

v2 adds Fixes tags to commit messages for when broken metrics were
first added. Adds a debug warning for division by zero in expr, and adds a workaround for id values in the expr test necessary for powerpc. It also fixes broken power8 and power9 metrics.

Quoted 8 lines

looks good to me

Jin Yao, is there a metric that's not working for you with this patchset applied?

thanks,
jirka

Let me look for a CLX for testing, but maybe need some time.

BTW, suppose this patchset can work well, does it mean we will change the json file format in future?

For example,

before:
cha@event\\=0x36\\\\\\

after:
cha@event\\=0x36\\

"\\\\" are removed.

If so, we need to change our event generation script.

Thanks
Jin Yao

Quoted 31 lines
Quoted 30 lines

Ian Rogers (11):
perf expr: unlimited escaped characters in a symbol perf metrics: fix parse errors in cascade lake metrics perf metrics: fix parse errors in skylake metrics perf expr: allow ',' to be an other token perf expr: increase max other
perf expr: parse numbers as doubles
perf expr: debug lex if debugging yacc
perf metrics: fix parse errors in power8 metrics perf metrics: fix parse errors in power9 metrics perf expr: print a debug message for division by zero perf test: add expr test for pmu metrics

  .../arch/powerpc/power8/metrics.json          |  4 +-
  .../arch/powerpc/power9/metrics.json          |  2 +-
  .../arch/x86/cascadelakex/clx-metrics.json    | 10 +-
  .../arch/x86/skylakex/skx-metrics.json        |  4 +-
  tools/perf/tests/builtin-test.c               |  5 +
  tools/perf/tests/expr.c                       | 96 ++++++++++++++++++-
  tools/perf/tests/tests.h                      |  2 +
  tools/perf/util/expr.c                        |  1 +
  tools/perf/util/expr.h                        |  2 +-
  tools/perf/util/expr.l                        | 16 ++--
  tools/perf/util/expr.y                        | 16 +++-
  11 files changed, 135 insertions(+), 23 deletions(-)
Signature · 1 line
2.26.2.303.gf8c07b1a785-goog
#33 lore raw
Quoted 32 lines · Yao

Hi Jiri,

Quoted 10 lines · Jiri Olsa
Quoted 8 lines · Ian Rogers

Add a test that all perf metrics (for your architecture) are parsable. Fix bugs in the expr parser and in x86 metrics. Untested on architectures other than x86.

v2 adds Fixes tags to commit messages for when broken metrics were
first added. Adds a debug warning for division by zero in expr, and adds a workaround for id values in the expr test necessary for powerpc. It also fixes broken power8 and power9 metrics.

Quoted 8 lines

looks good to me

Jin Yao, is there a metric that's not working for you with this patchset applied?

thanks,
jirka

Let me look for a CLX for testing, but maybe need some time.

BTW, suppose this patchset can work well, does it mean we will change the json file format in future?

For example,

before:
cha@event\\=0x36\\\\\\

Quoted 6 lines

after:
cha@event\\=0x36\\

"\\\\" are removed.

If so, we need to change our event generation script.

ok, maybe I got the wrong idea that the extra \\\\ were just superfluous, what was the actual error there? and what's the reason for that many '\' in there?

jirka

author reviewer
#34 lore raw
Quoted 37 lines · Jiri Olsa
Quoted 35 lines · Yao

Hi Jiri,

Quoted 10 lines · Jiri Olsa
Quoted 8 lines · Ian Rogers

Add a test that all perf metrics (for your architecture) are parsable. Fix bugs in the expr parser and in x86 metrics. Untested on architectures other than x86.

v2 adds Fixes tags to commit messages for when broken metrics were
first added. Adds a debug warning for division by zero in expr, and adds a workaround for id values in the expr test necessary for powerpc. It also fixes broken power8 and power9 metrics.

Quoted 8 lines

looks good to me

Jin Yao, is there a metric that's not working for you with this patchset applied?

thanks,
jirka

Let me look for a CLX for testing, but maybe need some time.

BTW, suppose this patchset can work well, does it mean we will change the json file format in future?

For example,

before:
cha@event\\=0x36\\\\\\

after:
cha@event\\=0x36\\

Quoted 7 lines
Quoted 3 lines

"\\\\" are removed.

If so, we need to change our event generation script.

ok, maybe I got the wrong idea that the extra \\\\ were just superfluous, what was the actual error there? and what's the reason for that many '\' in there?

I believe they are superfluous and break even before the flex change. I commented on it here with a reproduction of a parse events error on skylake:
https://lore.kernel.org/lkml/CAP-5=fUnWAycQehCJ9=btquV2c3DVDX+tTEc85H8py9Kfehq4w@mail.gmail.com/

Fixing the script that generates this would be great! With the test landed it should be much harder for this to be broken in the future.

Thanks,
Ian

Quoted 2 lines

jirka

#35 lore raw
Quoted 51 lines · Ian Rogers
Quoted 37 lines · Jiri Olsa
Quoted 35 lines · Yao

Hi Jiri,

Quoted 10 lines · Jiri Olsa
Quoted 8 lines · Ian Rogers

Add a test that all perf metrics (for your architecture) are parsable. Fix bugs in the expr parser and in x86 metrics. Untested on architectures other than x86.

v2 adds Fixes tags to commit messages for when broken metrics were
first added. Adds a debug warning for division by zero in expr, and adds a workaround for id values in the expr test necessary for powerpc. It also fixes broken power8 and power9 metrics.

Quoted 8 lines

looks good to me

Jin Yao, is there a metric that's not working for you with this patchset applied?

thanks,
jirka

Let me look for a CLX for testing, but maybe need some time.

BTW, suppose this patchset can work well, does it mean we will change the json file format in future?

For example,

before:
cha@event\\=0x36\\\\\\

after:
cha@event\\=0x36\\

Quoted 7 lines
Quoted 3 lines

"\\\\" are removed.

If so, we need to change our event generation script.

ok, maybe I got the wrong idea that the extra \\\\ were just superfluous, what was the actual error there? and what's the reason for that many '\' in there?

I believe they are superfluous and break even before the flex change. I commented on it here with a reproduction of a parse events error on skylake:
https://lore.kernel.org/lkml/CAP-5=fUnWAycQehCJ9=btquV2c3DVDX+tTEc85H8py9Kfehq4w@mail.gmail.com/

Quoted 6 lines

Fixing the script that generates this would be great! With the test landed it should be much harder for this to be broken in the future.

Thanks,
Ian

Tested on CLX and the error disappeared.

BTW, I will post the latest version of event list once the testing is completed.

Thanks
Jin Yao

Quoted 3 lines
Quoted 3 lines

jirka