perf, tools: Add support for skipping itrace instructions
PATCH 1/5 perf, tools: Add support for skipping itrace instructions
From: Andi Kleen <[email protected]>
When using perf script to look at PT traces it is often useful to ignore the initialization code at the beginning.
On larger traces which may have many millions of instructions in initialization code doing that in a pipeline can be very slow, with perf script spending a lot of CPU time calling printf and writing data.
This patch adds an extension to the --itrace argument that skips 'n' events (instructions, branches or transactions) at the beginning. This is much more efficient.
v2:
Add support for BTS (Adrian Hunter)
Document in itrace.txt
Fix branch check
Check transactions and instructions too
--- tools/perf/Documentation/intel-pt.txt | 7 +++++++ tools/perf/Documentation/itrace.txt | 8 ++++++++ tools/perf/util/auxtrace.c | 7 +++++++ tools/perf/util/auxtrace.h | 2 ++ tools/perf/util/intel-bts.c | 5 +++++ tools/perf/util/intel-pt.c | 22 ++++++++++++++++++++-- 6 files changed, 49 insertions(+), 2 deletions(-)
tools/perf/Documentation/intel-pt.txt +7 -0
@@ -672,6 +672,7 @@ The letters are: d create a debug log g synthesize a call chain (use with i or x) l synthesize last branch entries (use with i or x) + s skip initial number of events "Instructions" events look like they were recorded by "perf record -e instructions".
@@ -730,6 +731,12 @@ from one sample to the next. To disable trace decoding entirely, use the option --no-itrace. +It is also possible to skip events generated (instructions, branches, transactions) +at the beginning. This is useful to ignore initialization code. + + --itrace=i0nss1000000 + +skips the first million instructions. dump option -----------
tools/perf/Documentation/itrace.txt +8 -0
@@ -7,6 +7,7 @@ d create a debug log g synthesize a call chain (use with i or x) l synthesize last branch entries (use with i or x) + s skip initial number of events The default is all events i.e. the same as --itrace=ibxe
@@ -24,3 +25,10 @@ Also the number of last branch entries (default 64, max. 1024) for instructions or transactions events can be specified. + + It is also possible to skip events generated (instructions, branches, transactions) + at the beginning. This is useful to ignore initialization code. + + --itrace=i0nss1000000 + + skips the first million instructions.
tools/perf/util/auxtrace.c +7 -0
@@ -940,6 +940,7 @@ void itrace_synth_opts__set_default(struct itrace_synth_opts *synth_opts) synth_opts->period =PERF_ITRACE_DEFAULT_PERIOD ; synth_opts->callchain_sz =PERF_ITRACE_DEFAULT_CALLCHAIN_SZ ; synth_opts->last_branch_sz =PERF_ITRACE_DEFAULT_LAST_BRANCH_SZ ; + synth_opts->initial_skip =0 ; }/*
@@ -1064,6 +1065,12 @@ int itrace_parse_synth_opts(const struct option *opt, const char *str, synth_opts->last_branch_sz = val; }break ; +case 's' : + synth_opts->initial_skip =strtoul (p, &endptr,10 ); +if (p == endptr) + goto out_err; + p = endptr; +break ;case ' ' :case ',' :break ;
tools/perf/util/auxtrace.h +2 -0
@@ -68,6 +68,7 @@ enum itrace_period_type {* @last_branch_sz: branch context size * @period: 'instructions' events period * @period_type: 'instructions' events period type +* @initial_skip: skip N events at the beginning. */ struct itrace_synth_opts {bool set;
@@ -86,6 +87,7 @@ struct itrace_synth_opts {unsigned int last_branch_sz;unsigned long long period;enum itrace_period_type period_type; +unsigned long initial_skip; };/**
tools/perf/util/intel-bts.c +5 -0
@@ -66,6 +66,7 @@ struct intel_bts {u64 branches_id;size_t branches_event_size;bool synth_needs_swap; +unsigned long num_events; };struct intel_bts_queue {
@@ -275,6 +276,10 @@ static int intel_bts_synth_branch_sample(struct intel_bts_queue *btsq,union perf_event event;struct perf_sample sample = { .ip =0 , }; +if (bts->synth_opts.initial_skip && + bts->num_events++ <= bts->synth_opts.initial_skip) +return 0 ; + event.sample.header.type =PERF_RECORD_SAMPLE ; event.sample.header.misc =PERF_RECORD_MISC_USER ; event.sample.header.size =sizeof (struct perf_event_header );
tools/perf/util/intel-pt.c +20 -2
@@ -100,6 +100,8 @@ struct intel_pt {u64 cyc_bit;u64 noretcomp_bit;unsigned max_non_turbo_ratio; + +unsigned long num_events; };enum switch_state {
@@ -972,6 +974,10 @@ static int intel_pt_synth_branch_sample(struct intel_pt_queue *ptq)if (pt->branches_filter && !(pt->branches_filter & ptq->flags))return 0 ; +if (pt->synth_opts.initial_skip && + pt->num_events++ < pt->synth_opts.initial_skip) +return 0 ; + event->sample.header.type =PERF_RECORD_SAMPLE ; event->sample.header.misc =PERF_RECORD_MISC_USER ; event->sample.header.size =sizeof (struct perf_event_header );
@@ -1028,6 +1034,10 @@ static int intel_pt_synth_instruction_sample(struct intel_pt_queue *ptq)union perf_event *event = ptq->event_buf;struct perf_sample sample = { .ip =0 , }; +if (pt->synth_opts.initial_skip && + pt->num_events++ < pt->synth_opts.initial_skip) +return 0 ; + event->sample.header.type =PERF_RECORD_SAMPLE ; event->sample.header.misc =PERF_RECORD_MISC_USER ; event->sample.header.size =sizeof (struct perf_event_header );
@@ -1085,6 +1095,10 @@ static int intel_pt_synth_transaction_sample(struct intel_pt_queue *ptq)union perf_event *event = ptq->event_buf;struct perf_sample sample = { .ip =0 , }; +if (pt->synth_opts.initial_skip && + pt->num_events++ < pt->synth_opts.initial_skip) +return 0 ; + event->sample.header.type =PERF_RECORD_SAMPLE ; event->sample.header.misc =PERF_RECORD_MISC_USER ; event->sample.header.size =sizeof (struct perf_event_header );
@@ -1196,14 +1210,18 @@ static int intel_pt_sample(struct intel_pt_queue *ptq) ptq->have_sample = false;if (pt->sample_instructions && - (state->type &INTEL_PT_INSTRUCTION )) { + (state->type &INTEL_PT_INSTRUCTION ) && + (!pt->synth_opts.initial_skip || + pt->num_events++ >= pt->synth_opts.initial_skip)) { err =intel_pt_synth_instruction_sample (ptq);if (err)return err; }if (pt->sample_transactions && - (state->type &INTEL_PT_TRANSACTION )) { + (state->type &INTEL_PT_TRANSACTION ) && + (!pt->synth_opts.initial_skip || + pt->num_events++ >= pt->synth_opts.initial_skip)) { err =intel_pt_synth_transaction_sample (ptq);if (err)return err;
Signature · 1 line
2.5.5
PATCH 2/5 perf, tools: Add probing for udev86
From: Andi Kleen <[email protected]>
Add autoprobing for the udev86 disassembler library.
--- tools/build/Makefile.feature | 6 ++++-- tools/build/feature/Makefile | 8 ++++++-- tools/build/feature/test-all.c | 5 +++++ tools/build/feature/test-udis86.c | 8 ++++++++ tools/perf/config/Makefile | 5 +++++ 5 files changed, 28 insertions(+), 4 deletions(-) create mode 100644 tools/build/feature/test-udis86.c
tools/build/Makefile.feature +4 -2
@@ -55,7 +55,8 @@ FEATURE_TESTS_BASIC := \ zlib \ lzma \ get_cpuid \ - bpf + bpf \ + udis86 # FEATURE_TESTS_BASIC + FEATURE_TESTS_EXTRA is the complete list # of all feature tests
@@ -94,7 +95,8 @@ FEATURE_DISPLAY ?= \ zlib \ lzma \ get_cpuid \ - bpf + bpf \ + udis86 # Set FEATURE_CHECK_(C|LD)FLAGS-all for all FEATURE_TESTS features. # If in the future we need per-feature checks/flags for features not
tools/build/feature/Makefile +6 -2
@@ -36,7 +36,8 @@ FILES= \ test-zlib.bin \ test-lzma.bin \ test-bpf.bin \ - test-get_cpuid.bin + test-get_cpuid.bin \ + test-udis86.bin FILES := $(addprefix $(OUTPUT),$(FILES))
@@ -51,7 +52,7 @@ __BUILD = $(CC) $(CFLAGS) -Wall -Werror -o $@ $(patsubst %.bin,%.c,$(@F)) $(LDFL ############################### $(OUTPUT)test-all.bin: - $(BUILD) -fstack-protector-all -O2 -D_FORTIFY_SOURCE=2 -ldw -lelf -lnuma -lelf -laudit -I/usr/include/slang -lslang $(shell $(PKG_CONFIG) --libs --cflags gtk+-2.0 2>/dev/null) $(FLAGS_PERL_EMBED) $(FLAGS_PYTHON_EMBED) -DPACKAGE='"perf"' -lbfd -ldl -lz -llzma + $(BUILD) -fstack-protector-all -O2 -D_FORTIFY_SOURCE=2 -ldw -lelf -lnuma -lelf -laudit -I/usr/include/slang -lslang $(shell $(PKG_CONFIG) --libs --cflags gtk+-2.0 2>/dev/null) $(FLAGS_PERL_EMBED) $(FLAGS_PYTHON_EMBED) -DPACKAGE='"perf"' -lbfd -ldl -lz -llzma -ludis86 $(OUTPUT)test-hello.bin: $(BUILD)
@@ -97,6 +98,9 @@ $(OUTPUT)test-numa_num_possible_cpus.bin: $(OUTPUT)test-libunwind.bin: $(BUILD) -lelf +$(OUTPUT)test-udis86.bin: + $(BUILD) -ludis86 + $(OUTPUT)test-libunwind-debug-frame.bin: $(BUILD) -lelf
tools/build/feature/test-all.c +5 -0
@@ -133,6 +133,10 @@# include "test-libcrypto.c" #undef main +#define main main_test_udis86 +# include "test-udis86.c" +#endif +int main (int argc,char *argv[]) {main_test_libpython ();
@@ -163,6 +167,7 @@ int main(int argc, char *argv[])main_test_get_cpuid ();main_test_bpf ();main_test_libcrypto (); +main_test_udis86 ();return 0 ; }
tools/build/feature/test-udis86.c new file +8 -0
@@ -0,0 +1,8 @@ +#include <udis86.h> + +int main (void ) +{ +ud_t ud; +ud_init (&ud); +return 0 ; +}
tools/perf/config/Makefile +5 -0
@@ -587,6 +587,11 @@ ifneq ($(filter -lbfd,$(EXTLIBS)),) CFLAGS += -DHAVE_LIBBFD_SUPPORT endif +ifeq ($(feature-udis86), 1) + CFLAGS += -DHAVE_UDIS86 + EXTLIBS += -ludis86 +endif + ifndef NO_ZLIB ifeq ($(feature-zlib), 1) CFLAGS += -DHAVE_ZLIB_SUPPORT
Signature · 1 line
2.5.5
PATCH 3/5 perf, tools, script: Add support for printing assembler
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
When dumping PT traces with perf script it is very useful to see the assembler for each sample, so that it is easily possible to follow the control flow.
As using objdump is difficult and inefficient from perf script this patch uses the udis86 library to implement assembler output. The library can be downloaded from http://udis86.sourceforge.net/
The library is probed as an external dependency in the usual way. Then perf script calls into it when needed, and handles callbacks to resolve symbols.
% perf record -e intel_pt//u true
% perf script -F sym,symoff,ip,asm --itrace=i0ns | head
7fc7188b4190 _start+0x0 mov %rsp, %rdi
7fc7188b4193 _start+0x3 call _dl_start
7fc7188b7710 _dl_start+0x0 push %rbp
7fc7188b7711 _dl_start+0x1 mov %rsp, %rbp
7fc7188b7714 _dl_start+0x4 push %r15
7fc7188b7716 _dl_start+0x6 push %r14
7fc7188b7718 _dl_start+0x8 push %r13
7fc7188b771a _dl_start+0xa push %r12
7fc7188b771c _dl_start+0xc mov %rdi, %r12
7fc7188b771f _dl_start+0xf push %rbxCurrent issues:
- Some jump references do not get resolved to symbols.
- udis86 release does not support STAC/CLAC, which are used in the kernel, but there is a pending patch for it.
Cc: [email protected]
v2: Fix address resolution. Port to latest acme/perf/core
--- tools/perf/Documentation/perf-script.txt | 4 +- tools/perf/builtin-script.c | 107 +++++++++++++++++++++++++++++-- 2 files changed, 105 insertions(+), 6 deletions(-)
tools/perf/Documentation/perf-script.txt +3 -1
@@ -116,7 +116,7 @@ OPTIONS --fields:: Comma separated list of fields to print. Options are: comm, tid, pid, time, cpu, event, trace, ip, sym, dso, addr, symoff, - srcline, period, iregs, brstack, brstacksym, flags. + srcline, period, iregs, brstack, brstacksym, flags, asm. Field list can be prepended with the type, trace, sw or hw, to indicate to which event type the field list applies. e.g., -f sw:comm,tid,time,ip,sym and -f trace:time,cpu,trace
@@ -185,6 +185,8 @@ OPTIONS The brstacksym is identical to brstack, except that the FROM and TO addresses are printed in a symbolic form if possible. + When asm is specified the assembler instruction of each sample is printed in disassembled form. + -k:: --vmlinux=<file>:: vmlinux pathname
tools/perf/builtin-script.c +102 -5
173 lines not on this page · raw message
Signature · 1 line
2.5.5
PATCH 4/5 perf, tools, script: Add brstackasm output for branch stacks
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
Implement printing full disassembled sequences for branch stacks in perf script. This allows to directly print hot paths for individual samples, together with branch misprediction and even cycle count information.
% perf record -b ...
% perf script -F brstackasm
...
00007f0668d54e88 movsx (%rsi), %ecx
00007f0668d54e8b lea -0x30(%rcx), %eax
00007f0668d54e8e cmp $0x9, %al
00007f0668d54e90 jbe 0x68d54eaf
00007f0668d54e92 cmp %cl, %dl
00007f0668d54e94 jnz 0x68d54eb5
00007f0668d54e96 add $0x1, %rdi
00007f0668d54e9a movsx (%rdi), %edx
00007f0668d54e9d add $0x1, %rsi
00007f0668d54ea1 test %dl, %dl
00007f0668d54ea3 jnz _dl_cache_libcmp+11 # PRED 21 cycles
00007f0668d54dfb lea -0x30(%rdx), %eax
00007f0668d54dfe cmp $0x9, %al
00007f0668d54e00 ja _dl_cache_libcmp+152 # PRED 2 cycles
00007f0668d54e88 movsx (%rsi), %ecx
00007f0668d54e8b lea -0x30(%rcx), %eax
00007f0668d54e8e cmp $0x9, %al
00007f0668d54e90 jbe 0x68d54eaf
00007f0668d54e92 cmp %cl, %dl
00007f0668d54e94 jnz 0x68d54eb5 # PRED 3 cycles
00007f0668d54eb5 movsx %dl, %eax
00007f0668d54eb8 sub %ecx, %eax
00007f0668d54eba ret # PRED 1 cycles
00007f0668d54fae test %eax, %eax
00007f0668d54fb0 jz _dl_load_cache_lookup+688
00007f0668d54fb6 jns 0x68d54f70
00007f0668d54fb8 lea 0x1(%r14), %ebx
00007f0668d54fbc cmp %r15d, %ebx
00007f0668d54fbf nop
00007f0668d54fc0 jle 0x68d54f79 # PRED 2 cyclesOpen issues:
- Occasionally the path does not reach up to the sample IP, as the LBRs may be freezed earlier. Use precise events to avoid that.
v2: Remove bogus hunk. Document --max-blocks. Fix some printfs. Port to latest tree.
--- tools/perf/Documentation/perf-script.txt | 14 ++- tools/perf/builtin-script.c | 183 +++++++++++++++++++++++++++++++ 2 files changed, 195 insertions(+), 2 deletions(-)
tools/perf/Documentation/perf-script.txt +12 -2
@@ -116,7 +116,7 @@ OPTIONS --fields:: Comma separated list of fields to print. Options are: comm, tid, pid, time, cpu, event, trace, ip, sym, dso, addr, symoff, - srcline, period, iregs, brstack, brstacksym, flags, asm. + srcline, period, iregs, brstack, brstacksym, flags, asm, brstackasm Field list can be prepended with the type, trace, sw or hw, to indicate to which event type the field list applies. e.g., -f sw:comm,tid,time,ip,sym and -f trace:time,cpu,trace
@@ -176,17 +176,24 @@ OPTIONS i.e., -f "" is not allowed. The brstack output includes branch related information with raw addresses using the - /v/v/v/v/ syntax in the following order: + /v/v/v/v/cycles syntax in the following order: FROM: branch source instruction TO : branch target instruction M/P/-: M=branch target mispredicted or branch direction was mispredicted, P=target predicted or direction predicted, -=not supported X/- : X=branch inside a transactional region, -=not in transaction region or not supported A/- : A=TSX abort entry, -=not aborted region or not supported + cycles The brstacksym is identical to brstack, except that the FROM and TO addresses are printed in a symbolic form if possible. When asm is specified the assembler instruction of each sample is printed in disassembled form. + When brstackasm is specified the full assembler sequences of branch blocks for each sample + is printed (a branch block is a sequence of instructions not containing taken branches). + This is the full execution path leading to the sample. This is only supported when the + sample was recorded with perf record -b or -j any. + The maximum number of branch blocks to print can be configured with the --max-blocks option. + -k:: --vmlinux=<file>:: vmlinux pathname
@@ -268,6 +275,9 @@ include::itrace.txt[] --force:: Don't do ownership validation. +--max-blocks=N: + Maximum number of branch blocks to print with -F brstackasm + SEE ALSO -------- linkperf:perf-record[1], linkperf:perf-script-perl[1],
tools/perf/builtin-script.c +183 -0
225 lines not on this page · raw message
Signature · 1 line
2.5.5
PATCH 5/5 perf, tools, report: Add srcline_from/to branch sort keys
In reply to Andi KleenFrom: Andi Kleen <[email protected]>
Add srcline_from and srcline_to branch sort keys that allow to show the source lines of a branch. That makes it much easier to track down where particular branches happen in the program, for example to examine branch mispredictions, or to associate it with cycle counts:
% perf record -b -e cycles:p ./tcall
% perf report --sort srcline_from,srcline_to,mispredict
...
15.10% tcall.c:18 tcall.c:10 N
14.83% tcall.c:11 tcall.c:5 N
14.12% tcall.c:7 tcall.c:12 N
14.04% tcall.c:12 tcall.c:5 N
12.42% tcall.c:17 tcall.c:18 N
12.39% tcall.c:7 tcall.c:13 N
12.27% tcall.c:13 tcall.c:17 N...
% perf report --sort srcline_from,srcline_to,cycles
...
17.12% tcall.c:18 tcall.c:11 1
17.01% tcall.c:12 tcall.c:6 1
16.98% tcall.c:11 tcall.c:6 1
15.91% tcall.c:17 tcall.c:18 1
6.38% tcall.c:7 tcall.c:17 7
4.80% tcall.c:7 tcall.c:12 8
4.21% tcall.c:7 tcall.c:17 8
2.67% tcall.c:7 tcall.c:12 7
2.62% tcall.c:7 tcall.c:12 10
2.10% tcall.c:7 tcall.c:17 9
1.58% tcall.c:7 tcall.c:12 6
1.44% tcall.c:7 tcall.c:12 5
1.38% tcall.c:7 tcall.c:12 9
1.06% tcall.c:7 tcall.c:17 13
1.05% tcall.c:7 tcall.c:12 4
1.01% tcall.c:7 tcall.c:17 6Open issues:
- Some kernel symbols get misresolved.
--- tools/perf/Documentation/perf-report.txt | 3 +- tools/perf/util/hist.c | 9 ++++ tools/perf/util/hist.h | 2 + tools/perf/util/sort.c | 84 ++++++++++++++++++++++++++++++++ tools/perf/util/sort.h | 2 + tools/perf/util/symbol.h | 2 + 6 files changed, 101 insertions(+), 1 deletion(-)
tools/perf/Documentation/perf-report.txt +2 -1
@@ -103,12 +103,13 @@ OPTIONS If --branch-stack option is used, following sort keys are also available: - dso_from, dso_to, symbol_from, symbol_to, mispredict. - dso_from: name of library or module branched from - dso_to: name of library or module branched to - symbol_from: name of function branched from - symbol_to: name of function branched to + - srcline_from: source file and line branched from + - srcline_to: source file and line branched to - mispredict: "N" for predicted branch, "Y" for mispredicted branch - in_tx: branch in TSX transaction - abort: TSX transaction abort.
tools/perf/util/hist.c +9 -0
@@ -117,6 +117,13 @@ void hists__calc_col_len(struct hists *hists, struct hist_entry *h)hists__new_col_len (hists,HISTC_SYMBOL_TO , symlen);hists__set_unres_dso_col_len (hists,HISTC_DSO_TO ); } + +if (h->branch_info->srcline_from) +hists__new_col_len (hists,HISTC_SRCLINE_FROM , +strlen (h->branch_info->srcline_from)); +if (h->branch_info->srcline_to) +hists__new_col_len (hists,HISTC_SRCLINE_TO , +strlen (h->branch_info->srcline_to)); }if (h->mem_info) {
@@ -1042,6 +1049,8 @@ void hist_entry__delete(struct hist_entry *he)if (he->branch_info) {map__zput (he->branch_info->from.map);map__zput (he->branch_info->to.map); +free_srcline (he->branch_info->srcline_from); +free_srcline (he->branch_info->srcline_to);zfree (&he->branch_info); }
tools/perf/util/hist.h +2 -0
@@ -52,6 +52,8 @@ enum hist_column {HISTC_MEM_IADDR_SYMBOL ,HISTC_TRANSACTION ,HISTC_CYCLES , +HISTC_SRCLINE_FROM , +HISTC_SRCLINE_TO ,HISTC_TRACE , HISTC_NR_COLS, /* Last entry */ };
tools/perf/util/sort.c +84 -0
96 lines not on this page · raw message
tools/perf/util/sort.h +2 -0
@@ -222,6 +222,8 @@ enum sort_type {SORT_ABORT ,SORT_IN_TX ,SORT_CYCLES , +SORT_SRCLINE_FROM , +SORT_SRCLINE_TO ,/* memory mode specific sort keys */ __SORT_MEMORY_MODE,
tools/perf/util/symbol.h +2 -0
@@ -177,6 +177,8 @@ struct branch_info {struct addr_map_symbol from;struct addr_map_symbol to;struct branch_flags flags; +char *srcline_from; +char *srcline_to; };struct mem_info {
Signature · 1 line
2.5.5
Arnaldo Carvalho de Melo
maintainerEm Mon, Mar 28, 2016 at 10:45:38AM -0700, Andi Kleen escreveu:
Quoted 10 lines
From: Andi Kleen <[email protected]>
When using perf script to look at PT traces it is often useful to ignore the initialization code at the beginning.
On larger traces which may have many millions of instructions in initialization code doing that in a pipeline can be very slow, with perf script spending a lot of CPU time calling printf and writing data.
Quoted 5 lines
This patch adds an extension to the --itrace argument that skips 'n' events (instructions, branches or transactions) at the beginning. This is much more efficient.
Cc Adrian Hunter <[email protected]>
Adrian, are you ok now? Can I have your Acked-by?
- Arnaldo
Quoted 200 lines
v2:
Add support for BTS (Adrian Hunter)
Document in itrace.txt
Fix branch check
Check transactions and instructions tooSigned-off-by Andi Kleen <[email protected]>--- tools/perf/Documentation/intel-pt.txt | 7 +++++++ tools/perf/Documentation/itrace.txt | 8 ++++++++ tools/perf/util/auxtrace.c | 7 +++++++ tools/perf/util/auxtrace.h | 2 ++ tools/perf/util/intel-bts.c | 5 +++++ tools/perf/util/intel-pt.c | 22 ++++++++++++++++++++-- 6 files changed, 49 insertions(+), 2 deletions(-)tools/perf/Documentation/intel-pt.txt +7
-0@@ -672,6 +672,7 @@ The letters are: d create a debug log g synthesize a call chain (use with i or x) l synthesize last branch entries (use with i or x) + s skip initial number of events "Instructions" events look like they were recorded by "perf record -e instructions".@@ -730,6 +731,12 @@ from one sample to the next. To disable trace decoding entirely, use the option --no-itrace. +It is also possible to skip events generated (instructions, branches, transactions) +at the beginning. This is useful to ignore initialization code. + + --itrace=i0nss1000000 + +skips the first million instructions. dump option -----------tools/perf/Documentation/itrace.txt +8
-0@@ -7,6 +7,7 @@ d create a debug log g synthesize a call chain (use with i or x) l synthesize last branch entries (use with i or x) + s skip initial number of events The default is all events i.e. the same as --itrace=ibxe@@ -24,3 +25,10 @@ Also the number of last branch entries (default 64, max. 1024) for instructions or transactions events can be specified. + + It is also possible to skip events generated (instructions, branches, transactions) + at the beginning. This is useful to ignore initialization code. + + --itrace=i0nss1000000 + + skips the first million instructions.tools/perf/util/auxtrace.c +7
-0@@ -940,6 +940,7 @@ void itrace_synth_opts__set_default(struct itrace_synth_opts *synth_opts) synth_opts->period =PERF_ITRACE_DEFAULT_PERIOD ; synth_opts->callchain_sz =PERF_ITRACE_DEFAULT_CALLCHAIN_SZ ; synth_opts->last_branch_sz =PERF_ITRACE_DEFAULT_LAST_BRANCH_SZ ; + synth_opts->initial_skip =0 ; }/* @@ -1064,6 +1065,12 @@ int itrace_parse_synth_opts(const struct option *opt, const char *str, synth_opts->last_branch_sz = val; }break ; +case 's' : + synth_opts->initial_skip =strtoul (p, &endptr,10 ); +if (p == endptr) + goto out_err; + p = endptr; +break ;case ' ' :case ',' :break ;tools/perf/util/auxtrace.h +2
-0@@ -68,6 +68,7 @@ enum itrace_period_type {* @last_branch_sz: branch context size * @period: 'instructions' events period * @period_type: 'instructions' events period type +* @initial_skip: skip N events at the beginning. */ struct itrace_synth_opts {bool set;@@ -86,6 +87,7 @@ struct itrace_synth_opts {unsigned int last_branch_sz;unsigned long long period;enum itrace_period_type period_type; +unsigned long initial_skip; };/** tools/perf/util/intel-bts.c +5
-0@@ -66,6 +66,7 @@ struct intel_bts {u64 branches_id;size_t branches_event_size;bool synth_needs_swap; +unsigned long num_events; };struct intel_bts_queue {@@ -275,6 +276,10 @@ static int intel_bts_synth_branch_sample(struct intel_bts_queue *btsq,union perf_event event;struct perf_sample sample = { .ip =0 , }; +if (bts->synth_opts.initial_skip && + bts->num_events++ <= bts->synth_opts.initial_skip) +return 0 ; + event.sample.header.type =PERF_RECORD_SAMPLE ; event.sample.header.misc =PERF_RECORD_MISC_USER ; event.sample.header.size =sizeof (struct perf_event_header );tools/perf/util/intel-pt.c +20
-2@@ -100,6 +100,8 @@ struct intel_pt {u64 cyc_bit;u64 noretcomp_bit;unsigned max_non_turbo_ratio; + +unsigned long num_events; };enum switch_state {@@ -972,6 +974,10 @@ static int intel_pt_synth_branch_sample(struct intel_pt_queue *ptq)if (pt->branches_filter && !(pt->branches_filter & ptq->flags))return 0 ; +if (pt->synth_opts.initial_skip && + pt->num_events++ < pt->synth_opts.initial_skip) +return 0 ; + event->sample.header.type =PERF_RECORD_SAMPLE ; event->sample.header.misc =PERF_RECORD_MISC_USER ; event->sample.header.size =sizeof (struct perf_event_header );@@ -1028,6 +1034,10 @@ static int intel_pt_synth_instruction_sample(struct intel_pt_queue *ptq)union perf_event *event = ptq->event_buf;struct perf_sample sample = { .ip =0 , }; +if (pt->synth_opts.initial_skip && + pt->num_events++ < pt->synth_opts.initial_skip) +return 0 ; + event->sample.header.type =PERF_RECORD_SAMPLE ; event->sample.header.misc =PERF_RECORD_MISC_USER ; event->sample.header.size =sizeof (struct perf_event_header );@@ -1085,6 +1095,10 @@ static int intel_pt_synth_transaction_sample(struct intel_pt_queue *ptq)union perf_event *event = ptq->event_buf;struct perf_sample sample = { .ip =0 , }; +if (pt->synth_opts.initial_skip && + pt->num_events++ < pt->synth_opts.initial_skip) +return 0 ; + event->sample.header.type =PERF_RECORD_SAMPLE ; event->sample.header.misc =PERF_RECORD_MISC_USER ; event->sample.header.size =sizeof (struct perf_event_header );@@ -1196,14 +1210,18 @@ static int intel_pt_sample(struct intel_pt_queue *ptq) ptq->have_sample = false;if (pt->sample_instructions && - (state->type &INTEL_PT_INSTRUCTION )) { + (state->type &INTEL_PT_INSTRUCTION ) && + (!pt->synth_opts.initial_skip || + pt->num_events++ >= pt->synth_opts.initial_skip)) { err =intel_pt_synth_instruction_sample (ptq);if (err)return err; }if (pt->sample_transactions && - (state->type &INTEL_PT_TRANSACTION )) { + (state->type &INTEL_PT_TRANSACTION ) && + (!pt->synth_opts.initial_skip || + pt->num_events++ >= pt->synth_opts.initial_skip)) { err =intel_pt_synth_transaction_sample (ptq);if (err)return err;Signature · 1 line
2.5.5
Quoted 11 lines · Arnaldo Carvalho de Melo
Em Mon, Mar 28, 2016 at 10:45:38AM -0700, Andi Kleen escreveu:
Quoted 10 lines
From: Andi Kleen <[email protected]>
When using perf script to look at PT traces it is often useful to ignore the initialization code at the beginning.
On larger traces which may have many millions of instructions in initialization code doing that in a pipeline can be very slow, with perf script spending a lot of CPU time calling printf and writing data.
Quoted 7 lines
Quoted 5 lines
This patch adds an extension to the --itrace argument that skips 'n' events (instructions, branches or transactions) at the beginning. This is much more efficient.
Cc Adrian Hunter <[email protected]>Adrian, are you ok now? Can I have your Acked-by?
The last hunk isn't actually needed, but apart from that:
Quoted 193 lines
- Arnaldo
Quoted 190 lines
v2:
Add support for BTS (Adrian Hunter)
Document in itrace.txt
Fix branch check
Check transactions and instructions tooSigned-off-by Andi Kleen <[email protected]>--- tools/perf/Documentation/intel-pt.txt | 7 +++++++ tools/perf/Documentation/itrace.txt | 8 ++++++++ tools/perf/util/auxtrace.c | 7 +++++++ tools/perf/util/auxtrace.h | 2 ++ tools/perf/util/intel-bts.c | 5 +++++ tools/perf/util/intel-pt.c | 22 ++++++++++++++++++++-- 6 files changed, 49 insertions(+), 2 deletions(-)tools/perf/Documentation/intel-pt.txt +7
-0@@ -672,6 +672,7 @@ The letters are: d create a debug log g synthesize a call chain (use with i or x) l synthesize last branch entries (use with i or x) + s skip initial number of events "Instructions" events look like they were recorded by "perf record -e instructions".@@ -730,6 +731,12 @@ from one sample to the next. To disable trace decoding entirely, use the option --no-itrace. +It is also possible to skip events generated (instructions, branches, transactions) +at the beginning. This is useful to ignore initialization code. + + --itrace=i0nss1000000 + +skips the first million instructions. dump option -----------tools/perf/Documentation/itrace.txt +8
-0@@ -7,6 +7,7 @@ d create a debug log g synthesize a call chain (use with i or x) l synthesize last branch entries (use with i or x) + s skip initial number of events The default is all events i.e. the same as --itrace=ibxe@@ -24,3 +25,10 @@ Also the number of last branch entries (default 64, max. 1024) for instructions or transactions events can be specified. + + It is also possible to skip events generated (instructions, branches, transactions) + at the beginning. This is useful to ignore initialization code. + + --itrace=i0nss1000000 + + skips the first million instructions.tools/perf/util/auxtrace.c +7
-0@@ -940,6 +940,7 @@ void itrace_synth_opts__set_default(struct itrace_synth_opts *synth_opts) synth_opts->period =PERF_ITRACE_DEFAULT_PERIOD ; synth_opts->callchain_sz =PERF_ITRACE_DEFAULT_CALLCHAIN_SZ ; synth_opts->last_branch_sz =PERF_ITRACE_DEFAULT_LAST_BRANCH_SZ ; + synth_opts->initial_skip =0 ; }/* @@ -1064,6 +1065,12 @@ int itrace_parse_synth_opts(const struct option *opt, const char *str, synth_opts->last_branch_sz = val; }break ; +case 's' : + synth_opts->initial_skip =strtoul (p, &endptr,10 ); +if (p == endptr) + goto out_err; + p = endptr; +break ;case ' ' :case ',' :break ;tools/perf/util/auxtrace.h +2
-0@@ -68,6 +68,7 @@ enum itrace_period_type {* @last_branch_sz: branch context size * @period: 'instructions' events period * @period_type: 'instructions' events period type +* @initial_skip: skip N events at the beginning. */ struct itrace_synth_opts {bool set;@@ -86,6 +87,7 @@ struct itrace_synth_opts {unsigned int last_branch_sz;unsigned long long period;enum itrace_period_type period_type; +unsigned long initial_skip; };/** tools/perf/util/intel-bts.c +5
-0@@ -66,6 +66,7 @@ struct intel_bts {u64 branches_id;size_t branches_event_size;bool synth_needs_swap; +unsigned long num_events; };struct intel_bts_queue {@@ -275,6 +276,10 @@ static int intel_bts_synth_branch_sample(struct intel_bts_queue *btsq,union perf_event event;struct perf_sample sample = { .ip =0 , }; +if (bts->synth_opts.initial_skip && + bts->num_events++ <= bts->synth_opts.initial_skip) +return 0 ; + event.sample.header.type =PERF_RECORD_SAMPLE ; event.sample.header.misc =PERF_RECORD_MISC_USER ; event.sample.header.size =sizeof (struct perf_event_header );tools/perf/util/intel-pt.c +17
-1@@ -100,6 +100,8 @@ struct intel_pt {u64 cyc_bit;u64 noretcomp_bit;unsigned max_non_turbo_ratio; + +unsigned long num_events; };enum switch_state {@@ -972,6 +974,10 @@ static int intel_pt_synth_branch_sample(struct intel_pt_queue *ptq)if (pt->branches_filter && !(pt->branches_filter & ptq->flags))return 0 ; +if (pt->synth_opts.initial_skip && + pt->num_events++ < pt->synth_opts.initial_skip) +return 0 ; + event->sample.header.type =PERF_RECORD_SAMPLE ; event->sample.header.misc =PERF_RECORD_MISC_USER ; event->sample.header.size =sizeof (struct perf_event_header );@@ -1028,6 +1034,10 @@ static int intel_pt_synth_instruction_sample(struct intel_pt_queue *ptq)union perf_event *event = ptq->event_buf;struct perf_sample sample = { .ip =0 , }; +if (pt->synth_opts.initial_skip && + pt->num_events++ < pt->synth_opts.initial_skip) +return 0 ; + event->sample.header.type =PERF_RECORD_SAMPLE ; event->sample.header.misc =PERF_RECORD_MISC_USER ; event->sample.header.size =sizeof (struct perf_event_header );@@ -1085,6 +1095,10 @@ static int intel_pt_synth_transaction_sample(struct intel_pt_queue *ptq)union perf_event *event = ptq->event_buf;struct perf_sample sample = { .ip =0 , }; +if (pt->synth_opts.initial_skip && + pt->num_events++ < pt->synth_opts.initial_skip) +return 0 ; + event->sample.header.type =PERF_RECORD_SAMPLE ; event->sample.header.misc =PERF_RECORD_MISC_USER ; event->sample.header.size =sizeof (struct perf_event_header );@@ -1196,14 +1210,18 @@ static int intel_pt_sample(struct intel_pt_queue *ptq) ptq->have_sample = false;if (pt->sample_instructions && - (state->type &INTEL_PT_INSTRUCTION )) { + (state->type &INTEL_PT_INSTRUCTION ) && + (!pt->synth_opts.initial_skip || + pt->num_events++ >= pt->synth_opts.initial_skip)) { err =intel_pt_synth_instruction_sample (ptq);if (err)return err; }
Quoted 8 lines
tools/perf/util/intel-pt.c static int intel_pt_sample(struct intel_pt_queue *ptq) · line 1227
Quoted 8 lines
tools/perf/util/intel-pt.c +3
-1if (pt->sample_transactions && - (state->type &INTEL_PT_TRANSACTION )) { + (state->type &INTEL_PT_TRANSACTION ) && + (!pt->synth_opts.initial_skip || + pt->num_events++ >= pt->synth_opts.initial_skip)) { err =intel_pt_synth_transaction_sample (ptq);if (err)return err;
This last hunk (@@ -1196,14 +1210,18 @@) isn't needed since intel_pt_synth_instruction_sample() and intel_pt_synth_transaction_sample() check initial_skip themselves.
tip-bot for Andi Kleen
merge bot[tip:perf/core] perf tools: Add support for skipping itrace instructions
In reply to Andi Kleenperf tools: Add support for skipping itrace instructions
Commit-ID: d1706b39f0af6901ab2a5e2ebb210b53c1a5bdc7 Gitweb: http://git.kernel.org/tip/d1706b39f0af6901ab2a5e2ebb210b53c1a5bdc7 Author: Andi Kleen <[email protected]> AuthorDate: Mon, 28 Mar 2016 10:45:38 -0700 Committer: Arnaldo Carvalho de Melo <[email protected]> CommitDate: Wed, 30 Mar 2016 11:14:09 -0300
perf tools: Add support for skipping itrace instructions
When using 'perf script' to look at PT traces it is often useful to ignore the initialization code at the beginning.
On larger traces which may have many millions of instructions in initialization code doing that in a pipeline can be very slow, with perf script spending a lot of CPU time calling printf and writing data.
This patch adds an extension to the --itrace argument that skips 'n' events (instructions, branches or transactions) at the beginning. This is much more efficient.
v2:
Add support for BTS (Adrian Hunter)
Document in itrace.txt
Fix branch check
Check transactions and instructions too
Committer note:
To test intel_pt one needs to make sure VT-x isn't active, i.e. stopping KVM guests on the test machine, as described by Andi Kleen at http://lkml.kernel.org/r/[email protected]
--- tools/perf/Documentation/intel-pt.txt | 7 +++++++ tools/perf/Documentation/itrace.txt | 8 ++++++++ tools/perf/util/auxtrace.c | 7 +++++++ tools/perf/util/auxtrace.h | 2 ++ tools/perf/util/intel-bts.c | 5 +++++ tools/perf/util/intel-pt.c | 22 ++++++++++++++++++++-- 6 files changed, 49 insertions(+), 2 deletions(-)
tools/perf/Documentation/intel-pt.txt +7 -0
@@ -672,6 +672,7 @@ The letters are: d create a debug log g synthesize a call chain (use with i or x) l synthesize last branch entries (use with i or x) + s skip initial number of events "Instructions" events look like they were recorded by "perf record -e instructions".
@@ -730,6 +731,12 @@ from one sample to the next. To disable trace decoding entirely, use the option --no-itrace. +It is also possible to skip events generated (instructions, branches, transactions) +at the beginning. This is useful to ignore initialization code. + + --itrace=i0nss1000000 + +skips the first million instructions. dump option -----------
tools/perf/Documentation/itrace.txt +8 -0
@@ -7,6 +7,7 @@ d create a debug log g synthesize a call chain (use with i or x) l synthesize last branch entries (use with i or x) + s skip initial number of events The default is all events i.e. the same as --itrace=ibxe
@@ -24,3 +25,10 @@ Also the number of last branch entries (default 64, max. 1024) for instructions or transactions events can be specified. + + It is also possible to skip events generated (instructions, branches, transactions) + at the beginning. This is useful to ignore initialization code. + + --itrace=i0nss1000000 + + skips the first million instructions.
tools/perf/util/auxtrace.c +7 -0
@@ -940,6 +940,7 @@ void itrace_synth_opts__set_default(struct itrace_synth_opts *synth_opts) synth_opts->period =PERF_ITRACE_DEFAULT_PERIOD ; synth_opts->callchain_sz =PERF_ITRACE_DEFAULT_CALLCHAIN_SZ ; synth_opts->last_branch_sz =PERF_ITRACE_DEFAULT_LAST_BRANCH_SZ ; + synth_opts->initial_skip =0 ; }/*
@@ -1064,6 +1065,12 @@ int itrace_parse_synth_opts(const struct option *opt, const char *str, synth_opts->last_branch_sz = val; }break ; +case 's' : + synth_opts->initial_skip =strtoul (p, &endptr,10 ); +if (p == endptr) + goto out_err; + p = endptr; +break ;case ' ' :case ',' :break ;
tools/perf/util/auxtrace.h +2 -0
@@ -68,6 +68,7 @@ enum itrace_period_type {* @last_branch_sz: branch context size * @period: 'instructions' events period * @period_type: 'instructions' events period type +* @initial_skip: skip N events at the beginning. */ struct itrace_synth_opts {bool set;
@@ -86,6 +87,7 @@ struct itrace_synth_opts {unsigned int last_branch_sz;unsigned long long period;enum itrace_period_type period_type; +unsigned long initial_skip; };/**
tools/perf/util/intel-bts.c +5 -0
@@ -66,6 +66,7 @@ struct intel_bts {u64 branches_id;size_t branches_event_size;bool synth_needs_swap; +unsigned long num_events; };struct intel_bts_queue {
@@ -275,6 +276,10 @@ static int intel_bts_synth_branch_sample(struct intel_bts_queue *btsq,union perf_event event;struct perf_sample sample = { .ip =0 , }; +if (bts->synth_opts.initial_skip && + bts->num_events++ <= bts->synth_opts.initial_skip) +return 0 ; + event.sample.header.type =PERF_RECORD_SAMPLE ; event.sample.header.misc =PERF_RECORD_MISC_USER ; event.sample.header.size =sizeof (struct perf_event_header );
tools/perf/util/intel-pt.c +20 -2
@@ -100,6 +100,8 @@ struct intel_pt {u64 cyc_bit;u64 noretcomp_bit;unsigned max_non_turbo_ratio; + +unsigned long num_events; };enum switch_state {
@@ -972,6 +974,10 @@ static int intel_pt_synth_branch_sample(struct intel_pt_queue *ptq)if (pt->branches_filter && !(pt->branches_filter & ptq->flags))return 0 ; +if (pt->synth_opts.initial_skip && + pt->num_events++ < pt->synth_opts.initial_skip) +return 0 ; + event->sample.header.type =PERF_RECORD_SAMPLE ; event->sample.header.misc =PERF_RECORD_MISC_USER ; event->sample.header.size =sizeof (struct perf_event_header );
@@ -1029,6 +1035,10 @@ static int intel_pt_synth_instruction_sample(struct intel_pt_queue *ptq)union perf_event *event = ptq->event_buf;struct perf_sample sample = { .ip =0 , }; +if (pt->synth_opts.initial_skip && + pt->num_events++ < pt->synth_opts.initial_skip) +return 0 ; + event->sample.header.type =PERF_RECORD_SAMPLE ; event->sample.header.misc =PERF_RECORD_MISC_USER ; event->sample.header.size =sizeof (struct perf_event_header );
@@ -1087,6 +1097,10 @@ static int intel_pt_synth_transaction_sample(struct intel_pt_queue *ptq)union perf_event *event = ptq->event_buf;struct perf_sample sample = { .ip =0 , }; +if (pt->synth_opts.initial_skip && + pt->num_events++ < pt->synth_opts.initial_skip) +return 0 ; + event->sample.header.type =PERF_RECORD_SAMPLE ; event->sample.header.misc =PERF_RECORD_MISC_USER ; event->sample.header.size =sizeof (struct perf_event_header );
@@ -1199,14 +1213,18 @@ static int intel_pt_sample(struct intel_pt_queue *ptq) ptq->have_sample = false;if (pt->sample_instructions && - (state->type &INTEL_PT_INSTRUCTION )) { + (state->type &INTEL_PT_INSTRUCTION ) && + (!pt->synth_opts.initial_skip || + pt->num_events++ >= pt->synth_opts.initial_skip)) { err =intel_pt_synth_instruction_sample (ptq);if (err)return err; }if (pt->sample_transactions && - (state->type &INTEL_PT_TRANSACTION )) { + (state->type &INTEL_PT_TRANSACTION ) && + (!pt->synth_opts.initial_skip || + pt->num_events++ >= pt->synth_opts.initial_skip)) { err =intel_pt_synth_transaction_sample (ptq);if (err)return err;