From 69fb09f6ccdb2f070557fd1f4c56c4d646694c8e Mon Sep 17 00:00:00 2001 From: Jin Yao Date: Fri, 7 Jul 2017 13:06:34 +0800 Subject: perf annotate: Check for fused instructions Macro fusion merges two instructions to a single micro-op. Intel core platform performs this hardware optimization under limited circumstances. For example, CMP + JCC can be "fused" and executed /retired together. While with sampling this can result in the sample sometimes being on the JCC and sometimes on the CMP. So for the fused instruction pair, they could be considered together. On Nehalem, fused instruction pairs: cmp/test + jcc. On other new CPU: cmp/test/add/sub/and/inc/dec + jcc. This patch adds an x86-specific function which checks if 2 instructions are in a "fused" pair. For non-x86 arch, the function is just NULL. Changelog: v4: Move the CPU model checking to symbol__disassemble and save the CPU family/model in arch structure. It avoids checking every time when jump arrow printed. v3: Add checking for Nehalem (CMP, TEST). For other newer Intel CPUs just check it by default (CMP, TEST, ADD, SUB, AND, INC, DEC). v2: Remove the original weak function. Arnaldo points out that doing it as a weak function that will be overridden by the host arch doesn't work. So now it's implemented as an arch-specific function. Committer fix: Do not access evsel->evlist->env->cpuid, ->env can be null, introduce perf_evsel__env_cpuid(), just like perf_evsel__env_arch(), also used in this function call. The original patch was segfaulting 'perf top' + annotation. But this essentially disables this fused instructions augmentation in 'perf top', the right thing is to get the cpuid from the running kernel, left for a later patch tho. Signed-off-by: Yao Jin Tested-by: Arnaldo Carvalho de Melo Cc: Alexander Shishkin Cc: Andi Kleen Cc: Jiri Olsa Cc: Kan Liang Cc: Peter Zijlstra Link: http://lkml.kernel.org/r/1499403995-19857-2-git-send-email-yao.jin@linux.intel.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/ui/gtk/annotate.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools/perf/ui/gtk/annotate.c') diff --git a/tools/perf/ui/gtk/annotate.c b/tools/perf/ui/gtk/annotate.c index d903fd493416..87e3760624f2 100644 --- a/tools/perf/ui/gtk/annotate.c +++ b/tools/perf/ui/gtk/annotate.c @@ -169,7 +169,7 @@ static int symbol__gtk_annotate(struct symbol *sym, struct map *map, return -1; err = symbol__disassemble(sym, map, perf_evsel__env_arch(evsel), - 0, NULL); + 0, NULL, NULL); if (err) { char msg[BUFSIZ]; symbol__strerror_disassemble(sym, map, err, msg, sizeof(msg)); -- cgit From 896bccd3cb8d95cbc565687715516009c5169e71 Mon Sep 17 00:00:00 2001 From: Taeung Song Date: Thu, 20 Jul 2017 06:36:45 +0900 Subject: perf annotate: Introduce struct sym_hist_entry struct sym_hist has addr[] but it should have not only number of samples but also the sample period. So use new struct symhist_entry to pave the way to have that. Committer notes: This initial patch will only introduce the struct sym_hist_entry and use only the nr_samples member, which makes the code clearer and paves the way to save the period as well. Signed-off-by: Taeung Song Suggested-by: Arnaldo Carvalho de Melo Cc: Jiri Olsa Cc: Namhyung Kim Link: http://lkml.kernel.org/r/1500500205-16553-1-git-send-email-treeze.taeung@gmail.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/ui/browsers/annotate.c | 6 +++--- tools/perf/ui/gtk/annotate.c | 4 ++-- tools/perf/util/annotate.c | 45 ++++++++++++++++++++------------------- tools/perf/util/annotate.h | 9 ++++++-- 4 files changed, 35 insertions(+), 29 deletions(-) (limited to 'tools/perf/ui/gtk/annotate.c') diff --git a/tools/perf/ui/browsers/annotate.c b/tools/perf/ui/browsers/annotate.c index 6794a8bec404..dbe4e630b90f 100644 --- a/tools/perf/ui/browsers/annotate.c +++ b/tools/perf/ui/browsers/annotate.c @@ -450,14 +450,14 @@ static void annotate_browser__calc_percent(struct annotate_browser *browser, next = disasm__get_next_ip_line(¬es->src->source, pos); for (i = 0; i < browser->nr_events; i++) { - u64 nr_samples; + struct sym_hist_entry sample; bpos->samples[i].percent = disasm__calc_percent(notes, evsel->idx + i, pos->offset, next ? next->offset : len, - &path, &nr_samples); - bpos->samples[i].nr = nr_samples; + &path, &sample); + bpos->samples[i].nr = sample.nr_samples; if (max_percent < bpos->samples[i].percent) max_percent = bpos->samples[i].percent; diff --git a/tools/perf/ui/gtk/annotate.c b/tools/perf/ui/gtk/annotate.c index 87e3760624f2..d736fd57ab9b 100644 --- a/tools/perf/ui/gtk/annotate.c +++ b/tools/perf/ui/gtk/annotate.c @@ -34,10 +34,10 @@ static int perf_gtk__get_percent(char *buf, size_t size, struct symbol *sym, return 0; symhist = annotation__histogram(symbol__annotation(sym), evidx); - if (!symbol_conf.event_group && !symhist->addr[dl->offset]) + if (!symbol_conf.event_group && !symhist->addr[dl->offset].nr_samples) return 0; - percent = 100.0 * symhist->addr[dl->offset] / symhist->sum; + percent = 100.0 * symhist->addr[dl->offset].nr_samples / symhist->sum; markup = perf_gtk__get_percent_color(percent); if (markup) diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c index 1742510f0120..c3829555ce1c 100644 --- a/tools/perf/util/annotate.c +++ b/tools/perf/util/annotate.c @@ -610,10 +610,10 @@ int symbol__alloc_hist(struct symbol *sym) size_t sizeof_sym_hist; /* Check for overflow when calculating sizeof_sym_hist */ - if (size > (SIZE_MAX - sizeof(struct sym_hist)) / sizeof(u64)) + if (size > (SIZE_MAX - sizeof(struct sym_hist)) / sizeof(struct sym_hist_entry)) return -1; - sizeof_sym_hist = (sizeof(struct sym_hist) + size * sizeof(u64)); + sizeof_sym_hist = (sizeof(struct sym_hist) + size * sizeof(struct sym_hist_entry)); /* Check for overflow in zalloc argument */ if (sizeof_sym_hist > (SIZE_MAX - sizeof(*notes->src)) @@ -714,11 +714,11 @@ static int __symbol__inc_addr_samples(struct symbol *sym, struct map *map, offset = addr - sym->start; h = annotation__histogram(notes, evidx); h->sum++; - h->addr[offset]++; + h->addr[offset].nr_samples++; pr_debug3("%#" PRIx64 " %s: period++ [addr: %#" PRIx64 ", %#" PRIx64 ", evidx=%d] => %" PRIu64 "\n", sym->start, sym->name, - addr, addr - sym->start, evidx, h->addr[offset]); + addr, addr - sym->start, evidx, h->addr[offset].nr_samples); return 0; } @@ -928,11 +928,12 @@ struct disasm_line *disasm__get_next_ip_line(struct list_head *head, struct disa } double disasm__calc_percent(struct annotation *notes, int evidx, s64 offset, - s64 end, const char **path, u64 *nr_samples) + s64 end, const char **path, struct sym_hist_entry *sample) { struct source_line *src_line = notes->src->lines; double percent = 0.0; - *nr_samples = 0; + + sample->nr_samples = 0; if (src_line) { size_t sizeof_src_line = sizeof(*src_line) + @@ -946,7 +947,7 @@ double disasm__calc_percent(struct annotation *notes, int evidx, s64 offset, *path = src_line->path; percent += src_line->samples[evidx].percent; - *nr_samples += src_line->samples[evidx].nr; + sample->nr_samples += src_line->samples[evidx].nr; offset++; } } else { @@ -954,10 +955,10 @@ double disasm__calc_percent(struct annotation *notes, int evidx, s64 offset, unsigned int hits = 0; while (offset < end) - hits += h->addr[offset++]; + hits += h->addr[offset++].nr_samples; if (h->sum) { - *nr_samples = hits; + sample->nr_samples = hits; percent = 100.0 * hits / h->sum; } } @@ -1057,10 +1058,10 @@ static int disasm_line__print(struct disasm_line *dl, struct symbol *sym, u64 st if (dl->offset != -1) { const char *path = NULL; - u64 nr_samples; double percent, max_percent = 0.0; double *ppercents = &percent; - u64 *psamples = &nr_samples; + struct sym_hist_entry sample; + struct sym_hist_entry *psamples = &sample; int i, nr_percent = 1; const char *color; struct annotation *notes = symbol__annotation(sym); @@ -1074,7 +1075,7 @@ static int disasm_line__print(struct disasm_line *dl, struct symbol *sym, u64 st if (perf_evsel__is_group_event(evsel)) { nr_percent = evsel->nr_members; ppercents = calloc(nr_percent, sizeof(double)); - psamples = calloc(nr_percent, sizeof(u64)); + psamples = calloc(nr_percent, sizeof(struct sym_hist_entry)); if (ppercents == NULL || psamples == NULL) { return -1; } @@ -1085,10 +1086,10 @@ static int disasm_line__print(struct disasm_line *dl, struct symbol *sym, u64 st notes->src->lines ? i : evsel->idx + i, offset, next ? next->offset : (s64) len, - &path, &nr_samples); + &path, &sample); ppercents[i] = percent; - psamples[i] = nr_samples; + psamples[i] = sample; if (percent > max_percent) max_percent = percent; } @@ -1126,12 +1127,12 @@ static int disasm_line__print(struct disasm_line *dl, struct symbol *sym, u64 st for (i = 0; i < nr_percent; i++) { percent = ppercents[i]; - nr_samples = psamples[i]; + sample = psamples[i]; color = get_percent_color(percent); if (symbol_conf.show_total_period) color_fprintf(stdout, color, " %7" PRIu64, - nr_samples); + sample.nr_samples); else color_fprintf(stdout, color, " %7.2f", percent); } @@ -1147,7 +1148,7 @@ static int disasm_line__print(struct disasm_line *dl, struct symbol *sym, u64 st if (ppercents != &percent) free(ppercents); - if (psamples != &nr_samples) + if (psamples != &sample) free(psamples); } else if (max_lines && printed >= max_lines) @@ -1702,7 +1703,7 @@ static int symbol__get_source_line(struct symbol *sym, struct map *map, double percent = 0.0; h = annotation__histogram(notes, evidx + k); - nr_samples = h->addr[i]; + nr_samples = h->addr[i].nr_samples; if (h->sum) percent = 100.0 * nr_samples / h->sum; @@ -1773,9 +1774,9 @@ static void symbol__annotate_hits(struct symbol *sym, struct perf_evsel *evsel) u64 len = symbol__size(sym), offset; for (offset = 0; offset < len; ++offset) - if (h->addr[offset] != 0) + if (h->addr[offset].nr_samples != 0) printf("%*" PRIx64 ": %" PRIu64 "\n", BITS_PER_LONG / 2, - sym->start + offset, h->addr[offset]); + sym->start + offset, h->addr[offset].nr_samples); printf("%*s: %" PRIu64 "\n", BITS_PER_LONG / 2, "h->sum", h->sum); } @@ -1878,8 +1879,8 @@ void symbol__annotate_decay_histogram(struct symbol *sym, int evidx) h->sum = 0; for (offset = 0; offset < len; ++offset) { - h->addr[offset] = h->addr[offset] * 7 / 8; - h->sum += h->addr[offset]; + h->addr[offset].nr_samples = h->addr[offset].nr_samples * 7 / 8; + h->sum += h->addr[offset].nr_samples; } } diff --git a/tools/perf/util/annotate.h b/tools/perf/util/annotate.h index bac698d7cc6a..3a176633b324 100644 --- a/tools/perf/util/annotate.h +++ b/tools/perf/util/annotate.h @@ -74,16 +74,21 @@ static inline bool disasm_line__has_offset(const struct disasm_line *dl) return dl->ops.target.offset_avail; } +struct sym_hist_entry { + u64 nr_samples; + u64 period; +}; + void disasm_line__free(struct disasm_line *dl); struct disasm_line *disasm__get_next_ip_line(struct list_head *head, struct disasm_line *pos); int disasm_line__scnprintf(struct disasm_line *dl, char *bf, size_t size, bool raw); size_t disasm__fprintf(struct list_head *head, FILE *fp); double disasm__calc_percent(struct annotation *notes, int evidx, s64 offset, - s64 end, const char **path, u64 *nr_samples); + s64 end, const char **path, struct sym_hist_entry *sample); struct sym_hist { u64 sum; - u64 addr[0]; + struct sym_hist_entry addr[0]; }; struct cyc_hist { -- cgit From 8158683da3d30e0346275702a8e08f2b22726c45 Mon Sep 17 00:00:00 2001 From: Taeung Song Date: Thu, 20 Jul 2017 06:36:51 +0900 Subject: perf annotate: Rename 'sum' to 'nr_samples' in struct sym_hist To make it more clear that it is the sum of all the nr_samples fields in the addr[] entries, i.e.: sym_hist->nr_samples = sum(sym_hist->addr[0 .. symbol__size(sym)]->nr_samples) Committer notes: Taeung had renamed it to total_samples, but using nr_samples, as in the added explanation above, looks clearer and establishes the direct connection, making clear it is about the _number_ of samples. Signed-off-by: Taeung Song Cc: Jiri Olsa Cc: Namhyung Kim Link: http://lkml.kernel.org/r/1500500211-16599-1-git-send-email-treeze.taeung@gmail.com Signed-off-by: Arnaldo Carvalho de Melo --- tools/perf/ui/gtk/annotate.c | 2 +- tools/perf/util/annotate.c | 26 +++++++++++++------------- tools/perf/util/annotate.h | 2 +- 3 files changed, 15 insertions(+), 15 deletions(-) (limited to 'tools/perf/ui/gtk/annotate.c') diff --git a/tools/perf/ui/gtk/annotate.c b/tools/perf/ui/gtk/annotate.c index d736fd57ab9b..02176193f427 100644 --- a/tools/perf/ui/gtk/annotate.c +++ b/tools/perf/ui/gtk/annotate.c @@ -37,7 +37,7 @@ static int perf_gtk__get_percent(char *buf, size_t size, struct symbol *sym, if (!symbol_conf.event_group && !symhist->addr[dl->offset].nr_samples) return 0; - percent = 100.0 * symhist->addr[dl->offset].nr_samples / symhist->sum; + percent = 100.0 * symhist->addr[dl->offset].nr_samples / symhist->nr_samples; markup = perf_gtk__get_percent_color(percent); if (markup) diff --git a/tools/perf/util/annotate.c b/tools/perf/util/annotate.c index c3829555ce1c..58c6b63ff049 100644 --- a/tools/perf/util/annotate.c +++ b/tools/perf/util/annotate.c @@ -713,7 +713,7 @@ static int __symbol__inc_addr_samples(struct symbol *sym, struct map *map, offset = addr - sym->start; h = annotation__histogram(notes, evidx); - h->sum++; + h->nr_samples++; h->addr[offset].nr_samples++; pr_debug3("%#" PRIx64 " %s: period++ [addr: %#" PRIx64 ", %#" PRIx64 @@ -957,9 +957,9 @@ double disasm__calc_percent(struct annotation *notes, int evidx, s64 offset, while (offset < end) hits += h->addr[offset++].nr_samples; - if (h->sum) { + if (h->nr_samples) { sample->nr_samples = hits; - percent = 100.0 * hits / h->sum; + percent = 100.0 * hits / h->nr_samples; } } @@ -1672,19 +1672,19 @@ static int symbol__get_source_line(struct symbol *sym, struct map *map, struct sym_hist *h = annotation__histogram(notes, evidx); struct rb_root tmp_root = RB_ROOT; int nr_pcnt = 1; - u64 h_sum = h->sum; + u64 nr_samples = h->nr_samples; size_t sizeof_src_line = sizeof(struct source_line); if (perf_evsel__is_group_event(evsel)) { for (i = 1; i < evsel->nr_members; i++) { h = annotation__histogram(notes, evidx + i); - h_sum += h->sum; + nr_samples += h->nr_samples; } nr_pcnt = evsel->nr_members; sizeof_src_line += (nr_pcnt - 1) * sizeof(src_line->samples); } - if (!h_sum) + if (!nr_samples) return 0; src_line = notes->src->lines = calloc(len, sizeof_src_line); @@ -1694,7 +1694,7 @@ static int symbol__get_source_line(struct symbol *sym, struct map *map, start = map__rip_2objdump(map, sym->start); for (i = 0; i < len; i++) { - u64 offset, nr_samples; + u64 offset; double percent_max = 0.0; src_line->nr_pcnt = nr_pcnt; @@ -1704,8 +1704,8 @@ static int symbol__get_source_line(struct symbol *sym, struct map *map, h = annotation__histogram(notes, evidx + k); nr_samples = h->addr[i].nr_samples; - if (h->sum) - percent = 100.0 * nr_samples / h->sum; + if (h->nr_samples) + percent = 100.0 * nr_samples / h->nr_samples; if (percent > percent_max) percent_max = percent; @@ -1777,7 +1777,7 @@ static void symbol__annotate_hits(struct symbol *sym, struct perf_evsel *evsel) if (h->addr[offset].nr_samples != 0) printf("%*" PRIx64 ": %" PRIu64 "\n", BITS_PER_LONG / 2, sym->start + offset, h->addr[offset].nr_samples); - printf("%*s: %" PRIu64 "\n", BITS_PER_LONG / 2, "h->sum", h->sum); + printf("%*s: %" PRIu64 "\n", BITS_PER_LONG / 2, "h->nr_samples", h->nr_samples); } int symbol__annotate_printf(struct symbol *sym, struct map *map, @@ -1813,7 +1813,7 @@ int symbol__annotate_printf(struct symbol *sym, struct map *map, width *= evsel->nr_members; graph_dotted_len = printf(" %-*.*s| Source code & Disassembly of %s for %s (%" PRIu64 " samples)\n", - width, width, "Percent", d_filename, evsel_name, h->sum); + width, width, "Percent", d_filename, evsel_name, h->nr_samples); printf("%-*.*s----\n", graph_dotted_len, graph_dotted_len, graph_dotted_line); @@ -1877,10 +1877,10 @@ void symbol__annotate_decay_histogram(struct symbol *sym, int evidx) struct sym_hist *h = annotation__histogram(notes, evidx); int len = symbol__size(sym), offset; - h->sum = 0; + h->nr_samples = 0; for (offset = 0; offset < len; ++offset) { h->addr[offset].nr_samples = h->addr[offset].nr_samples * 7 / 8; - h->sum += h->addr[offset].nr_samples; + h->nr_samples += h->addr[offset].nr_samples; } } diff --git a/tools/perf/util/annotate.h b/tools/perf/util/annotate.h index 3a176633b324..e8c246ec53df 100644 --- a/tools/perf/util/annotate.h +++ b/tools/perf/util/annotate.h @@ -87,7 +87,7 @@ double disasm__calc_percent(struct annotation *notes, int evidx, s64 offset, s64 end, const char **path, struct sym_hist_entry *sample); struct sym_hist { - u64 sum; + u64 nr_samples; struct sym_hist_entry addr[0]; }; -- cgit