haraldwolff
/
qemu-patch-raspberry4
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

qapi: introduce x-query-jit QMP command 

This is a counterpart to the HMP "info jit" command. It is being
added with an "x-" prefix because this QMP command is intended as an
ad hoc debugging tool and will thus not be modelled in QAPI as fully
structured data, nor will it have long term guaranteed stability.
The existing HMP command is rewritten to call the QMP command.

Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
Signed-off-by: Daniel P. Berrangé <berrange@redhat.com>
staging
Daniel P. Berrangé 2021-09-08 10:35:43 +01:00
parent 91f2fa7045
commit 3a841ab53f
8 changed files with 140 additions and 101 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

37
accel/tcg/cpu-exec.c
View File

@ -20,6 +20,9 @@
#include "qemu/osdep.h"
#include "qemu-common.h"
#include "qemu/qemu-print.h"
#include "qapi/error.h"
#include "qapi/qapi-commands-machine.h"
#include "qapi/type-helpers.h"
#include "hw/core/tcg-cpu-ops.h"
#include "trace.h"
#include "disas/disas.h"
@ -38,6 +41,7 @@
#include "exec/cpu-all.h"
#include "sysemu/cpu-timers.h"
#include "sysemu/replay.h"
#include "sysemu/tcg.h"
#include "exec/helper-proto.h"
#include "tb-hash.h"
#include "tb-context.h"
@ -1028,23 +1032,38 @@ void tcg_exec_unrealizefn(CPUState *cpu)
#ifndef CONFIG_USER_ONLY
void dump_drift_info(void)
void dump_drift_info(GString *buf)
{
if (!icount_enabled()) {
return;
}
qemu_printf("Host - Guest clock %"PRIi64" ms\n",
(cpu_get_clock() - icount_get()) / SCALE_MS);
g_string_append_printf(buf, "Host - Guest clock %"PRIi64" ms\n",
(cpu_get_clock() - icount_get()) / SCALE_MS);
if (icount_align_option) {
qemu_printf("Max guest delay %"PRIi64" ms\n",
-max_delay / SCALE_MS);
qemu_printf("Max guest advance %"PRIi64" ms\n",
max_advance / SCALE_MS);
g_string_append_printf(buf, "Max guest delay %"PRIi64" ms\n",
-max_delay / SCALE_MS);
g_string_append_printf(buf, "Max guest advance %"PRIi64" ms\n",
max_advance / SCALE_MS);
} else {
qemu_printf("Max guest delay NA\n");
qemu_printf("Max guest advance NA\n");
g_string_append_printf(buf, "Max guest delay NA\n");
g_string_append_printf(buf, "Max guest advance NA\n");
}
}
HumanReadableText *qmp_x_query_jit(Error **errp)
{
g_autoptr(GString) buf = g_string_new("");
if (!tcg_enabled()) {
error_setg(errp, "JIT information is only available with accel=tcg");
return NULL;
}
dump_exec_info(buf);
dump_drift_info(buf);
return human_readable_text_from_str(buf);
}
#endif /* !CONFIG_USER_ONLY */

15
accel/tcg/hmp.c
View File

@ -1,20 +1,11 @@
#include "qemu/osdep.h"
#include "qemu/error-report.h"
#include "qapi/error.h"
#include "qapi/qapi-commands-machine.h"
#include "exec/exec-all.h"
#include "monitor/monitor.h"
#include "sysemu/tcg.h"
static void hmp_info_jit(Monitor *mon, const QDict *qdict)
{
if (!tcg_enabled()) {
error_report("JIT information is only available with accel=tcg");
return;
}
dump_exec_info();
dump_drift_info();
}
static void hmp_info_opcount(Monitor *mon, const QDict *qdict)
{
dump_opcount_info();
@ -22,7 +13,7 @@ static void hmp_info_opcount(Monitor *mon, const QDict *qdict)
static void hmp_tcg_register(void)
{
monitor_register_hmp("jit", true, hmp_info_jit);
monitor_register_hmp_info_hrt("jit", qmp_x_query_jit);
monitor_register_hmp("opcount", true, hmp_info_opcount);
}

80
accel/tcg/translate-all.c
View File

@ -1991,7 +1991,7 @@ void cpu_io_recompile(CPUState *cpu, uintptr_t retaddr)
cpu_loop_exit_noexc(cpu);
}
static void print_qht_statistics(struct qht_stats hst)
static void print_qht_statistics(struct qht_stats hst, GString *buf)
{
uint32_t hgram_opts;
size_t hgram_bins;
@ -2000,9 +2000,11 @@ static void print_qht_statistics(struct qht_stats hst)
if (!hst.head_buckets) {
return;
}
qemu_printf("TB hash buckets %zu/%zu (%0.2f%% head buckets used)\n",
hst.used_head_buckets, hst.head_buckets,
(double)hst.used_head_buckets / hst.head_buckets * 100);
g_string_append_printf(buf, "TB hash buckets %zu/%zu "
"(%0.2f%% head buckets used)\n",
hst.used_head_buckets, hst.head_buckets,
(double)hst.used_head_buckets /
hst.head_buckets * 100);
hgram_opts = QDIST_PR_BORDER | QDIST_PR_LABELS;
hgram_opts |= QDIST_PR_100X | QDIST_PR_PERCENT;
@ -2010,8 +2012,9 @@ static void print_qht_statistics(struct qht_stats hst)
hgram_opts |= QDIST_PR_NODECIMAL;
}
hgram = qdist_pr(&hst.occupancy, 10, hgram_opts);
qemu_printf("TB hash occupancy %0.2f%% avg chain occ. Histogram: %s\n",
qdist_avg(&hst.occupancy) * 100, hgram);
g_string_append_printf(buf, "TB hash occupancy %0.2f%% avg chain occ. "
"Histogram: %s\n",
qdist_avg(&hst.occupancy) * 100, hgram);
g_free(hgram);
hgram_opts = QDIST_PR_BORDER | QDIST_PR_LABELS;
@ -2023,8 +2026,9 @@ static void print_qht_statistics(struct qht_stats hst)
hgram_opts |= QDIST_PR_NODECIMAL | QDIST_PR_NOBINRANGE;
}
hgram = qdist_pr(&hst.chain, hgram_bins, hgram_opts);
qemu_printf("TB hash avg chain %0.3f buckets. Histogram: %s\n",
qdist_avg(&hst.chain), hgram);
g_string_append_printf(buf, "TB hash avg chain %0.3f buckets. "
"Histogram: %s\n",
qdist_avg(&hst.chain), hgram);
g_free(hgram);
}
@ -2061,7 +2065,7 @@ static gboolean tb_tree_stats_iter(gpointer key, gpointer value, gpointer data)
return false;
}
void dump_exec_info(void)
void dump_exec_info(GString *buf)
{
struct tb_tree_stats tst = {};
struct qht_stats hst;
@ -2070,44 +2074,48 @@ void dump_exec_info(void)
tcg_tb_foreach(tb_tree_stats_iter, &tst);
nb_tbs = tst.nb_tbs;
/* XXX: avoid using doubles ? */
qemu_printf("Translation buffer state:\n");
g_string_append_printf(buf, "Translation buffer state:\n");
/*
* Report total code size including the padding and TB structs;
* otherwise users might think "-accel tcg,tb-size" is not honoured.
* For avg host size we use the precise numbers from tb_tree_stats though.
*/
qemu_printf("gen code size %zu/%zu\n",
tcg_code_size(), tcg_code_capacity());
qemu_printf("TB count %zu\n", nb_tbs);
qemu_printf("TB avg target size %zu max=%zu bytes\n",
nb_tbs ? tst.target_size / nb_tbs : 0,
tst.max_target_size);
qemu_printf("TB avg host size %zu bytes (expansion ratio: %0.1f)\n",
nb_tbs ? tst.host_size / nb_tbs : 0,
tst.target_size ? (double)tst.host_size / tst.target_size : 0);
qemu_printf("cross page TB count %zu (%zu%%)\n", tst.cross_page,
nb_tbs ? (tst.cross_page * 100) / nb_tbs : 0);
qemu_printf("direct jump count %zu (%zu%%) (2 jumps=%zu %zu%%)\n",
tst.direct_jmp_count,
nb_tbs ? (tst.direct_jmp_count * 100) / nb_tbs : 0,
tst.direct_jmp2_count,
nb_tbs ? (tst.direct_jmp2_count * 100) / nb_tbs : 0);
g_string_append_printf(buf, "gen code size %zu/%zu\n",
tcg_code_size(), tcg_code_capacity());
g_string_append_printf(buf, "TB count %zu\n", nb_tbs);
g_string_append_printf(buf, "TB avg target size %zu max=%zu bytes\n",
nb_tbs ? tst.target_size / nb_tbs : 0,
tst.max_target_size);
g_string_append_printf(buf, "TB avg host size %zu bytes "
"(expansion ratio: %0.1f)\n",
nb_tbs ? tst.host_size / nb_tbs : 0,
tst.target_size ?
(double)tst.host_size / tst.target_size : 0);
g_string_append_printf(buf, "cross page TB count %zu (%zu%%)\n",
tst.cross_page,
nb_tbs ? (tst.cross_page * 100) / nb_tbs : 0);
g_string_append_printf(buf, "direct jump count %zu (%zu%%) "
"(2 jumps=%zu %zu%%)\n",
tst.direct_jmp_count,
nb_tbs ? (tst.direct_jmp_count * 100) / nb_tbs : 0,
tst.direct_jmp2_count,
nb_tbs ? (tst.direct_jmp2_count * 100) / nb_tbs : 0);
qht_statistics_init(&tb_ctx.htable, &hst);
print_qht_statistics(hst);
print_qht_statistics(hst, buf);
qht_statistics_destroy(&hst);
qemu_printf("\nStatistics:\n");
qemu_printf("TB flush count %u\n",
qatomic_read(&tb_ctx.tb_flush_count));
qemu_printf("TB invalidate count %u\n",
qatomic_read(&tb_ctx.tb_phys_invalidate_count));
g_string_append_printf(buf, "\nStatistics:\n");
g_string_append_printf(buf, "TB flush count %u\n",
qatomic_read(&tb_ctx.tb_flush_count));
g_string_append_printf(buf, "TB invalidate count %u\n",
qatomic_read(&tb_ctx.tb_phys_invalidate_count));
tlb_flush_counts(&flush_full, &flush_part, &flush_elide);
qemu_printf("TLB full flushes %zu\n", flush_full);
qemu_printf("TLB partial flushes %zu\n", flush_part);
qemu_printf("TLB elided flushes %zu\n", flush_elide);
tcg_dump_info();
g_string_append_printf(buf, "TLB full flushes %zu\n", flush_full);
g_string_append_printf(buf, "TLB partial flushes %zu\n", flush_part);
g_string_append_printf(buf, "TLB elided flushes %zu\n", flush_elide);
tcg_dump_info(buf);
}
void dump_opcount_info(void)

4
include/exec/cpu-all.h
View File

@ -429,9 +429,9 @@ static inline bool tlb_hit(target_ulong tlb_addr, target_ulong addr)
#ifdef CONFIG_TCG
/* accel/tcg/cpu-exec.c */
void dump_drift_info(void);
void dump_drift_info(GString *buf);
/* accel/tcg/translate-all.c */
void dump_exec_info(void);
void dump_exec_info(GString *buf);
void dump_opcount_info(void);
#endif /* CONFIG_TCG */

2
include/tcg/tcg.h
View File

@ -937,7 +937,7 @@ int tcg_check_temp_count(void);
#endif
int64_t tcg_cpu_exec_time(void);
void tcg_dump_info(void);
void tcg_dump_info(GString *buf);
void tcg_dump_op_count(void);
#define TCG_CT_CONST 1 /* any constant of register size */

13
qapi/machine.json
View File

@ -1424,6 +1424,19 @@
{ 'command': 'x-query-irq',
'returns': 'HumanReadableText' }
##
# @x-query-jit:
#
# Query TCG compiler statistics
#
# Returns: TCG compiler statistics
#
# Since: 6.2
##
{ 'command': 'x-query-jit',
'returns': 'HumanReadableText',
'if': 'CONFIG_TCG' }
##
# @x-query-numa:
#

88
tcg/tcg.c
View File

@ -4383,7 +4383,7 @@ int tcg_gen_code(TCGContext *s, TranslationBlock *tb)
}
#ifdef CONFIG_PROFILER
void tcg_dump_info(void)
void tcg_dump_info(GString *buf)
{
TCGProfile prof = {};
const TCGProfile *s;
@ -4397,53 +4397,59 @@ void tcg_dump_info(void)
tb_div_count = tb_count ? tb_count : 1;
tot = s->interm_time + s->code_time;
qemu_printf("JIT cycles %" PRId64 " (%0.3f s at 2.4 GHz)\n",
tot, tot / 2.4e9);
qemu_printf("translated TBs %" PRId64 " (aborted=%" PRId64
" %0.1f%%)\n",
tb_count, s->tb_count1 - tb_count,
(double)(s->tb_count1 - s->tb_count)
/ (s->tb_count1 ? s->tb_count1 : 1) * 100.0);
qemu_printf("avg ops/TB %0.1f max=%d\n",
(double)s->op_count / tb_div_count, s->op_count_max);
qemu_printf("deleted ops/TB %0.2f\n",
(double)s->del_op_count / tb_div_count);
qemu_printf("avg temps/TB %0.2f max=%d\n",
(double)s->temp_count / tb_div_count, s->temp_count_max);
qemu_printf("avg host code/TB %0.1f\n",
(double)s->code_out_len / tb_div_count);
qemu_printf("avg search data/TB %0.1f\n",
(double)s->search_out_len / tb_div_count);
g_string_append_printf(buf, "JIT cycles %" PRId64
" (%0.3f s at 2.4 GHz)\n",
tot, tot / 2.4e9);
g_string_append_printf(buf, "translated TBs %" PRId64
" (aborted=%" PRId64 " %0.1f%%)\n",
tb_count, s->tb_count1 - tb_count,
(double)(s->tb_count1 - s->tb_count)
/ (s->tb_count1 ? s->tb_count1 : 1) * 100.0);
g_string_append_printf(buf, "avg ops/TB %0.1f max=%d\n",
(double)s->op_count / tb_div_count, s->op_count_max);
g_string_append_printf(buf, "deleted ops/TB %0.2f\n",
(double)s->del_op_count / tb_div_count);
g_string_append_printf(buf, "avg temps/TB %0.2f max=%d\n",
(double)s->temp_count / tb_div_count,
s->temp_count_max);
g_string_append_printf(buf, "avg host code/TB %0.1f\n",
(double)s->code_out_len / tb_div_count);
g_string_append_printf(buf, "avg search data/TB %0.1f\n",
(double)s->search_out_len / tb_div_count);
qemu_printf("cycles/op %0.1f\n",
s->op_count ? (double)tot / s->op_count : 0);
qemu_printf("cycles/in byte %0.1f\n",
s->code_in_len ? (double)tot / s->code_in_len : 0);
qemu_printf("cycles/out byte %0.1f\n",
s->code_out_len ? (double)tot / s->code_out_len : 0);
qemu_printf("cycles/search byte %0.1f\n",
s->search_out_len ? (double)tot / s->search_out_len : 0);
g_string_append_printf(buf, "cycles/op %0.1f\n",
s->op_count ? (double)tot / s->op_count : 0);
g_string_append_printf(buf, "cycles/in byte %0.1f\n",
s->code_in_len ? (double)tot / s->code_in_len : 0);
g_string_append_printf(buf, "cycles/out byte %0.1f\n",
s->code_out_len ? (double)tot / s->code_out_len : 0);
g_string_append_printf(buf, "cycles/search byte %0.1f\n",
s->search_out_len ?
(double)tot / s->search_out_len : 0);
if (tot == 0) {
tot = 1;
}
qemu_printf(" gen_interm time %0.1f%%\n",
(double)s->interm_time / tot * 100.0);
qemu_printf(" gen_code time %0.1f%%\n",
(double)s->code_time / tot * 100.0);
qemu_printf("optim./code time %0.1f%%\n",
(double)s->opt_time / (s->code_time ? s->code_time : 1)
* 100.0);
qemu_printf("liveness/code time %0.1f%%\n",
(double)s->la_time / (s->code_time ? s->code_time : 1) * 100.0);
qemu_printf("cpu_restore count %" PRId64 "\n",
s->restore_count);
qemu_printf(" avg cycles %0.1f\n",
s->restore_count ? (double)s->restore_time / s->restore_count : 0);
g_string_append_printf(buf, " gen_interm time %0.1f%%\n",
(double)s->interm_time / tot * 100.0);
g_string_append_printf(buf, " gen_code time %0.1f%%\n",
(double)s->code_time / tot * 100.0);
g_string_append_printf(buf, "optim./code time %0.1f%%\n",
(double)s->opt_time / (s->code_time ?
s->code_time : 1)
* 100.0);
g_string_append_printf(buf, "liveness/code time %0.1f%%\n",
(double)s->la_time / (s->code_time ?
s->code_time : 1) * 100.0);
g_string_append_printf(buf, "cpu_restore count %" PRId64 "\n",
s->restore_count);
g_string_append_printf(buf, " avg cycles %0.1f\n",
s->restore_count ?
(double)s->restore_time / s->restore_count : 0);
}
#else
void tcg_dump_info(void)
void tcg_dump_info(GString *buf)
{
qemu_printf("[TCG profiler not compiled]\n");
g_string_append_printf(buf, "[TCG profiler not compiled]\n");
}
#endif

2
tests/qtest/qmp-cmd-test.c
View File

@ -51,6 +51,8 @@ static int query_error_class(const char *cmd)
#endif
/* Only valid with a USB bus added */
{ "x-query-usb", ERROR_CLASS_GENERIC_ERROR },
/* Only valid with accel=tcg */
{ "x-query-jit", ERROR_CLASS_GENERIC_ERROR },
{ NULL, -1 }
};
int i;