On Sat, 14 Feb 2026, Marc-André Lureau wrote:
On Fri, Feb 13, 2026 at 4:50 PM Mark Cave-Ayland
<[email protected]> wrote:
On 27/01/2026 18:25, [email protected] wrote:
From: Marc-André Lureau <[email protected]>
Use more correct GTimer, specific for each backend, with a trace event.
Signed-off-by: Marc-André Lureau <[email protected]>
---
audio/audio_int.h | 1 +
audio/audio-mixeng-be.c | 25 +++++++++----------------
audio/trace-events | 1 +
3 files changed, 11 insertions(+), 16 deletions(-)
diff --git a/audio/audio_int.h b/audio/audio_int.h
index dd5f2220d75..e0ed6af67ed 100644
--- a/audio/audio_int.h
+++ b/audio/audio_int.h
@@ -224,6 +224,7 @@ struct AudioMixengBackend {
Audiodev *dev;
QEMUTimer *ts;
+ GTimer *run_timer;
QLIST_HEAD (hw_in_listhead, HWVoiceIn) hw_head_in;
QLIST_HEAD (hw_out_listhead, HWVoiceOut) hw_head_out;
QLIST_HEAD (cap_listhead, CaptureVoiceOut) cap_head;
diff --git a/audio/audio-mixeng-be.c b/audio/audio-mixeng-be.c
index 343f6ec181e..1b2c1fafa87 100644
--- a/audio/audio-mixeng-be.c
+++ b/audio/audio-mixeng-be.c
@@ -24,13 +24,13 @@
#include "system/replay.h"
#include "system/runstate.h"
#include "trace.h"
+#include "trace/control.h"
#define AUDIO_CAP "audio"
#include "audio_int.h"
/* #define DEBUG_OUT */
/* #define DEBUG_CAPTURE */
-/* #define DEBUG_POLL */
#define SW_NAME(sw) (sw)->name ? (sw)->name : "unknown"
@@ -1269,22 +1269,12 @@ void audio_run(AudioMixengBackend *s, const char *msg)
audio_run_in(s);
audio_run_capture(s);
-#ifdef DEBUG_POLL
- {
- static double prevtime;
- double currtime;
- struct timeval tv;
-
- if (gettimeofday (&tv, NULL)) {
- perror ("audio_run: gettimeofday");
- return;
- }
-
- currtime = tv.tv_sec + tv.tv_usec * 1e-6;
- dolog ("Elapsed since last %s: %f\n", msg, currtime - prevtime);
- prevtime = currtime;
+ if (trace_event_get_state(TRACE_AUDIO_RUN_POLL)) {
+ /* Convert seconds to microseconds for trace event */
+ int64_t elapsed_us = g_timer_elapsed(s->run_timer, NULL) * 1000000;
Not being familiar with g_timer_elapsed() I had a look at the glib docs,
and it seems that g_timer_elapsed() returns a gdouble as opposed to an
int64_t?
yes, as seconds. So it is multiplied by 1000000 to get back usec, as before
There's #define NANOSECONDS_PER_SECOND 1000000000LL in qemu/timer.h. Do we
need a MICROSECONDS_PER_SECOND constant there (or defined here) to make it
more obvious?
Regards,
BALATON Zoltan
