AI/rocm-systems
2
0
Vork 0
Code Kwesties Pull-aanvragen Actions Paketten Projecten Publicaties Wiki Activiteit

[SDK] Added retry for signal wait (#494)

Bladeren bron 
Added retry for signal wait

Co-authored-by: Benjamin Welton <bewelton@amd.com>
This commit is contained in:
Welton, Benjamin
2025-07-08 23:28:02 -07:00
gecommit door GitHub
bovenliggende 30e239d1b7
commit 2fa98c4d14
1 gewijzigde bestanden met toevoegingen van 13 en 7 verwijderingen
Download Patch-bestand Download Diff-bestand Expand all files Collapse all files
source/lib/rocprofiler-sdk/hsa/queue.cpp
+13 -7
Bestand weergeven
@@ -569,15 +569,21 @@ Queue::Queue(const AgentCache& agent,
counters::submitPacket(_intercept_queue, &pkt);
constexpr auto timeout_hint =
std::chrono::duration_cast<std::chrono::nanoseconds>(std::chrono::seconds{1});
if(core_api.hsa_signal_wait_relaxed_fn(completion,
HSA_SIGNAL_CONDITION_EQ,
0,
timeout_hint.count(),
HSA_WAIT_STATE_ACTIVE) != 0)
hsa_signal_value_t val;
for(int i = 0; i < 3; i++)
{
ROCP_FATAL << "Could not set agent to be profiled";
val = core_api.hsa_signal_wait_scacquire_fn(completion,
HSA_SIGNAL_CONDITION_EQ,
0,
timeout_hint.count(),
HSA_WAIT_STATE_ACTIVE);
if(val == 0)
{
core_api.hsa_signal_destroy_fn(completion);
return;
}
}
core_api.hsa_signal_destroy_fn(completion);
ROCP_FATAL << "Could not set agent to be profiled - Signal Value: " << val;
});
}