<html><head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body>
<p><br>
</p>
<div class="moz-cite-prefix">On 2024-07-17 16:10, Felix Kuehling
wrote:<br>
</div>
<blockquote type="cite" cite="mid:2239cfa0-6f41-4d8a-9915-0ec1829e248d@amd.com">
<blockquote type="cite" style="color: #007cff;">@@ -603,8 +606,6
@@ struct queue {
<br>
void *gang_ctx_bo;
<br>
uint64_t gang_ctx_gpu_addr;
<br>
void *gang_ctx_cpu_ptr;
<br>
-
<br>
- struct amdgpu_bo *wptr_bo;
<br>
</blockquote>
<br>
If the wptr_bo_gart is GART-mapped and freed in the same place as
the gang_ctx_bo, then maybe it makes sense to keep the two
together in this structure. It also avoids having two different
reference to the same BO in the same queue_properties structure
above.
<br>
</blockquote>
<p>Yes, agree it makes sense to keep it inside struct queue and
rename to wptr_bo_gart for GART mapping and unmapping. Add wptr_bo
to struct queue_properties for queue wptr acquire and release.</p>
<p>Regards,</p>
<p>Philip<br>
</p>
<blockquote type="cite" cite="mid:2239cfa0-6f41-4d8a-9915-0ec1829e248d@amd.com">
<br>
Other than that, this patch looks good to me.
<br>
<br>
Regards,
<br>
Felix
<br>
<br>
<br>
<blockquote type="cite" style="color: #007cff;"> };
<br>
enum KFD_MQD_TYPE {
</blockquote>
</blockquote>
</body>
</html>