The existing gpu_mem_total tracepoint provides GPU drivers a uniform way
to report the per-process and system-wide GPU memory usage. This
tracepoint reports a single total of the GPU private allocations and the
imported memory. [1]
To allow distinguishing GPU private vs imported memory, add an
imported_size field to the gpu_mem_total tracepoint. GPU drivers can use
this new field to report the per-process and global GPU-imported memory
in a uniform way.
User space tools can detect and handle the old vs new gpu_mem_total
format via the gpu_mem/gpu_mem_total/format file.
[1] https://lore.kernel.org/r/[email protected]/
Signed-off-by: Kalesh Singh <[email protected]>
---
Changes in v2:
- Add imported_size field to track imported memory instead of a
separate tracepoint, per Ørjan
v1 of this patch can be found at:
https://lore.kernel.org/r/[email protected]/
include/trace/events/gpu_mem.h | 17 +++++++++++------
1 file changed, 11 insertions(+), 6 deletions(-)
diff --git a/include/trace/events/gpu_mem.h b/include/trace/events/gpu_mem.h
index 26d871f96e94..ae6fab6bab7b 100644
--- a/include/trace/events/gpu_mem.h
+++ b/include/trace/events/gpu_mem.h
@@ -15,7 +15,7 @@
/*
* The gpu_memory_total event indicates that there's an update to either the
- * global or process total gpu memory counters.
+ * global or process total and imported gpu memory counters.
*
* This event should be emitted whenever the kernel device driver allocates,
* frees, imports, unimports memory in the GPU addressable space.
@@ -24,31 +24,36 @@
*
* @pid: Put 0 for global total, while positive pid for process total.
*
- * @size: Size of the allocation in bytes.
+ * @size: Total size of allocated and imported memory in bytes.
+ *
+ * @imported_size: Total size of imported memory in bytes.
*
*/
TRACE_EVENT(gpu_mem_total,
- TP_PROTO(uint32_t gpu_id, uint32_t pid, uint64_t size),
+ TP_PROTO(uint32_t gpu_id, uint32_t pid, uint64_t size, uint64_t imported_size),
- TP_ARGS(gpu_id, pid, size),
+ TP_ARGS(gpu_id, pid, size, imported_size),
TP_STRUCT__entry(
__field(uint32_t, gpu_id)
__field(uint32_t, pid)
__field(uint64_t, size)
+ __field(uint64_t, imported_size)
),
TP_fast_assign(
__entry->gpu_id = gpu_id;
__entry->pid = pid;
__entry->size = size;
+ __entry->imported_size = imported_size;
),
- TP_printk("gpu_id=%u pid=%u size=%llu",
+ TP_printk("gpu_id=%u pid=%u size=%llu imported_size=%llu",
__entry->gpu_id,
__entry->pid,
- __entry->size)
+ __entry->size,
+ __entry->imported_size)
);
#endif /* _TRACE_GPU_MEM_H */
base-commit: a2824f19e6065a0d3735acd9fe7155b104e7edf5
--
2.33.0.rc1.237.g0d66db33f3-goog