habanalabs: add uapi to retrieve device utilization

Users and sysadmins usually want to know what is the device utilization as a level 0 indication if they are efficiently using the device. Add a new opcode to the INFO IOCTL that will return the device utilization over the last period of 100-1000ms. The return value is 0-100, representing as percentage the total utilization rate.

Oded Gabbay committed 7 years ago

Commit 75b3cb21 parent 413cf577 files · +222 −23Browse files

Showing the changes to include/uapi (1 of 7 files). Show all

Unified · Split

include/uapi/misc/habanalabs.h+34−15

@@ -77,22 +77,29 @@ enum hl_device_status { · 76 unmodified lines
77	77
78	78	/* Opcode for management ioctl
79	79	*
80		− * HW_IP_INFO - Receive information about different IP blocks in the
81		− * device.
82		− * HL_INFO_HW_EVENTS - Receive an array describing how many times each event
83		− * occurred since the last hard reset.
84		− * HL_INFO_DRAM_USAGE - Retrieve the dram usage inside the device and of the
85		− * specific context. This is relevant only for GOYA device.
86		− * HL_INFO_HW_IDLE - Retrieve information about the idle status of each
87		− * internal engine.
	80	+ * HW_IP_INFO - Receive information about different IP blocks in the
	81	+ * device.
	82	+ * HL_INFO_HW_EVENTS - Receive an array describing how many times each event
	83	+ * occurred since the last hard reset.
	84	+ * HL_INFO_DRAM_USAGE - Retrieve the dram usage inside the device and of the
	85	+ * specific context. This is relevant only for devices
	86	+ * where the dram is managed by the kernel driver
	87	+ * HL_INFO_HW_IDLE - Retrieve information about the idle status of each
	88	+ * internal engine.
88	89	* HL_INFO_DEVICE_STATUS - Retrieve the device's status. This opcode doesn't
89	90	* require an open context.
	91	+ * HL_INFO_DEVICE_UTILIZATION - Retrieve the total utilization of the device
	92	+ * over the last period specified by the user.
	93	+ * The period can be between 100ms to 1s, in
	94	+ * resolution of 100ms. The return value is a
	95	+ * percentage of the utilization rate.
90	96	*/
91		−#define HL_INFO_HW_IP_INFO 0
92		−#define HL_INFO_HW_EVENTS 1
93		−#define HL_INFO_DRAM_USAGE 2
94		−#define HL_INFO_HW_IDLE 3
95		−#define HL_INFO_DEVICE_STATUS 4
	97	+#define HL_INFO_HW_IP_INFO 0
	98	+#define HL_INFO_HW_EVENTS 1
	99	+#define HL_INFO_DRAM_USAGE 2
	100	+#define HL_INFO_HW_IDLE 3
	101	+#define HL_INFO_DEVICE_STATUS 4
	102	+#define HL_INFO_DEVICE_UTILIZATION 6
96	103
97	104	#define HL_INFO_VERSION_MAX_LEN 128
98	105
@@ -134,6 +141,11 @@ struct hl_info_device_status { · 35 unmodified lines
134	141	__u32 pad;
135	142	};
136	143
	144	+struct hl_info_device_utilization {
	145	+ __u32 utilization;
	146	+ __u32 pad;
	147	+};
	148	+
137	149	struct hl_info_args {
138	150	/* Location of relevant struct in userspace */
139	151	__u64 return_pointer;
@@ -149,8 +161,15 @@ struct hl_info_args { · 9 unmodified lines
149	161	/* HL_INFO_* */
150	162	__u32 op;
151	163
152		− /* Context ID - Currently not in use */
153		− __u32 ctx_id;
	164	+ union {
	165	+ /* Context ID - Currently not in use */
	166	+ __u32 ctx_id;
	167	+ /* Period value for utilization rate (100ms - 1000ms, in 100ms
	168	+ * resolution.
	169	+ */
	170	+ __u32 period_ms;
	171	+ };
	172	+
154	173	__u32 pad;
155	174	};
156	175