[PATCH v2 1/2] drm/amd/pm: Add gpu_metrics_v1_7
Wang, Yang(Kevin)
KevinYang.Wang at amd.com
Wed Nov 13 08:35:15 UTC 2024
[AMD Official Use Only - AMD Internal Distribution Only]
-----Original Message-----
From: amd-gfx <amd-gfx-bounces at lists.freedesktop.org> On Behalf Of Asad Kamal
Sent: Wednesday, November 13, 2024 16:25
To: amd-gfx at lists.freedesktop.org; Lazar, Lijo <Lijo.Lazar at amd.com>
Cc: Ma, Le <Le.Ma at amd.com>; Zhang, Hawking <Hawking.Zhang at amd.com>; Zhang, Morris <Shiwu.Zhang at amd.com>; Kamal, Asad <Asad.Kamal at amd.com>; Poag, Charis <Charis.Poag at amd.com>; Cheung, Donald <donald.cheung at amd.com>; Khatir, Sepehr <sepehr.khatir at amd.com>; Oliveira, Daniel <Daniel.Oliveira at amd.com>
Subject: [PATCH v2 1/2] drm/amd/pm: Add gpu_metrics_v1_7
Add new gpu_metrics_v1_7 to acquire xgmi link status, application counter and max vram bandwidth
v2: Use gpu_metrics_v1_7 for SMU_v_13_0_6 (Lijo)
Signed-off-by: Asad Kamal <asad.kamal at amd.com>
Reviewed-by: Lijo Lazar <lijo.lazar at amd.com>
---
.../gpu/drm/amd/include/kgd_pp_interface.h | 110 ++++++++++++++++++
.../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 8 +-
drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c | 3 +
3 files changed, 117 insertions(+), 4 deletions(-)
diff --git a/drivers/gpu/drm/amd/include/kgd_pp_interface.h b/drivers/gpu/drm/amd/include/kgd_pp_interface.h
index ee8170cda1d7..a95935cc3d76 100644
--- a/drivers/gpu/drm/amd/include/kgd_pp_interface.h
+++ b/drivers/gpu/drm/amd/include/kgd_pp_interface.h
@@ -364,6 +364,17 @@ struct amdgpu_xcp_metrics {
u64 gfx_busy_acc[MAX_XCC];
};
+struct amdgpu_xcp_metrics_v1_1 {
+ /* Utilization Instantaneous (%) */
+ u32 gfx_busy_inst[MAX_XCC];
+ u16 jpeg_busy[NUM_JPEG_ENG];
+ u16 vcn_busy[NUM_VCN];
+ /* Utilization Accumulated (%) */
+ u64 gfx_busy_acc[MAX_XCC];
+ /* Total App Clock Counter Accumulated */
+ u64 gfx_below_host_limit_acc[MAX_XCC];
+};
[kevin]:
It is better to change the type to 'uint64/32/16_t' (stdint.h) to indicate that it is a uapi interface.
With fixed,
Reviewed-by: Yang Wang <kevinyang.wang at amd.com>
Best Regards,
Kevin
+
struct amd_pm_funcs {
/* export for dpm on ci and si */
int (*pre_set_power_state)(void *handle); @@ -979,6 +990,105 @@ struct gpu_metrics_v1_6 {
uint32_t pcie_lc_perf_other_end_recovery;
};
+struct gpu_metrics_v1_7 {
+ struct metrics_table_header common_header;
+
+ /* Temperature (Celsius) */
+ uint16_t temperature_hotspot;
+ uint16_t temperature_mem;
+ uint16_t temperature_vrsoc;
+
+ /* Power (Watts) */
+ uint16_t curr_socket_power;
+
+ /* Utilization (%) */
+ uint16_t average_gfx_activity;
+ uint16_t average_umc_activity; // memory controller
+
+ /* VRAM max bandwidthi (in GB/sec) at max memory clock */
+ uint64_t mem_max_bandwidth;
+
+ /* Energy (15.259uJ (2^-16) units) */
+ uint64_t energy_accumulator;
+
+ /* Driver attached timestamp (in ns) */
+ uint64_t system_clock_counter;
+
+ /* Accumulation cycle counter */
+ uint32_t accumulation_counter;
+
+ /* Accumulated throttler residencies */
+ uint32_t prochot_residency_acc;
+ uint32_t ppt_residency_acc;
+ uint32_t socket_thm_residency_acc;
+ uint32_t vr_thm_residency_acc;
+ uint32_t hbm_thm_residency_acc;
+
+ /* Clock Lock Status. Each bit corresponds to clock instance */
+ uint32_t gfxclk_lock_status;
+
+ /* Link width (number of lanes) and speed (in 0.1 GT/s) */
+ uint16_t pcie_link_width;
+ uint16_t pcie_link_speed;
+
+ /* XGMI bus width and bitrate (in Gbps) */
+ uint16_t xgmi_link_width;
+ uint16_t xgmi_link_speed;
+
+ /* Utilization Accumulated (%) */
+ uint32_t gfx_activity_acc;
+ uint32_t mem_activity_acc;
+
+ /*PCIE accumulated bandwidth (GB/sec) */
+ uint64_t pcie_bandwidth_acc;
+
+ /*PCIE instantaneous bandwidth (GB/sec) */
+ uint64_t pcie_bandwidth_inst;
+
+ /* PCIE L0 to recovery state transition accumulated count */
+ uint64_t pcie_l0_to_recov_count_acc;
+
+ /* PCIE replay accumulated count */
+ uint64_t pcie_replay_count_acc;
+
+ /* PCIE replay rollover accumulated count */
+ uint64_t pcie_replay_rover_count_acc;
+
+ /* PCIE NAK sent accumulated count */
+ uint32_t pcie_nak_sent_count_acc;
+
+ /* PCIE NAK received accumulated count */
+ uint32_t pcie_nak_rcvd_count_acc;
+
+ /* XGMI accumulated data transfer size(KiloBytes) */
+ uint64_t xgmi_read_data_acc[NUM_XGMI_LINKS];
+ uint64_t xgmi_write_data_acc[NUM_XGMI_LINKS];
+
+ /* XGMI link status(active/inactive) */
+ uint16_t xgmi_link_status[NUM_XGMI_LINKS];
+
+ uint16_t padding;
+
+ /* PMFW attached timestamp (10ns resolution) */
+ uint64_t firmware_timestamp;
+
+ /* Current clocks (Mhz) */
+ uint16_t current_gfxclk[MAX_GFX_CLKS];
+ uint16_t current_socclk[MAX_CLKS];
+ uint16_t current_vclk0[MAX_CLKS];
+ uint16_t current_dclk0[MAX_CLKS];
+ uint16_t current_uclk;
+
+ /* Number of current partition */
+ uint16_t num_partition;
+
+ /* XCP metrics stats */
+ struct amdgpu_xcp_metrics_v1_1 xcp_stats[NUM_XCP];
+
+ /* PCIE other end recovery counter */
+ uint32_t pcie_lc_perf_other_end_recovery;
+};
+
/*
* gpu_metrics_v2_0 is not recommended as it's not naturally aligned.
* Use gpu_metrics_v2_1 or later instead.
diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c
index fa30a9e1f27a..11ecaa62f419 100644
--- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c
+++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c
@@ -370,7 +370,7 @@ static int smu_v13_0_6_tables_init(struct smu_context *smu)
return -ENOMEM;
smu_table->metrics_time = 0;
- smu_table->gpu_metrics_table_size = sizeof(struct gpu_metrics_v1_6);
+ smu_table->gpu_metrics_table_size = sizeof(struct gpu_metrics_v1_7);
smu_table->gpu_metrics_table =
kzalloc(smu_table->gpu_metrics_table_size, GFP_KERNEL);
if (!smu_table->gpu_metrics_table) {
@@ -2321,8 +2321,8 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table {
bool per_inst, smu_13_0_6_per_inst, smu_13_0_14_per_inst, apu_per_inst;
struct smu_table_context *smu_table = &smu->smu_table;
- struct gpu_metrics_v1_6 *gpu_metrics =
- (struct gpu_metrics_v1_6 *)smu_table->gpu_metrics_table;
+ struct gpu_metrics_v1_7 *gpu_metrics =
+ (struct gpu_metrics_v1_7 *)smu_table->gpu_metrics_table;
bool flag = smu_v13_0_6_is_unified_metrics(smu);
int ret = 0, xcc_id, inst, i, j, k, idx;
struct amdgpu_device *adev = smu->adev; @@ -2341,7 +2341,7 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table
metrics_a = (MetricsTableA_t *)metrics_x;
- smu_cmn_init_soft_gpu_metrics(gpu_metrics, 1, 6);
+ smu_cmn_init_soft_gpu_metrics(gpu_metrics, 1, 7);
gpu_metrics->temperature_hotspot =
SMUQ10_ROUND(GET_METRIC_FIELD(MaxSocketTemperature, flag)); diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c
index fd2aa949538e..a9e69d321f8c 100644
--- a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c
+++ b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c
@@ -1081,6 +1081,9 @@ void smu_cmn_init_soft_gpu_metrics(void *table, uint8_t frev, uint8_t crev)
case METRICS_VERSION(1, 6):
structure_size = sizeof(struct gpu_metrics_v1_6);
break;
+ case METRICS_VERSION(1, 7):
+ structure_size = sizeof(struct gpu_metrics_v1_7);
+ break;
case METRICS_VERSION(2, 0):
structure_size = sizeof(struct gpu_metrics_v2_0);
break;
--
2.46.0
More information about the amd-gfx
mailing list