ai_core_utilization (AI Core Instruction Proportion)

The timeline information of the AI Core instruction proportion data is displayed at the AI Core Utilization layer in the msprof_*.json file, and the summary information is summarized in the ai_core_utilization_*.csv file.

Applicable Products

For details about the mapping between AI processors and Ascend products, see Ascend Product Overview.

AI Processor Type	Supported
Ascend 950PR/Ascend 950DT	√
Atlas A3 training products/Atlas A3 inference products	√
Atlas A2 training products/Atlas A2 inference products	√
Atlas 200I/500 A2 inference products	√
Atlas inference products	√
Atlas training products	√
Ascend 950PR/Ascend 950DT	√

AI Core Instruction Proportion Data in msprof_*.json

The following shows the file content format.

Figure 1 AI Core utilization layer

**Table 1** Field description
Field	Description
Average	Mean value.
Core <ID>	Core ID.
utilization(%)	Percentage of total execution cycles (counting from the first operator instruction executed by the AI Core to the completion of the last instruction executed) of a task on the AI Core in the current sampling period.

ai_core_utilization_*.csv File

The file content is formatted as follows.

Figure 2 ai_core_utilization (example only)

The file display result varies according to the value of --aic-metrics. The table describes the fields.

Supported fields may vary by product. Please refer to the actual result files for the final list of fields.
The following fields are generated when --task-time is set to l1 and --aic-mode is set to sample-based. If --task-time is set to l0, this field is not profiled and N/A is displayed. The generated data is controlled by the aic_metrics parameter.

**Table 2** Field description (**PipeUtilization**)
Field	Description
vec_ratio	Ratio of cycles taken to execute Vector instructions to the total cycles. For the Atlas 200I/500 A2 inference products, this field is not supported and defaults to N/A.The Atlas A2 training products/Atlas A2 inference products does not support this field.The Atlas A3 training products/Atlas A3 inference products does not support this field.
mac_ratio	Ratio of cycles taken to execute Cube instructions to the total cycles.
scalar_ratio	Ratio of cycles taken to execute Scalar instructions to the total cycles.
mte1_ratio	Ratio of cycles taken to execute MTE1 instructions (L1-to-L0A/L0B transfer) to the total cycles.
mte2_ratio	Ratio of cycles taken to execute MTE2 instructions (DDR-to-AI Core transfert) to the total cycles.
mte3_ratio	Ratio of cycles taken to execute MTE3 instructions (AI Core-to-DDR transfer) to the total cycles. The Atlas A2 training products/Atlas A2 inference products does not support this field.The Atlas A3 training products/Atlas A3 inference products does not support this field.
icache_miss_rate	iCache is the L2 cache reserved for instructions. If the value of icache_miss_rate is high, the AI Core reads instructions at a low efficiency.
fixpipe_ratio	Ratio of cycles taken to execute fixpipe instructions (L0C-to-OUT/L1 transfer) to the total cycles.
memory_bound	AI Core memory bound, calculated as: mte2_ratio/max(mac_ratio, vec_ratio). If the calculation result is less than 1, there is no memory bottleneck. If the result exceeds 1, the AI Core spends most of its time on memory data movement rather than computation during task execution. A larger value indicates a more severe memory bottleneck. The Atlas A2 training products/Atlas A2 inference products does not support this field.The Atlas A3 training products/Atlas A3 inference products does not support this field.

**Table 3** Field description (**ArithmeticUtilization**)
Field	Description
mac_fp16_ratio	Ratio of cycles taken to execute Cube fp16 instructions to the total cycles.
mac_int8_ratio	Ratio of cycles taken to execute Cube int8 instructions to the total cycles.
vec_fp32_ratio	Ratio of cycles taken to execute vec fp32 instructions to the total cycles. For the Atlas 200I/500 A2 inference products, this field is not supported and defaults to N/A.The Ascend 950PR/Ascend 950DT does not support this field.
vec_fp16_ratio	Ratio of cycles taken to execute vec fp16 instructions to the total cycles. For the Atlas 200I/500 A2 inference products, this field is not supported and defaults to N/A.The Ascend 950PR/Ascend 950DT does not support this field.
vec_int32_ratio	Ratio of cycles taken to execute vec int32 instructions to the total cycles. For the Atlas 200I/500 A2 inference products, this field is not supported and defaults to N/A.The Ascend 950PR/Ascend 950DT does not support this field.
vec_misc_ratio	Ratio of cycles taken to execute vec misc instructions to the total cycles. For the Atlas 200I/500 A2 inference products, this field is not supported and defaults to N/A.The Ascend 950PR/Ascend 950DT does not support this field.
cube_fops	Floating-point operations (FLOPs, that is, fops in this field) of the Cube type, indicating the computation amount. This field can be used to measure the complexity of an algorithm or model.
vector_fops	Floating-point operations (FLOPs, that is, fops in this field) of the Vector type, indicating the computation amount. This field can be used to measure the complexity of an algorithm or model. The Ascend 950PR/Ascend 950DT does not support this field.

**Table 4** Field description (**Memory**)
Field	Description
ub_read_bw(GB/s)	UB read bandwidth, in GB/s. For the Atlas 200I/500 A2 inference products, this field is not supported and defaults to N/A.
ub_write_bw(GB/s)	UB write bandwidth, in GB/s. For the Atlas 200I/500 A2 inference products, this field is not supported and defaults to N/A.
l1_read_bw(GB/s)	L1 read bandwidth, in GB/s.
l1_write_bw(GB/s)	L1 write bandwidth, in GB/s.
l2_read_bw	L2 read bandwidth, in GB/s. The Ascend 950PR/Ascend 950DT does not support this field.
l2_write_bw	L2 write bandwidth, in GB/s. For the Atlas 200I/500 A2 inference products, this field is not supported and defaults to N/A.The Ascend 950PR/Ascend 950DT does not support this field.
main_mem_read_bw(GB/s)	Main memory read bandwidth, in GB/s.
main_mem_write_bw(GB/s)	Main memory write bandwidth, in GB/s. For the Atlas 200I/500 A2 inference products, this field is not supported and defaults to N/A.

**Table 5** Field description (**MemoryL0**)
Field	Description
l0a_read_bw(GB/s)	l0a read bandwidth, in GB/s.
l0a_write_bw(GB/s)	l0a write bandwidth, in GB/s.
l0b_read_bw(GB/s)	l0b read bandwidth, in GB/s.
l0b_write_bw(GB/s)	l0b write bandwidth, in GB/s.
l0c_read_bw(GB/s)	Bandwidth rate for Vector to read data from L0C, in GB/s.
l0c_write_bw(GB/s)	Bandwidth rate for Vector to write data to L0C, in GB/s. This field is not supported by the Ascend 950PR/Ascend 950DT.
l0c_read_bw_cube(GB/s)	Bandwidth rate for Cube to read data from L0C, in GB/s.
l0c_write_bw_cube(GB/s)	Bandwidth rate for Cube to write data to L0C, in GB/s.
Note: Data about the MemoryL0 metric of the AI Vector Core is 0.

**Table 6** Field description (**MemoryUB**)
Field	Description
ub_read_bw_vector(GB/s)	Bandwidth rate for Vector to read data from UB, in GB/s.
ub_write_bw_vector(GB/s)	Bandwidth rate for Vector to write data to UB, in GB/s.
ub_read_bw_scalar(GB/s)	Bandwidth rate for Scalar to read data from UB, in GB/s.
ub_write_bw_scalar(GB/s)	Bandwidth rate for Scalar to write data to UB, in GB/s.
ub_fixp2ub_write_bw(GB/s)	Write bandwidth of Vector fixpipe to UB (excluding the UB backpressure), in GB/s. Only the Ascend 950PR/Ascend 950DT supports this field.

**Table 7** Field description (**ResourceConflictRatio**)
Field	Description
vec_bankgroup_cflt_ratio	Ratio of cycles taken to execute vec_bankgroup_stall_cycles instructions to the total cycles. Improper block stride settings in Vector instructions can lead to bank group conflicts. This field is not supported by the Atlas 200I/500 A2 inference products and defaults to N/A.This field is not supported by the Ascend 950PR/Ascend 950DT.
vec_bank_cflt_ratio	Ratio of cycles taken to execute vec_bank_stall_cycles instructions to the total cycles. Improper read/write pointer addresses for Vector instruction operands can lead to bank conflicts. This field is not supported by the Atlas 200I/500 A2 inference products and defaults to N/A.
vec_resc_cflt_ratio	Ratio of cycles taken to execute vec_resc_cflt_ratio instructions to the total cycles. If an operator involves multiple compute units, ensure that they are concurrently scheduled. When a compute unit is working, but the operator logic still delivers instructions to it, the overall computing power is not fully utilized. This field is not supported by the Atlas 200I/500 A2 inference products and defaults to N/A.

**Table 8** Field description (**L2Cache**)
Field	Description
write_cache_hit	Write cache hits. The Ascend 950PR/Ascend 950DT does not support this field.
write_cache_miss_allocate	Cache re-allocations upon write cache missing. The Ascend 950PR/Ascend 950DT does not support this field.
r*_read_cache_hit	Read cache hits in the r* channel. The Ascend 950PR/Ascend 950DT does not support this field.
r*_read_cache_miss_allocate	Cache re-allocations upon read cache missing in the r* channel. The Ascend 950PR/Ascend 950DT does not support this field.
read_local_l2_hit	Read cache hits. Only the Ascend 950PR/Ascend 950DT supports this field.
read_local_l2_miss	Read cache misses. Only the Ascend 950PR/Ascend 950DT supports this field.
read_local_l2_victim	Number of read cache misses that trigger cache victimization. Only the Ascend 950PR/Ascend 950DT supports this field.
write_local_l2_hit	Write cache hits. Only the Ascend 950PR/Ascend 950DT supports this field.
write_local_l2_miss	Write cache misses. Only the Ascend 950PR/Ascend 950DT supports this field.
write_local_l2_victim	Number of write cache misses that trigger cache victimization. Only the Ascend 950PR/Ascend 950DT supports this field.
Availability: Atlas A2 training products/Atlas A2 inference products Atlas A3 training products/Atlas A3 inference products Ascend 950PR/Ascend 950DT Atlas 200I/500 A2 inference products

**Table 9** Field description (**MemoryAccess**)
Field	Description
read_main_memory_datas(KB)	Amount of data read from the on-chip memory, in KB.
write_main_memory_datas(KB)	Amount of data written to the on-chip memory, in KB.
gm_to_l1_datas(KB)	Amount of data transferred from GM to L1, in KB.
l0c_to_l1_datas(KB)	Amount of data transferred from L0C to L1, in KB.
l0c_to_gm_datas(KB)	Amount of data transferred from L0C to GM, in KB.
gm_to_ub_datas(KB)	Amount of data transferred from GM to UB, in KB.
ub_to_gm_datas(KB)	Amount of data transferred from UB to GM, in KB.
Availability: Atlas A2 training products/Atlas A2 inference products Atlas A3 training products/Atlas A3 inference products

Parent topic: Profile Data File References