Memory (Memory Read/Write Bandwidth Rate)

The memory read/write bandwidth rate data collected from the UB/L1/L2/main memory is saved in Memory.csv. For details, see the field description in the following table.

The unit GB/s indicates that 1 GB of data is transmitted per second.

Atlas A3 Training Products/Atlas A3 Inference Products and Atlas A2 Training Products/Atlas A2 Inference Products

Figure 1 Memory.csv file
See the following table for more details.
Table 1 Field description

Field

Description

block_id

Number of running task blocks, which corresponds to the number of cores configured during task running.

sub_block_id

Name and sequence number of each block used for task running.

aic_time(us)

Execution time of each AI Cube Core compute unit after the task is allocated to the unit, in μs.

aic_total_cycles

Total number of cycles executed on each AI Cube Core compute unit after the task is allocated to the unit.

aiv_time(us)

Execution time of each AI Vector Core compute unit after the task is allocated to the unit, in μs.

aiv_total_cycles

Total number of cycles executed on each AI Vector Core compute unit after the task is allocated to the unit.

aiv_ub_to_gm_bw(GB/s)

Bandwidth of data written to GM by UB corresponding to the total cycles, in GB/s.

aiv_gm_to_ub_bw(GB/s)

Bandwidth of data written to UB by GM corresponding to the total cycles, in GB/s.

aic_l1_read_bw(GB/s)

Bandwidth of data read from all other units by the L1 unit in this operator corresponding to the total cycles, in GB/s.

aic_l1_write_bw(GB/s)

Bandwidth of data written to all other units by the L1 unit in this operator corresponding to the total cycles, in GB/s.

ai*_main_mem_read_bw(GB/s)

Bandwidth of data read from all other units by the main memory corresponding to the total cycles, in GB/s.

ai*_main_mem_write_bw(GB/s)

Bandwidth of data written to all other units by the main memory corresponding to the total cycles, in GB/s.

aic_mte1_instructions

Number of MTE1 instructions.

aic_mte1_ratio

Ratio of cycles taken to execute MTE1 instructions to the total cycles.

ai*_mte2_instructions

Number of MTE2 instructions.

ai*_mte2_ratio

Ratio of cycles taken to execute MTE2 instructions to the total cycles.

ai*_mte3_instructions

Number of MTE3 instructions.

ai*_mte3_ratio

Ratio of cycles taken to execute MTE3 instructions to the total cycles.

read_main_memory_datas(KB)

Total amount of data read from the main memory.

write_main_memory_datas(KB)

Total amount of data written to the main memory.

GM_to_L1_datas(KB)

Amount of data transferred from GM to L1.

L1_to_GM_datas(KB)(estimate)

Estimated amount of data transferred from L1 to GM.

L0C_to_L1_datas(KB)

Amount of data transferred from L0C to L1.

L0C_to_GM_datas(KB)

Amount of data transferred from L0C to GM.

GM_to_UB_datas(KB)

Amount of data transferred from GM to UB.

UB_to_GM_datas(KB)

Amount of data transferred from UB to GM.

GM_to_L1_bw_usage_rate(%)

Bandwidth usage of the channel from GM to L1.

L1_to_GM_bw_usage_rate(%)(estimate)

Estimated bandwidth usage of the channel from L1 to GM.

L0C_to_L1_bw_usage_rate(%)

Bandwidth usage of the channel from L0C to L1.

L0C_to_GM_bw_usage_rate(%)

Bandwidth usage of the channel from L0C to GM.

GM_to_UB_bw_usage_rate(%)

Bandwidth usage of the channel from GM to UB.

UB_to_GM_bw_usage_rate(%)

Bandwidth usage of the channel from UB to GM.

Atlas Inference Products

Figure 2 Memory.csv file
See the following table for more details.
Table 2 Field description

Field

Description

aic_time(us)

Execution time of each AI Core compute unit after the task is allocated to the unit, in μs.

aic_total_cycles

Total number of cycles executed on each AI Core compute unit after the task is allocated to the unit.

aic_ub_to_gm_bw(GB/s)

Bandwidth of data written to GM by UB corresponding to the total cycles, in GB/s.

aic_gm_to_ub_bw(GB/s)

Bandwidth of data written to UB by GM corresponding to the total cycles, in GB/s.

aic_l1_read_bw(GB/s)

Bandwidth of data read from all other units by the L1 unit in this operator corresponding to the total cycles, in GB/s.

aic_l1_write_bw(GB/s)

Bandwidth of data written to all other units by the L1 unit in this operator corresponding to the total cycles, in GB/s.

aic_main_mem_read_bw(GB/s)

Bandwidth of data read from all other units by the main memory corresponding to the total cycles, in GB/s.

aic_main_mem_write_bw(GB/s)

Bandwidth of data written to all other units by the main memory corresponding to the total cycles, in GB/s.

aic_mte1_instructions

Number of MTE1 instructions.

aic_mte1_ratio

Ratio of cycles taken to execute MTE1 instructions to the total cycles.

aic_mte2_instructions

Number of MTE2 instructions.

aic_mte2_ratio

Ratio of cycles taken to execute MTE2 instructions to the total cycles.

aic_mte3_instructions

Number of MTE3 instructions.

aic_mte3_ratio

Ratio of cycles taken to execute MTE3 instructions to the total cycles.