Total Time Avg:

{{avgTime}}

FLOPS utilization:

HBM bandwidth utilization:

On-Chip Read bandwidth utilization:

On-Chip Write bandwidth utilization:

{{flopsRate}}

bf16 normalized FLOP rate (per core):

{{bf16FlopsRate}}

{{bandwidths[memBwType.MEM_BW_TYPE_HBM_RW]}}

{{bandwidths[memBwType.MEM_BW_TYPE_SRAM_RD]}}

{{bandwidths[memBwType.MEM_BW_TYPE_SRAM_WR]}}

XLA expression:

{{expression}}

Long Op Name:

{{provenance}}

Xprof Kernel Metadata:

Total Time Sum:

{{rawTimeMs}}

Occurrences:

{{occurrences}}

Performance information for individual fused operations is not available.

Select items within this category for performace details.

Layout:

Hardware Computation Size:

{{computationPrimitiveSize}}

Utilization may have been capped at 100% as accurate bandwidth could not be determined.

Source:

{{sourceTopLine}}

Source Stack:

{{sourceStack}}