FPGA AI Suite: Compiler Reference Manual

ID 768972
Date 9/06/2024
Public

A newer version of this document is available. Customers should click here to go to the newest version.

Document Table of Contents

4.7. Analyzer Tool Options (dla_compiler Command Options)

--fanalyze-performance Option

Syntax
--fanalyze-performance
Description
When this option is set, the compiler estimates the performance of the given graph on the specified architecture. The performance report provides the estimated kernel throughput for each network graph provided.

The compiler also estimates the average memory bandwidth and the memory requirements. The estimated memory requirement is typically an underestimate because it excludes the effects of memory fragmentation.

An example performance report is as follows:
Graph reported throughput: 12.345678
Parsed As
(bool)

--fdump-performance-report Option

Syntax
--fdump-performance-report=<dump_performance>
Description
Optional output file for the performance estimate.

This option also enables the --fanalyze-performance option.

Parsed As
(string)

--fanalyze-area Option

Syntax
--fanalyze-area
Description
Set this option to have the compiler estimate the area of the specified architecture.
Parsed As
(bool)

--fdump-area-report Option

Syntax
--fdump-area-report=<dump_area>
Description
Optional output file for the area estimate.

This option also enables the --fanalyze-area option.

Parsed As
(string)

--fdisplay-device Option

Syntax
--fdisplay-device
Description
The option enables displaying the device (for example, CPU or FPGA) for each subgraph for all networks.
Parsed As
(bool)