Add evp_mac perftool

[andrewkdinh]

Summary

Adds a CLI tool evp_mac that computes an HMAC of random data using SHA-256 digest.
Runs for 5 seconds and prints the average execution time per encryption.

Fixes openssl/project#1837

Features

• Four modes of operation:
  • evp_shared (default): Use EVP API and allow shared data between computations
  • evp_isolated: Use EVP API and don't allow shared data between computations
  • deprecated_shared: Use legacy API and allow shared data between computations
  • deprecated_isolated: Use legacy API and don't allow shared data between computations
• Configurable number of times to update
• Terse output for easier CI automation (-t)
• Configurable thread count

TODO: Since freeze functionality hasn't been properly added yet (see openssl/project#1832 & openssl/openssl#29433), support for freezing the context store has not been added yet.

Usage

$ ./evp_mac -h
Usage: evp_mac [-h] [-t] [-o operation] [-u update-times] [-V] thread-count
-h - print this help output
-t - terse output
-o operation - mode of operation. One of [evp_isolated, evp_shared, deprecated_isolated, deprecated_shared] (default: evp_shared)
-u update-times - times to update (default: 1)
-V - print version information and exit
thread-count - number of threads

$ ./evp_mac -o evp_shared 10 # evp_shared operation mode, 10 threads
Average time per computation: 0.207903us

$ ./evp_mac -o evp_isolated 10 # now using evp_isolated mode
Average time per computation: 7.918923us

Findings

Initial results showed that evp_isolated mode generally runs about 30x slower than evp_shared mode.

[npajkovsky]

nit: It's better to have size_t count locally, and update it after the test.

[andrewkdinh]

Why is it better to do it like that?

[npajkovsky]

The local counter has less overhead than accessing a global array each iteration.