[Article suggestion] Performance benchmark of LUKS2 with integrity on different CPUs

[hardfalcon]

Hi,

I'd be interested in a performance benchmark of LUKS2 with enabled integrity on different CPUs, more specifically in cryptsetup configurations like

cryptsetup --type luks2 --cipher aegis128-random --integrity aead --key-size 128 --sector-size 4096 luksFormat /dev/nvme0n1p2

vs.

cryptsetup --type luks2 --cipher aes-xts-plain64 --integrity hmac-sha1 --key-size 128 --sector-size 4096 luksFormat /dev/nvme0n1p2

vs.

cryptsetup --type luks2 --cipher aes-xts-plain64 --integrity hmac-sha256 --key-size 128 --sector-size 4096 luksFormat /dev/nvme0n1p2

vs.

cryptsetup --type luks2 --cipher xchacha12,aes-adiantum-plain64 --integrity hmac-sha1 --key-size 128 --sector-size 4096 luksFormat /dev/nvme0n1p2

vs.

cryptsetup --type luks2 --cipher xchacha12,aes-adiantum-plain64 --integrity hmac-sha256 --key-size 128 --sector-size 4096 luksFormat /dev/nvme0n1p2

vs.

cryptsetup --type luks2 --cipher xchacha20,aes-adiantum-plain64 --integrity hmac-sha1 --key-size 128 --sector-size 4096 luksFormat /dev/nvme0n1p2

vs.

cryptsetup --type luks2 --cipher xchacha20,aes-adiantum-plain64 --integrity hmac-sha256 --key-size 128 --sector-size 4096 luksFormat /dev/nvme0n1p2

I'd like to know how big the performance impact of SHA extensions in recent Intel and AMD CPUs is, and if using those extensions does degrade the performance of AES-NI.