Intel Dramatically Speeds Up NSS With AVX2
Intel has managed to dramatically speed-up Network Security Services (NSS) for the new Haswell (and forthcoming Broadwell) processors that boast AVX2 instruction set support.
Yesterday a patch surfaced by Intel's Shay Gueron on the Mozilla bug tracker for dramatically boosting the performance of NSS. "It provides an efficient and constant-time implementation of modular exponent function, using the AVX2 instructions set, and achieves high performance on Intel 4th Generation Core Processors...Applying this 'vectorized' algorithm to modular exponentiation improves the performance of: DH1024, DH2048, RSA2048 sign and verify, RSA1024 verify, JPAKE and DSA1024 with DSA2048 sign and verify."
The NSS libraries performance is improved by over 2x for RSA2048 and DSA2048 signing while the other signing and verifying algorithms also see a great speed-up by taking advantage of Advanced Vector Extensions 2.
For the Linux scoop on Intel's Haswell processors, see Assessing The Current Intel Haswell Linux Experience.
Yesterday a patch surfaced by Intel's Shay Gueron on the Mozilla bug tracker for dramatically boosting the performance of NSS. "It provides an efficient and constant-time implementation of modular exponent function, using the AVX2 instructions set, and achieves high performance on Intel 4th Generation Core Processors...Applying this 'vectorized' algorithm to modular exponentiation improves the performance of: DH1024, DH2048, RSA2048 sign and verify, RSA1024 verify, JPAKE and DSA1024 with DSA2048 sign and verify."
The NSS libraries performance is improved by over 2x for RSA2048 and DSA2048 signing while the other signing and verifying algorithms also see a great speed-up by taking advantage of Advanced Vector Extensions 2.
For the Linux scoop on Intel's Haswell processors, see Assessing The Current Intel Haswell Linux Experience.
16 Comments