Some test codes for vector capabilities targeted in EPI (SVE, RISC-V V)
Cryptographic codes here are 'standalone' versions of implementations designed for the Supercop benchmark https://bench.cr.yp.to/supercop.html, and include some of the Supercop source code. They also include some objects file generated from Supercop (cpucycles.o, kernelrandombytes.o) to simplify usage.
All codes currently use ACLE (Arm C Language Extension) for SVE, and EPI Builtins for V (https://repo.hca.bsc.es/gitlab/rferrer/epi-builtins-ref/).
-
aes256ctrstandalone-rvv: AES-256 in counter mode, for RISC-V with V extension. This doesn't use any AES-specific instructions, but vectorize a scalar AES-256 (specifically, the implementation from
supercop/crypto_core/aes256encrypt/dolbeau/std-4ft). -
aes256ctrstandalone-sve: same as above, but for Arm SVE.
-
aes256encryptstandalone-rvv: AES-256 encryption vectorized per-block for RISC-V with V extension, same idea as aes256ctrstandalone-rvv above minus the counter mode.
-
chacha20standalone-rvv: Chacha20 vectorized for RISC-V with V extension, basically the RISC-V version of the SVE version below.
-
chacha20standalone-sve: Chacha20 vectorized for Arm SVE. This is a standalone version of
supercop/crypto_stream/chacha20/dolbeau/arm-sve.
The first part of the output (an hexadecimal string) should match the ckechsum from the benchmark, to for instance the output of chacha20_small should match the content of supercop/crypto_stream/chacha20/checksumsmall.
-
simplefft16-rvv: 16-points in-register double-precision DFT/FFT (so requires a 1024+ bits vector that can fit 16 double-precision values). Multiple implementations of the algorithm, both as DFT and FFT. This includes some code generated by the FFTW3 codelet generator.
-
SVE/RVV-supporting code for FFTW3 can be found in https://github.com/rdolbeau/fftw3, see the wiki there for the relevant branches.
This work has partly been done as part of the European Processor Initiative project.
The European Processor Initiative (EPI) (FPA: 800928) has received funding from the European Union's Horizon 2020 research and innovation programme under grant agreement EPI-SGA1: 826647