2024 version is faster

[g4m613r]

     _ec_jacobi_add1: 4,83M it/s ~ 1,24s
     _ec_jacobi_add2: 4,12M it/s ~ 1,46s
     _ec_jacobi_dbl1: 4,19M it/s ~ 1,43s
     _ec_jacobi_dbl2: 6,24M it/s ~ 0,96s
       ec_jacobi_mul: 0,01M it/s ~ 0,81s
       ec_gtable_mul: 0,21M it/s ~ 2,40s
       ec_affine_add: 0,26M it/s ~ 1,95s
       ec_affine_dbl: 0,26M it/s ~ 1,96s
   _fe_modinv_binpow: 0,15M it/s ~ 0,66s
   _fe_modinv_addchn: 0,26M it/s ~ 0,38s
              addr33: 3,98M it/s ~ 1,26s
              addr65: 3,34M it/s ~ 1,50s

[user@ryzen ecloop-main]$ ./ecloopnew bench
     _ec_jacobi_add1: 3,19M it/s ~ 1,88s
     _ec_jacobi_add2: 2,81M it/s ~ 2,14s
     _ec_jacobi_dbl1: 3,15M it/s ~ 1,91s
     _ec_jacobi_dbl2: 4,59M it/s ~ 1,31s
       ec_jacobi_mul: 0,01M it/s ~ 1,12s
       ec_gtable_mul: 0,15M it/s ~ 3,27s
       ec_affine_add: 0,19M it/s ~ 2,59s
       ec_affine_dbl: 0,19M it/s ~ 2,60s
   _fe_modinv_binpow: 0,11M it/s ~ 0,95s
   _fe_modinv_addchn: 100,00M it/s ~ 0,00s
              addr33: 5000,00M it/s ~ 0,00s
              addr65: 5000,00M it/s ~ 0,00s

[vladkens]

Hi, @g4m613r. If you are using Linux try to compile with clang. The previous version of ecloop used clang by default, now it uses cc (the default compiler on the system).

> gcc --version
gcc (Ubuntu 11.4.0-1ubuntu1~22.04) 11.4.0

> gcc -O3 -march=native -pthread -lpthread main.c -o ecloop && ./ecloop bench
     _ec_jacobi_add1: 1.95M it/s ~ 3.08s
     _ec_jacobi_add2: 1.71M it/s ~ 3.51s
     _ec_jacobi_dbl1: 2.00M it/s ~ 3.00s
     _ec_jacobi_dbl2: 2.97M it/s ~ 2.02s
       ec_jacobi_mul: 0.01M it/s ~ 1.81s
       ec_gtable_mul: 0.09M it/s ~ 5.47s
       ec_affine_add: 0.12M it/s ~ 4.10s
       ec_affine_dbl: 0.12M it/s ~ 4.10s
   _fe_modinv_binpow: 0.06M it/s ~ 1.59s
   _fe_modinv_addchn: 0.12M it/s ~ 0.80s
              addr33: 3.84M it/s ~ 1.30s
              addr65: 3.50M it/s ~ 1.43s

> clang --version
Ubuntu clang version 14.0.0-1ubuntu1.1

> clang -O3 -march=native -pthread -lpthread main.c -o ecloop && ./ecloop bench
     _ec_jacobi_add1: 3.25M it/s ~ 1.84s
     _ec_jacobi_add2: 2.82M it/s ~ 2.13s
     _ec_jacobi_dbl1: 3.13M it/s ~ 1.92s
     _ec_jacobi_dbl2: 4.87M it/s ~ 1.23s
       ec_jacobi_mul: 0.01M it/s ~ 1.10s
       ec_gtable_mul: 0.14M it/s ~ 3.68s
       ec_affine_add: 0.19M it/s ~ 2.68s
       ec_affine_dbl: 0.18M it/s ~ 2.72s
   _fe_modinv_binpow: 0.10M it/s ~ 1.02s
   _fe_modinv_addchn: 0.19M it/s ~ 0.53s
              addr33: 3.68M it/s ~ 1.36s
              addr65: 3.25M it/s ~ 1.54s

---

I added to Makefile option to select cc. You can run it like:

make build CC=gcc
make build CC=clang
make build CC=/usr/bin/gcc

Or with any other command:

make bench CC=clang

[g4m613r]

[user@ryzen 11111]$ ./ecloop_2024 bench
     _ec_jacobi_add1: 4,86M it/s ~ 1,23s
     _ec_jacobi_add2: 4,15M it/s ~ 1,45s
     _ec_jacobi_dbl1: 4,24M it/s ~ 1,42s
     _ec_jacobi_dbl2: 6,27M it/s ~ 0,96s
       ec_jacobi_mul: 0,01M it/s ~ 0,80s
       ec_gtable_mul: 0,22M it/s ~ 2,31s
       ec_affine_add: 0,26M it/s ~ 1,94s
       ec_affine_dbl: 0,26M it/s ~ 1,95s
   _fe_modinv_binpow: 0,15M it/s ~ 0,65s
   _fe_modinv_addchn: 0,26M it/s ~ 0,38s
              addr33: 4,00M it/s ~ 1,25s
              addr65: 3,36M it/s ~ 1,49s

[user@ryzen 11111]$ ./ecloop_2025_new bench
     _ec_jacobi_add1: 4,91M it/s ~ 1,22s
     _ec_jacobi_add2: 4,14M it/s ~ 1,45s
     _ec_jacobi_dbl1: 4,20M it/s ~ 1,43s
     _ec_jacobi_dbl2: 6,28M it/s ~ 0,95s
       ec_jacobi_mul: 0,01M it/s ~ 0,80s
       ec_gtable_mul: 0,21M it/s ~ 2,37s
       ec_affine_add: 0,26M it/s ~ 1,95s
       ec_affine_dbl: 0,26M it/s ~ 1,95s
   _fe_modinv_binpow: 0,15M it/s ~ 0,66s
   _fe_modinv_addchn: 0,26M it/s ~ 0,38s
              addr33: 3,98M it/s ~ 1,26s
              addr65: 3,36M it/s ~ 1,49s

😘 ✌️