lvshuo
d414b2702a
erasure_code: optimize RVV implementation
...
The ISA-L EC code has been written using RVV vector instructions and the minimum multiplication table,
resulting in a performance improvement of over 10 times compared to the existing implementation.
Signed-off-by: Shuo Lv <lv.shuo@sanechips.com.cn>
2025-07-11 15:55:57 +02:00
sunyuechi
eb130eaf6b
erasure_code: R-V V ec_encode_data
...
banana_f3:
rvv:
erasure_code_encode_warm: runtime = 3065696 usecs, bandwidth 108 MB in 3.0657 sec = 35.37 MB/s
erasure_code_decode_warm: runtime = 3001213 usecs, bandwidth 136 MB in 3.0012 sec = 45.47 MB/s
c:
erasure_code_encode_warm: runtime = 3002512 usecs, bandwidth 52 MB in 3.0025 sec = 17.34 MB/s
erasure_code_decode_warm: runtime = 3065235 usecs, bandwidth 57 MB in 3.0652 sec = 18.69 MB/s
Signed-off-by: sunyuechi <sunyuechi@iscas.ac.cn>
2025-05-01 17:44:19 +01:00
sunyuechi
c5d75f1e27
erasure_code: R-V V gf_vect_dot_prod
...
banana_f3:
rvv: gf_vect_dot_prod_warm: runtime = 3062964 usecs, bandwidth 490 MB in 3.0630 sec = 160.25 MB/s
c: gf_vect_dot_prod_warm: runtime = 3000581 usecs, bandwidth 173 MB in 3.0006 sec = 57.69 MB/s
Signed-off-by: sunyuechi <sunyuechi@iscas.ac.cn>
2025-05-01 17:44:19 +01:00
sunyuechi
0a68e9434a
erasure_code: R-V V gf_vect_mul
...
banana_f3:
rvv: gf_vect_mul_warm: runtime = 3062541 usecs, bandwidth 1889 MB in 3.0625 sec = 616.84 MB/s
c: gf_vect_mul_warm: runtime = 3062014 usecs, bandwidth 285 MB in 3.0620 sec = 93.29 MB/s
Signed-off-by: sunyuechi <sunyuechi@iscas.ac.cn>
2025-05-01 17:44:19 +01:00