SM4 optimization for non-asm mode

[zzl360]

This patch use table-lookup borrow from aes in crypto/aes/aes_core.c.

Test on my PC(AMD Ryzen Threadripper 3990X 64-Core Processor),

before and after optimization:

debug mode:

Before: type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes SM4-CBC 40101.14k 41453.80k 42073.86k 42174.81k 42216.11k 42227.03k SM4-ECB 41222.60k 42074.88k 42673.66k 42868.05k 42896.04k 42844.16k SM4-CTR 35867.22k 36874.47k 37004.97k 37083.82k 37052.42k 37076.99k

After: type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes SM4-CBC 47273.51k 48957.40k 49665.19k 49810.77k 49859.24k 49834.67k SM4-ECB 48100.01k 49323.34k 50224.04k 50273.28k 50533.72k 50730.12k SM4-CTR 41352.64k 42621.29k 42971.22k 43061.59k 43089.92k 43100.84k

non-debug mode:

Before: type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes SM4-CBC 141596.59k 145102.93k 146794.50k 146540.89k 146650.45k 146877.10k SM4-ECB 144774.71k 155106.28k 158166.36k 158279.00k 158520.66k 159280.97k SM4-CTR 138021.10k 141577.60k 142493.53k 142736.38k 142852.10k 143125.16k

After: type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes SM4-CBC 142016.95k 150068.48k 152238.25k 152773.97k 153094.83k 152027.14k SM4-ECB 148842.94k 159919.87k 163628.37k 164515.84k 164697.43k 164790.27k SM4-CTR 141774.23k 146206.89k 147470.25k 147816.28k 146770.60k 148346.20k

[jntass]

2%-3%的性能提升。我看到你给openssl也提了PR，审核通过的话我们后续也会同步

[zzl360]

upstream merged