Openssl riscv optimization：Patch series for SM2, AES and Montgomery multiplication assembly optimization · Pull Request !456 · src-openEuler/openssl

1.SM2 optimization framework
ARM optimization for the SM2 algorithm: foundational framework modifications
riscv perl module

2.AES implementation based on Zvkned
Backport (riscv): Add AES implementation based on Zvkned. Also, further optimizes AES-128-CBC large data block decryption, changing the original single-block decryption loop to multi-block decryption loop processing, and the performance is improved.

Test Verification
```
make test

All tests successful.
Files=251, Tests=2890, 11489 wallclock secs (199.68 usr 48.53 sys + 7152.96 cusr 3700.49 csys = 11101.66 CPU)
Result: PASS
```
Performance Improvement

```
openssl speed -elapsed -seconds 3 -evp aes-128-cbc
```
benchmark results run on qemu_openEuler_24.03 (LTS-SP2):

| Test Item | Baseline  |   Optimized   | Improvement |
| --------- | --------  | ------------- | ----------- |
| 16bytes   | 8870.93k  |  14277.99k    |    161%     |
| 64bytes   | 16342.55k |  31712.83k    |    194%     |
| 256bytes  | 21510.55k |  47318.53k    |    220%     |
| 1024bytes | 21868.54k |  51585.02k    |    236%     |
| 8192bytes | 22986.59k |  54946.28k    |    239%     |
| 16384bytes| 20687.53k |  54744.41k    |    265%     |

```
openssl speed -decrypt -elapsed -seconds 3 -evp aes-128-cbc
```
benchmark results run on qemu_openEuler_24.03 (LTS-SP2):

| Test Item | Baseline  |   Optimized   | Improvement |
| --------- | --------  | ------------- | ----------- |
| 16bytes   | 11584.89k |  11581.33k    |    100%     |
| 64bytes   | 19876.37k |  22127.49k    |    111%     |
| 256bytes  | 22693.42k |  34000.21k    |    150%     |
| 1024bytes | 23544.15k |  38399.66k    |    163%     |
| 8192bytes | 25815.72k |  40498.52k    |    157%     |
| 16384bytes| 27044.52k |  39953.01k    |    148%     |

3.sm2 optimization implementation in riscv asm

Add SM2 implementation in generic riscv64 asm
    
    This PR adds the support of SM2 implementation for RISC-V 64 platform with base integer instruction set.
    
    Testing
    For testing, Built-in test suite was run using user-mode qemu. All tests pass.
    First, configuring the build
    
    $ CC="riscv64-unknown-linux-gnu-gcc" ./Configure linux64-riscv64 no-afalgeng
    $ make -j <num threads>
    Next, I ran the built-in test suite using make test. The summary is below:
    
    All tests successful.
    FFiles=327, Tests=3842, 8418 wallclock secs (99.68 usr 26.07 sys + 3654.55 cusr 3638.32 csys = 7418.62 CPU)
    Result: PASS
    
4.Implement Montgomery multiplication assembly optimization for RV64GC

This PR introduces the RV64GC-optimized assembly implementation for Montgomery multiplication, and significantly accelerating BN modular exponentiation and RSA operations.

Test Verification
```
make test

All tests successful.
Files=251, Tests=2890, 7753 wallclock secs (178.34 usr 51.01 sys + 4480.51 cusr 2970.78 csys = 7680.64 CPU)
Result: PASS
```
Performance Improvement

```
openssl speed rsa2048
```
benchmark results run on qemu_openEuler_24.03 (LTS-SP2):

| Test Item | Baseline | ASM Optimized | Optimized/Baseline |
| --------- | -------- | ------------- | -----------------  |
| sign/s    | 43.6     | 115.3         | 264.45%            |
| verify/s  | 2399.3   | 4426.9        | 184.51%            |

Signed-off-by: liuqingtao <liu.qingtao2@zte.com.cn>

src-openEuler/openssl .gitee-modal { width: 500px !important; }

搜索帮助

src-openEuler/openssl