Add assembly optimized code for OpenSSL on powerpc, powerpc64 and powerpc64le

[pkubaj]

Summary

1. https://github.com/openssl/openssl/commit/34ab13b7d8e3e723adb60be8142e38b7c9cd382a

needs to be merged for ELFv2 support on big-endian.

2. crypto/openssl/crypto/ppccap.c needs to be patched.

Same reason as in https://github.com/openssl/openssl/pull/17082.

After committing to main, please also MFC to stable/13. It should be possible to do it without any changes. MFC'ing to stable/12 won't be possible unless this patch is modified and may prove straight impossible, given that stable/12 uses GCC 4.2.

Test Plan

Test on powerpc, powerpc64 and powerpc64le.

I test using "openssl speed -evp aes-256-gcm"
powerpc64le unpatched:
Doing aes-256-gcm for 3s on 16 size blocks: 9552470 aes-256-gcm's in 3.01s
Doing aes-256-gcm for 3s on 64 size blocks: 3083878 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 256 size blocks: 830906 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 1024 size blocks: 211122 aes-256-gcm's in 3.00s
Doing aes-256-gcm for 3s on 8192 size blocks: 26861 aes-256-gcm's in 3.05s
Doing aes-256-gcm for 3s on 16384 size blocks: 13415 aes-256-gcm's in 3.05s
OpenSSL 1.1.1k-freebsd 24 Aug 2021
built on: reproducible build, date unspecified
options:bn(64,64) rc4(int) des(int) aes(partial) idea(int) blowfish(ptr)
compiler: clang
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes
aes-256-gcm 50814.18k 65448.52k 70536.60k 72062.98k 72220.00k 72136.65k

powerpc64le patched:
Doing aes-256-gcm for 3s on 16 size blocks: 19873877 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 64 size blocks: 12461931 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 256 size blocks: 9315680 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 1024 size blocks: 3705172 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 8192 size blocks: 545984 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 16384 size blocks: 275305 aes-256-gcm's in 3.02s
OpenSSL 1.1.1k-freebsd 24 Aug 2021
built on: reproducible build, date unspecified
options:bn(64,64) rc4(int) des(int) aes(partial) idea(int) blowfish(ptr)
compiler: clang
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes
aes-256-gcm 105172.35k 264477.04k 788775.72k 1254894.84k 1483175.44k 1491877.08k

powerpc64 unpatched:
Doing aes-256-gcm for 3s on 16 size blocks: 7070177 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 64 size blocks: 2484593 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 256 size blocks: 673707 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 1024 size blocks: 171983 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 8192 size blocks: 21644 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 16384 size blocks: 10819 aes-256-gcm's in 3.02s
OpenSSL 1.1.1k-freebsd 24 Aug 2021
built on: reproducible build, date unspecified
options:bn(64,64) rc4(int) des(int) aes(partial) idea(int) blowfish(ptr)
compiler: clang
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes
aes-256-gcm 37415.30k 52730.02k 57044.01k 58248.46k 58644.39k 58780.02k

powerpc64 patched:
Doing aes-256-gcm for 3s on 16 size blocks: 15075576 aes-256-gcm's in 3.04s
Doing aes-256-gcm for 3s on 64 size blocks: 10145628 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 256 size blocks: 7837733 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 1024 size blocks: 3106231 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 8192 size blocks: 461837 aes-256-gcm's in 3.05s
Doing aes-256-gcm for 3s on 16384 size blocks: 231623 aes-256-gcm's in 3.05s
OpenSSL 1.1.1k-freebsd 24 Aug 2021
built on: reproducible build, date unspecified
options:bn(64,64) rc4(int) des(int) aes(partial) idea(int) blowfish(ptr)
compiler: clang
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes
aes-256-gcm 79369.61k 215318.61k 663635.23k 1054766.61k 1241721.01k 1245509.33k

powerpc unpatched:
Doing aes-256-gcm for 3s on 16 size blocks: 5233246 aes-256-gcm's in 3.01s
Doing aes-256-gcm for 3s on 64 size blocks: 1972493 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 256 size blocks: 543703 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 1024 size blocks: 139295 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 8192 size blocks: 17494 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 16384 size blocks: 8752 aes-256-gcm's in 3.02s
OpenSSL 1.1.1k-freebsd 25 Mar 2021
built on: reproducible build, date unspecified
options:bn(64,32) rc4(int) des(long) aes(partial) idea(int) blowfish(ptr)
compiler: clang
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes
aes-256-gcm 27838.15k 41861.82k 46036.33k 47299.67k 47399.97k 47549.93k

powerpc patched:
Doing aes-256-gcm for 3s on 16 size blocks: 11680952 aes-256-gcm's in 3.01s
Doing aes-256-gcm for 3s on 64 size blocks: 8420502 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 256 size blocks: 6846209 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 1024 size blocks: 2946064 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 8192 size blocks: 454328 aes-256-gcm's in 3.02s
Doing aes-256-gcm for 3s on 16384 size blocks: 229675 aes-256-gcm's in 3.02s
OpenSSL 1.1.1k-freebsd 24 Aug 2021
built on: reproducible build, date unspecified
options:bn(64,32) rc4(int) des(long) aes(partial) idea(int) blowfish(ptr)
compiler: clang
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes
aes-256-gcm 62136.60k 178706.61k 581182.84k 997794.58k 1234190.25k 1244608.23k

[jhibbits]

Thanks for adding this!

[jkim]

I am okay with it as powerpc maintainers approved.