base OpenSSL is much, much slower than the package

[nia <nia%NetBSD.org@localhost>]

It almost seems like the asm implementations aren't being enabled
properly. Algorithms with no x86_64 asm (like blowfish) don't
shown a significant difference, whereas with AES and sha512 
the difference is huge.

What's going on?

r[nia]$ /usr/pkg/bin/openssl speed -evp aes-256-gcm
Doing aes-256-gcm for 3s on 16 size blocks: 29619193 aes-256-gcm's in 2.85s
Doing aes-256-gcm for 3s on 64 size blocks: 15947903 aes-256-gcm's in 2.74s
Doing aes-256-gcm for 3s on 256 size blocks: 9981367 aes-256-gcm's in 2.82s
Doing aes-256-gcm for 3s on 1024 size blocks: 3761836 aes-256-gcm's in 2.91s
Doing aes-256-gcm for 3s on 8192 size blocks: 526211 aes-256-gcm's in 2.64s
Doing aes-256-gcm for 3s on 16384 size blocks: 284726 aes-256-gcm's in 2.67s
OpenSSL 1.1.1g  21 Apr 2020
built on: Tue May  5 14:23:01 2020 UTC
options:bn(64,64) md2(char) rc4(16x,int) des(int) aes(partial) idea(int) blowfish(ptr)
compiler: cc -fPIC -pthread -Wa,--noexecstack -O2 -pipe -D_FORTIFY_SOURCE=2 -I/usr/include -O2 -pipe -DL_ENDIAN -DOPENSSL_PIC -DOPENSSL_CPUID_OBJ -DOPENSSL_IA32_SSE2 -DOPENSSL_BN_ASM_MONT -DOPENSSL_BN_ASM_MONT5 -DOPENSSL_BN_ASM_GF2m -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DKECCAK1600_ASM -DRC4_ASM -DMD5_ASM -DAESNI_ASM -DVPAES_ASM -DGHASH_ASM -DECP_NISTZ256_ASM -DX25519_ASM -DPOLY1305_ASM -D_THREAD_SAFE -D_REENTRANT -DNDEBUG -D_FORTIFY_SOURCE=2 -I/usr/include -I/usr/include
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     166283.19k   372505.76k   906109.91k  1323752.60k  1632848.68k  1747172.58k
r[nia]$ /usr/bin/openssl speed -evp aes-256-gcm
Doing aes-256-gcm for 3s on 16 size blocks: 6073816 aes-256-gcm's in 2.68s
Doing aes-256-gcm for 3s on 64 size blocks: 1624277 aes-256-gcm's in 2.54s
Doing aes-256-gcm for 3s on 256 size blocks: 412605 aes-256-gcm's in 2.61s
Doing aes-256-gcm for 3s on 1024 size blocks: 108895 aes-256-gcm's in 2.56s
Doing aes-256-gcm for 3s on 8192 size blocks: 13510 aes-256-gcm's in 2.58s
Doing aes-256-gcm for 3s on 16384 size blocks: 7508 aes-256-gcm's in 2.71s
OpenSSL 1.1.1g  21 Apr 2020
NetBSD 9.0_STABLE
options:bn(64,64) rc4(16x,int) des(int) aes(partial) idea(int) blowfish(ptr)
gcc version 7.4.0 (NetBSD nb3 20190319)
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      36261.59k    40926.66k    40470.07k    43558.00k    42896.87k    45391.54k