stupidgcm: cache C.EVP_chacha20_poly1305()

2% performance improvement, almost for free.

gocryptfs/internal/speed$ benchstat old new
name             old time/op   new time/op   delta
StupidXchacha-4   5.82µs ± 0%   5.68µs ± 0%  -2.37%  (p=0.008 n=5+5)

name             old speed     new speed     delta
StupidXchacha-4  704MB/s ± 0%  721MB/s ± 0%  +2.43%  (p=0.008 n=5+5)

1 files changed, 15 insertions, 1 deletions

diff --git a/internal/stupidgcm/chacha.go b/internal/stupidgcm/chacha.go
index 2e6e6e6..c90d721 100644
--- a/internal/stupidgcm/chacha.go
+++ b/internal/stupidgcm/chacha.go
@@ -21,6 +21,20 @@ type stupidChacha20poly1305 struct {
 // Verify that we satisfy the cipher.AEAD interface
 var _ cipher.AEAD = &stupidChacha20poly1305{}
 
+// _EVP_chacha20_poly1305 caches C.EVP_chacha20_poly1305() to avoid the Cgo call
+// overhead for each instantiation of NewChacha20poly1305.
+var _EVP_chacha20_poly1305 *C.EVP_CIPHER
+
+func init() {
+	_EVP_chacha20_poly1305 = C.EVP_chacha20_poly1305()
+}
+
+// NewChacha20poly1305 returns a new instance of the OpenSSL ChaCha20-Poly1305 AEAD
+// cipher ( https://www.openssl.org/docs/man1.1.1/man3/EVP_chacha20_poly1305.html ).
+//
+// gocryptfs only uses ChaCha20-Poly1305 as a building block for OpenSSL
+// XChaCha20-Poly1305. This function is hot because it gets called once for each
+// block by XChaCha20-Poly1305.
 func NewChacha20poly1305(key []byte) *stupidChacha20poly1305 {
 	if len(key) != chacha20poly1305.KeySize {
 		log.Panicf("Only %d-byte keys are supported, you passed %d bytes", chacha20poly1305.KeySize, len(key))
@@ -28,7 +42,7 @@ func NewChacha20poly1305(key []byte) *stupidChacha20poly1305 {
 	return &stupidChacha20poly1305{
 		stupidAEADCommon{
 			key:              append([]byte{}, key...), // private copy
-			openSSLEVPCipher: C.EVP_chacha20_poly1305(),
+			openSSLEVPCipher: _EVP_chacha20_poly1305,
 			nonceSize:        chacha20poly1305.NonceSize,
 		},
 	}