diff --git a/lib/arch/ssse3/enc_loop.c b/lib/arch/ssse3/enc_loop.c
index 87a50784..6de652e1 100644
--- a/lib/arch/ssse3/enc_loop.c
+++ b/lib/arch/ssse3/enc_loop.c
@@ -1,3 +1,22 @@
+static inline void
+enc_loop_ssse3_inner (const uint8_t **s, uint8_t **o)
+{
+	// Load input:
+	__m128i str = _mm_loadu_si128((__m128i *) *s);
+
+	// Reshuffle:
+	str = enc_reshuffle(str);
+
+	// Translate reshuffled bytes to the Base64 alphabet:
+	str = enc_translate(str);
+
+	// Store:
+	_mm_storeu_si128((__m128i *) *o, str);
+
+	*s += 12;
+	*o += 16;
+}
+
 static inline void
 enc_loop_ssse3 (const uint8_t **s, size_t *slen, uint8_t **o, size_t *olen)
 {
@@ -15,20 +34,34 @@ enc_loop_ssse3 (const uint8_t **s, size_t *slen, uint8_t **o, size_t *olen)
 	*olen += rounds * 16;	// 16 bytes produced per round
 
 	do {
-		// Load string:
-		__m128i str = _mm_loadu_si128((__m128i *) *s);
-
-		// Reshuffle:
-		str = enc_reshuffle(str);
-
-		// Translate reshuffled bytes to the Base64 alphabet:
-		str = enc_translate(str);
-
-		// Store:
-		_mm_storeu_si128((__m128i *) *o, str);
-
-		*s += 12;
-		*o += 16;
+		if (rounds >= 8) {
+			enc_loop_ssse3_inner(s, o);
+			enc_loop_ssse3_inner(s, o);
+			enc_loop_ssse3_inner(s, o);
+			enc_loop_ssse3_inner(s, o);
+			enc_loop_ssse3_inner(s, o);
+			enc_loop_ssse3_inner(s, o);
+			enc_loop_ssse3_inner(s, o);
+			enc_loop_ssse3_inner(s, o);
+			rounds -= 8;
+			continue;
+		}
+		if (rounds >= 4) {
+			enc_loop_ssse3_inner(s, o);
+			enc_loop_ssse3_inner(s, o);
+			enc_loop_ssse3_inner(s, o);
+			enc_loop_ssse3_inner(s, o);
+			rounds -= 4;
+			continue;
+		}
+		if (rounds >= 2) {
+			enc_loop_ssse3_inner(s, o);
+			enc_loop_ssse3_inner(s, o);
+			rounds -= 2;
+			continue;
+		}
+		enc_loop_ssse3_inner(s, o);
+		break;
 
-	} while (--rounds > 0);
+	} while (rounds > 0);
 }