diff --git a/src/se050_blake2s.c b/src/se050_blake2s.c
index 0803384..12ad37e 100644
--- a/src/se050_blake2s.c
+++ b/src/se050_blake2s.c
@@ -140,13 +140,9 @@ int se050_blake2s_update(se050_blake2s_ctx_t *ctx, const void *data, size_t len)
     if (!ctx || !data) return -1;
     if (len > 0) {
         size_t left = inner->buflen, fill = 64 - left;
-        if (len > fill) {
-            memcpy(inner->buf + left, in, fill);
-            inner->buflen = 0;
-            inner->t[0] += 64;
-            if (inner->t[0] < 64) inner->t[1]++;
-            blake2s_compress(inner, inner->buf);
-            in += fill; len -= fill;
+        
+        /* If buffer is empty, process full blocks directly */
+        if (left == 0) {
             while (len > 64) {
                 inner->t[0] += 64;
                 if (inner->t[0] < 64) inner->t[1]++;
@@ -154,6 +150,25 @@ int se050_blake2s_update(se050_blake2s_ctx_t *ctx, const void *data, size_t len)
                 in += 64; len -= 64;
             }
         }
+        /* If we can fill the buffer (including exact fill), do it */
+        else if (len >= fill) {
+            memcpy(inner->buf + left, in, fill);
+            inner->buflen = 0;
+            inner->t[0] += 64;
+            if (inner->t[0] < 64) inner->t[1]++;
+            blake2s_compress(inner, inner->buf);
+            in += fill; len -= fill;
+            
+            /* Process remaining full blocks */
+            while (len > 64) {
+                inner->t[0] += 64;
+                if (inner->t[0] < 64) inner->t[1]++;
+                blake2s_compress(inner, in);
+                in += 64; len -= 64;
+            }
+        }
+        
+        /* Store remaining data in buffer */
         memcpy(inner->buf + inner->buflen, in, len);
         inner->buflen += len;
     }