diff --git a/wolfcrypt/src/misc.c b/wolfcrypt/src/misc.c
index e69ad655b..7a9bcb02c 100644
--- a/wolfcrypt/src/misc.c
+++ b/wolfcrypt/src/misc.c
@@ -709,13 +709,23 @@ WC_MISC_STATIC WC_INLINE void w64SetLow32(w64wrapper *n, word32 low) {
 
 WC_MISC_STATIC WC_INLINE w64wrapper w64Add32(w64wrapper a, word32 b, byte *wrap)
 {
-    a.n = a.n + b;
+    a.n += b;
     if (a.n < b && wrap != NULL)
         *wrap = 1;
 
     return a;
 }
 
+WC_MISC_STATIC WC_INLINE w64wrapper w64Add(w64wrapper a, w64wrapper b,
+    byte *wrap)
+{
+    a.n += b.n;
+    if (a.n < b.n && wrap != NULL)
+        *wrap = 1;
+
+    return a;
+}
+
 WC_MISC_STATIC WC_INLINE w64wrapper w64Sub32(w64wrapper a, word32 b, byte *wrap)
 {
     if (a.n < b && wrap != NULL)
@@ -796,6 +806,13 @@ WC_MISC_STATIC WC_INLINE w64wrapper w64ShiftLeft(w64wrapper a, int shift)
     return a;
 }
 
+WC_MISC_STATIC WC_INLINE w64wrapper w64Mul(word32 a, word32 b)
+{
+    w64wrapper ret;
+    ret.n = (word64)a * (word64)b;
+    return ret;
+}
+
 #else
 
 WC_MISC_STATIC WC_INLINE void w64Increment(w64wrapper *n)
@@ -831,7 +848,7 @@ WC_MISC_STATIC WC_INLINE void w64SetLow32(w64wrapper *n, word32 low)
 
 WC_MISC_STATIC WC_INLINE w64wrapper w64Add32(w64wrapper a, word32 b, byte *wrap)
 {
-    a.n[1] = a.n[1] + b;
+    a.n[1] += b;
     if (a.n[1] < b) {
         a.n[0]++;
         if (wrap != NULL && a.n[0] == 0)
@@ -841,6 +858,24 @@ WC_MISC_STATIC WC_INLINE w64wrapper w64Add32(w64wrapper a, word32 b, byte *wrap)
     return a;
 }
 
+WC_MISC_STATIC WC_INLINE w64wrapper w64Add(w64wrapper a, w64wrapper b,
+    byte *wrap)
+{
+    a.n[1] += b.n[1];
+    if (a.n[1] < b.n[1]) {
+        a.n[0]++;
+        if (wrap != NULL && a.n[0] == 0)
+                *wrap = 1;
+    }
+
+    a.n[0] += b.n[0];
+    if (wrap != NULL && a.n[0] < b.n[0]) {
+        *wrap = 1;
+    }
+
+    return a;
+}
+
 WC_MISC_STATIC WC_INLINE w64wrapper w64Sub32(w64wrapper a, word32 b, byte *wrap)
 {
     byte _underflow = 0;
@@ -894,7 +929,7 @@ WC_MISC_STATIC WC_INLINE byte w64IsZero(w64wrapper a)
     return a.n[0] == 0 && a.n[1] == 0;
 }
 
-WC_MISC_STATIC WC_INLINE void c64toa(w64wrapper *a, byte *out)
+WC_MISC_STATIC WC_INLINE void c64toa(const w64wrapper *a, byte *out)
 {
 #ifdef BIG_ENDIAN_ORDER
     word32 *_out = (word32*)(out);
@@ -939,7 +974,7 @@ WC_MISC_STATIC WC_INLINE byte w64LT(w64wrapper a, w64wrapper b)
 WC_MISC_STATIC WC_INLINE w64wrapper w64ShiftRight(w64wrapper a, int shift)
 {
      if (shift < 32) {
-         a.n[1] = (a.n[1] >> shift) || (a.n[0] << (32 - shift));
+         a.n[1] = (a.n[1] >> shift) | (a.n[0] << (32 - shift));
          a.n[0] >>= shift;
      }
      else {
@@ -951,7 +986,7 @@ WC_MISC_STATIC WC_INLINE w64wrapper w64ShiftRight(w64wrapper a, int shift)
 WC_MISC_STATIC WC_INLINE w64wrapper w64ShiftLeft(w64wrapper a, int shift)
 {
      if (shift < 32) {
-         a.n[0] = (a.n[0] << shift) || (a.n[1] >> (32 - shift));
+         a.n[0] = (a.n[0] << shift) | (a.n[1] >> (32 - shift));
          a.n[1] <<= shift;
      }
      else {
@@ -961,6 +996,30 @@ WC_MISC_STATIC WC_INLINE w64wrapper w64ShiftLeft(w64wrapper a, int shift)
      return a;
 }
 
+WC_MISC_STATIC WC_INLINE w64wrapper w64Mul(word32 a, word32 b)
+{
+    w64wrapper ret;
+    word16 ltlA, ltlB, ltlC, ltlD;
+    word32 bigA, bigB, bigC, bigD;
+
+    ltlA = a & 0xFFFF;
+    ltlB = (a >> 16) & 0xFFFF;
+    ltlC = b & 0xFFFF;
+    ltlD = (b >> 16) & 0xFFFF;
+
+    bigA = (word32)ltlA * (word32)ltlC;
+    bigC = (word32)ltlB * (word32)ltlC;
+    bigD = (word32)ltlA * (word32)ltlD;
+    bigB = (word32)ltlB * (word32)ltlD;
+
+    ret = w64From32(0, bigB);
+    ret = w64ShiftLeft(ret, 16);
+    ret = w64Add32(ret, bigD, NULL);
+    ret = w64Add32(ret, bigC, NULL);
+    ret = w64ShiftLeft(ret, 16);
+    return w64Add32(ret, bigA, NULL);
+}
+
 #endif /* WORD64_AVAILABLE && !WOLFSSL_W64_WRAPPER_TEST */
 #endif /* WOLFSSL_W64_WRAPPER */
 
diff --git a/wolfcrypt/src/poly1305.c b/wolfcrypt/src/poly1305.c
index 730a5c2af..43528c23f 100644
--- a/wolfcrypt/src/poly1305.c
+++ b/wolfcrypt/src/poly1305.c
@@ -29,6 +29,13 @@ and Daniel J. Bernstein
 */
 
 
+/*
+ * WOLFSSL_W64_WRAPPER Uses wrappers around word64 types for a system that does
+ *                     not have word64 available. As expected it reduces
+ *                     performance. Benchmarks collected July 2024 show
+ *                     303.004 MiB/s with and 1874.194 MiB/s without.
+ */
+
 #ifdef HAVE_CONFIG_H
     #include <config.h>
 #endif
@@ -332,8 +339,22 @@ static int poly1305_blocks(Poly1305* ctx, const unsigned char *m,
     word32 r0,r1,r2,r3,r4;
     word32 s1,s2,s3,s4;
     word32 h0,h1,h2,h3,h4;
-    word64 d0,d1,d2,d3,d4;
     word32 c;
+#ifdef WOLFSSL_W64_WRAPPER
+    #ifdef WOLFSSL_SMALL_STACK
+    w64wrapper* d;
+
+    d = (w64wrapper*)XMALLOC(5 * sizeof(w64wrapper), NULL,
+        DYNAMIC_TYPE_TMP_BUFFER);
+    if (d == NULL) {
+        return MEMORY_E;
+    }
+    #else
+    w64wrapper d[5];
+    #endif
+#else
+    word64 d0,d1,d2,d3,d4;
+#endif
 
 
     r0 = ctx->r[0];
@@ -362,6 +383,41 @@ static int poly1305_blocks(Poly1305* ctx, const unsigned char *m,
         h4 += (U8TO32(m+12) >> 8) | hibit;
 
         /* h *= r */
+#ifdef WOLFSSL_W64_WRAPPER
+        {
+            w64wrapper tmp;
+
+            d[0] = w64Mul(h0, r0); tmp = w64Mul(h1, s4);
+            d[0] = w64Add(d[0], tmp, NULL); tmp = w64Mul(h2, s3);
+            d[0] = w64Add(d[0], tmp, NULL); tmp = w64Mul(h3, s2);
+            d[0] = w64Add(d[0], tmp, NULL); tmp = w64Mul(h4, s1);
+            d[0] = w64Add(d[0], tmp, NULL);
+
+            d[1] = w64Mul(h0, r1); tmp = w64Mul(h1, r0);
+            d[1] = w64Add(d[1], tmp, NULL); tmp = w64Mul(h2, s4);
+            d[1] = w64Add(d[1], tmp, NULL); tmp = w64Mul(h3, s3);
+            d[1] = w64Add(d[1], tmp, NULL); tmp = w64Mul(h4, s2);
+            d[1] = w64Add(d[1], tmp, NULL);
+
+            d[2] = w64Mul(h0, r2); tmp = w64Mul(h1, r1);
+            d[2] = w64Add(d[2], tmp, NULL); tmp = w64Mul(h2, r0);
+            d[2] = w64Add(d[2], tmp, NULL); tmp = w64Mul(h3, s4);
+            d[2] = w64Add(d[2], tmp, NULL); tmp = w64Mul(h4, s3);
+            d[2] = w64Add(d[2], tmp, NULL);
+
+            d[3] = w64Mul(h0, r3); tmp = w64Mul(h1, r2);
+            d[3] = w64Add(d[3], tmp, NULL); tmp = w64Mul(h2, r1);
+            d[3] = w64Add(d[3], tmp, NULL); tmp = w64Mul(h3, r0);
+            d[3] = w64Add(d[3], tmp, NULL); tmp = w64Mul(h4, s4);
+            d[3] = w64Add(d[3], tmp, NULL);
+
+            d[4] = w64Mul(h0, r4); tmp = w64Mul(h1, r3);
+            d[4] = w64Add(d[4], tmp, NULL); tmp = w64Mul(h2, r2);
+            d[4] = w64Add(d[4], tmp, NULL); tmp = w64Mul(h3, r1);
+            d[4] = w64Add(d[4], tmp, NULL); tmp = w64Mul(h4, r0);
+            d[4] = w64Add(d[4], tmp, NULL);
+        }
+#else
         d0 = ((word64)h0 * r0) + ((word64)h1 * s4) + ((word64)h2 * s3) +
              ((word64)h3 * s2) + ((word64)h4 * s1);
         d1 = ((word64)h0 * r1) + ((word64)h1 * r0) + ((word64)h2 * s4) +
@@ -372,13 +428,31 @@ static int poly1305_blocks(Poly1305* ctx, const unsigned char *m,
              ((word64)h3 * r0) + ((word64)h4 * s4);
         d4 = ((word64)h0 * r4) + ((word64)h1 * r3) + ((word64)h2 * r2) +
              ((word64)h3 * r1) + ((word64)h4 * r0);
+#endif
 
         /* (partial) h %= p */
+#ifdef WOLFSSL_W64_WRAPPER
+        c = w64GetLow32(w64ShiftRight(d[0], 26));
+        h0 = w64GetLow32(d[0]) & 0x3ffffff;
+        d[1] = w64Add32(d[1], c, NULL);
+        c = w64GetLow32(w64ShiftRight(d[1], 26));
+        h1 = w64GetLow32(d[1]) & 0x3ffffff;
+        d[2] = w64Add32(d[2], c, NULL);
+        c = w64GetLow32(w64ShiftRight(d[2], 26));
+        h2 = w64GetLow32(d[2]) & 0x3ffffff;
+        d[3] = w64Add32(d[3], c, NULL);
+        c = w64GetLow32(w64ShiftRight(d[3], 26));
+        h3 = w64GetLow32(d[3]) & 0x3ffffff;
+        d[4] = w64Add32(d[4], c, NULL);
+        c = w64GetLow32(w64ShiftRight(d[4], 26));
+        h4 = w64GetLow32(d[4]) & 0x3ffffff;
+#else
                       c = (word32)(d0 >> 26); h0 = (word32)d0 & 0x3ffffff;
         d1 += c;      c = (word32)(d1 >> 26); h1 = (word32)d1 & 0x3ffffff;
         d2 += c;      c = (word32)(d2 >> 26); h2 = (word32)d2 & 0x3ffffff;
         d3 += c;      c = (word32)(d3 >> 26); h3 = (word32)d3 & 0x3ffffff;
         d4 += c;      c = (word32)(d4 >> 26); h4 = (word32)d4 & 0x3ffffff;
+#endif
         h0 += c * 5;  c =  (h0 >> 26); h0 =                h0 & 0x3ffffff;
         h1 += c;
 
@@ -392,6 +466,10 @@ static int poly1305_blocks(Poly1305* ctx, const unsigned char *m,
     ctx->h[3] = h3;
     ctx->h[4] = h4;
 
+#if defined(WOLFSSL_W64_WRAPPER) && defined(WOLFSSL_SMALL_STACK)
+    XFREE(d, NULL, DYNAMIC_TYPE_TMP_BUFFER);
+#endif
+
     return 0;
 
 #endif /* end of 64 bit cpu blocks or 32 bit cpu */
@@ -517,7 +595,11 @@ int wc_Poly1305Final(Poly1305* ctx, byte* mac)
 
     word32 h0,h1,h2,h3,h4,c;
     word32 g0,g1,g2,g3,g4;
+#ifdef WOLFSSL_W64_WRAPPER
+    w64wrapper f;
+#else
     word64 f;
+#endif
     word32 mask;
 
 #endif
@@ -656,10 +738,31 @@ int wc_Poly1305Final(Poly1305* ctx, byte* mac)
     h3 = ((h3 >> 18) | (h4 <<  8)) & 0xffffffff;
 
     /* mac = (h + pad) % (2^128) */
+#ifdef WOLFSSL_W64_WRAPPER
+    f = w64From32(0, h0);
+    f = w64Add32(f, ctx->pad[0], NULL);
+    h0 = w64GetLow32(f);
+
+    f = w64ShiftRight(f, 32);
+    f = w64Add32(f, h1, NULL);
+    f = w64Add32(f, ctx->pad[1], NULL);
+    h1 = w64GetLow32(f);
+
+    f = w64ShiftRight(f, 32);
+    f = w64Add32(f, h2, NULL);
+    f = w64Add32(f, ctx->pad[2], NULL);
+    h2 = w64GetLow32(f);
+
+    f = w64ShiftRight(f, 32);
+    f = w64Add32(f, h3, NULL);
+    f = w64Add32(f, ctx->pad[3], NULL);
+    h3 = w64GetLow32(f);
+#else
     f = (word64)h0 + ctx->pad[0]            ; h0 = (word32)f;
     f = (word64)h1 + ctx->pad[1] + (f >> 32); h1 = (word32)f;
     f = (word64)h2 + ctx->pad[2] + (f >> 32); h2 = (word32)f;
     f = (word64)h3 + ctx->pad[3] + (f >> 32); h3 = (word32)f;
+#endif
 
     U32TO8(mac + 0, h0);
     U32TO8(mac + 4, h1);
diff --git a/wolfssl/wolfcrypt/misc.h b/wolfssl/wolfcrypt/misc.h
index 533faf2df..cc068db44 100644
--- a/wolfssl/wolfcrypt/misc.h
+++ b/wolfssl/wolfcrypt/misc.h
@@ -145,6 +145,7 @@ WOLFSSL_LOCAL word32 w64GetLow32(w64wrapper n);
 WOLFSSL_LOCAL word32 w64GetHigh32(w64wrapper n);
 WOLFSSL_LOCAL void w64SetLow32(w64wrapper *n, word32 low);
 WOLFSSL_LOCAL w64wrapper w64Add32(w64wrapper a, word32 b, byte *wrap);
+WOLFSSL_LOCAL w64wrapper w64Add(w64wrapper a, w64wrapper b, byte *wrap);
 WOLFSSL_LOCAL w64wrapper w64Sub32(w64wrapper a, word32 b, byte *wrap);
 WOLFSSL_LOCAL byte w64GT(w64wrapper a, w64wrapper b);
 WOLFSSL_LOCAL byte w64IsZero(w64wrapper a);
@@ -157,6 +158,7 @@ WOLFSSL_LOCAL w64wrapper w64Sub(w64wrapper a, w64wrapper b);
 WOLFSSL_LOCAL void w64Zero(w64wrapper *a);
 WOLFSSL_LOCAL w64wrapper w64ShiftRight(w64wrapper a, int shift);
 WOLFSSL_LOCAL w64wrapper w64ShiftLeft(w64wrapper a, int shift);
+WOLFSSL_LOCAL w64wrapper w64Mul(word32 a, word32 b);
 
 #else /* !NO_INLINE */
 
diff --git a/wolfssl/wolfcrypt/poly1305.h b/wolfssl/wolfcrypt/poly1305.h
index 7e598e859..68fc74ae5 100644
--- a/wolfssl/wolfcrypt/poly1305.h
+++ b/wolfssl/wolfcrypt/poly1305.h
@@ -57,7 +57,7 @@
 
 #if defined(USE_INTEL_POLY1305_SPEEDUP)
 #elif (defined(WC_HAS_SIZEOF_INT128_64BIT) || defined(WC_HAS_MSVC_64BIT) ||  \
-       defined(WC_HAS_GCC_4_4_64BIT))
+       defined(WC_HAS_GCC_4_4_64BIT)) && !defined(WOLFSSL_W64_WRAPPER_TEST)
 #define POLY130564
 #else
 #define POLY130532