diff --git a/Makefile b/Makefile
index 516b31e..3edd250 100644
--- a/Makefile
+++ b/Makefile
@@ -74,7 +74,7 @@ default: DEBUGFLAGS=
 default: lib xxhsum_and_links
 
 .PHONY: all
-all: lib xxhsum xxhsum_inlinedXXH dispatch
+all: lib xxhsum xxhsum_inlinedXXH
 
 ## xxhsum is the command line interface (CLI)
 xxhsum: xxhash.o xxhsum.o
@@ -84,6 +84,7 @@ xxhsum32: CFLAGS += -m32  ## generate CLI in 32-bits mode
 xxhsum32: xxhash.c xxhsum.c  ## do not generate object (avoid mixing different ABI)
 	$(CC) $(FLAGS) $^ $(LDFLAGS) -o $@$(EXT)
 
+## dispatch only works for x86/x64 systems
 dispatch: CPPFLAGS += -DXXHSUM_DISPATCH=1
 dispatch: xxhash.o xxh_x86dispatch.o xxhsum.c
 	$(CC) $(FLAGS) $^ $(LDFLAGS) -o $@$(EXT)
diff --git a/xxh3.h b/xxh3.h
index 2d9c269..d6291e9 100644
--- a/xxh3.h
+++ b/xxh3.h
@@ -1478,9 +1478,9 @@ XXH3_scrambleAcc_vsx(void* XXH_RESTRICT acc, const void* XXH_RESTRICT secret)
 
 XXH_FORCE_INLINE void
 XXH3_accumulate_512_scalar(void* XXH_RESTRICT acc,
-                    const void* XXH_RESTRICT input,
-                    const void* XXH_RESTRICT secret,
-                    XXH3_accWidth_e accWidth)
+                     const void* XXH_RESTRICT input,
+                     const void* XXH_RESTRICT secret,
+                     XXH3_accWidth_e accWidth)
 {
     XXH_ALIGN(XXH_ACC_ALIGN) xxh_u64* const xacc = (xxh_u64*) acc; /* presumed aligned */
     const xxh_u8* const xinput  = (const xxh_u8*) input;  /* no alignment restriction */
@@ -1994,18 +1994,20 @@ XXH3_consumeStripes(xxh_u64* XXH_RESTRICT acc,
                     size_t* XXH_RESTRICT nbStripesSoFarPtr, size_t nbStripesPerBlock,
                     const xxh_u8* XXH_RESTRICT input, size_t totalStripes,
                     const xxh_u8* XXH_RESTRICT secret, size_t secretLimit,
-                    XXH3_accWidth_e accWidth)
+                    XXH3_accWidth_e accWidth,
+                    XXH3_f_accumulate_512 f_acc512,
+                    XXH3_f_scrambleAcc f_scramble)
 {
     XXH_ASSERT(*nbStripesSoFarPtr < nbStripesPerBlock);
     if (nbStripesPerBlock - *nbStripesSoFarPtr <= totalStripes) {
         /* need a scrambling operation */
         size_t const nbStripes = nbStripesPerBlock - *nbStripesSoFarPtr;
-        XXH3_accumulate(acc, input, secret + nbStripesSoFarPtr[0] * XXH_SECRET_CONSUME_RATE, nbStripes, accWidth, XXH3_accumulate_512);
-        XXH3_scrambleAcc(acc, secret + secretLimit);
-        XXH3_accumulate(acc, input + nbStripes * XXH_STRIPE_LEN, secret, totalStripes - nbStripes, accWidth, XXH3_accumulate_512);
+        XXH3_accumulate(acc, input, secret + nbStripesSoFarPtr[0] * XXH_SECRET_CONSUME_RATE, nbStripes, accWidth, f_acc512);
+        f_scramble(acc, secret + secretLimit);
+        XXH3_accumulate(acc, input + nbStripes * XXH_STRIPE_LEN, secret, totalStripes - nbStripes, accWidth, f_acc512);
         *nbStripesSoFarPtr = totalStripes - nbStripes;
     } else {
-        XXH3_accumulate(acc, input, secret + nbStripesSoFarPtr[0] * XXH_SECRET_CONSUME_RATE, totalStripes, accWidth, XXH3_accumulate_512);
+        XXH3_accumulate(acc, input, secret + nbStripesSoFarPtr[0] * XXH_SECRET_CONSUME_RATE, totalStripes, accWidth, f_acc512);
         *nbStripesSoFarPtr += totalStripes;
     }
 }
@@ -2014,7 +2016,11 @@ XXH3_consumeStripes(xxh_u64* XXH_RESTRICT acc,
  * Both XXH3_64bits_update and XXH3_128bits_update use this routine.
  */
 XXH_FORCE_INLINE XXH_errorcode
-XXH3_update(XXH3_state_t* state, const xxh_u8* input, size_t len, XXH3_accWidth_e accWidth)
+XXH3_update(XXH3_state_t* state,
+            const xxh_u8* input, size_t len,
+            XXH3_accWidth_e accWidth,
+            XXH3_f_accumulate_512 f_acc512,
+            XXH3_f_scrambleAcc f_scramble)
 {
     if (input==NULL)
 #if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1)
@@ -2050,7 +2056,7 @@ XXH3_update(XXH3_state_t* state, const xxh_u8* input, size_t len, XXH3_accWidth_
                                &state->nbStripesSoFar, state->nbStripesPerBlock,
                                 state->buffer, XXH3_INTERNALBUFFER_STRIPES,
                                 secret, state->secretLimit,
-                                accWidth);
+                                accWidth, f_acc512, f_scramble);
             state->bufferedSize = 0;
         }
 
@@ -2062,7 +2068,7 @@ XXH3_update(XXH3_state_t* state, const xxh_u8* input, size_t len, XXH3_accWidth_
                                    &state->nbStripesSoFar, state->nbStripesPerBlock,
                                     input, XXH3_INTERNALBUFFER_STRIPES,
                                     secret, state->secretLimit,
-                                    accWidth);
+                                    accWidth, f_acc512, f_scramble);
                 input += XXH3_INTERNALBUFFER_SIZE;
             } while (input<=limit);
             /* for last partial stripe */
@@ -2081,7 +2087,8 @@ XXH3_update(XXH3_state_t* state, const xxh_u8* input, size_t len, XXH3_accWidth_
 XXH_PUBLIC_API XXH_errorcode
 XXH3_64bits_update(XXH3_state_t* state, const void* input, size_t len)
 {
-    return XXH3_update(state, (const xxh_u8*)input, len, XXH3_acc_64bits);
+    return XXH3_update(state, (const xxh_u8*)input, len,
+                       XXH3_acc_64bits, XXH3_accumulate_512, XXH3_scrambleAcc);
 }
 
 
@@ -2103,7 +2110,7 @@ XXH3_digest_long (XXH64_hash_t* acc,
                            &nbStripesSoFar, state->nbStripesPerBlock,
                             state->buffer, nbStripes,
                             secret, state->secretLimit,
-                            accWidth);
+                            accWidth, XXH3_accumulate_512, XXH3_scrambleAcc);
         if (state->bufferedSize % XXH_STRIPE_LEN) {  /* one last partial stripe */
             XXH3_accumulate_512(acc,
                                 state->buffer + state->bufferedSize - XXH_STRIPE_LEN,
@@ -2557,7 +2564,8 @@ XXH3_128bits_reset_withSeed(XXH3_state_t* statePtr, XXH64_hash_t seed)
 XXH_PUBLIC_API XXH_errorcode
 XXH3_128bits_update(XXH3_state_t* state, const void* input, size_t len)
 {
-    return XXH3_update(state, (const xxh_u8*)input, len, XXH3_acc_128bits);
+    return XXH3_update(state, (const xxh_u8*)input, len,
+                       XXH3_acc_128bits, XXH3_accumulate_512, XXH3_scrambleAcc);
 }
 
 XXH_PUBLIC_API XXH128_hash_t XXH3_128bits_digest (const XXH3_state_t* state)
diff --git a/xxh_x86dispatch.c b/xxh_x86dispatch.c
index dd99ec7..950ffe3 100644
--- a/xxh_x86dispatch.c
+++ b/xxh_x86dispatch.c
@@ -409,6 +409,24 @@ static const dispatchFunctions_s k_dispatch[NB_DISPATCHES] = {
         /* avx512 */ { XXHL64_default_avx512, XXHL64_seed_avx512, XXHL64_secret_avx512 }
 };
 
+typedef void (*XXH3_dispatchx86_accumulate_512)(void* XXH_RESTRICT acc, const void* XXH_RESTRICT input, const void* XXH_RESTRICT secret, XXH3_accWidth_e accWidth);
+typedef void (*XXH3_dispatchx86_scrambleAcc)(void* XXH_RESTRICT acc, const void* XXH_RESTRICT secret);
+
+typedef struct {
+    XXH3_dispatchx86_accumulate_512  accumulate_512;
+    XXH3_dispatchx86_scrambleAcc     scrambleAcc;
+} coreFunctions_s;
+
+static coreFunctions_s g_coreFunc = { NULL, NULL };
+
+static const coreFunctions_s k_coreFunc[NB_DISPATCHES] = {
+        /* scalar */ { XXH3_accumulate_512_scalar, XXH3_scrambleAcc_scalar },
+        /* sse2   */ { XXH3_accumulate_512_sse2,   XXH3_scrambleAcc_sse2 },
+        /* avx2   */ { XXH3_accumulate_512_avx2,   XXH3_scrambleAcc_avx2 },
+        /* avx512 */ { XXH3_accumulate_512_avx512, XXH3_scrambleAcc_avx512 },
+};
+
+
 
 static void setDispatch(void)
 {
@@ -421,6 +439,7 @@ static void setDispatch(void)
     assert(vecID != XXH_AVX2);
 #endif
     g_dispatch = k_dispatch[vecID];
+    g_coreFunc = k_coreFunc[vecID];
 }
 
 
@@ -465,3 +484,11 @@ XXH64_hash_t XXH3_64bits_withSecret_dispatch(const void* input, size_t len, cons
 {
     return XXH3_64bits_internal(input, len, 0, secret, secretLen, XXH3_hashLong_64b_withSecret_selection);
 }
+
+XXH_errorcode
+XXH3_64bits_update_dispatch(XXH3_state_t* state, const void* input, size_t len)
+{
+    if (g_coreFunc.accumulate_512 == NULL) setDispatch();
+    return XXH3_update(state, (const xxh_u8*)input, len,
+                       XXH3_acc_64bits, g_coreFunc.accumulate_512, g_coreFunc.scrambleAcc);
+}
diff --git a/xxh_x86dispatch.h b/xxh_x86dispatch.h
index d86f89f..d08c680 100644
--- a/xxh_x86dispatch.h
+++ b/xxh_x86dispatch.h
@@ -45,6 +45,7 @@ extern "C" {
 XXH64_hash_t XXH3_64bits_dispatch(const void* input, size_t len);
 XXH64_hash_t XXH3_64bits_withSeed_dispatch(const void* input, size_t len, XXH64_hash_t seed);
 XXH64_hash_t XXH3_64bits_withSecret_dispatch(const void* input, size_t len, const void* secret, size_t secretLen);
+XXH_errorcode XXH3_64bits_update_dispatch(XXH3_state_t* state, const void* input, size_t len);
 
 
 /* automatic replacement of XXH3 functions.
@@ -57,6 +58,8 @@ XXH64_hash_t XXH3_64bits_withSecret_dispatch(const void* input, size_t len, cons
 # define XXH3_64bits_withSeed XXH3_64bits_withSeed_dispatch
 # undef  XXH3_64bits_withSecret
 # define XXH3_64bits_withSecret XXH3_64bits_withSecret_dispatch
+# undef  XXH3_64bits_update
+# define XXH3_64bits_update XXH3_64bits_update_dispatch
 
 #endif /* XXH_DISPATCH_DISABLE_RENAME */