diff --git a/Makefile b/Makefile
index b5fb136..164853a 100644
--- a/Makefile
+++ b/Makefile
@@ -3,6 +3,7 @@ CPPFLAGS=	-DHAVE_KALLOC
 INCLUDES=
 OBJS=		kthread.o kalloc.o misc.o bseq.o sketch.o sdust.o options.o index.o chain.o align.o hit.o map.o format.o pe.o esterr.o splitidx.o ksw2_ll_sse.o
 OBJS_SSE=	ksw2_extz2_sse41.o ksw2_extd2_sse41.o ksw2_exts2_sse41.o ksw2_extz2_sse2.o ksw2_extd2_sse2.o ksw2_exts2_sse2.o
+DISPATCH_FLAG=-msse4.1
 PROG=		minimap2
 PROG_EXTRA=	sdust minimap2-lite
 LIBS=		-lm -lz -lpthread
@@ -14,9 +15,11 @@ ifeq ($(avx2),)
 	OBJS+=$(OBJS_SSE) ksw2_dispatch.o
 else
 	OBJS+=ksw2_extd2_avx2.o $(OBJS_SSE) ksw2_dispatch.o
+	DISPATCH_FLAG=-mavx2
 endif
 else
 	OBJS+=ksw2_extd2_avx512.o ksw2_extd2_avx2.o $(OBJS_SSE) ksw2_dispatch.o
+	DISPATCH_FLAG=-mavx512bw
 endif
 else                # if sse2only is defined
 	OBJS+=ksw2_extz2_sse.o ksw2_extd2_sse.o ksw2_exts2_sse.o
@@ -84,13 +87,8 @@ ksw2_exts2_sse41.o:ksw2_exts2_sse.c ksw2.h kalloc.h
 ksw2_exts2_sse2.o:ksw2_exts2_sse.c ksw2.h kalloc.h
 		$(CC) -c $(CFLAGS) -msse2 -mno-sse4.1 $(CPPFLAGS) -DKSW_CPU_DISPATCH -DKSW_SSE2_ONLY $(INCLUDES) $< -o $@
 
-ifeq ($(avx2),)
 ksw2_dispatch.o:ksw2_dispatch.c ksw2.h
-		$(CC) -c $(CFLAGS) -msse4.1 $(CPPFLAGS) -DKSW_CPU_DISPATCH $(INCLUDES) $< -o $@
-else
-ksw2_dispatch.o:ksw2_dispatch.c ksw2.h
-		$(CC) -c $(CFLAGS) -mavx2 $(CPPFLAGS) -DKSW_CPU_DISPATCH $(INCLUDES) $< -o $@
-endif
+		$(CC) -c $(CFLAGS) $(DISPATCH_FLAG) $(CPPFLAGS) -DKSW_CPU_DISPATCH $(INCLUDES) $< -o $@
 
 # NEON-specific targets on ARM
 
diff --git a/ksw2_extd2_avx512.c b/ksw2_extd2_avx512.c
index ed3bab3..05eae58 100644
--- a/ksw2_extd2_avx512.c
+++ b/ksw2_extd2_avx512.c
@@ -166,8 +166,8 @@ void ksw_extd2_avx512(void *km, int qlen, const uint8_t *query, int tlen, const
 				__dp_code_block2; // save u[] and v[]; update a, b, a2 and b2
 				_mm512_store_si512(&x[t],  _mm512_sub_epi8(_mm512_max_epi8(a,  zero_), qe_));
 				_mm512_store_si512(&y[t],  _mm512_sub_epi8(_mm512_max_epi8(b,  zero_), qe_));
-				_mm512_store_si512(&x2[t], _mm512_sub_epi8(_mm512_max_epi8(a2, zero_), qe_));
-				_mm512_store_si512(&y2[t], _mm512_sub_epi8(_mm512_max_epi8(b2, zero_), qe_));
+				_mm512_store_si512(&x2[t], _mm512_sub_epi8(_mm512_max_epi8(a2, zero_), qe2_));
+				_mm512_store_si512(&y2[t], _mm512_sub_epi8(_mm512_max_epi8(b2, zero_), qe2_));
 			}
 		} else if (!(flag&KSW_EZ_RIGHT)) { // gap left-alignment
 			__m512i *pr = p + (size_t)r * n_col_ - st_;
@@ -190,9 +190,9 @@ void ksw_extd2_avx512(void *km, int qlen, const uint8_t *query, int tlen, const
 				d = _mm512_or_si512(d, _mm512_maskz_set1_epi8(_mm512_cmpgt_epi8_mask(b,  zero_), 0x10)); // d = b  > 0? 1<<4 : 0
 				_mm512_store_si512(&y[t],  _mm512_sub_epi8(_mm512_max_epi8(b,  zero_), qe_));
 				d = _mm512_or_si512(d, _mm512_maskz_set1_epi8(_mm512_cmpgt_epi8_mask(a2, zero_), 0x20)); // d = a2 > 0? 1<<5 : 0
-				_mm512_store_si512(&x2[t], _mm512_sub_epi8(_mm512_max_epi8(a2, zero_), qe_));
+				_mm512_store_si512(&x2[t], _mm512_sub_epi8(_mm512_max_epi8(a2, zero_), qe2_));
 				d = _mm512_or_si512(d, _mm512_maskz_set1_epi8(_mm512_cmpgt_epi8_mask(b2, zero_), 0x40)); // d = b2 > 0? 1<<6 : 0
-				_mm512_store_si512(&y2[t], _mm512_sub_epi8(_mm512_max_epi8(b2, zero_), qe_));
+				_mm512_store_si512(&y2[t], _mm512_sub_epi8(_mm512_max_epi8(b2, zero_), qe2_));
 				_mm512_store_si512(&pr[t], d);
 			}
 		} else { // gap right-alignment
@@ -201,7 +201,7 @@ void ksw_extd2_avx512(void *km, int qlen, const uint8_t *query, int tlen, const
 			for (t = st_; t <= en_; ++t) {
 				__m512i d, z, a, b, a2, b2, xt1, x2t1, vt1, ut, tmp;
 				__dp_code_block1;
-				d = _mm512_mask_blend_epi8(_mm512_cmpgt_epi8_mask(z, a),  _mm512_setzero(), _mm512_set1_epi8(1));
+				d = _mm512_mask_blend_epi8(_mm512_cmpgt_epi8_mask(z, a),  zero_, _mm512_set1_epi8(1));
 				z = _mm512_max_epi8(z, a);
 				d = _mm512_mask_blend_epi8(_mm512_cmpgt_epi8_mask(z, b),  d, _mm512_set1_epi8(2));
 				z = _mm512_max_epi8(z, b);
@@ -216,9 +216,9 @@ void ksw_extd2_avx512(void *km, int qlen, const uint8_t *query, int tlen, const
 				d = _mm512_or_si512(d, _mm512_maskz_set1_epi8(_mm512_cmpge_epi8_mask(b,  zero_), 0x10)); // d = b  >= 0? 1<<4 : 0
 				_mm512_store_si512(&y[t],  _mm512_sub_epi8(_mm512_max_epi8(b,  zero_), qe_));
 				d = _mm512_or_si512(d, _mm512_maskz_set1_epi8(_mm512_cmpge_epi8_mask(a2, zero_), 0x20)); // d = a2 >= 0? 1<<5 : 0
-				_mm512_store_si512(&x2[t], _mm512_sub_epi8(_mm512_max_epi8(a2, zero_), qe_));
+				_mm512_store_si512(&x2[t], _mm512_sub_epi8(_mm512_max_epi8(a2, zero_), qe2_));
 				d = _mm512_or_si512(d, _mm512_maskz_set1_epi8(_mm512_cmpge_epi8_mask(b2, zero_), 0x40)); // d = b2 >= 0? 1<<6 : 0
-				_mm512_store_si512(&y2[t], _mm512_sub_epi8(_mm512_max_epi8(b2, zero_), qe_));
+				_mm512_store_si512(&y2[t], _mm512_sub_epi8(_mm512_max_epi8(b2, zero_), qe2_));
 				_mm512_store_si512(&pr[t], d);
 			}
 		}