Merge branch 'master' into avx

for every idx should map all input seqs

FAQ.md

@@ -0,0 +1,46 @@
+#### 1. Alignment different with option `-a` or `-c`?
+
+Without `-a`, `-c` or `--cs`, minimap2 only finds *approximate* mapping
+locations without detailed base alignment. In particular, the start and end
+positions of the alignment are impricise. With one of those options, minimap2
+will perform base alignment, which is generally more accurate but is much
+slower.
+
+#### 2. How to map Illumina short reads to noisy long reads?
+
+No good solutions. The better approach is to assemble short reads into contigs
+and then map noisy reads to contigs.
+
+#### 3. The output SAM doesn't have a header.
+
+By default, minimap2 indexes 4 billion reference bases (4Gb) in a batch and map
+all reads against each reference batch. Given a reference longer than 4Gb,
+minimap2 is unable to see all the sequences and thus can't produce a correct
+SAM header. In this case, minimap2 doesn't output any SAM header. There are two
+solutions to this issue. First, you may increase option `-I` to, for example,
+`-I8g` to index more reference bases in a batch. This is preferred if your
+machine has enough memory. Second, if your machines doesn't have enough memory
+to hold the reference index, you can use the `--split-prefix` option in a
+command line like:
+```sh
+minimap2 -ax map-ont --split-prefix=tmp ref.fa reads.fq
+```
+This second approach uses less memory, but it is slower and requires temporary
+disk space.
+
+#### 4. The output SAM is malformatted.
+
+This typically happens when you use nohup to wrap a minimap2 command line.
+Nohup is discouraged as it breaks piping. If you have to use nohup, please
+specify an output file with option `-o`.
+
+#### 5. How to output one alignment per read?
+
+You can use `--secondary=no` to suppress secondary alignments (aka multiple
+mappings), but you can't suppress supplementary alignment (aka split or
+chimeric alignment) this way. You can use samtools to filter out these
+alignments:
+```sh
+minimap2 -ax map-out ref.fa reads.fq | samtools view -F0x900
+```
+However, this is discouraged as supplementary alignment is informative.

Makefile

@@ -2,13 +2,25 @@ CFLAGS=		-g -Wall -O2 -Wc++-compat #-Wextra
 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
 
 ifeq ($(arm_neon),) # if arm_neon is not defined
 ifeq ($(sse2only),) # if sse2only is not defined
-	OBJS+=ksw2_extz2_sse41.o ksw2_extd2_sse41.o ksw2_exts2_sse41.o ksw2_extz2_sse2.o ksw2_extd2_sse2.o ksw2_exts2_sse2.o ksw2_dispatch.o
+ifeq ($(avx512),)
+ifeq ($(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
 endif
@@ -22,6 +34,16 @@ else				#if aarch64 is defined
 endif
 endif
 
+ifneq ($(asan),)
+	CFLAGS+=-fsanitize=address
+	LIBS+=-fsanitize=address
+endif
+
+ifneq ($(tsan),)
+	CFLAGS+=-fsanitize=thread
+	LIBS+=-fsanitize=thread
+endif
+
 .PHONY:all extra clean depend
 .SUFFIXES:.c .o
 
@@ -57,11 +79,17 @@ ksw2_extz2_sse41.o:ksw2_extz2_sse.c ksw2.h kalloc.h
 ksw2_extz2_sse2.o:ksw2_extz2_sse.c ksw2.h kalloc.h
 		$(CC) -c $(CFLAGS) -msse2 -mno-sse4.1 $(CPPFLAGS) -DKSW_CPU_DISPATCH -DKSW_SSE2_ONLY $(INCLUDES) $< -o $@
 
+ksw2_extd2_avx2.o:ksw2_extd2_sse.c ksw2.h kalloc.h
+		$(CC) -c $(CFLAGS) -mavx2 $(CPPFLAGS) -DKSW_CPU_DISPATCH $(INCLUDES) $< -o $@
+
+ksw2_extd2_avx512.o:ksw2_extd2_sse.c ksw2.h kalloc.h
+		$(CC) -c $(CFLAGS) -mavx512bw $(CPPFLAGS) -DKSW_CPU_DISPATCH $(INCLUDES) $< -o $@
+
 ksw2_extd2_sse41.o:ksw2_extd2_sse.c ksw2.h kalloc.h
-		$(CC) -c $(CFLAGS) -msse4.1 $(CPPFLAGS) -DKSW_CPU_DISPATCH $(INCLUDES) $< -o $@
+		$(CC) -c $(CFLAGS) -msse4.1 -mno-avx2 $(CPPFLAGS) -DKSW_CPU_DISPATCH $(INCLUDES) $< -o $@
 
 ksw2_extd2_sse2.o:ksw2_extd2_sse.c ksw2.h kalloc.h
-		$(CC) -c $(CFLAGS) -msse2 -mno-sse4.1 $(CPPFLAGS) -DKSW_CPU_DISPATCH -DKSW_SSE2_ONLY $(INCLUDES) $< -o $@
+		$(CC) -c $(CFLAGS) -msse2 -mno-sse4.1 -mno-avx2 $(CPPFLAGS) -DKSW_CPU_DISPATCH -DKSW_SSE2_ONLY $(INCLUDES) $< -o $@
 
 ksw2_exts2_sse41.o:ksw2_exts2_sse.c ksw2.h kalloc.h
 		$(CC) -c $(CFLAGS) -msse4.1 $(CPPFLAGS) -DKSW_CPU_DISPATCH $(INCLUDES) $< -o $@
@@ -70,7 +98,7 @@ 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 $@
 
 ksw2_dispatch.o:ksw2_dispatch.c ksw2.h
-		$(CC) -c $(CFLAGS) -msse4.1 $(CPPFLAGS) -DKSW_CPU_DISPATCH $(INCLUDES) $< -o $@
+		$(CC) -c $(CFLAGS) $(DISPATCH_FLAG) $(CPPFLAGS) -DKSW_CPU_DISPATCH $(INCLUDES) $< -o $@
 
 # NEON-specific targets on ARM
 

README.md

@@ -315,9 +315,10 @@ highlighted in bold. The description may help to tune minimap2 parameters.
 ### <a name="help"></a>Getting help
 
 Manpage [minimap2.1][manpage] provides detailed description of minimap2
-command line options and optional tags. If you encounter bugs or have further
-questions or requests, you can raise an issue at the [issue page][issue].
-There is not a specific mailing list for the time being.
+command line options and optional tags. The [FAQ](FAQ.md) page answers several
+frequently asked questions. If you encounter bugs or have further questions or
+requests, you can raise an issue at the [issue page][issue].  There is not a
+specific mailing list for the time being.
 
 ### <a name="cite"></a>Citing minimap2
 

chain.c

@@ -155,8 +155,8 @@ mm128_t *mm_chain_dp(int max_dist_x, int max_dist_y, int bw, int max_skip, int m
 		memcpy(&a[k], &b[w[i].y>>32], n * sizeof(mm128_t));
 		k += n;
 	}
-	memcpy(u, u2, n_u * 8);
-	memcpy(b, a, k * sizeof(mm128_t)); // write _a_ to _b_ and deallocate _a_ because _a_ is oversized, sometimes a lot
+	if (n_u) memcpy(u, u2, n_u * 8);
+	if (k) memcpy(b, a, k * sizeof(mm128_t)); // write _a_ to _b_ and deallocate _a_ because _a_ is oversized, sometimes a lot
 	kfree(km, a); kfree(km, w); kfree(km, u2);
 	return b;
 }

example.c

@@ -35,6 +35,8 @@ int main(int argc, char *argv[])
 	while ((mi = mm_idx_reader_read(r, n_threads)) != 0) { // traverse each part of the index
 		mm_mapopt_update(&mopt, mi); // this sets the maximum minimizer occurrence; TODO: set a better default in mm_mapopt_init()!
 		mm_tbuf_t *tbuf = mm_tbuf_init(); // thread buffer; for multi-threading, allocate one tbuf for each thread
+		gzrewind(f);
+		kseq_rewind(ks);
 		while (kseq_read(ks) >= 0) { // each kseq_read() call reads one query sequence
 			mm_reg1_t *reg;
 			int j, i, n_reg;

format.c

@@ -79,11 +79,11 @@ static char *mm_escape(char *s)
 	return s;
 }
 
-static void sam_write_rg_line(kstring_t *str, const char *s)
+static int sam_write_rg_line(kstring_t *str, const char *s)
 {
 	char *p, *q, *r, *rg_line = 0;
 	memset(mm_rg_id, 0, 256);
-	if (s == 0) return;
+	if (s == 0) return 0;
 	if (strstr(s, "@RG") != s) {
 		if (mm_verbose >= 1) fprintf(stderr, "[ERROR] the read group line is not started with @RG\n");
 		goto err_set_rg;
@@ -108,20 +108,23 @@ static void sam_write_rg_line(kstring_t *str, const char *s)
 	for (q = p, r = mm_rg_id; *q && *q != '\t' && *q != '\n'; ++q)
 		*r++ = *q;
 	mm_sprintf_lite(str, "%s\n", rg_line);
+	return 0;
 
 err_set_rg:
 	free(rg_line);
+	return -1;
 }
 
-void mm_write_sam_hdr(const mm_idx_t *idx, const char *rg, const char *ver, int argc, char *argv[])
+int mm_write_sam_hdr(const mm_idx_t *idx, const char *rg, const char *ver, int argc, char *argv[])
 {
 	kstring_t str = {0,0,0};
+	int ret = 0;
 	if (idx) {
 		uint32_t i;
 		for (i = 0; i < idx->n_seq; ++i)
 			mm_sprintf_lite(&str, "@SQ\tSN:%s\tLN:%d\n", idx->seq[i].name, idx->seq[i].len);
 	}
-	if (rg) sam_write_rg_line(&str, rg);
+	if (rg) ret = sam_write_rg_line(&str, rg);
 	mm_sprintf_lite(&str, "@PG\tID:minimap2\tPN:minimap2");
 	if (ver) mm_sprintf_lite(&str, "\tVN:%s", ver);
 	if (argc > 1) {
@@ -132,6 +135,7 @@ void mm_write_sam_hdr(const mm_idx_t *idx, const char *rg, const char *ver, int
 	}
 	mm_err_puts(str.s);
 	free(str.s);
+	return ret;
 }
 
 static void write_cs_core(kstring_t *s, const uint8_t *tseq, const uint8_t *qseq, const mm_reg1_t *r, char *tmp, int no_iden, int write_tag)

index.c

@@ -117,8 +117,8 @@ void mm_idx_stat(const mm_idx_t *mi)
 				if (kh_key(h, k)&1) ++n1;
 			}
 	}
-	fprintf(stderr, "[M::%s::%.3f*%.2f] distinct minimizers: %d (%.2f%% are singletons); average occurrences: %.3lf; average spacing: %.3lf\n",
-			__func__, realtime() - mm_realtime0, cputime() / (realtime() - mm_realtime0), n, 100.0*n1/n, (double)sum / n, (double)len / sum);
+	fprintf(stderr, "[M::%s::%.3f*%.2f] distinct minimizers: %d (%.2f%% are singletons); average occurrences: %.3lf; average spacing: %.3lf; total length: %ld\n",
+			__func__, realtime() - mm_realtime0, cputime() / (realtime() - mm_realtime0), n, 100.0*n1/n, (double)sum / n, (double)len / sum, (long)len);
 }
 
 int mm_idx_index_name(mm_idx_t *mi)

kalloc.c

@@ -18,15 +18,14 @@
  *       |                           |                 |                               |
  *       p=p->ptr->ptr->ptr->ptr     p->ptr            p->ptr->ptr                     p->ptr->ptr->ptr
  */
-
-#define MIN_CORE_SIZE 0x80000
-
 typedef struct header_t {
 	size_t size;
 	struct header_t *ptr;
 } header_t;
 
 typedef struct {
+	void *par;
+	size_t min_core_size;
 	header_t base, *loop_head, *core_head; /* base is a zero-sized block always kept in the loop */
 } kmem_t;
 
@@ -36,31 +35,39 @@ static void panic(const char *s)
 	abort();
 }
 
-void *km_init(void)
+void *km_init2(void *km_par, size_t min_core_size)
 {
-	return calloc(1, sizeof(kmem_t));
+	kmem_t *km;
+	km = (kmem_t*)kcalloc(km_par, 1, sizeof(kmem_t));
+	km->par = km_par;
+	km->min_core_size = min_core_size > 0? min_core_size : 0x80000;
+	return (void*)km;
 }
 
+void *km_init(void) { return km_init2(0, 0); }
+
 void km_destroy(void *_km)
 {
 	kmem_t *km = (kmem_t*)_km;
+	void *km_par;
 	header_t *p, *q;
 	if (km == NULL) return;
+	km_par = km->par;
 	for (p = km->core_head; p != NULL;) {
 		q = p->ptr;
-		free(p);
+		kfree(km_par, p);
 		p = q;
 	}
-	free(km);
+	kfree(km_par, km);
 }
 
 static header_t *morecore(kmem_t *km, size_t nu)
 {
 	header_t *q;
 	size_t bytes, *p;
-	nu = (nu + 1 + (MIN_CORE_SIZE - 1)) / MIN_CORE_SIZE * MIN_CORE_SIZE; /* the first +1 for core header */
+	nu = (nu + 1 + (km->min_core_size - 1)) / km->min_core_size * km->min_core_size; /* the first +1 for core header */
 	bytes = nu * sizeof(header_t);
-	q = (header_t*)malloc(bytes);
+	q = (header_t*)kmalloc(km->par, bytes);
 	if (!q) panic("[morecore] insufficient memory");
 	q->ptr = km->core_head, q->size = nu, km->core_head = q;
 	p = (size_t*)(q + 1);
@@ -125,7 +132,7 @@ void *kmalloc(void *_km, size_t n_bytes)
 
 	if (n_bytes == 0) return 0;
 	if (km == NULL) return malloc(n_bytes);
-	n_units = (n_bytes + sizeof(size_t) + sizeof(header_t) - 1) / sizeof(header_t) + 1;
+	n_units = (n_bytes + sizeof(size_t) + sizeof(header_t) - 1) / sizeof(header_t); /* header+n_bytes requires at least this number of units */
 
 	if (!(q = km->loop_head)) /* the first time when kmalloc() is called, intialize it */
 		q = km->loop_head = km->base.ptr = &km->base;
@@ -160,18 +167,18 @@ void *kcalloc(void *_km, size_t count, size_t size)
 void *krealloc(void *_km, void *ap, size_t n_bytes) // TODO: this can be made more efficient in principle
 {
 	kmem_t *km = (kmem_t*)_km;
-	size_t n_units, *p, *q;
+	size_t cap, *p, *q;
 
 	if (n_bytes == 0) {
 		kfree(km, ap); return 0;
 	}
 	if (km == NULL) return realloc(ap, n_bytes);
 	if (ap == NULL) return kmalloc(km, n_bytes);
-	n_units = (n_bytes + sizeof(size_t) + sizeof(header_t) - 1) / sizeof(header_t);
 	p = (size_t*)ap - 1;
-	if (*p >= n_units) return ap; /* TODO: this prevents shrinking */
+	cap = (*p) * sizeof(header_t) - sizeof(size_t);
+	if (cap >= n_bytes) return ap; /* TODO: this prevents shrinking */
 	q = (size_t*)kmalloc(km, n_bytes);
-	memcpy(q, ap, (*p - 1) * sizeof(header_t));
+	memcpy(q, ap, cap);
 	kfree(km, ap);
 	return q;
 }

kalloc.h

@@ -17,6 +17,7 @@ void *kcalloc(void *km, size_t count, size_t size);
 void kfree(void *km, void *ptr);
 
 void *km_init(void);
+void *km_init2(void *km_par, size_t min_core_size);
 void km_destroy(void *km);
 void km_stat(const void *_km, km_stat_t *s);
 
@@ -24,4 +25,13 @@ void km_stat(const void *_km, km_stat_t *s);
 }
 #endif
 
+#define KMALLOC(km, ptr, len) ((ptr) = (__typeof__(ptr))kmalloc((km), (len) * sizeof(*(ptr))))
+#define KCALLOC(km, ptr, len) ((ptr) = (__typeof__(ptr))kcalloc((km), (len), sizeof(*(ptr))))
+#define KREALLOC(km, ptr, len) ((ptr) = (__typeof__(ptr))krealloc((km), (ptr), (len) * sizeof(*(ptr))))
+
+#define KEXPAND(km, a, m) do { \
+		(m) = (m) >= 4? (m) + ((m)>>1) : 16; \
+		KREALLOC((km), (a), (m)); \
+	} while (0)
+
 #endif

ksw2_dispatch.c

@@ -2,15 +2,16 @@
 #include <stdlib.h>
 #include "ksw2.h"
 
-#define SIMD_SSE     0x1
-#define SIMD_SSE2    0x2
-#define SIMD_SSE3    0x4
-#define SIMD_SSSE3   0x8
-#define SIMD_SSE4_1  0x10
-#define SIMD_SSE4_2  0x20
-#define SIMD_AVX     0x40
-#define SIMD_AVX2    0x80
-#define SIMD_AVX512F 0x100
+#define SIMD_SSE      0x1
+#define SIMD_SSE2     0x2
+#define SIMD_SSE3     0x4
+#define SIMD_SSSE3    0x8
+#define SIMD_SSE4_1   0x10
+#define SIMD_SSE4_2   0x20
+#define SIMD_AVX      0x40
+#define SIMD_AVX2     0x80
+#define SIMD_AVX512F  0x100
+#define SIMD_AVX512BW 0x200
 
 #ifndef _MSC_VER
 // adapted from https://github.com/01org/linux-sgx/blob/master/common/inc/internal/linux/cpuid_gnu.h
@@ -48,6 +49,7 @@ static int x86_simd(void)
 		__cpuidex(cpuid, 7, 0);
 		if (cpuid[1]>>5 &1) flag |= SIMD_AVX2;
 		if (cpuid[1]>>16&1) flag |= SIMD_AVX512F;
+		if (cpuid[1]>>30&1) flag |= SIMD_AVX512BW;
 	}
 	return flag;
 }
@@ -71,7 +73,21 @@ void ksw_extd2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 				   int8_t q, int8_t e, int8_t q2, int8_t e2, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez);
 	extern void ksw_extd2_sse41(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
 				   int8_t q, int8_t e, int8_t q2, int8_t e2, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez);
+	extern void ksw_extd2_avx2(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
+				   int8_t q, int8_t e, int8_t q2, int8_t e2, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez);
+	extern void ksw_extd2_avx512(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
+				   int8_t q, int8_t e, int8_t q2, int8_t e2, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez);
 	if (ksw_simd < 0) ksw_simd = x86_simd();
+#if defined(__AVX512BW__)
+	if (ksw_simd & SIMD_AVX512BW)
+		ksw_extd2_avx512(km, qlen, query, tlen, target, m, mat, q, e, q2, e2, w, zdrop, end_bonus, flag, ez);
+	else
+#endif
+#if defined(__AVX2__)
+	if (ksw_simd & SIMD_AVX2)
+		ksw_extd2_avx2(km, qlen, query, tlen, target, m, mat, q, e, q2, e2, w, zdrop, end_bonus, flag, ez);
+	else
+#endif
 	if (ksw_simd & SIMD_SSE4_1)
 		ksw_extd2_sse41(km, qlen, query, tlen, target, m, mat, q, e, q2, e2, w, zdrop, end_bonus, flag, ez);
 	else if (ksw_simd & SIMD_SSE2)

ksw2_extd2_sse.c

@@ -4,7 +4,27 @@
 #include "ksw2.h"
 
 #ifdef __SSE2__
+
+#if defined(__AVX512BW__)
+#include <immintrin.h>
+#define SIMD_INT __m512i
+#define SIMD_SHIFT 6
+#define simd_func(func) _mm512_##func
+#define simd_funcw(func) _mm512_##func##_si512
+
+#elif defined(__AVX2__)
+#include <immintrin.h>
+#define SIMD_INT __m256i
+#define SIMD_SHIFT 5
+#define simd_func(func) _mm256_##func
+#define simd_funcw(func) _mm256_##func##_si256
+
+#elif defined(__SSE2__)
 #include <emmintrin.h>
+#define SIMD_INT __m128i
+#define SIMD_SHIFT 4
+#define simd_func(func) _mm_##func
+#define simd_funcw(func) _mm_##func##_si128
 
 #ifdef KSW_SSE2_ONLY
 #undef __SSE4_1__
@@ -13,12 +33,39 @@
 #ifdef __SSE4_1__
 #include <smmintrin.h>
 #endif
+#endif // defined(__SSE2__)
+
+#define SIMD_WIDTH (1<<SIMD_SHIFT)
+
+
+#if !defined(__AVX512BW__)
+#if defined(__AVX2__)
+static inline __m256i simd_slli_1(__m256i x)
+{
+	return _mm256_insert_epi8(_mm256_slli_si256(x, 1), _mm256_extract_epi8(x, 15), 16);
+}
+static inline __m256i simd_srli_last(__m256i x)
+{
+	return _mm256_insert_epi8(_mm256_setzero_si256(), _mm256_extract_epi8(x, 31), 0);
+}
+#elif defined(__SSE2__)
+static inline __m128i simd_slli_1(__m128i x) { return _mm_slli_si128(x, 1); }
+static inline __m128i simd_srli_last(__m128i x) { return _mm_srli_si128(x, 15); }
+#endif
+#endif // ~__AVX512BW__
+
 
 #ifdef KSW_CPU_DISPATCH
-#ifdef __SSE4_1__
+#if defined(__AVX512BW__)
+void ksw_extd2_avx512(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
+				   int8_t q, int8_t e, int8_t q2, int8_t e2, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez)
+#elif defined(__AVX2__)
+void ksw_extd2_avx2(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
+				   int8_t q, int8_t e, int8_t q2, int8_t e2, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez)
+#elif defined(__SSE4_1__)
 void ksw_extd2_sse41(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
 				   int8_t q, int8_t e, int8_t q2, int8_t e2, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez)
-#else
+#elif defined(__SSE2__)
 void ksw_extd2_sse2(void *km, int qlen, const uint8_t *query, int tlen, const uint8_t *target, int8_t m, const int8_t *mat,
 				   int8_t q, int8_t e, int8_t q2, int8_t e2, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez)
 #endif
@@ -27,64 +74,91 @@ void ksw_extd2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 				   int8_t q, int8_t e, int8_t q2, int8_t e2, int w, int zdrop, int end_bonus, int flag, ksw_extz_t *ez)
 #endif // ~KSW_CPU_DISPATCH
 {
+#if defined(__AVX512BW__)
 #define __dp_code_block1 \
-	z = _mm_load_si128(&s[t]); \
-	xt1 = _mm_load_si128(&x[t]);                     /* xt1 <- x[r-1][t..t+15] */ \
-	tmp = _mm_srli_si128(xt1, 15);                   /* tmp <- x[r-1][t+15] */ \
-	xt1 = _mm_or_si128(_mm_slli_si128(xt1, 1), x1_); /* xt1 <- x[r-1][t-1..t+14] */ \
+	z = _mm512_load_si512(&s[t]); \
+	tmp = _mm512_loadu_si512((uint8_t*)&x[t] - 1); \
+	xt1 = _mm512_mask_blend_epi8(1, tmp, x1_); \
+	x1_ = _mm512_maskz_set1_epi8(1, *((uint8_t*)&x[t] + 63)); \
+	tmp = _mm512_loadu_si512((uint8_t*)&v[t] - 1); \
+	vt1 = _mm512_mask_blend_epi8(1, tmp, v1_); \
+	v1_ = _mm512_maskz_set1_epi8(1, *((uint8_t*)&v[t] + 63)); \
+	a = _mm512_add_epi8(xt1, vt1); \
+	ut = _mm512_load_si512(&u[t]); \
+	b = _mm512_add_epi8(_mm512_load_si512(&y[t]), ut); \
+	tmp = _mm512_loadu_si512((uint8_t*)&x2[t] - 1); \
+	x2t1 = _mm512_mask_blend_epi8(1, tmp, x21_); \
+	x21_ = _mm512_maskz_set1_epi8(1, *((uint8_t*)&x2[t] + 63)); \
+	a2= _mm512_add_epi8(x2t1, vt1); \
+	b2= _mm512_add_epi8(_mm512_load_si512(&y2[t]), ut);
+#else
+#define __dp_code_block1 \
+	z = simd_funcw(load)(&s[t]); \
+	xt1 = simd_funcw(load)(&x[t]);                        /* xt1 <- x[r-1][t..t+15] */ \
+	tmp = simd_srli_last(xt1);                            /* tmp <- x[r-1][t+15] */ \
+	xt1 = simd_funcw(or)(simd_slli_1(xt1), x1_);          /* xt1 <- x[r-1][t-1..t+14] */ \
 	x1_ = tmp; \
-	vt1 = _mm_load_si128(&v[t]);                     /* vt1 <- v[r-1][t..t+15] */ \
-	tmp = _mm_srli_si128(vt1, 15);                   /* tmp <- v[r-1][t+15] */ \
-	vt1 = _mm_or_si128(_mm_slli_si128(vt1, 1), v1_); /* vt1 <- v[r-1][t-1..t+14] */ \
+	vt1 = simd_funcw(load)(&v[t]);                        /* vt1 <- v[r-1][t..t+15] */ \
+	tmp = simd_srli_last(vt1);                            /* tmp <- v[r-1][t+15] */ \
+	vt1 = simd_funcw(or)(simd_slli_1(vt1), v1_);          /* vt1 <- v[r-1][t-1..t+14] */ \
 	v1_ = tmp; \
-	a = _mm_add_epi8(xt1, vt1);                      /* a <- x[r-1][t-1..t+14] + v[r-1][t-1..t+14] */ \
-	ut = _mm_load_si128(&u[t]);                      /* ut <- u[t..t+15] */ \
-	b = _mm_add_epi8(_mm_load_si128(&y[t]), ut);     /* b <- y[r-1][t..t+15] + u[r-1][t..t+15] */ \
-	x2t1= _mm_load_si128(&x2[t]); \
-	tmp = _mm_srli_si128(x2t1, 15); \
-	x2t1= _mm_or_si128(_mm_slli_si128(x2t1, 1), x21_); \
+	a = simd_func(add_epi8)(xt1, vt1);                    /* a <- x[r-1][t-1..t+14] + v[r-1][t-1..t+14] */ \
+	ut = simd_funcw(load)(&u[t]);                         /* ut <- u[t..t+15] */ \
+	b = simd_func(add_epi8)(simd_funcw(load)(&y[t]), ut); /* b <- y[r-1][t..t+15] + u[r-1][t..t+15] */ \
+	x2t1= simd_funcw(load)(&x2[t]); \
+	tmp = simd_srli_last(x2t1); \
+	x2t1= simd_funcw(or)(simd_slli_1(x2t1), x21_); \
 	x21_= tmp; \
-	a2= _mm_add_epi8(x2t1, vt1); \
-	b2= _mm_add_epi8(_mm_load_si128(&y2[t]), ut);
+	a2= simd_func(add_epi8)(x2t1, vt1); \
+	b2= simd_func(add_epi8)(simd_funcw(load)(&y2[t]), ut);
+#endif // ~__AVX512BW__
 
 #define __dp_code_block2 \
-	_mm_store_si128(&u[t], _mm_sub_epi8(z, vt1));    /* u[r][t..t+15] <- z - v[r-1][t-1..t+14] */ \
-	_mm_store_si128(&v[t], _mm_sub_epi8(z, ut));     /* v[r][t..t+15] <- z - u[r-1][t..t+15] */ \
-	tmp = _mm_sub_epi8(z, q_); \
-	a = _mm_sub_epi8(a, tmp); \
-	b = _mm_sub_epi8(b, tmp); \
-	tmp = _mm_sub_epi8(z, q2_); \
-	a2= _mm_sub_epi8(a2, tmp); \
-	b2= _mm_sub_epi8(b2, tmp);
+	simd_funcw(store)(&u[t], simd_func(sub_epi8)(z, vt1));/* u[r][t..t+15] <- z - v[r-1][t-1..t+14] */ \
+	simd_funcw(store)(&v[t], simd_func(sub_epi8)(z, ut)); /* v[r][t..t+15] <- z - u[r-1][t..t+15] */ \
+	tmp = simd_func(sub_epi8)(z, q_); \
+	a = simd_func(sub_epi8)(a, tmp); \
+	b = simd_func(sub_epi8)(b, tmp); \
+	tmp = simd_func(sub_epi8)(z, q2_); \
+	a2= simd_func(sub_epi8)(a2, tmp); \
+	b2= simd_func(sub_epi8)(b2, tmp);
 
 	int r, t, qe = q + e, n_col_, *off = 0, *off_end = 0, tlen_, qlen_, last_st, last_en, wl, wr, max_sc, min_sc, long_thres, long_diff;
 	int with_cigar = !(flag&KSW_EZ_SCORE_ONLY), approx_max = !!(flag&KSW_EZ_APPROX_MAX);
 	int32_t *H = 0, H0 = 0, last_H0_t = 0;
 	uint8_t *qr, *sf, *mem, *mem2 = 0;
-	__m128i q_, q2_, qe_, qe2_, zero_, sc_mch_, sc_mis_, m1_, sc_N_;
-	__m128i *u, *v, *x, *y, *x2, *y2, *s, *p = 0;
+	SIMD_INT q_, q2_, qe_, qe2_, zero_, sc_mch_, sc_mis_, m1_, sc_N_, mask1_;
+	SIMD_INT *u, *v, *x, *y, *x2, *y2, *s, *p = 0;
 
 	ksw_reset_extz(ez);
 	if (m <= 1 || qlen <= 0 || tlen <= 0) return;
 
 	if (q2 + e2 < q + e) t = q, q = q2, q2 = t, t = e, e = e2, e2 = t; // make sure q+e no larger than q2+e2
 
-	zero_   = _mm_set1_epi8(0);
-	q_      = _mm_set1_epi8(q);
-	q2_     = _mm_set1_epi8(q2);
-	qe_     = _mm_set1_epi8(q + e);
-	qe2_    = _mm_set1_epi8(q2 + e2);
-	sc_mch_ = _mm_set1_epi8(mat[0]);
-	sc_mis_ = _mm_set1_epi8(mat[1]);
-	sc_N_   = mat[m*m-1] == 0? _mm_set1_epi8(-e2) : _mm_set1_epi8(mat[m*m-1]);
-	m1_     = _mm_set1_epi8(m - 1); // wildcard
+	zero_   = simd_func(set1_epi8)(0);
+	q_      = simd_func(set1_epi8)(q);
+	q2_     = simd_func(set1_epi8)(q2);
+	qe_     = simd_func(set1_epi8)(q + e);
+	qe2_    = simd_func(set1_epi8)(q2 + e2);
+	sc_mch_ = simd_func(set1_epi8)(mat[0]);
+	sc_mis_ = simd_func(set1_epi8)(mat[1]);
+	sc_N_   = mat[m*m-1] == 0? simd_func(set1_epi8)(-e2) : simd_func(set1_epi8)(mat[m*m-1]);
+	m1_     = simd_func(set1_epi8)(m - 1); // wildcard
+
+#if defined(__AVX512BW__)
+	mask1_ = _mm512_maskz_set1_epi8(1, 0xff);
+#elif defined(__AVX2__)
+	mask1_ = _mm256_setr_epi32(0xff, 0, 0, 0, 0, 0, 0, 0);
+#elif defined(__SSE2__)
+	mask1_ = _mm_setr_epi32(0xff, 0, 0, 0);
+#endif
 
 	if (w < 0) w = tlen > qlen? tlen : qlen;
 	wl = wr = w;
-	tlen_ = (tlen + 15) / 16;
+	tlen_ = (tlen + SIMD_WIDTH - 1) / SIMD_WIDTH;
 	n_col_ = qlen < tlen? qlen : tlen;
-	n_col_ = ((n_col_ < w + 1? n_col_ : w + 1) + 15) / 16 + 1;
-	qlen_ = (qlen + 15) / 16;
+	n_col_ = ((n_col_ < w + 1? n_col_ : w + 1) + SIMD_WIDTH - 1) / SIMD_WIDTH + 1;
+	qlen_ = (qlen + SIMD_WIDTH - 1) / SIMD_WIDTH;
 	for (t = 1, max_sc = mat[0], min_sc = mat[1]; t < m * m; ++t) {
 		max_sc = max_sc > mat[t]? max_sc : mat[t];
 		min_sc = min_sc < mat[t]? min_sc : mat[t];
@@ -96,23 +170,23 @@ void ksw_extd2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 		++long_thres;
 	long_diff = long_thres * (e - e2) - (q2 - q) - e2;
 
-	mem = (uint8_t*)kcalloc(km, tlen_ * 8 + qlen_ + 1, 16);
-	u = (__m128i*)(((size_t)mem + 15) >> 4 << 4); // 16-byte aligned
+	mem = (uint8_t*)kcalloc(km, tlen_ * 8 + qlen_ + 1, SIMD_WIDTH);
+	u = (SIMD_INT*)(((size_t)mem + SIMD_WIDTH - 1) >> SIMD_SHIFT << SIMD_SHIFT); // 16-byte aligned
 	v = u + tlen_, x = v + tlen_, y = x + tlen_, x2 = y + tlen_, y2 = x2 + tlen_;
-	s = y2 + tlen_, sf = (uint8_t*)(s + tlen_), qr = sf + tlen_ * 16;
-	memset(u,  -q  - e,  tlen_ * 16);
-	memset(v,  -q  - e,  tlen_ * 16);
-	memset(x,  -q  - e,  tlen_ * 16);
-	memset(y,  -q  - e,  tlen_ * 16);
-	memset(x2, -q2 - e2, tlen_ * 16);
-	memset(y2, -q2 - e2, tlen_ * 16);
+	s = y2 + tlen_, sf = (uint8_t*)(s + tlen_), qr = sf + tlen_ * SIMD_WIDTH;
+	memset(u,  -q  - e,  tlen_ * SIMD_WIDTH);
+	memset(v,  -q  - e,  tlen_ * SIMD_WIDTH);
+	memset(x,  -q  - e,  tlen_ * SIMD_WIDTH);
+	memset(y,  -q  - e,  tlen_ * SIMD_WIDTH);
+	memset(x2, -q2 - e2, tlen_ * SIMD_WIDTH);
+	memset(y2, -q2 - e2, tlen_ * SIMD_WIDTH);
 	if (!approx_max) {
-		H = (int32_t*)kmalloc(km, tlen_ * 16 * 4);
-		for (t = 0; t < tlen_ * 16; ++t) H[t] = KSW_NEG_INF;
+		H = (int32_t*)kmalloc(km, tlen_ * SIMD_WIDTH * 4);
+		for (t = 0; t < tlen_ * SIMD_WIDTH; ++t) H[t] = KSW_NEG_INF;
 	}
 	if (with_cigar) {
-		mem2 = (uint8_t*)kmalloc(km, ((size_t)(qlen + tlen - 1) * n_col_ + 1) * 16);
-		p = (__m128i*)(((size_t)mem2 + 15) >> 4 << 4);
+		mem2 = (uint8_t*)kmalloc(km, ((size_t)(qlen + tlen - 1) * n_col_ + 1) * SIMD_WIDTH);
+		p = (SIMD_INT*)(((size_t)mem2 + SIMD_WIDTH - 1) >> SIMD_SHIFT << SIMD_SHIFT);
 		off = (int*)kmalloc(km, (qlen + tlen - 1) * sizeof(int) * 2);
 		off_end = off + qlen + tlen - 1;
 	}
@@ -125,7 +199,7 @@ void ksw_extd2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 		int8_t x1, x21, v1;
 		uint8_t *qrr = qr + (qlen - 1 - r);
 		int8_t *u8 = (int8_t*)u, *v8 = (int8_t*)v, *x8 = (int8_t*)x, *x28 = (int8_t*)x2;
-		__m128i x1_, x21_, v1_;
+		SIMD_INT x1_, x21_, v1_;
 		// find the boundaries
 		if (st < r - qlen + 1) st = r - qlen + 1;
 		if (en > r) en = r;
@@ -136,7 +210,7 @@ void ksw_extd2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 			break;
 		}
 		st0 = st, en0 = en;
-		st = st / 16 * 16, en = (en + 16) / 16 * 16 - 1;
+		st = st / SIMD_WIDTH * SIMD_WIDTH, en = (en + SIMD_WIDTH) / SIMD_WIDTH * SIMD_WIDTH - 1;
 		// set boundary conditions
 		if (st > 0) {
 			if (st - 1 >= last_st && st - 1 <= last_en) {
@@ -155,47 +229,53 @@ void ksw_extd2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 		}
 		// loop fission: set scores first
 		if (!(flag & KSW_EZ_GENERIC_SC)) {
-			for (t = st0; t <= en0; t += 16) {
-				__m128i sq, st, tmp, mask;
-				sq = _mm_loadu_si128((__m128i*)&sf[t]);
-				st = _mm_loadu_si128((__m128i*)&qrr[t]);
-				mask = _mm_or_si128(_mm_cmpeq_epi8(sq, m1_), _mm_cmpeq_epi8(st, m1_));
-				tmp = _mm_cmpeq_epi8(sq, st);
-#ifdef __SSE4_1__
-				tmp = _mm_blendv_epi8(sc_mis_, sc_mch_, tmp);
-				tmp = _mm_blendv_epi8(tmp,     sc_N_,   mask);
-#else
+			for (t = st0; t <= en0; t += SIMD_WIDTH) {
+				SIMD_INT sq, st, tmp;
+				sq = simd_funcw(loadu)((SIMD_INT*)&sf[t]);
+				st = simd_funcw(loadu)((SIMD_INT*)&qrr[t]);
+#if defined(__AVX512BW__)
+				__mmask64 mask = _mm512_cmpeq_epi8_mask(sq, m1_) | _mm512_cmpeq_epi8_mask(st, m1_);
+				tmp = _mm512_mask_blend_epi8(_mm512_cmpeq_epi8_mask(sq, st), sc_mis_, sc_mch_);
+				tmp = _mm512_mask_blend_epi8(mask, tmp, sc_N_);
+#elif defined(__SSE4_1__) || defined(__AVX2__)
+				SIMD_INT mask = simd_funcw(or)(simd_func(cmpeq_epi8)(sq, m1_), simd_func(cmpeq_epi8)(st, m1_));
+				tmp = simd_func(cmpeq_epi8)(sq, st);
+				tmp = simd_func(blendv_epi8)(sc_mis_, sc_mch_, tmp);
+				tmp = simd_func(blendv_epi8)(tmp,     sc_N_,   mask);
+#elif defined(__SSE2__) // emulate blendv
+				SIMD_INT mask = simd_funcw(or)(simd_func(cmpeq_epi8)(sq, m1_), simd_func(cmpeq_epi8)(st, m1_));
+				tmp = simd_func(cmpeq_epi8)(sq, st);
 				tmp = _mm_or_si128(_mm_andnot_si128(tmp,  sc_mis_), _mm_and_si128(tmp,  sc_mch_));
 				tmp = _mm_or_si128(_mm_andnot_si128(mask, tmp),     _mm_and_si128(mask, sc_N_));
 #endif
-				_mm_storeu_si128((__m128i*)((int8_t*)s + t), tmp);
+				simd_funcw(storeu)((SIMD_INT*)((int8_t*)s + t), tmp);
 			}
 		} else {
 			for (t = st0; t <= en0; ++t)
 				((uint8_t*)s)[t] = mat[sf[t] * m + qrr[t]];
 		}
 		// core loop
-		x1_  = _mm_cvtsi32_si128((uint8_t)x1);
-		x21_ = _mm_cvtsi32_si128((uint8_t)x21);
-		v1_  = _mm_cvtsi32_si128((uint8_t)v1);
-		st_ = st / 16, en_ = en / 16;
+		x1_  = simd_funcw(and)(simd_func(set1_epi8)((uint8_t)x1),  mask1_);
+		x21_ = simd_funcw(and)(simd_func(set1_epi8)((uint8_t)x21), mask1_);
+		v1_  = simd_funcw(and)(simd_func(set1_epi8)((uint8_t)v1),  mask1_);
+		st_ = st / SIMD_WIDTH, en_ = en / SIMD_WIDTH;
 		assert(en_ - st_ + 1 <= n_col_);
 		if (!with_cigar) { // score only
 			for (t = st_; t <= en_; ++t) {
-				__m128i z, a, b, a2, b2, xt1, x2t1, vt1, ut, tmp;
+				SIMD_INT z, a, b, a2, b2, xt1, x2t1, vt1, ut, tmp;
 				__dp_code_block1;
-#ifdef __SSE4_1__
-				z = _mm_max_epi8(z, a);
-				z = _mm_max_epi8(z, b);
-				z = _mm_max_epi8(z, a2);
-				z = _mm_max_epi8(z, b2);
-				z = _mm_min_epi8(z, sc_mch_);
+#if defined(__SSE4_1__) || defined(__AVX2__) || defined(__AVX512BW__)
+				z = simd_func(max_epi8)(z, a);
+				z = simd_func(max_epi8)(z, b);
+				z = simd_func(max_epi8)(z, a2);
+				z = simd_func(max_epi8)(z, b2);
+				z = simd_func(min_epi8)(z, sc_mch_);
 				__dp_code_block2; // save u[] and v[]; update a, b, a2 and b2
-				_mm_store_si128(&x[t],  _mm_sub_epi8(_mm_max_epi8(a,  zero_), qe_));
-				_mm_store_si128(&y[t],  _mm_sub_epi8(_mm_max_epi8(b,  zero_), qe_));
-				_mm_store_si128(&x2[t], _mm_sub_epi8(_mm_max_epi8(a2, zero_), qe2_));
-				_mm_store_si128(&y2[t], _mm_sub_epi8(_mm_max_epi8(b2, zero_), qe2_));
-#else
+				simd_funcw(store)(&x[t],  simd_func(sub_epi8)(simd_func(max_epi8)(a,  zero_), qe_));
+				simd_funcw(store)(&y[t],  simd_func(sub_epi8)(simd_func(max_epi8)(b,  zero_), qe_));
+				simd_funcw(store)(&x2[t], simd_func(sub_epi8)(simd_func(max_epi8)(a2, zero_), qe2_));
+				simd_funcw(store)(&y2[t], simd_func(sub_epi8)(simd_func(max_epi8)(b2, zero_), qe2_));
+#elif defined(__SSE2__)
 				tmp = _mm_cmpgt_epi8(a,  z);
 				z = _mm_or_si128(_mm_andnot_si128(tmp, z), _mm_and_si128(tmp, a));
 				tmp = _mm_cmpgt_epi8(b,  z);
@@ -218,22 +298,42 @@ void ksw_extd2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 #endif
 			}
 		} else if (!(flag&KSW_EZ_RIGHT)) { // gap left-alignment
-			__m128i *pr = p + (size_t)r * n_col_ - st_;
+			SIMD_INT *pr = p + (size_t)r * n_col_ - st_;
 			off[r] = st, off_end[r] = en;
 			for (t = st_; t <= en_; ++t) {
-				__m128i d, z, a, b, a2, b2, xt1, x2t1, vt1, ut, tmp;
+				SIMD_INT d, z, a, b, a2, b2, xt1, x2t1, vt1, ut, tmp;
 				__dp_code_block1;
-#ifdef __SSE4_1__
-				d = _mm_and_si128(_mm_cmpgt_epi8(a, z), _mm_set1_epi8(1));       // d = a  > z? 1 : 0
-				z = _mm_max_epi8(z, a);
-				d = _mm_blendv_epi8(d, _mm_set1_epi8(2), _mm_cmpgt_epi8(b,  z)); // d = b  > z? 2 : d
-				z = _mm_max_epi8(z, b);
-				d = _mm_blendv_epi8(d, _mm_set1_epi8(3), _mm_cmpgt_epi8(a2, z)); // d = a2 > z? 3 : d
-				z = _mm_max_epi8(z, a2);
-				d = _mm_blendv_epi8(d, _mm_set1_epi8(4), _mm_cmpgt_epi8(b2, z)); // d = a2 > z? 3 : d
-				z = _mm_max_epi8(z, b2);
-				z = _mm_min_epi8(z, sc_mch_);
-#else // we need to emulate SSE4.1 intrinsics _mm_max_epi8() and _mm_blendv_epi8()
+#if defined(__AVX512BW__)
+				d = _mm512_maskz_set1_epi8(_mm512_cmpgt_epi8_mask(a,  z), 1);
+				z = _mm512_max_epi8(z, a);
+				d = _mm512_mask_blend_epi8(_mm512_cmpgt_epi8_mask(b,  z), d, _mm512_set1_epi8(2));
+				z = _mm512_max_epi8(z, b);
+				d = _mm512_mask_blend_epi8(_mm512_cmpgt_epi8_mask(a2, z), d, _mm512_set1_epi8(3));
+				z = _mm512_max_epi8(z, a2);
+				d = _mm512_mask_blend_epi8(_mm512_cmpgt_epi8_mask(b2, z), d, _mm512_set1_epi8(4));
+				z = _mm512_max_epi8(z, b2);
+				z = _mm512_min_epi8(z, sc_mch_);
+				__dp_code_block2;
+				d = _mm512_or_si512(d, _mm512_maskz_set1_epi8(_mm512_cmpgt_epi8_mask(a,  zero_), 0x08)); // d = a  > 0? 1<<3 : 0
+				_mm512_store_si512(&x[t],  _mm512_sub_epi8(_mm512_max_epi8(a,  zero_), qe_));
+				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_), 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_), qe2_));
+#else
+#if defined(__SSE4_1__) || defined(__AVX2__)
+				d = simd_funcw(and)(simd_func(cmpgt_epi8)(a, z), simd_func(set1_epi8)(1));       // d = a  > z? 1 : 0
+				z = simd_func(max_epi8)(z, a);
+				d = simd_func(blendv_epi8)(d, simd_func(set1_epi8)(2), simd_func(cmpgt_epi8)(b,  z)); // d = b  > z? 2 : d
+				z = simd_func(max_epi8)(z, b);
+				d = simd_func(blendv_epi8)(d, simd_func(set1_epi8)(3), simd_func(cmpgt_epi8)(a2, z)); // d = a2 > z? 3 : d
+				z = simd_func(max_epi8)(z, a2);
+				d = simd_func(blendv_epi8)(d, simd_func(set1_epi8)(4), simd_func(cmpgt_epi8)(b2, z)); // d = a2 > z? 3 : d
+				z = simd_func(max_epi8)(z, b2);
+				z = simd_func(min_epi8)(z, sc_mch_);
+#elif defined(__SSE2__) // emulate SSE4.1 intrinsics _mm_max_epi8() and _mm_blendv_epi8()
 				tmp = _mm_cmpgt_epi8(a,  z);
 				d = _mm_and_si128(tmp, _mm_set1_epi8(1));
 				z = _mm_or_si128(_mm_andnot_si128(tmp, z), _mm_and_si128(tmp, a));
@@ -248,39 +348,60 @@ void ksw_extd2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 				z = _mm_or_si128(_mm_andnot_si128(tmp, z), _mm_and_si128(tmp, b2));
 				tmp = _mm_cmplt_epi8(sc_mch_, z);
 				z = _mm_or_si128(_mm_and_si128(tmp, sc_mch_), _mm_andnot_si128(tmp, z));
-#endif
+#endif // ~__SSE2__
 				__dp_code_block2;
-				tmp = _mm_cmpgt_epi8(a, zero_);
-				_mm_store_si128(&x[t],  _mm_sub_epi8(_mm_and_si128(tmp, a),  qe_));
-				d = _mm_or_si128(d, _mm_and_si128(tmp, _mm_set1_epi8(0x08))); // d = a > 0? 1<<3 : 0
-				tmp = _mm_cmpgt_epi8(b, zero_);
-				_mm_store_si128(&y[t],  _mm_sub_epi8(_mm_and_si128(tmp, b),  qe_));
-				d = _mm_or_si128(d, _mm_and_si128(tmp, _mm_set1_epi8(0x10))); // d = b > 0? 1<<4 : 0
-				tmp = _mm_cmpgt_epi8(a2, zero_);
-				_mm_store_si128(&x2[t], _mm_sub_epi8(_mm_and_si128(tmp, a2), qe2_));
-				d = _mm_or_si128(d, _mm_and_si128(tmp, _mm_set1_epi8(0x20))); // d = a > 0? 1<<5 : 0
-				tmp = _mm_cmpgt_epi8(b2, zero_);
-				_mm_store_si128(&y2[t], _mm_sub_epi8(_mm_and_si128(tmp, b2), qe2_));
-				d = _mm_or_si128(d, _mm_and_si128(tmp, _mm_set1_epi8(0x40))); // d = b > 0? 1<<6 : 0
-				_mm_store_si128(&pr[t], d);
+				tmp = simd_func(cmpgt_epi8)(a, zero_);
+				simd_funcw(store)(&x[t],  simd_func(sub_epi8)(simd_funcw(and)(tmp, a),  qe_));
+				d = simd_funcw(or)(d, simd_funcw(and)(tmp, simd_func(set1_epi8)(0x08))); // d = a > 0? 1<<3 : 0
+				tmp = simd_func(cmpgt_epi8)(b, zero_);
+				simd_funcw(store)(&y[t],  simd_func(sub_epi8)(simd_funcw(and)(tmp, b),  qe_));
+				d = simd_funcw(or)(d, simd_funcw(and)(tmp, simd_func(set1_epi8)(0x10))); // d = b > 0? 1<<4 : 0
+				tmp = simd_func(cmpgt_epi8)(a2, zero_);
+				simd_funcw(store)(&x2[t], simd_func(sub_epi8)(simd_funcw(and)(tmp, a2), qe2_));
+				d = simd_funcw(or)(d, simd_funcw(and)(tmp, simd_func(set1_epi8)(0x20))); // d = a > 0? 1<<5 : 0
+				tmp = simd_func(cmpgt_epi8)(b2, zero_);
+				simd_funcw(store)(&y2[t], simd_func(sub_epi8)(simd_funcw(and)(tmp, b2), qe2_));
+				d = simd_funcw(or)(d, simd_funcw(and)(tmp, simd_func(set1_epi8)(0x40))); // d = b > 0? 1<<6 : 0
+#endif // ~__AVX512BW__
+				simd_funcw(store)(&pr[t], d);
 			}
 		} else { // gap right-alignment
-			__m128i *pr = p + (size_t)r * n_col_ - st_;
+			SIMD_INT *pr = p + (size_t)r * n_col_ - st_;
 			off[r] = st, off_end[r] = en;
 			for (t = st_; t <= en_; ++t) {
-				__m128i d, z, a, b, a2, b2, xt1, x2t1, vt1, ut, tmp;
+				SIMD_INT d, z, a, b, a2, b2, xt1, x2t1, vt1, ut, tmp;
 				__dp_code_block1;
-#ifdef __SSE4_1__
-				d = _mm_andnot_si128(_mm_cmpgt_epi8(z, a), _mm_set1_epi8(1));    // d = z > a?  0 : 1
-				z = _mm_max_epi8(z, a);
-				d = _mm_blendv_epi8(_mm_set1_epi8(2), d, _mm_cmpgt_epi8(z, b));  // d = z > b?  d : 2
-				z = _mm_max_epi8(z, b);
-				d = _mm_blendv_epi8(_mm_set1_epi8(3), d, _mm_cmpgt_epi8(z, a2)); // d = z > a2? d : 3
-				z = _mm_max_epi8(z, a2);
-				d = _mm_blendv_epi8(_mm_set1_epi8(4), d, _mm_cmpgt_epi8(z, b2)); // d = z > b2? d : 4
-				z = _mm_max_epi8(z, b2);
-				z = _mm_min_epi8(z, sc_mch_);
-#else // we need to emulate SSE4.1 intrinsics _mm_max_epi8() and _mm_blendv_epi8()
+#if defined(__AVX512BW__)
+				d = _mm512_maskz_set1_epi8(_mm512_cmpge_epi8_mask(a,  z), 1);
+				z = _mm512_max_epi8(z, a);
+				d = _mm512_mask_blend_epi8(_mm512_cmpge_epi8_mask(b,  z), d, _mm512_set1_epi8(2));
+				z = _mm512_max_epi8(z, b);
+				d = _mm512_mask_blend_epi8(_mm512_cmpge_epi8_mask(a2, z), d, _mm512_set1_epi8(3));
+				z = _mm512_max_epi8(z, a2);
+				d = _mm512_mask_blend_epi8(_mm512_cmpge_epi8_mask(b2, z), d, _mm512_set1_epi8(4));
+				z = _mm512_max_epi8(z, b2);
+				z = _mm512_min_epi8(z, sc_mch_);
+				__dp_code_block2;
+				d = _mm512_or_si512(d, _mm512_maskz_set1_epi8(_mm512_cmpge_epi8_mask(a,  zero_), 0x08)); // d = a  >= 0? 1<<3 : 0
+				_mm512_store_si512(&x[t],  _mm512_sub_epi8(_mm512_max_epi8(a,  zero_), qe_));
+				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_), 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_), qe2_));
+#else
+#if defined(__SSE4_1__) || defined(__AVX2__)
+				d = simd_funcw(andnot)(simd_func(cmpgt_epi8)(z, a), simd_func(set1_epi8)(1));    // d = z > a?  0 : 1
+				z = simd_func(max_epi8)(z, a);
+				d = simd_func(blendv_epi8)(simd_func(set1_epi8)(2), d, simd_func(cmpgt_epi8)(z, b));  // d = z > b?  d : 2
+				z = simd_func(max_epi8)(z, b);
+				d = simd_func(blendv_epi8)(simd_func(set1_epi8)(3), d, simd_func(cmpgt_epi8)(z, a2)); // d = z > a2? d : 3
+				z = simd_func(max_epi8)(z, a2);
+				d = simd_func(blendv_epi8)(simd_func(set1_epi8)(4), d, simd_func(cmpgt_epi8)(z, b2)); // d = z > b2? d : 4
+				z = simd_func(max_epi8)(z, b2);
+				z = simd_func(min_epi8)(z, sc_mch_);
+#elif defined(__SSE2__)
 				tmp = _mm_cmpgt_epi8(z, a);
 				d = _mm_andnot_si128(tmp, _mm_set1_epi8(1));
 				z = _mm_or_si128(_mm_and_si128(tmp, z), _mm_andnot_si128(tmp, a));
@@ -295,52 +416,64 @@ void ksw_extd2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 				z = _mm_or_si128(_mm_and_si128(tmp, z), _mm_andnot_si128(tmp, b2));
 				tmp = _mm_cmplt_epi8(sc_mch_, z);
 				z = _mm_or_si128(_mm_and_si128(tmp, sc_mch_), _mm_andnot_si128(tmp, z));
-#endif
+#endif // ~__SSE2__
 				__dp_code_block2;
-				tmp = _mm_cmpgt_epi8(zero_, a);
-				_mm_store_si128(&x[t],  _mm_sub_epi8(_mm_andnot_si128(tmp, a),  qe_));
-				d = _mm_or_si128(d, _mm_andnot_si128(tmp, _mm_set1_epi8(0x08))); // d = a > 0? 1<<3 : 0
-				tmp = _mm_cmpgt_epi8(zero_, b);
-				_mm_store_si128(&y[t],  _mm_sub_epi8(_mm_andnot_si128(tmp, b),  qe_));
-				d = _mm_or_si128(d, _mm_andnot_si128(tmp, _mm_set1_epi8(0x10))); // d = b > 0? 1<<4 : 0
-				tmp = _mm_cmpgt_epi8(zero_, a2);
-				_mm_store_si128(&x2[t], _mm_sub_epi8(_mm_andnot_si128(tmp, a2), qe2_));
-				d = _mm_or_si128(d, _mm_andnot_si128(tmp, _mm_set1_epi8(0x20))); // d = a > 0? 1<<5 : 0
-				tmp = _mm_cmpgt_epi8(zero_, b2);
-				_mm_store_si128(&y2[t], _mm_sub_epi8(_mm_andnot_si128(tmp, b2), qe2_));
-				d = _mm_or_si128(d, _mm_andnot_si128(tmp, _mm_set1_epi8(0x40))); // d = b > 0? 1<<6 : 0
-				_mm_store_si128(&pr[t], d);
+				tmp = simd_func(cmpgt_epi8)(zero_, a);
+				simd_funcw(store)(&x[t],  simd_func(sub_epi8)(simd_funcw(andnot)(tmp, a),  qe_));
+				d = simd_funcw(or)(d, simd_funcw(andnot)(tmp, simd_func(set1_epi8)(0x08))); // d = a > 0? 1<<3 : 0
+				tmp = simd_func(cmpgt_epi8)(zero_, b);
+				simd_funcw(store)(&y[t],  simd_func(sub_epi8)(simd_funcw(andnot)(tmp, b),  qe_));
+				d = simd_funcw(or)(d, simd_funcw(andnot)(tmp, simd_func(set1_epi8)(0x10))); // d = b > 0? 1<<4 : 0
+				tmp = simd_func(cmpgt_epi8)(zero_, a2);
+				simd_funcw(store)(&x2[t], simd_func(sub_epi8)(simd_funcw(andnot)(tmp, a2), qe2_));
+				d = simd_funcw(or)(d, simd_funcw(andnot)(tmp, simd_func(set1_epi8)(0x20))); // d = a > 0? 1<<5 : 0
+				tmp = simd_func(cmpgt_epi8)(zero_, b2);
+				simd_funcw(store)(&y2[t], simd_func(sub_epi8)(simd_funcw(andnot)(tmp, b2), qe2_));
+				d = simd_funcw(or)(d, simd_funcw(andnot)(tmp, simd_func(set1_epi8)(0x40))); // d = b > 0? 1<<6 : 0
+#endif // ~__AVX512BW__
+				simd_funcw(store)(&pr[t], d);
 			}
 		}
 		if (!approx_max) { // find the exact max with a 32-bit score array
 			int32_t max_H, max_t;
 			// compute H[], max_H and max_t
 			if (r > 0) {
-				int32_t HH[4], tt[4], en1 = st0 + (en0 - st0) / 4 * 4, i;
-				__m128i max_H_, max_t_;
+				int32_t HH[SIMD_WIDTH/4], tt[SIMD_WIDTH/4], en1 = st0 + (en0 - st0) / (SIMD_WIDTH/4) * (SIMD_WIDTH/4), i;
+				SIMD_INT max_H_, max_t_;
 				max_H = H[en0] = en0 > 0? H[en0-1] + u8[en0] : H[en0] + v8[en0]; // special casing the last element
 				max_t = en0;
-				max_H_ = _mm_set1_epi32(max_H);
-				max_t_ = _mm_set1_epi32(max_t);
-				for (t = st0; t < en1; t += 4) { // this implements: H[t]+=v8[t]-qe; if(H[t]>max_H) max_H=H[t],max_t=t;
-					__m128i H1, tmp, t_;
-					H1 = _mm_loadu_si128((__m128i*)&H[t]);
+				max_H_ = simd_func(set1_epi32)(max_H);
+				max_t_ = simd_func(set1_epi32)(max_t);
+				for (t = st0; t < en1; t += SIMD_WIDTH/4) { // this implements: H[t]+=v8[t]; if(H[t]>max_H) max_H=H[t],max_t=t;
+					SIMD_INT H1, t_;
+					H1 = simd_funcw(loadu)((SIMD_INT*)&H[t]);
+#if defined(__AVX512BW__)
+					t_ = _mm512_cvtepi8_epi32(_mm_loadu_si128((__m128i*)&v8[t]));
+#elif defined(__AVX2__)
+					t_ = _mm256_setr_epi32(v8[t], v8[t+1], v8[t+2], v8[t+3], v8[t+4], v8[t+5], v8[t+6], v8[t+7]);
+#elif defined(__SSE2__)
 					t_ = _mm_setr_epi32(v8[t], v8[t+1], v8[t+2], v8[t+3]);
-					H1 = _mm_add_epi32(H1, t_);
-					_mm_storeu_si128((__m128i*)&H[t], H1);
-					t_ = _mm_set1_epi32(t);
-					tmp = _mm_cmpgt_epi32(H1, max_H_);
-#ifdef __SSE4_1__
-					max_H_ = _mm_blendv_epi8(max_H_, H1, tmp);
-					max_t_ = _mm_blendv_epi8(max_t_, t_, tmp);
-#else
-					max_H_ = _mm_or_si128(_mm_and_si128(tmp, H1), _mm_andnot_si128(tmp, max_H_));
-					max_t_ = _mm_or_si128(_mm_and_si128(tmp, t_), _mm_andnot_si128(tmp, max_t_));
+#endif
+					H1 = simd_func(add_epi32)(H1, t_);
+					simd_funcw(storeu)((SIMD_INT*)&H[t], H1);
+					t_ = simd_func(set1_epi32)(t);
+#if defined(__AVX512BW__)
+					__mmask64 tmp = _mm512_cmpgt_epi32_mask(H1, max_H_);
+					max_H_ = _mm512_mask_blend_epi32(tmp, max_H_, H1);
+					max_t_ = _mm512_mask_blend_epi32(tmp, max_t_, t_);
+#elif defined(__SSE4_1__) || defined(__AVX2__)
+					SIMD_INT tmp = simd_func(cmpgt_epi32)(H1, max_H_);
+					max_H_ = simd_func(blendv_epi8)(max_H_, H1, tmp);
+					max_t_ = simd_func(blendv_epi8)(max_t_, t_, tmp);
+#elif defined(__SSE2__)
+					SIMD_INT tmp = simd_func(cmpgt_epi32)(H1, max_H_);
+					max_H_ = simd_funcw(or)(simd_funcw(and)(tmp, H1), simd_funcw(andnot)(tmp, max_H_));
+					max_t_ = simd_funcw(or)(simd_funcw(and)(tmp, t_), simd_funcw(andnot)(tmp, max_t_));
 #endif
 				}
-				_mm_storeu_si128((__m128i*)HH, max_H_);
-				_mm_storeu_si128((__m128i*)tt, max_t_);
-				for (i = 0; i < 4; ++i)
+				simd_funcw(storeu)((SIMD_INT*)HH, max_H_);
+				simd_funcw(storeu)((SIMD_INT*)tt, max_t_);
+				for (i = 0; i < SIMD_WIDTH/4; ++i)
 					if (max_H < HH[i]) max_H = HH[i], max_t = tt[i] + i;
 				for (; t < en0; ++t) { // for the rest of values that haven't been computed with SSE
 					H[t] += (int32_t)v8[t];
@@ -381,12 +514,12 @@ void ksw_extd2_sse(void *km, int qlen, const uint8_t *query, int tlen, const uin
 	if (with_cigar) { // backtrack
 		int rev_cigar = !!(flag & KSW_EZ_REV_CIGAR);
 		if (!ez->zdropped && !(flag&KSW_EZ_EXTZ_ONLY)) {
-			ksw_backtrack(km, 1, rev_cigar, 0, (uint8_t*)p, off, off_end, n_col_*16, tlen-1, qlen-1, &ez->m_cigar, &ez->n_cigar, &ez->cigar);
+			ksw_backtrack(km, 1, rev_cigar, 0, (uint8_t*)p, off, off_end, n_col_*SIMD_WIDTH, tlen-1, qlen-1, &ez->m_cigar, &ez->n_cigar, &ez->cigar);
 		} else if (!ez->zdropped && (flag&KSW_EZ_EXTZ_ONLY) && ez->mqe + end_bonus > (int)ez->max) {
 			ez->reach_end = 1;
-			ksw_backtrack(km, 1, rev_cigar, 0, (uint8_t*)p, off, off_end, n_col_*16, ez->mqe_t, qlen-1, &ez->m_cigar, &ez->n_cigar, &ez->cigar);
+			ksw_backtrack(km, 1, rev_cigar, 0, (uint8_t*)p, off, off_end, n_col_*SIMD_WIDTH, ez->mqe_t, qlen-1, &ez->m_cigar, &ez->n_cigar, &ez->cigar);
 		} else if (ez->max_t >= 0 && ez->max_q >= 0) {
-			ksw_backtrack(km, 1, rev_cigar, 0, (uint8_t*)p, off, off_end, n_col_*16, ez->max_t, ez->max_q, &ez->m_cigar, &ez->n_cigar, &ez->cigar);
+			ksw_backtrack(km, 1, rev_cigar, 0, (uint8_t*)p, off, off_end, n_col_*SIMD_WIDTH, ez->max_t, ez->max_q, &ez->m_cigar, &ez->n_cigar, &ez->cigar);
 		}
 		kfree(km, mem2); kfree(km, off);
 	}

main.c

@@ -1,12 +1,13 @@
 #include <stdlib.h>
 #include <stdio.h>
 #include <string.h>
+#include <errno.h>
 #include "bseq.h"
 #include "minimap.h"
 #include "mmpriv.h"
 #include "ketopt.h"
 
-#define MM_VERSION "2.17-r941"
+#define MM_VERSION "2.17-r963-dirty"
 
 #ifdef __linux__
 #include <sys/resource.h>
@@ -172,7 +173,7 @@ int main(int argc, char *argv[])
 		else if (c == 'o') {
 			if (strcmp(o.arg, "-") != 0) {
 				if (freopen(o.arg, "wb", stdout) == NULL) {
-					fprintf(stderr, "[ERROR]\033[1;31m failed to write the output to file '%s'\033[0m\n", o.arg);
+					fprintf(stderr, "[ERROR]\033[1;31m failed to write the output to file '%s'\033[0m: %s\n", o.arg, strerror(errno));
 					exit(1);
 				}
 			}
@@ -208,6 +209,7 @@ int main(int argc, char *argv[])
 		else if (c == 337) opt.max_sw_mat = mm_parse_num(o.arg); // --cap-sw-mat
 		else if (c == 338) opt.max_qlen = mm_parse_num(o.arg); // --max-qlen
 		else if (c == 340) junc_bed = o.arg; // --junc-bed
+		else if (c == 341) opt.junc_bonus = atoi(o.arg); // --junc-bonus
 		else if (c == 342) opt.flag |= MM_F_SAM_HIT_ONLY; // --sam-hit-only
 		else if (c == 314) { // --frag
 			yes_or_no(&opt, MM_F_FRAG_MODE, o.longidx, o.arg, 1);
@@ -322,11 +324,11 @@ int main(int argc, char *argv[])
 		fprintf(fp_help, "    --version    show version number\n");
 		fprintf(fp_help, "  Preset:\n");
 		fprintf(fp_help, "    -x STR       preset (always applied before other options; see minimap2.1 for details) []\n");
-		fprintf(fp_help, "                 - map-pb/map-ont: PacBio/Nanopore vs reference mapping\n");
-		fprintf(fp_help, "                 - ava-pb/ava-ont: PacBio/Nanopore read overlap\n");
-		fprintf(fp_help, "                 - asm5/asm10/asm20: asm-to-ref mapping, for ~0.1/1/5%% sequence divergence\n");
-		fprintf(fp_help, "                 - splice: long-read spliced alignment\n");
-		fprintf(fp_help, "                 - sr: genomic short-read mapping\n");
+		fprintf(fp_help, "                 - map-pb/map-ont - PacBio/Nanopore vs reference mapping\n");
+		fprintf(fp_help, "                 - ava-pb/ava-ont - PacBio/Nanopore read overlap\n");
+		fprintf(fp_help, "                 - asm5/asm10/asm20 - asm-to-ref mapping, for ~0.1/1/5%% sequence divergence\n");
+		fprintf(fp_help, "                 - splice/splice:hq - long-read/Pacbio-CCS spliced alignment\n");
+		fprintf(fp_help, "                 - sr - genomic short-read mapping\n");
 		fprintf(fp_help, "\nSee `man ./minimap2.1' for detailed description of these and other advanced command-line options.\n");
 		return fp_help == stdout? 0 : 1;
 	}
@@ -337,7 +339,7 @@ int main(int argc, char *argv[])
 	}
 	idx_rdr = mm_idx_reader_open(argv[o.ind], &ipt, fnw);
 	if (idx_rdr == 0) {
-		fprintf(stderr, "[ERROR] failed to open file '%s'\n", argv[o.ind]);
+		fprintf(stderr, "[ERROR] failed to open file '%s': %s\n", argv[o.ind], strerror(errno));
 		return 1;
 	}
 	if (!idx_rdr->is_idx && fnw == 0 && argc - o.ind < 2) {
@@ -355,13 +357,19 @@ int main(int argc, char *argv[])
 			return 1;
 		}
 		if ((opt.flag & MM_F_OUT_SAM) && idx_rdr->n_parts == 1) {
+			int ret;
 			if (mm_idx_reader_eof(idx_rdr)) {
-				mm_write_sam_hdr(mi, rg, MM_VERSION, argc, argv);
+				ret = mm_write_sam_hdr(mi, rg, MM_VERSION, argc, argv);
 			} else {
-				mm_write_sam_hdr(0, rg, MM_VERSION, argc, argv);
+				ret = mm_write_sam_hdr(0, rg, MM_VERSION, argc, argv);
 				if (opt.split_prefix == 0 && mm_verbose >= 2)
 					fprintf(stderr, "[WARNING]\033[1;31m For a multi-part index, no @SQ lines will be outputted. Please use --split-prefix.\033[0m\n");
 			}
+			if (ret != 0) {
+				mm_idx_destroy(mi);
+				mm_idx_reader_close(idx_rdr);
+				return 1;
+			}
 		}
 		if (mm_verbose >= 3)
 			fprintf(stderr, "[M::%s::%.3f*%.2f] loaded/built the index for %d target sequence(s)\n",
@@ -384,7 +392,7 @@ int main(int argc, char *argv[])
 		mm_split_merge(argc - (o.ind + 1), (const char**)&argv[o.ind + 1], &opt, n_parts);
 
 	if (fflush(stdout) == EOF) {
-		fprintf(stderr, "[ERROR] failed to write the results\n");
+		perror("[ERROR] failed to write the results");
 		exit(EXIT_FAILURE);
 	}
 

map.c

@@ -1,6 +1,7 @@
 #include <stdlib.h>
 #include <string.h>
 #include <assert.h>
+#include <errno.h>
 #include "kthread.h"
 #include "kvec.h"
 #include "kalloc.h"
@@ -622,7 +623,7 @@ static mm_bseq_file_t **open_bseqs(int n, const char **fn)
 	for (i = 0; i < n; ++i) {
 		if ((fp[i] = mm_bseq_open(fn[i])) == 0) {
 			if (mm_verbose >= 1)
-				fprintf(stderr, "ERROR: failed to open file '%s'\n", fn[i]);
+				fprintf(stderr, "ERROR: failed to open file '%s': %s\n", fn[i], strerror(errno));
 			for (j = 0; j < i; ++j)
 				mm_bseq_close(fp[j]);
 			free(fp);

minimap2.1

@@ -638,6 +638,7 @@ s2	i	Chaining score of the best secondary chain
 NM	i	Total number of mismatches and gaps in the alignment
 MD	Z	To generate the ref sequence in the alignment
 AS	i	DP alignment score
+SA	Z	List of other supplementary alignments
 ms	i	DP score of the max scoring segment in the alignment
 nn	i	Number of ambiguous bases in the alignment
 ts	A	Transcript strand (splice mode only)

misc.c

@@ -125,7 +125,7 @@ void mm_err_puts(const char *str)
 	int ret;
 	ret = puts(str);
 	if (ret == EOF) {
-		fprintf(stderr, "[ERROR] failed to write the results\n");
+		perror("[ERROR] failed to write the results");
 		exit(EXIT_FAILURE);
 	}
 }
@@ -135,7 +135,7 @@ void mm_err_fwrite(const void *p, size_t size, size_t nitems, FILE *fp)
 	int ret;
 	ret = fwrite(p, size, nitems, fp);
 	if (ret == EOF) {
-		fprintf(stderr, "[ERROR] failed to write data\n");
+		perror("[ERROR] failed to write data");
 		exit(EXIT_FAILURE);
 	}
 }
@@ -145,7 +145,7 @@ void mm_err_fread(void *p, size_t size, size_t nitems, FILE *fp)
 	int ret;
 	ret = fread(p, size, nitems, fp);
 	if (ret == EOF) {
-		fprintf(stderr, "[ERROR] failed to read data\n");
+		perror("[ERROR] failed to read data");
 		exit(EXIT_FAILURE);
 	}
 }

misc/paftools.js

@@ -1,6 +1,6 @@
 #!/usr/bin/env k8
 
-var paftools_version = '2.17-r941';
+var paftools_version = '2.17-r949-dirty';
 
 /*****************************
  ***** Library functions *****
@@ -1509,6 +1509,7 @@ function paf_gff2bed(args)
 
 	var colors = {
 		'protein_coding':'0,128,255',
+		'mRNA':'0,128,255',
 		'lincRNA':'0,192,0',
 		'snRNA':'0,192,0',
 		'miRNA':'0,192,0',
@@ -1541,8 +1542,8 @@ function paf_gff2bed(args)
 		print(a[0][0], st, en, name, 1000, a[0][3], cds_st, cds_en, color, a.length, sizes.join(",") + ",", starts.join(",") + ",");
 	}
 
-	var re_gtf = /(transcript_id|transcript_type|transcript_biotype|gene_name|transcript_name) "([^"]+)";/g;
-	var re_gff3 = /(transcript_id|transcript_type|transcript_biotype|gene_name|transcript_name)=([^;]+)/g;
+	var re_gtf = /\b(transcript_id|transcript_type|transcript_biotype|gene_name|gene_id|gbkey|transcript_name) "([^"]+)";/g;
+	var re_gff3 = /\b(transcript_id|transcript_type|transcript_biotype|gene_name|gene_id|gbkey|transcript_name)=([^;]+)/g;
 	var buf = new Bytes();
 	var file = args[getopt.ind] == '-'? new File() : new File(args[getopt.ind]);
 
@@ -1559,19 +1560,19 @@ function paf_gff2bed(args)
 		if (t[2] != "CDS" && t[2] != "exon") continue;
 		t[3] = parseInt(t[3]) - 1;
 		t[4] = parseInt(t[4]);
-		var id = null, type = "", gname = "N/A", biotype = "", m, tname = "N/A";
+		var id = null, type = "", name = "N/A", biotype = "", m, tname = "N/A";
 		while ((m = re_gtf.exec(t[8])) != null) {
 			if (m[1] == "transcript_id") id = m[2];
 			else if (m[1] == "transcript_type") type = m[2];
-			else if (m[1] == "transcript_biotype") biotype = m[2];
-			else if (m[1] == "gene_name") name = m[2];
+			else if (m[1] == "transcript_biotype" || m[1] == "gbkey") biotype = m[2];
+			else if (m[1] == "gene_name" || m[1] == "gene_id") name = m[2];
 			else if (m[1] == "transcript_name") tname = m[2];
 		}
 		while ((m = re_gff3.exec(t[8])) != null) {
 			if (m[1] == "transcript_id") id = m[2];
 			else if (m[1] == "transcript_type") type = m[2];
-			else if (m[1] == "transcript_biotype") biotype = m[2];
-			else if (m[1] == "gene_name") name = m[2];
+			else if (m[1] == "transcript_biotype" || m[1] == "gbkey") biotype = m[2];
+			else if (m[1] == "gene_name" || m[1] == "gene_id") name = m[2];
 			else if (m[1] == "transcript_name") tname = m[2];
 		}
 		if (type == "" && biotype != "") type = biotype;

mmpriv.h

@@ -59,7 +59,7 @@ uint32_t ks_ksmall_uint32_t(size_t n, uint32_t arr[], size_t kk);
 
 void mm_sketch(void *km, const char *str, int len, int w, int k, uint32_t rid, int is_hpc, mm128_v *p);
 
-void mm_write_sam_hdr(const mm_idx_t *mi, const char *rg, const char *ver, int argc, char *argv[]);
+int mm_write_sam_hdr(const mm_idx_t *mi, const char *rg, const char *ver, int argc, char *argv[]);
 void mm_write_paf(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r, void *km, int opt_flag);
 void mm_write_paf3(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r, void *km, int opt_flag, int rep_len);
 void mm_write_sam(kstring_t *s, const mm_idx_t *mi, const mm_bseq1_t *t, const mm_reg1_t *r, int n_regs, const mm_reg1_t *regs);

python/mappy.pyx

@@ -113,6 +113,7 @@ cdef class Aligner:
 	cdef cmappy.mm_mapopt_t map_opt
 
 	def __cinit__(self, fn_idx_in=None, preset=None, k=None, w=None, min_cnt=None, min_chain_score=None, min_dp_score=None, bw=None, best_n=None, n_threads=3, fn_idx_out=None, max_frag_len=None, extra_flags=None, seq=None, scoring=None):
+		self._idx = NULL
 		cmappy.mm_set_opt(NULL, &self.idx_opt, &self.map_opt) # set the default options
 		if preset is not None:
 			cmappy.mm_set_opt(str.encode(preset), &self.idx_opt, &self.map_opt) # apply preset
@@ -170,6 +171,7 @@ cdef class Aligner:
 		cdef void *km
 		cdef cmappy.mm_mapopt_t map_opt
 
+		if self._idx == NULL: return
 		map_opt = self.map_opt
 		if max_frag_len is not None: map_opt.max_frag_len = max_frag_len
 		if extra_flags is not None: map_opt.flag |= extra_flags
@@ -186,27 +188,36 @@ cdef class Aligner:
 			_seq2 = seq2 if isinstance(seq2, bytes) else seq2.encode()
 			regs = cmappy.mm_map_aux(self._idx, _seq, _seq2, &n_regs, b._b, &map_opt)
 
-		for i in range(n_regs):
-			cmappy.mm_reg2hitpy(self._idx, &regs[i], &h)
-			cigar, _cs, _MD = [], '', ''
-			for k in range(h.n_cigar32): # convert the 32-bit CIGAR encoding to Python array
-				c = h.cigar32[k]
-				cigar.append([c>>4, c&0xf])
-			if cs or MD: # generate the cs and/or the MD tag, if requested
-				if cs:
-					l_cs_str = cmappy.mm_gen_cs(km, &cs_str, &m_cs_str, self._idx, &regs[i], _seq, 1)
-					_cs = cs_str[:l_cs_str] if isinstance(cs_str, str) else cs_str[:l_cs_str].decode()
-				if MD:
-					l_cs_str = cmappy.mm_gen_MD(km, &cs_str, &m_cs_str, self._idx, &regs[i], _seq)
-					_MD = cs_str[:l_cs_str] if isinstance(cs_str, str) else cs_str[:l_cs_str].decode()
-			yield Alignment(h.ctg, h.ctg_len, h.ctg_start, h.ctg_end, h.strand, h.qry_start, h.qry_end, h.mapq, cigar, h.is_primary, h.mlen, h.blen, h.NM, h.trans_strand, h.seg_id, _cs, _MD)
-			cmappy.mm_free_reg1(&regs[i])
-		free(regs)
-		free(cs_str)
+		try:
+			i = 0
+			while i < n_regs:
+				cmappy.mm_reg2hitpy(self._idx, &regs[i], &h)
+				cigar, _cs, _MD = [], '', ''
+				for k in range(h.n_cigar32): # convert the 32-bit CIGAR encoding to Python array
+					c = h.cigar32[k]
+					cigar.append([c>>4, c&0xf])
+				if cs or MD: # generate the cs and/or the MD tag, if requested
+					if cs:
+						l_cs_str = cmappy.mm_gen_cs(km, &cs_str, &m_cs_str, self._idx, &regs[i], _seq, 1)
+						_cs = cs_str[:l_cs_str] if isinstance(cs_str, str) else cs_str[:l_cs_str].decode()
+					if MD:
+						l_cs_str = cmappy.mm_gen_MD(km, &cs_str, &m_cs_str, self._idx, &regs[i], _seq)
+						_MD = cs_str[:l_cs_str] if isinstance(cs_str, str) else cs_str[:l_cs_str].decode()
+				yield Alignment(h.ctg, h.ctg_len, h.ctg_start, h.ctg_end, h.strand, h.qry_start, h.qry_end, h.mapq, cigar, h.is_primary, h.mlen, h.blen, h.NM, h.trans_strand, h.seg_id, _cs, _MD)
+				cmappy.mm_free_reg1(&regs[i])
+				i += 1
+		finally:
+			while i < n_regs:
+				cmappy.mm_free_reg1(&regs[i])
+				i += 1
+			free(regs)
+			free(cs_str)
 
 	def seq(self, str name, int start=0, int end=0x7fffffff):
 		cdef int l
-		cdef char *s = cmappy.mappy_fetch_seq(self._idx, name.encode(), start, end, &l)
+		cdef char *s
+		if self._idx == NULL: return
+		s = cmappy.mappy_fetch_seq(self._idx, name.encode(), start, end, &l)
 		if l == 0: return None
 		r = s[:l] if isinstance(s, str) else s[:l].decode()
 		free(s)
@@ -224,6 +235,7 @@ cdef class Aligner:
 	@property
 	def seq_names(self):
 		cdef char *p
+		if self._idx == NULL: return
 		sn = []
 		for i in range(self._idx.n_seq):
 			p = self._idx.seq[i].name

splitidx.c

@@ -2,6 +2,7 @@
 #include <assert.h>
 #include <stdlib.h>
 #include <stdio.h>
+#include <errno.h>
 #include "mmpriv.h"
 
 FILE *mm_split_init(const char *prefix, const mm_idx_t *mi)
@@ -13,15 +14,15 @@ FILE *mm_split_init(const char *prefix, const mm_idx_t *mi)
 	sprintf(fn, "%s.%.4d.tmp", prefix, mi->index);
 	if ((fp = fopen(fn, "wb")) == NULL) {
 		if (mm_verbose >= 1)
-			fprintf(stderr, "[ERROR]\033[1;31m failed to write to temporary file '%s'\033[0m\n", fn);
+			fprintf(stderr, "[ERROR]\033[1;31m failed to write to temporary file '%s'\033[0m: %s\n", fn, strerror(errno));
 		exit(1);
 	}
 	mm_err_fwrite(&k, 4, 1, fp);
 	mm_err_fwrite(&mi->n_seq, 4, 1, fp);
 	for (i = 0; i < mi->n_seq; ++i) {
-		uint8_t l;
+		uint32_t l;
 		l = strlen(mi->seq[i].name);
-		mm_err_fwrite(&l, 1, 1, fp);
+		mm_err_fwrite(&l, 1, 4, fp);
 		mm_err_fwrite(mi->seq[i].name, 1, l, fp);
 		mm_err_fwrite(&mi->seq[i].len, 4, 1, fp);
 	}
@@ -41,7 +42,7 @@ mm_idx_t *mm_split_merge_prep(const char *prefix, int n_splits, FILE **fp, uint3
 		sprintf(fn, "%s.%.4d.tmp", prefix, i);
 		if ((fp[i] = fopen(fn, "rb")) == 0) {
 			if (mm_verbose >= 1)
-				fprintf(stderr, "ERROR: failed to open temporary file '%s'\n", fn);
+				fprintf(stderr, "ERROR: failed to open temporary file '%s': %s\n", fn, strerror(errno));
 			for (j = 0; j < i; ++j)
 				fclose(fp[j]);
 			free(fn);
@@ -60,8 +61,8 @@ mm_idx_t *mm_split_merge_prep(const char *prefix, int n_splits, FILE **fp, uint3
 	for (i = j = 0; i < n_splits; ++i) {
 		uint32_t k;
 		for (k = 0; k < n_seq_part[i]; ++k, ++j) {
-			uint8_t l;
-			mm_err_fread(&l, 1, 1, fp[i]);
+			uint32_t l;
+			mm_err_fread(&l, 1, 4, fp[i]);
 			mi->seq[j].name = (char*)calloc(l + 1, 1);
 			mm_err_fread(mi->seq[j].name, 1, l, fp[i]);
 			mm_err_fread(&mi->seq[j].len, 4, 1, fp[i]);