From 3847654b17a3123a084f3a3dea375e00355846a8 Mon Sep 17 00:00:00 2001
From: Ilya Grebnov <Ilya.Grebnov@gmail.com>
Date: Tue, 18 Apr 2023 22:18:47 -0700
Subject: [PATCH] Version 2.7.2 Fixed out-of-bound memory access issue for
 large inputs.

---
 CHANGES                        |  3 ++
 Makefile                       | 54 ------------------------
 README.md                      |  2 +
 VERSION                        |  2 +-
 src/libsais.c => libsais.c     | 76 +++++++++++++++++-----------------
 src/libsais.h => libsais.h     |  4 +-
 src/libsais16.c => libsais16.c | 76 +++++++++++++++++-----------------
 src/libsais16.h => libsais16.h |  4 +-
 src/libsais64.c => libsais64.c | 76 +++++++++++++++++-----------------
 src/libsais64.h => libsais64.h |  4 +-
 10 files changed, 126 insertions(+), 175 deletions(-)
 delete mode 100644 Makefile
 rename src/libsais.c => libsais.c (99%)
 rename src/libsais.h => libsais.h (99%)
 rename src/libsais16.c => libsais16.c (99%)
 rename src/libsais16.h => libsais16.h (99%)
 rename src/libsais64.c => libsais64.c (99%)
 rename src/libsais64.h => libsais64.h (99%)

diff --git a/CHANGES b/CHANGES
index c41b978..ba816ad 100644
--- a/CHANGES
+++ b/CHANGES
@@ -1,3 +1,6 @@
+Changes in 2.7.2 (April 18, 2023)
+- Fixed out-of-bound memory access issue for large inputs (libsais64).
+
 Changes in 2.7.1 (June 19, 2022)
 - Improved cache coherence for ARMv8 architecture.
 
diff --git a/Makefile b/Makefile
deleted file mode 100644
index 82480b7..0000000
--- a/Makefile
+++ /dev/null
@@ -1,54 +0,0 @@
-PROJECT=sais
-PLIBNAME=lib$(PROJECT)
-PVER=2.7.1
-PSOVER=2
-ifeq ($(OS),Windows_NT)
-  PLIBSTATIC=$(PROJECT).a
-  PLIBSHARED=$(PROJECT)-$(PVER).dll
-else
-  PLIBSTATIC=$(PLIBNAME).a
-  PLIBSHARED=$(PLIBNAME).so.$(PSOVER)
-endif
-PLIBS=$(PLIBSTATIC) $(PLIBSHARED)
-CC=gcc
-CFLAGS?=-Wall -O2
-LDFLAGS?=-lm
-AR?=ar
-INSTALL?=install
-RM?=rm -f
-RMD?=$(RM) -r
-PREFIX?=/usr/local
-SRCS=src
-DOCS?=share/doc/$(LIBNAME)
-LIBS?=lib
-INCLUDES?=include
-MANS?=man/man1
-
-all: $(PLIBS)
-
-$(SRCS)/$(PLIBNAME).o: $(SRCS)/$(PLIBNAME).c
-	$(CC) $(CFLAGS) -c -o $@ $^
-
-$(PLIBSTATIC): $(SRCS)/$(PLIBNAME).o
-	$(AR) rcs $@ $^
-
-$(PLIBSHARED): $(SRCS)/$(PLIBNAME).o
-	$(CC) $(CFLAGS) -shared -Wl,-soname,$@ $^ -o $@
-
-clean:
-	$(RM) $(SRCS)/$(PLIBNAME).o $(PLIBS)
-
-install:
-	$(INSTALL) -d $(PREFIX)/$(LIBS)
-	$(INSTALL) -d $(PREFIX)/$(INCLUDES)
-	$(INSTALL) -d $(PREFIX)/$(MANS)
-	$(INSTALL) -d $(PREFIX)/$(DOCS)
-	$(INSTALL) -m 0644 $(PLIBS) $(PREFIX)/$(LIBS)
-	$(INSTALL) -m 0644 $(SRCS)/$(PLIBNAME).h $(PREFIX)/$(INCLUDES)
-	$(INSTALL) -m 0644 CHANGES LICENSE README.md VERSION $(PREFIX)/$(DOCS)
-
-uninstall:
-	$(RM) $(PREFIX)/$(LIBS)/$(PLIBSTATIC)
-	$(RM) $(PREFIX)/$(LIBS)/$(PLIBSHARED)
-	$(RM) $(PREFIX)/$(INCLUDES)/$(SRCS)/$(PLIBNAME).h
-	$(RMD) $(PREFIX)/$(DOCS)
diff --git a/README.md b/README.md
index c869a9c..83f993b 100644
--- a/README.md
+++ b/README.md
@@ -20,6 +20,8 @@ The libsais provides simple C99 API to construct suffix array and Burrows-Wheele
 The libsais is released under the [Apache License Version 2.0](LICENSE "Apache license")
 
 ## Changes
+* April 18, 2023 (2.7.2)
+  *  Fixed out-of-bound memory access issue for large inputs (libsais64).
 * June 19, 2022 (2.7.1)
   * Improved cache coherence for ARMv8 architecture.
 * April 12, 2022 (2.7.0)
diff --git a/VERSION b/VERSION
index 5588ae8..fbafd6b 100644
--- a/VERSION
+++ b/VERSION
@@ -1 +1 @@
-2.7.1
\ No newline at end of file
+2.7.2
\ No newline at end of file
diff --git a/src/libsais.c b/libsais.c
similarity index 99%
rename from src/libsais.c
rename to libsais.c
index fcbfe7e..6d4c0e1 100644
--- a/src/libsais.c
+++ b/libsais.c
@@ -1126,7 +1126,7 @@ static void libsais_count_and_gather_compacted_lms_suffixes_32s_2k_fs_omp(const
         else
         {
             fast_sint_t bucket_size       = 2 * (fast_sint_t)k;
-            fast_sint_t bucket_stride     = libsais_get_bucket_stride(buckets - &SA[n + n], bucket_size, omp_num_threads);
+            fast_sint_t bucket_stride     = libsais_get_bucket_stride(buckets - &SA[(fast_sint_t)n + (fast_sint_t)n], bucket_size, omp_num_threads);
 
             {
                 thread_state[omp_thread_num].state.position = omp_block_start + omp_block_size;
@@ -1310,7 +1310,7 @@ static sa_sint_t libsais_count_and_gather_lms_suffixes_32s_2k_omp(const sa_sint_
 static void libsais_count_and_gather_compacted_lms_suffixes_32s_2k_omp(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
 #if defined(_OPENMP)
-    sa_sint_t max_threads = (sa_sint_t)((buckets - &SA[n + n]) / ((2 * (fast_sint_t)k + 15) & (-16))); if (max_threads > threads) { max_threads = threads; }
+    sa_sint_t max_threads = (sa_sint_t)((buckets - &SA[(fast_sint_t)n + (fast_sint_t)n]) / ((2 * (fast_sint_t)k + 15) & (-16))); if (max_threads > threads) { max_threads = threads; }
     if (max_threads > 1 && n >= 65536 && n / k >= 2)
     {
         if (max_threads > n / 8 / k) { max_threads = n / 8 / k; }
@@ -1381,8 +1381,8 @@ static void libsais_initialize_buckets_start_and_end_8u(sa_sint_t * RESTRICT buc
 
 static void libsais_initialize_buckets_start_and_end_32s_6k(sa_sint_t k, sa_sint_t * RESTRICT buckets)
 {
-    sa_sint_t * RESTRICT bucket_start = &buckets[4 * k];
-    sa_sint_t * RESTRICT bucket_end   = &buckets[5 * k];
+    sa_sint_t * RESTRICT bucket_start = &buckets[4 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT bucket_end   = &buckets[5 * (fast_sint_t)k];
 
     fast_sint_t i, j; sa_sint_t sum = 0;
     for (i = BUCKETS_INDEX4(0, 0), j = 0; i <= BUCKETS_INDEX4((fast_sint_t)k - 1, 0); i += BUCKETS_INDEX4(1, 0), j += 1)
@@ -1395,8 +1395,8 @@ static void libsais_initialize_buckets_start_and_end_32s_6k(sa_sint_t k, sa_sint
 
 static void libsais_initialize_buckets_start_and_end_32s_4k(sa_sint_t k, sa_sint_t * RESTRICT buckets)
 {
-    sa_sint_t * RESTRICT bucket_start = &buckets[2 * k];
-    sa_sint_t * RESTRICT bucket_end   = &buckets[3 * k];
+    sa_sint_t * RESTRICT bucket_start = &buckets[2 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT bucket_end   = &buckets[3 * (fast_sint_t)k];
 
     fast_sint_t i, j; sa_sint_t sum = 0;
     for (i = BUCKETS_INDEX2(0, 0), j = 0; i <= BUCKETS_INDEX2((fast_sint_t)k - 1, 0); i += BUCKETS_INDEX2(1, 0), j += 1)
@@ -1501,7 +1501,7 @@ static sa_sint_t libsais_initialize_buckets_for_lms_suffixes_radix_sort_32s_6k(c
     }
 
     {
-        sa_sint_t * RESTRICT temp_bucket = &buckets[4 * k];
+        sa_sint_t * RESTRICT temp_bucket = &buckets[4 * (fast_sint_t)k];
 
         fast_sint_t i, j; sa_sint_t sum = 0;
         for (i = BUCKETS_INDEX4(0, 0), j = 0; i <= BUCKETS_INDEX4((fast_sint_t)k - 1, 0); i += BUCKETS_INDEX4(1, 0), j += 1)
@@ -1515,8 +1515,8 @@ static sa_sint_t libsais_initialize_buckets_for_lms_suffixes_radix_sort_32s_6k(c
 
 static void libsais_initialize_buckets_for_radix_and_partial_sorting_32s_4k(const sa_sint_t * RESTRICT T, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t first_lms_suffix)
 {
-    sa_sint_t * RESTRICT bucket_start = &buckets[2 * k];
-    sa_sint_t * RESTRICT bucket_end   = &buckets[3 * k];
+    sa_sint_t * RESTRICT bucket_start = &buckets[2 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT bucket_end   = &buckets[3 * (fast_sint_t)k];
 
     buckets[BUCKETS_INDEX2(T[first_lms_suffix], 0)]++;
     buckets[BUCKETS_INDEX2(T[first_lms_suffix], 1)]--;
@@ -2062,7 +2062,7 @@ static void libsais_initialize_buckets_for_partial_sorting_8u(const uint8_t * RE
 
 static void libsais_initialize_buckets_for_partial_sorting_32s_6k(const sa_sint_t * RESTRICT T, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t first_lms_suffix, sa_sint_t left_suffixes_count)
 {
-    sa_sint_t * RESTRICT temp_bucket = &buckets[4 * k];
+    sa_sint_t * RESTRICT temp_bucket = &buckets[4 * (fast_sint_t)k];
 
     fast_sint_t i, j; sa_sint_t sum0 = left_suffixes_count + 1, sum1 = 0, sum2 = 0;
     for (first_lms_suffix = T[first_lms_suffix], i = BUCKETS_INDEX4(0, 0), j = BUCKETS_INDEX2(0, 0); i <= BUCKETS_INDEX4((fast_sint_t)first_lms_suffix - 1, 0); i += BUCKETS_INDEX4(1, 0), j += BUCKETS_INDEX2(1, 0))
@@ -2350,8 +2350,8 @@ static sa_sint_t libsais_partial_sorting_scan_left_to_right_32s_4k(const sa_sint
 {
     const fast_sint_t prefetch_distance = 32;
 
-    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * k];
-    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * k];
+    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * (fast_sint_t)k];
 
     fast_sint_t i, j;
     for (i = omp_block_start, j = omp_block_start + omp_block_size - 2 * prefetch_distance - 1; i < j; i += 2)
@@ -2523,8 +2523,8 @@ static sa_sint_t libsais_partial_sorting_scan_left_to_right_32s_4k_block_sort(co
 {
     const fast_sint_t prefetch_distance = 32;
 
-    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * k];
-    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * k];
+    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * (fast_sint_t)k];
 
     fast_sint_t i, j, omp_block_end = omp_block_start + omp_block_size;
     for (i = omp_block_start, j = omp_block_end - prefetch_distance - 1; i < j; i += 2)
@@ -2782,8 +2782,8 @@ static sa_sint_t libsais_partial_sorting_scan_left_to_right_32s_6k_omp(const sa_
 
 static sa_sint_t libsais_partial_sorting_scan_left_to_right_32s_4k_omp(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t d, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
-    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * k];
-    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * k];
+    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * (fast_sint_t)k];
 
     SA[induction_bucket[T[n - 1]]++] = (n - 1) | ((sa_sint_t)(T[n - 2] < T[n - 1]) << (SAINT_BIT - 1)) | SUFFIX_GROUP_MARKER;
     distinct_names[BUCKETS_INDEX2(T[n - 1], T[n - 2] < T[n - 1])] = ++d;
@@ -2871,7 +2871,7 @@ static void libsais_partial_sorting_shift_markers_32s_6k_omp(sa_sint_t * RESTRIC
 {
     const fast_sint_t prefetch_distance = 32;
 
-    const sa_sint_t * RESTRICT temp_bucket = &buckets[4 * k];
+    const sa_sint_t * RESTRICT temp_bucket = &buckets[4 * (fast_sint_t)k];
     
     fast_sint_t c;
 
@@ -2923,7 +2923,7 @@ static void libsais_partial_sorting_shift_markers_32s_4k(sa_sint_t * RESTRICT SA
 
 static void libsais_partial_sorting_shift_buckets_32s_6k(sa_sint_t k, sa_sint_t * RESTRICT buckets)
 {
-    sa_sint_t * RESTRICT temp_bucket = &buckets[4 * k];
+    sa_sint_t * RESTRICT temp_bucket = &buckets[4 * (fast_sint_t)k];
 
     fast_sint_t i;
     for (i = BUCKETS_INDEX2(0, 0); i <= BUCKETS_INDEX2((fast_sint_t)k - 1, 0); i += BUCKETS_INDEX2(1, 0))
@@ -3179,8 +3179,8 @@ static sa_sint_t libsais_partial_sorting_scan_right_to_left_32s_4k(const sa_sint
 {
     const fast_sint_t prefetch_distance = 32;
 
-    sa_sint_t * RESTRICT induction_bucket = &buckets[3 * k];
-    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * k];
+    sa_sint_t * RESTRICT induction_bucket = &buckets[3 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * (fast_sint_t)k];
 
     fast_sint_t i, j;
     for (i = omp_block_start + omp_block_size - 1, j = omp_block_start + 2 * prefetch_distance + 1; i >= j; i -= 2)
@@ -3352,8 +3352,8 @@ static sa_sint_t libsais_partial_sorting_scan_right_to_left_32s_4k_block_sort(co
 {
     const fast_sint_t prefetch_distance = 32;
 
-    sa_sint_t * RESTRICT induction_bucket = &buckets[3 * k];
-    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * k];
+    sa_sint_t * RESTRICT induction_bucket = &buckets[3 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * (fast_sint_t)k];
 
     fast_sint_t i, j;
     for (i = omp_block_start + omp_block_size - 1, j = omp_block_start + prefetch_distance + 1; i >= j; i -= 2)
@@ -3832,8 +3832,8 @@ static void libsais_induce_partial_order_32s_4k_omp(const sa_sint_t * RESTRICT T
 
 static void libsais_induce_partial_order_32s_2k_omp(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
-    libsais_partial_sorting_scan_left_to_right_32s_1k_omp(T, SA, n, &buckets[1 * k], threads, thread_state);
-    libsais_partial_sorting_scan_right_to_left_32s_1k_omp(T, SA, n, &buckets[0 * k], threads, thread_state);
+    libsais_partial_sorting_scan_left_to_right_32s_1k_omp(T, SA, n, &buckets[1 * (fast_sint_t)k], threads, thread_state);
+    libsais_partial_sorting_scan_right_to_left_32s_1k_omp(T, SA, n, &buckets[0 * (fast_sint_t)k], threads, thread_state);
     libsais_partial_sorting_gather_lms_suffixes_32s_1k_omp(SA, n, threads, thread_state);
 }
 
@@ -4371,7 +4371,7 @@ static void libsais_place_lms_suffixes_interval_8u(sa_sint_t * RESTRICT SA, sa_s
 
 static void libsais_place_lms_suffixes_interval_32s_4k(sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t m, const sa_sint_t * RESTRICT buckets)
 {
-    const sa_sint_t * RESTRICT bucket_end = &buckets[3 * k];
+    const sa_sint_t * RESTRICT bucket_end = &buckets[3 * (fast_sint_t)k];
 
     fast_sint_t c, j = n;
     for (c = (fast_sint_t)k - 2; c >= 0; --c)
@@ -4448,7 +4448,7 @@ static void libsais_place_lms_suffixes_interval_32s_1k(const sa_sint_t * RESTRIC
 
 static void libsais_place_lms_suffixes_histogram_32s_6k(sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t m, const sa_sint_t * RESTRICT buckets)
 {
-    const sa_sint_t * RESTRICT bucket_end = &buckets[5 * k];
+    const sa_sint_t * RESTRICT bucket_end = &buckets[5 * (fast_sint_t)k];
 
     fast_sint_t c, j = n;
     for (c = (fast_sint_t)k - 2; c >= 0; --c)
@@ -4471,7 +4471,7 @@ static void libsais_place_lms_suffixes_histogram_32s_6k(sa_sint_t * RESTRICT SA,
 
 static void libsais_place_lms_suffixes_histogram_32s_4k(sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t m, const sa_sint_t * RESTRICT buckets)
 {
-    const sa_sint_t * RESTRICT bucket_end = &buckets[3 * k];
+    const sa_sint_t * RESTRICT bucket_end = &buckets[3 * (fast_sint_t)k];
 
     fast_sint_t c, j = n;
     for (c = (fast_sint_t)k - 2; c >= 0; --c)
@@ -5830,20 +5830,20 @@ static sa_sint_t libsais_induce_final_order_8u_omp(const uint8_t * RESTRICT T, s
 
 static void libsais_induce_final_order_32s_6k(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
-    libsais_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[4 * k], threads, thread_state);
-    libsais_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[5 * k], threads, thread_state);
+    libsais_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[4 * (fast_sint_t)k], threads, thread_state);
+    libsais_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[5 * (fast_sint_t)k], threads, thread_state);
 }
 
 static void libsais_induce_final_order_32s_4k(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
-    libsais_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[2 * k], threads, thread_state);
-    libsais_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[3 * k], threads, thread_state);
+    libsais_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[2 * (fast_sint_t)k], threads, thread_state);
+    libsais_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[3 * (fast_sint_t)k], threads, thread_state);
 }
 
 static void libsais_induce_final_order_32s_2k(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
-    libsais_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[1 * k], threads, thread_state);
-    libsais_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[0 * k], threads, thread_state);
+    libsais_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[1 * (fast_sint_t)k], threads, thread_state);
+    libsais_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[0 * (fast_sint_t)k], threads, thread_state);
 }
 
 static void libsais_induce_final_order_32s_1k(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
@@ -6266,7 +6266,7 @@ static sa_sint_t libsais_main_32s(sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT S
     if (k > 0 && fs / k >= 6)
     {
         sa_sint_t alignment = (fs - 1024) / k >= 6 ? 1024 : 16;
-        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 6 ? (sa_sint_t *)libsais_align_up(&SA[n + fs - 6 * k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 6 * k];
+        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 6 ? (sa_sint_t *)libsais_align_up(&SA[n + fs - 6 * (fast_sint_t)k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 6 * (fast_sint_t)k];
 
         sa_sint_t m = libsais_count_and_gather_lms_suffixes_32s_4k_omp(T, SA, n, k, buckets, threads, thread_state);
         if (m > 1)
@@ -6276,8 +6276,8 @@ static sa_sint_t libsais_main_32s(sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT S
             sa_sint_t first_lms_suffix    = SA[n - m];
             sa_sint_t left_suffixes_count = libsais_initialize_buckets_for_lms_suffixes_radix_sort_32s_6k(T, k, buckets, first_lms_suffix);
 
-            libsais_radix_sort_lms_suffixes_32s_6k_omp(T, SA, n, m, &buckets[4 * k], threads, thread_state);
-            libsais_radix_sort_set_markers_32s_6k_omp(SA, k, &buckets[4 * k], threads);
+            libsais_radix_sort_lms_suffixes_32s_6k_omp(T, SA, n, m, &buckets[4 * (fast_sint_t)k], threads, thread_state);
+            libsais_radix_sort_set_markers_32s_6k_omp(SA, k, &buckets[4 * (fast_sint_t)k], threads);
 
             if (threads > 1 && n >= 65536) { memset(&SA[(fast_sint_t)n - (fast_sint_t)m], 0, (size_t)m * sizeof(sa_sint_t)); }
 
@@ -6319,7 +6319,7 @@ static sa_sint_t libsais_main_32s(sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT S
     else if (k > 0 && fs / k >= 4)
     {
         sa_sint_t alignment = (fs - 1024) / k >= 4 ? 1024 : 16;
-        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 4 ? (sa_sint_t *)libsais_align_up(&SA[n + fs - 4 * k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 4 * k];
+        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 4 ? (sa_sint_t *)libsais_align_up(&SA[n + fs - 4 * (fast_sint_t)k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 4 * (fast_sint_t)k];
 
         sa_sint_t m = libsais_count_and_gather_lms_suffixes_32s_2k_omp(T, SA, n, k, buckets, threads, thread_state);
         if (m > 1)
@@ -6363,7 +6363,7 @@ static sa_sint_t libsais_main_32s(sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT S
     else if (k > 0 && fs / k >= 2)
     {
         sa_sint_t alignment = (fs - 1024) / k >= 2 ? 1024 : 16;
-        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 2 ? (sa_sint_t *)libsais_align_up(&SA[n + fs - 2 * k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 2 * k];
+        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 2 ? (sa_sint_t *)libsais_align_up(&SA[n + fs - 2 * (fast_sint_t)k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 2 * (fast_sint_t)k];
 
         sa_sint_t m = libsais_count_and_gather_lms_suffixes_32s_2k_omp(T, SA, n, k, buckets, threads, thread_state);
         if (m > 1)
diff --git a/src/libsais.h b/libsais.h
similarity index 99%
rename from src/libsais.h
rename to libsais.h
index c931701..f259fd0 100644
--- a/src/libsais.h
+++ b/libsais.h
@@ -26,8 +26,8 @@ Please see the file LICENSE for full copyright information.
 
 #define LIBSAIS_VERSION_MAJOR   2
 #define LIBSAIS_VERSION_MINOR   7
-#define LIBSAIS_VERSION_PATCH   1
-#define LIBSAIS_VERSION_STRING  "2.7.1"
+#define LIBSAIS_VERSION_PATCH   2
+#define LIBSAIS_VERSION_STRING  "2.7.2"
 
 #ifdef __cplusplus
 extern "C" {
diff --git a/src/libsais16.c b/libsais16.c
similarity index 99%
rename from src/libsais16.c
rename to libsais16.c
index 2c2baca..381db0e 100644
--- a/src/libsais16.c
+++ b/libsais16.c
@@ -1104,7 +1104,7 @@ static void libsais16_count_and_gather_compacted_lms_suffixes_32s_2k_fs_omp(cons
         else
         {
             fast_sint_t bucket_size       = 2 * (fast_sint_t)k;
-            fast_sint_t bucket_stride     = libsais16_get_bucket_stride(buckets - &SA[n + n], bucket_size, omp_num_threads);
+            fast_sint_t bucket_stride     = libsais16_get_bucket_stride(buckets - &SA[(fast_sint_t)n + (fast_sint_t)n], bucket_size, omp_num_threads);
 
             {
                 thread_state[omp_thread_num].state.position = omp_block_start + omp_block_size;
@@ -1288,7 +1288,7 @@ static sa_sint_t libsais16_count_and_gather_lms_suffixes_32s_2k_omp(const sa_sin
 static void libsais16_count_and_gather_compacted_lms_suffixes_32s_2k_omp(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
 #if defined(_OPENMP)
-    sa_sint_t max_threads = (sa_sint_t)((buckets - &SA[n + n]) / ((2 * (fast_sint_t)k + 15) & (-16))); if (max_threads > threads) { max_threads = threads; }
+    sa_sint_t max_threads = (sa_sint_t)((buckets - &SA[(fast_sint_t)n + (fast_sint_t)n]) / ((2 * (fast_sint_t)k + 15) & (-16))); if (max_threads > threads) { max_threads = threads; }
     if (max_threads > 1 && n >= 65536 && n / k >= 2)
     {
         if (max_threads > n / 8 / k) { max_threads = n / 8 / k; }
@@ -1359,8 +1359,8 @@ static void libsais16_initialize_buckets_start_and_end_16u(sa_sint_t * RESTRICT
 
 static void libsais16_initialize_buckets_start_and_end_32s_6k(sa_sint_t k, sa_sint_t * RESTRICT buckets)
 {
-    sa_sint_t * RESTRICT bucket_start = &buckets[4 * k];
-    sa_sint_t * RESTRICT bucket_end   = &buckets[5 * k];
+    sa_sint_t * RESTRICT bucket_start = &buckets[4 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT bucket_end   = &buckets[5 * (fast_sint_t)k];
 
     fast_sint_t i, j; sa_sint_t sum = 0;
     for (i = BUCKETS_INDEX4(0, 0), j = 0; i <= BUCKETS_INDEX4((fast_sint_t)k - 1, 0); i += BUCKETS_INDEX4(1, 0), j += 1)
@@ -1373,8 +1373,8 @@ static void libsais16_initialize_buckets_start_and_end_32s_6k(sa_sint_t k, sa_si
 
 static void libsais16_initialize_buckets_start_and_end_32s_4k(sa_sint_t k, sa_sint_t * RESTRICT buckets)
 {
-    sa_sint_t * RESTRICT bucket_start = &buckets[2 * k];
-    sa_sint_t * RESTRICT bucket_end   = &buckets[3 * k];
+    sa_sint_t * RESTRICT bucket_start = &buckets[2 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT bucket_end   = &buckets[3 * (fast_sint_t)k];
 
     fast_sint_t i, j; sa_sint_t sum = 0;
     for (i = BUCKETS_INDEX2(0, 0), j = 0; i <= BUCKETS_INDEX2((fast_sint_t)k - 1, 0); i += BUCKETS_INDEX2(1, 0), j += 1)
@@ -1479,7 +1479,7 @@ static sa_sint_t libsais16_initialize_buckets_for_lms_suffixes_radix_sort_32s_6k
     }
 
     {
-        sa_sint_t * RESTRICT temp_bucket = &buckets[4 * k];
+        sa_sint_t * RESTRICT temp_bucket = &buckets[4 * (fast_sint_t)k];
 
         fast_sint_t i, j; sa_sint_t sum = 0;
         for (i = BUCKETS_INDEX4(0, 0), j = 0; i <= BUCKETS_INDEX4((fast_sint_t)k - 1, 0); i += BUCKETS_INDEX4(1, 0), j += 1)
@@ -1493,8 +1493,8 @@ static sa_sint_t libsais16_initialize_buckets_for_lms_suffixes_radix_sort_32s_6k
 
 static void libsais16_initialize_buckets_for_radix_and_partial_sorting_32s_4k(const sa_sint_t * RESTRICT T, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t first_lms_suffix)
 {
-    sa_sint_t * RESTRICT bucket_start = &buckets[2 * k];
-    sa_sint_t * RESTRICT bucket_end   = &buckets[3 * k];
+    sa_sint_t * RESTRICT bucket_start = &buckets[2 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT bucket_end   = &buckets[3 * (fast_sint_t)k];
 
     buckets[BUCKETS_INDEX2(T[first_lms_suffix], 0)]++;
     buckets[BUCKETS_INDEX2(T[first_lms_suffix], 1)]--;
@@ -2040,7 +2040,7 @@ static void libsais16_initialize_buckets_for_partial_sorting_16u(const uint16_t
 
 static void libsais16_initialize_buckets_for_partial_sorting_32s_6k(const sa_sint_t * RESTRICT T, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t first_lms_suffix, sa_sint_t left_suffixes_count)
 {
-    sa_sint_t * RESTRICT temp_bucket = &buckets[4 * k];
+    sa_sint_t * RESTRICT temp_bucket = &buckets[4 * (fast_sint_t)k];
 
     fast_sint_t i, j; sa_sint_t sum0 = left_suffixes_count + 1, sum1 = 0, sum2 = 0;
     for (first_lms_suffix = T[first_lms_suffix], i = BUCKETS_INDEX4(0, 0), j = BUCKETS_INDEX2(0, 0); i <= BUCKETS_INDEX4((fast_sint_t)first_lms_suffix - 1, 0); i += BUCKETS_INDEX4(1, 0), j += BUCKETS_INDEX2(1, 0))
@@ -2328,8 +2328,8 @@ static sa_sint_t libsais16_partial_sorting_scan_left_to_right_32s_4k(const sa_si
 {
     const fast_sint_t prefetch_distance = 32;
 
-    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * k];
-    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * k];
+    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * (fast_sint_t)k];
 
     fast_sint_t i, j;
     for (i = omp_block_start, j = omp_block_start + omp_block_size - 2 * prefetch_distance - 1; i < j; i += 2)
@@ -2501,8 +2501,8 @@ static sa_sint_t libsais16_partial_sorting_scan_left_to_right_32s_4k_block_sort(
 {
     const fast_sint_t prefetch_distance = 32;
 
-    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * k];
-    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * k];
+    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * (fast_sint_t)k];
 
     fast_sint_t i, j, omp_block_end = omp_block_start + omp_block_size;
     for (i = omp_block_start, j = omp_block_end - prefetch_distance - 1; i < j; i += 2)
@@ -2760,8 +2760,8 @@ static sa_sint_t libsais16_partial_sorting_scan_left_to_right_32s_6k_omp(const s
 
 static sa_sint_t libsais16_partial_sorting_scan_left_to_right_32s_4k_omp(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t d, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
-    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * k];
-    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * k];
+    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * (fast_sint_t)k];
 
     SA[induction_bucket[T[n - 1]]++] = (n - 1) | ((sa_sint_t)(T[n - 2] < T[n - 1]) << (SAINT_BIT - 1)) | SUFFIX_GROUP_MARKER;
     distinct_names[BUCKETS_INDEX2(T[n - 1], T[n - 2] < T[n - 1])] = ++d;
@@ -2849,7 +2849,7 @@ static void libsais16_partial_sorting_shift_markers_32s_6k_omp(sa_sint_t * RESTR
 {
     const fast_sint_t prefetch_distance = 32;
 
-    const sa_sint_t * RESTRICT temp_bucket = &buckets[4 * k];
+    const sa_sint_t * RESTRICT temp_bucket = &buckets[4 * (fast_sint_t)k];
     
     fast_sint_t c;
 
@@ -2901,7 +2901,7 @@ static void libsais16_partial_sorting_shift_markers_32s_4k(sa_sint_t * RESTRICT
 
 static void libsais16_partial_sorting_shift_buckets_32s_6k(sa_sint_t k, sa_sint_t * RESTRICT buckets)
 {
-    sa_sint_t * RESTRICT temp_bucket = &buckets[4 * k];
+    sa_sint_t * RESTRICT temp_bucket = &buckets[4 * (fast_sint_t)k];
 
     fast_sint_t i;
     for (i = BUCKETS_INDEX2(0, 0); i <= BUCKETS_INDEX2((fast_sint_t)k - 1, 0); i += BUCKETS_INDEX2(1, 0))
@@ -3157,8 +3157,8 @@ static sa_sint_t libsais16_partial_sorting_scan_right_to_left_32s_4k(const sa_si
 {
     const fast_sint_t prefetch_distance = 32;
 
-    sa_sint_t * RESTRICT induction_bucket = &buckets[3 * k];
-    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * k];
+    sa_sint_t * RESTRICT induction_bucket = &buckets[3 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * (fast_sint_t)k];
 
     fast_sint_t i, j;
     for (i = omp_block_start + omp_block_size - 1, j = omp_block_start + 2 * prefetch_distance + 1; i >= j; i -= 2)
@@ -3330,8 +3330,8 @@ static sa_sint_t libsais16_partial_sorting_scan_right_to_left_32s_4k_block_sort(
 {
     const fast_sint_t prefetch_distance = 32;
 
-    sa_sint_t * RESTRICT induction_bucket = &buckets[3 * k];
-    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * k];
+    sa_sint_t * RESTRICT induction_bucket = &buckets[3 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * (fast_sint_t)k];
 
     fast_sint_t i, j;
     for (i = omp_block_start + omp_block_size - 1, j = omp_block_start + prefetch_distance + 1; i >= j; i -= 2)
@@ -3810,8 +3810,8 @@ static void libsais16_induce_partial_order_32s_4k_omp(const sa_sint_t * RESTRICT
 
 static void libsais16_induce_partial_order_32s_2k_omp(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
-    libsais16_partial_sorting_scan_left_to_right_32s_1k_omp(T, SA, n, &buckets[1 * k], threads, thread_state);
-    libsais16_partial_sorting_scan_right_to_left_32s_1k_omp(T, SA, n, &buckets[0 * k], threads, thread_state);
+    libsais16_partial_sorting_scan_left_to_right_32s_1k_omp(T, SA, n, &buckets[1 * (fast_sint_t)k], threads, thread_state);
+    libsais16_partial_sorting_scan_right_to_left_32s_1k_omp(T, SA, n, &buckets[0 * (fast_sint_t)k], threads, thread_state);
     libsais16_partial_sorting_gather_lms_suffixes_32s_1k_omp(SA, n, threads, thread_state);
 }
 
@@ -4349,7 +4349,7 @@ static void libsais16_place_lms_suffixes_interval_16u(sa_sint_t * RESTRICT SA, s
 
 static void libsais16_place_lms_suffixes_interval_32s_4k(sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t m, const sa_sint_t * RESTRICT buckets)
 {
-    const sa_sint_t * RESTRICT bucket_end = &buckets[3 * k];
+    const sa_sint_t * RESTRICT bucket_end = &buckets[3 * (fast_sint_t)k];
 
     fast_sint_t c, j = n;
     for (c = (fast_sint_t)k - 2; c >= 0; --c)
@@ -4426,7 +4426,7 @@ static void libsais16_place_lms_suffixes_interval_32s_1k(const sa_sint_t * RESTR
 
 static void libsais16_place_lms_suffixes_histogram_32s_6k(sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t m, const sa_sint_t * RESTRICT buckets)
 {
-    const sa_sint_t * RESTRICT bucket_end = &buckets[5 * k];
+    const sa_sint_t * RESTRICT bucket_end = &buckets[5 * (fast_sint_t)k];
 
     fast_sint_t c, j = n;
     for (c = (fast_sint_t)k - 2; c >= 0; --c)
@@ -4449,7 +4449,7 @@ static void libsais16_place_lms_suffixes_histogram_32s_6k(sa_sint_t * RESTRICT S
 
 static void libsais16_place_lms_suffixes_histogram_32s_4k(sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t m, const sa_sint_t * RESTRICT buckets)
 {
-    const sa_sint_t * RESTRICT bucket_end = &buckets[3 * k];
+    const sa_sint_t * RESTRICT bucket_end = &buckets[3 * (fast_sint_t)k];
 
     fast_sint_t c, j = n;
     for (c = (fast_sint_t)k - 2; c >= 0; --c)
@@ -5808,20 +5808,20 @@ static sa_sint_t libsais16_induce_final_order_16u_omp(const uint16_t * RESTRICT
 
 static void libsais16_induce_final_order_32s_6k(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
-    libsais16_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[4 * k], threads, thread_state);
-    libsais16_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[5 * k], threads, thread_state);
+    libsais16_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[4 * (fast_sint_t)k], threads, thread_state);
+    libsais16_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[5 * (fast_sint_t)k], threads, thread_state);
 }
 
 static void libsais16_induce_final_order_32s_4k(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
-    libsais16_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[2 * k], threads, thread_state);
-    libsais16_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[3 * k], threads, thread_state);
+    libsais16_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[2 * (fast_sint_t)k], threads, thread_state);
+    libsais16_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[3 * (fast_sint_t)k], threads, thread_state);
 }
 
 static void libsais16_induce_final_order_32s_2k(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
-    libsais16_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[1 * k], threads, thread_state);
-    libsais16_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[0 * k], threads, thread_state);
+    libsais16_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[1 * (fast_sint_t)k], threads, thread_state);
+    libsais16_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[0 * (fast_sint_t)k], threads, thread_state);
 }
 
 static void libsais16_induce_final_order_32s_1k(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
@@ -6244,7 +6244,7 @@ static sa_sint_t libsais16_main_32s(sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT
     if (k > 0 && fs / k >= 6)
     {
         sa_sint_t alignment = (fs - 1024) / k >= 6 ? 1024 : 16;
-        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 6 ? (sa_sint_t *)libsais16_align_up(&SA[n + fs - 6 * k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 6 * k];
+        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 6 ? (sa_sint_t *)libsais16_align_up(&SA[n + fs - 6 * (fast_sint_t)k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 6 * (fast_sint_t)k];
 
         sa_sint_t m = libsais16_count_and_gather_lms_suffixes_32s_4k_omp(T, SA, n, k, buckets, threads, thread_state);
         if (m > 1)
@@ -6254,8 +6254,8 @@ static sa_sint_t libsais16_main_32s(sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT
             sa_sint_t first_lms_suffix    = SA[n - m];
             sa_sint_t left_suffixes_count = libsais16_initialize_buckets_for_lms_suffixes_radix_sort_32s_6k(T, k, buckets, first_lms_suffix);
 
-            libsais16_radix_sort_lms_suffixes_32s_6k_omp(T, SA, n, m, &buckets[4 * k], threads, thread_state);
-            libsais16_radix_sort_set_markers_32s_6k_omp(SA, k, &buckets[4 * k], threads);
+            libsais16_radix_sort_lms_suffixes_32s_6k_omp(T, SA, n, m, &buckets[4 * (fast_sint_t)k], threads, thread_state);
+            libsais16_radix_sort_set_markers_32s_6k_omp(SA, k, &buckets[4 * (fast_sint_t)k], threads);
 
             if (threads > 1 && n >= 65536) { memset(&SA[(fast_sint_t)n - (fast_sint_t)m], 0, (size_t)m * sizeof(sa_sint_t)); }
 
@@ -6297,7 +6297,7 @@ static sa_sint_t libsais16_main_32s(sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT
     else if (k > 0 && fs / k >= 4)
     {
         sa_sint_t alignment = (fs - 1024) / k >= 4 ? 1024 : 16;
-        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 4 ? (sa_sint_t *)libsais16_align_up(&SA[n + fs - 4 * k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 4 * k];
+        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 4 ? (sa_sint_t *)libsais16_align_up(&SA[n + fs - 4 * (fast_sint_t)k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 4 * (fast_sint_t)k];
 
         sa_sint_t m = libsais16_count_and_gather_lms_suffixes_32s_2k_omp(T, SA, n, k, buckets, threads, thread_state);
         if (m > 1)
@@ -6341,7 +6341,7 @@ static sa_sint_t libsais16_main_32s(sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT
     else if (k > 0 && fs / k >= 2)
     {
         sa_sint_t alignment = (fs - 1024) / k >= 2 ? 1024 : 16;
-        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 2 ? (sa_sint_t *)libsais16_align_up(&SA[n + fs - 2 * k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 2 * k];
+        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 2 ? (sa_sint_t *)libsais16_align_up(&SA[n + fs - 2 * (fast_sint_t)k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 2 * (fast_sint_t)k];
 
         sa_sint_t m = libsais16_count_and_gather_lms_suffixes_32s_2k_omp(T, SA, n, k, buckets, threads, thread_state);
         if (m > 1)
diff --git a/src/libsais16.h b/libsais16.h
similarity index 99%
rename from src/libsais16.h
rename to libsais16.h
index 11bc94f..2874632 100644
--- a/src/libsais16.h
+++ b/libsais16.h
@@ -26,8 +26,8 @@ Please see the file LICENSE for full copyright information.
 
 #define LIBSAIS16_VERSION_MAJOR   2
 #define LIBSAIS16_VERSION_MINOR   7
-#define LIBSAIS16_VERSION_PATCH   1
-#define LIBSAIS16_VERSION_STRING  "2.7.1"
+#define LIBSAIS16_VERSION_PATCH   2
+#define LIBSAIS16_VERSION_STRING  "2.7.2"
 
 #ifdef __cplusplus
 extern "C" {
diff --git a/src/libsais64.c b/libsais64.c
similarity index 99%
rename from src/libsais64.c
rename to libsais64.c
index 72786f6..0e4defc 100644
--- a/src/libsais64.c
+++ b/libsais64.c
@@ -1096,7 +1096,7 @@ static void libsais64_count_and_gather_compacted_lms_suffixes_32s_2k_fs_omp(cons
         else
         {
             fast_sint_t bucket_size       = 2 * (fast_sint_t)k;
-            fast_sint_t bucket_stride     = libsais64_get_bucket_stride(buckets - &SA[n + n], bucket_size, omp_num_threads);
+            fast_sint_t bucket_stride     = libsais64_get_bucket_stride(buckets - &SA[(fast_sint_t)n + (fast_sint_t)n], bucket_size, omp_num_threads);
 
             {
                 thread_state[omp_thread_num].state.position = omp_block_start + omp_block_size;
@@ -1280,7 +1280,7 @@ static sa_sint_t libsais64_count_and_gather_lms_suffixes_32s_2k_omp(const sa_sin
 static void libsais64_count_and_gather_compacted_lms_suffixes_32s_2k_omp(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
 #if defined(_OPENMP)
-    sa_sint_t max_threads = (sa_sint_t)((buckets - &SA[n + n]) / ((2 * (fast_sint_t)k + 15) & (-16))); if (max_threads > threads) { max_threads = threads; }
+    sa_sint_t max_threads = (sa_sint_t)((buckets - &SA[(fast_sint_t)n + (fast_sint_t)n]) / ((2 * (fast_sint_t)k + 15) & (-16))); if (max_threads > threads) { max_threads = threads; }
     if (max_threads > 1 && n >= 65536 && n / k >= 2)
     {
         if (max_threads > n / 8 / k) { max_threads = n / 8 / k; }
@@ -1351,8 +1351,8 @@ static void libsais64_initialize_buckets_start_and_end_8u(sa_sint_t * RESTRICT b
 
 static void libsais64_initialize_buckets_start_and_end_32s_6k(sa_sint_t k, sa_sint_t * RESTRICT buckets)
 {
-    sa_sint_t * RESTRICT bucket_start = &buckets[4 * k];
-    sa_sint_t * RESTRICT bucket_end   = &buckets[5 * k];
+    sa_sint_t * RESTRICT bucket_start = &buckets[4 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT bucket_end   = &buckets[5 * (fast_sint_t)k];
 
     fast_sint_t i, j; sa_sint_t sum = 0;
     for (i = BUCKETS_INDEX4(0, 0), j = 0; i <= BUCKETS_INDEX4((fast_sint_t)k - 1, 0); i += BUCKETS_INDEX4(1, 0), j += 1)
@@ -1365,8 +1365,8 @@ static void libsais64_initialize_buckets_start_and_end_32s_6k(sa_sint_t k, sa_si
 
 static void libsais64_initialize_buckets_start_and_end_32s_4k(sa_sint_t k, sa_sint_t * RESTRICT buckets)
 {
-    sa_sint_t * RESTRICT bucket_start = &buckets[2 * k];
-    sa_sint_t * RESTRICT bucket_end   = &buckets[3 * k];
+    sa_sint_t * RESTRICT bucket_start = &buckets[2 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT bucket_end   = &buckets[3 * (fast_sint_t)k];
 
     fast_sint_t i, j; sa_sint_t sum = 0;
     for (i = BUCKETS_INDEX2(0, 0), j = 0; i <= BUCKETS_INDEX2((fast_sint_t)k - 1, 0); i += BUCKETS_INDEX2(1, 0), j += 1)
@@ -1471,7 +1471,7 @@ static sa_sint_t libsais64_initialize_buckets_for_lms_suffixes_radix_sort_32s_6k
     }
 
     {
-        sa_sint_t * RESTRICT temp_bucket = &buckets[4 * k];
+        sa_sint_t * RESTRICT temp_bucket = &buckets[4 * (fast_sint_t)k];
 
         fast_sint_t i, j; sa_sint_t sum = 0;
         for (i = BUCKETS_INDEX4(0, 0), j = 0; i <= BUCKETS_INDEX4((fast_sint_t)k - 1, 0); i += BUCKETS_INDEX4(1, 0), j += 1)
@@ -1485,8 +1485,8 @@ static sa_sint_t libsais64_initialize_buckets_for_lms_suffixes_radix_sort_32s_6k
 
 static void libsais64_initialize_buckets_for_radix_and_partial_sorting_32s_4k(const sa_sint_t * RESTRICT T, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t first_lms_suffix)
 {
-    sa_sint_t * RESTRICT bucket_start = &buckets[2 * k];
-    sa_sint_t * RESTRICT bucket_end   = &buckets[3 * k];
+    sa_sint_t * RESTRICT bucket_start = &buckets[2 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT bucket_end   = &buckets[3 * (fast_sint_t)k];
 
     buckets[BUCKETS_INDEX2(T[first_lms_suffix], 0)]++;
     buckets[BUCKETS_INDEX2(T[first_lms_suffix], 1)]--;
@@ -2032,7 +2032,7 @@ static void libsais64_initialize_buckets_for_partial_sorting_8u(const uint8_t *
 
 static void libsais64_initialize_buckets_for_partial_sorting_32s_6k(const sa_sint_t * RESTRICT T, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t first_lms_suffix, sa_sint_t left_suffixes_count)
 {
-    sa_sint_t * RESTRICT temp_bucket = &buckets[4 * k];
+    sa_sint_t * RESTRICT temp_bucket = &buckets[4 * (fast_sint_t)k];
 
     fast_sint_t i, j; sa_sint_t sum0 = left_suffixes_count + 1, sum1 = 0, sum2 = 0;
     for (first_lms_suffix = T[first_lms_suffix], i = BUCKETS_INDEX4(0, 0), j = BUCKETS_INDEX2(0, 0); i <= BUCKETS_INDEX4((fast_sint_t)first_lms_suffix - 1, 0); i += BUCKETS_INDEX4(1, 0), j += BUCKETS_INDEX2(1, 0))
@@ -2320,8 +2320,8 @@ static sa_sint_t libsais64_partial_sorting_scan_left_to_right_32s_4k(const sa_si
 {
     const fast_sint_t prefetch_distance = 32;
 
-    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * k];
-    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * k];
+    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * (fast_sint_t)k];
 
     fast_sint_t i, j;
     for (i = omp_block_start, j = omp_block_start + omp_block_size - 2 * prefetch_distance - 1; i < j; i += 2)
@@ -2493,8 +2493,8 @@ static sa_sint_t libsais64_partial_sorting_scan_left_to_right_32s_4k_block_sort(
 {
     const fast_sint_t prefetch_distance = 32;
 
-    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * k];
-    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * k];
+    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * (fast_sint_t)k];
 
     fast_sint_t i, j, omp_block_end = omp_block_start + omp_block_size;
     for (i = omp_block_start, j = omp_block_end - prefetch_distance - 1; i < j; i += 2)
@@ -2752,8 +2752,8 @@ static sa_sint_t libsais64_partial_sorting_scan_left_to_right_32s_6k_omp(const s
 
 static sa_sint_t libsais64_partial_sorting_scan_left_to_right_32s_4k_omp(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t d, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
-    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * k];
-    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * k];
+    sa_sint_t * RESTRICT induction_bucket = &buckets[2 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * (fast_sint_t)k];
 
     SA[induction_bucket[T[n - 1]]++] = (n - 1) | ((sa_sint_t)(T[n - 2] < T[n - 1]) << (SAINT_BIT - 1)) | SUFFIX_GROUP_MARKER;
     distinct_names[BUCKETS_INDEX2(T[n - 1], T[n - 2] < T[n - 1])] = ++d;
@@ -2841,7 +2841,7 @@ static void libsais64_partial_sorting_shift_markers_32s_6k_omp(sa_sint_t * RESTR
 {
     const fast_sint_t prefetch_distance = 32;
 
-    const sa_sint_t * RESTRICT temp_bucket = &buckets[4 * k];
+    const sa_sint_t * RESTRICT temp_bucket = &buckets[4 * (fast_sint_t)k];
     
     fast_sint_t c;
 
@@ -2893,7 +2893,7 @@ static void libsais64_partial_sorting_shift_markers_32s_4k(sa_sint_t * RESTRICT
 
 static void libsais64_partial_sorting_shift_buckets_32s_6k(sa_sint_t k, sa_sint_t * RESTRICT buckets)
 {
-    sa_sint_t * RESTRICT temp_bucket = &buckets[4 * k];
+    sa_sint_t * RESTRICT temp_bucket = &buckets[4 * (fast_sint_t)k];
 
     fast_sint_t i;
     for (i = BUCKETS_INDEX2(0, 0); i <= BUCKETS_INDEX2((fast_sint_t)k - 1, 0); i += BUCKETS_INDEX2(1, 0))
@@ -3149,8 +3149,8 @@ static sa_sint_t libsais64_partial_sorting_scan_right_to_left_32s_4k(const sa_si
 {
     const fast_sint_t prefetch_distance = 32;
 
-    sa_sint_t * RESTRICT induction_bucket = &buckets[3 * k];
-    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * k];
+    sa_sint_t * RESTRICT induction_bucket = &buckets[3 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * (fast_sint_t)k];
 
     fast_sint_t i, j;
     for (i = omp_block_start + omp_block_size - 1, j = omp_block_start + 2 * prefetch_distance + 1; i >= j; i -= 2)
@@ -3322,8 +3322,8 @@ static sa_sint_t libsais64_partial_sorting_scan_right_to_left_32s_4k_block_sort(
 {
     const fast_sint_t prefetch_distance = 32;
 
-    sa_sint_t * RESTRICT induction_bucket = &buckets[3 * k];
-    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * k];
+    sa_sint_t * RESTRICT induction_bucket = &buckets[3 * (fast_sint_t)k];
+    sa_sint_t * RESTRICT distinct_names   = &buckets[0 * (fast_sint_t)k];
 
     fast_sint_t i, j;
     for (i = omp_block_start + omp_block_size - 1, j = omp_block_start + prefetch_distance + 1; i >= j; i -= 2)
@@ -3802,8 +3802,8 @@ static void libsais64_induce_partial_order_32s_4k_omp(const sa_sint_t * RESTRICT
 
 static void libsais64_induce_partial_order_32s_2k_omp(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
-    libsais64_partial_sorting_scan_left_to_right_32s_1k_omp(T, SA, n, &buckets[1 * k], threads, thread_state);
-    libsais64_partial_sorting_scan_right_to_left_32s_1k_omp(T, SA, n, &buckets[0 * k], threads, thread_state);
+    libsais64_partial_sorting_scan_left_to_right_32s_1k_omp(T, SA, n, &buckets[1 * (fast_sint_t)k], threads, thread_state);
+    libsais64_partial_sorting_scan_right_to_left_32s_1k_omp(T, SA, n, &buckets[0 * (fast_sint_t)k], threads, thread_state);
     libsais64_partial_sorting_gather_lms_suffixes_32s_1k_omp(SA, n, threads, thread_state);
 }
 
@@ -4341,7 +4341,7 @@ static void libsais64_place_lms_suffixes_interval_8u(sa_sint_t * RESTRICT SA, sa
 
 static void libsais64_place_lms_suffixes_interval_32s_4k(sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t m, const sa_sint_t * RESTRICT buckets)
 {
-    const sa_sint_t * RESTRICT bucket_end = &buckets[3 * k];
+    const sa_sint_t * RESTRICT bucket_end = &buckets[3 * (fast_sint_t)k];
 
     fast_sint_t c, j = n;
     for (c = (fast_sint_t)k - 2; c >= 0; --c)
@@ -4418,7 +4418,7 @@ static void libsais64_place_lms_suffixes_interval_32s_1k(const sa_sint_t * RESTR
 
 static void libsais64_place_lms_suffixes_histogram_32s_6k(sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t m, const sa_sint_t * RESTRICT buckets)
 {
-    const sa_sint_t * RESTRICT bucket_end = &buckets[5 * k];
+    const sa_sint_t * RESTRICT bucket_end = &buckets[5 * (fast_sint_t)k];
 
     fast_sint_t c, j = n;
     for (c = (fast_sint_t)k - 2; c >= 0; --c)
@@ -4441,7 +4441,7 @@ static void libsais64_place_lms_suffixes_histogram_32s_6k(sa_sint_t * RESTRICT S
 
 static void libsais64_place_lms_suffixes_histogram_32s_4k(sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t m, const sa_sint_t * RESTRICT buckets)
 {
-    const sa_sint_t * RESTRICT bucket_end = &buckets[3 * k];
+    const sa_sint_t * RESTRICT bucket_end = &buckets[3 * (fast_sint_t)k];
 
     fast_sint_t c, j = n;
     for (c = (fast_sint_t)k - 2; c >= 0; --c)
@@ -5800,20 +5800,20 @@ static sa_sint_t libsais64_induce_final_order_8u_omp(const uint8_t * RESTRICT T,
 
 static void libsais64_induce_final_order_32s_6k(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
-    libsais64_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[4 * k], threads, thread_state);
-    libsais64_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[5 * k], threads, thread_state);
+    libsais64_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[4 * (fast_sint_t)k], threads, thread_state);
+    libsais64_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[5 * (fast_sint_t)k], threads, thread_state);
 }
 
 static void libsais64_induce_final_order_32s_4k(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
-    libsais64_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[2 * k], threads, thread_state);
-    libsais64_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[3 * k], threads, thread_state);
+    libsais64_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[2 * (fast_sint_t)k], threads, thread_state);
+    libsais64_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[3 * (fast_sint_t)k], threads, thread_state);
 }
 
 static void libsais64_induce_final_order_32s_2k(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
 {
-    libsais64_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[1 * k], threads, thread_state);
-    libsais64_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[0 * k], threads, thread_state);
+    libsais64_final_sorting_scan_left_to_right_32s_omp(T, SA, n, &buckets[1 * (fast_sint_t)k], threads, thread_state);
+    libsais64_final_sorting_scan_right_to_left_32s_omp(T, SA, n, &buckets[0 * (fast_sint_t)k], threads, thread_state);
 }
 
 static void libsais64_induce_final_order_32s_1k(const sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT SA, sa_sint_t n, sa_sint_t k, sa_sint_t * RESTRICT buckets, sa_sint_t threads, LIBSAIS_THREAD_STATE * RESTRICT thread_state)
@@ -6302,7 +6302,7 @@ static sa_sint_t libsais64_main_32s(sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT
     if (k > 0 && fs / k >= 6)
     {
         sa_sint_t alignment = (fs - 1024) / k >= 6 ? 1024 : 16;
-        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 6 ? (sa_sint_t *)libsais64_align_up(&SA[n + fs - 6 * k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 6 * k];
+        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 6 ? (sa_sint_t *)libsais64_align_up(&SA[n + fs - 6 * (fast_sint_t)k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 6 * (fast_sint_t)k];
 
         sa_sint_t m = libsais64_count_and_gather_lms_suffixes_32s_4k_omp(T, SA, n, k, buckets, threads, thread_state);
         if (m > 1)
@@ -6312,8 +6312,8 @@ static sa_sint_t libsais64_main_32s(sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT
             sa_sint_t first_lms_suffix    = SA[n - m];
             sa_sint_t left_suffixes_count = libsais64_initialize_buckets_for_lms_suffixes_radix_sort_32s_6k(T, k, buckets, first_lms_suffix);
 
-            libsais64_radix_sort_lms_suffixes_32s_6k_omp(T, SA, n, m, &buckets[4 * k], threads, thread_state);
-            libsais64_radix_sort_set_markers_32s_6k_omp(SA, k, &buckets[4 * k], threads);
+            libsais64_radix_sort_lms_suffixes_32s_6k_omp(T, SA, n, m, &buckets[4 * (fast_sint_t)k], threads, thread_state);
+            libsais64_radix_sort_set_markers_32s_6k_omp(SA, k, &buckets[4 * (fast_sint_t)k], threads);
 
             if (threads > 1 && n >= 65536) { memset(&SA[(fast_sint_t)n - (fast_sint_t)m], 0, (size_t)m * sizeof(sa_sint_t)); }
 
@@ -6355,7 +6355,7 @@ static sa_sint_t libsais64_main_32s(sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT
     else if (k > 0 && fs / k >= 4)
     {
         sa_sint_t alignment = (fs - 1024) / k >= 4 ? 1024 : 16;
-        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 4 ? (sa_sint_t *)libsais64_align_up(&SA[n + fs - 4 * k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 4 * k];
+        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 4 ? (sa_sint_t *)libsais64_align_up(&SA[n + fs - 4 * (fast_sint_t)k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 4 * (fast_sint_t)k];
 
         sa_sint_t m = libsais64_count_and_gather_lms_suffixes_32s_2k_omp(T, SA, n, k, buckets, threads, thread_state);
         if (m > 1)
@@ -6399,7 +6399,7 @@ static sa_sint_t libsais64_main_32s(sa_sint_t * RESTRICT T, sa_sint_t * RESTRICT
     else if (k > 0 && fs / k >= 2)
     {
         sa_sint_t alignment = (fs - 1024) / k >= 2 ? 1024 : 16;
-        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 2 ? (sa_sint_t *)libsais64_align_up(&SA[n + fs - 2 * k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 2 * k];
+        sa_sint_t * RESTRICT buckets = (fs - alignment) / k >= 2 ? (sa_sint_t *)libsais64_align_up(&SA[n + fs - 2 * (fast_sint_t)k - alignment], (size_t)alignment * sizeof(sa_sint_t)) : &SA[n + fs - 2 * (fast_sint_t)k];
 
         sa_sint_t m = libsais64_count_and_gather_lms_suffixes_32s_2k_omp(T, SA, n, k, buckets, threads, thread_state);
         if (m > 1)
diff --git a/src/libsais64.h b/libsais64.h
similarity index 99%
rename from src/libsais64.h
rename to libsais64.h
index 9384974..786a76c 100644
--- a/src/libsais64.h
+++ b/libsais64.h
@@ -26,8 +26,8 @@ Please see the file LICENSE for full copyright information.
 
 #define LIBSAIS64_VERSION_MAJOR   2
 #define LIBSAIS64_VERSION_MINOR   7
-#define LIBSAIS64_VERSION_PATCH   1
-#define LIBSAIS64_VERSION_STRING  "2.7.1"
+#define LIBSAIS64_VERSION_PATCH   2
+#define LIBSAIS64_VERSION_STRING  "2.7.2"
 
 #ifdef __cplusplus
 extern "C" {