--- a/configure

+++ b/configure

@@ -115,6 +115,7 @@

       echo '    [--static] [--64] [--libdir=LIBDIR] [--sharedlibdir=LIBDIR]' | tee -a configure.log

       echo '    [--includedir=INCLUDEDIR] [--archs="-arch i386 -arch x86_64"]' | tee -a configure.log

       echo '    [--dfltcc]' | tee -a configure.log

+      echo '    [--simd-slide-hash]' | tee -a configure.log

         exit 0 ;;

     -p*=* | --prefix=*) prefix=`echo $1 | sed 's/.*=//'`; shift ;;

     -e*=* | --eprefix=*) exec_prefix=`echo $1 | sed 's/.*=//'`; shift ;;

@@ -144,6 +145,11 @@

 	    PIC_OBJC="$PIC_OBJC dfltcc.lo"

       shift

       ;; 

+    --simd-slide-hash)

+	    OBJC="$OBJC slide_avx2.o slide_sse.o"

+	    PIC_OBJC="$PIC_OBJC slide_avx2.lo slide_sse.lo"

+      shift

+      ;; 

     *)

       echo "unknown option: $1" | tee -a configure.log

       echo "$0 --help for help" | tee -a configure.log

--- a/Makefile.in

+++ b/Makefile.in

@@ -152,6 +152,22 @@

 	$(CC) $(SFLAGS) $(ZINC) -DPIC -c -o objs/dfltcc.o $(SRCDIR)contrib/s390/dfltcc.c

 	-@mv objs/dfltcc.o $@

+slide_sse.o: $(SRCDIR)slide_sse.c

+	$(CC) $(CFLAGS) $(ZINC) -msse2 -c -o $@ $(SRCDIR)slide_sse.c

+

+slide_sse.lo: $(SRCDIR)slide_sse.c

+	-@mkdir objs 2>/dev/null || test -d objs

+	$(CC) $(SFLAGS) $(ZINC) -DPIC -msse2 -c -o objs/slide_sse.o $(SRCDIR)slide_sse.c

+	-@mv objs/slide_sse.o $@

+

+slide_avx2.o: $(SRCDIR)slide_avx2.c

+	$(CC) $(CFLAGS) $(ZINC) -mavx2 -c -o $@ $(SRCDIR)slide_avx2.c

+

+slide_avx2.lo: $(SRCDIR)slide_avx2.c

+	-@mkdir objs 2>/dev/null || test -d objs

+	$(CC) $(SFLAGS) $(ZINC) -DPIC -mavx2 -c -o objs/slide_avx2.o $(SRCDIR)slide_avx2.c

+	-@mv objs/slide_avx2.o $@

+

 crc32_test.o: $(SRCDIR)test/crc32_test.c $(SRCDIR)zlib.h zconf.h

 	$(CC) $(CFLAGS) $(ZINCOUT) -c -o $@ $(SRCDIR)test/crc32_test.c

--- /dev/null

+++ b/slide_sse.c

@@ -0,0 +1,47 @@

+/*

+ * SSE optimized hash slide

+ *

+ * Copyright (C) 2017 Intel Corporation

+ * Authors:

+ *   Arjan van de Ven	<arjan@linux.intel.com>

+ *   Jim Kukunas	<james.t.kukunas@linux.intel.com>

+ *

+ * For conditions of distribution and use, see copyright notice in zlib.h

+ */

+#include "deflate.h"

+#include <immintrin.h>

+

+void slide_hash_sse(deflate_state *s)

+{

+    unsigned n;

+    Posf *p;

+    uInt wsize = s->w_size;

+    z_const __m128i xmm_wsize = _mm_set1_epi16(s->w_size);

+

+    n = s->hash_size;

+    p = &s->head[n] - 8;

+    do {

+        __m128i value, result;

+

+	value = _mm_loadu_si128((__m128i *)p);

+	result= _mm_subs_epu16(value, xmm_wsize);

+	_mm_storeu_si128((__m128i *)p, result);

+	p -= 8;

+	n -= 8;

+    } while (n > 0);

+

+#ifndef FASTEST

+    n = wsize;

+    p = &s->prev[n] - 8;

+    do {

+        __m128i value, result;

+

+	value = _mm_loadu_si128((__m128i *)p);

+	result= _mm_subs_epu16(value, xmm_wsize);

+	_mm_storeu_si128((__m128i *)p, result);

+

+	p -= 8;

+	n -= 8;

+    } while (n > 0);

+#endif

+}

--- /dev/null

+++ b/slide_avx2.c

@@ -0,0 +1,44 @@

+/*

+ * AVX2 optimized hash slide

+ *

+ * Copyright (C) 2020 Intel Corporation

+ *

+ * For conditions of distribution and use, see copyright notice in zlib.h

+ */

+#include "deflate.h"

+#include <immintrin.h>

+

+void slide_hash_avx2(deflate_state *s)

+{

+    unsigned n;

+    Posf *p;

+    uInt wsize = s->w_size;

+    z_const __m256i ymm_wsize = _mm256_set1_epi16(s->w_size);

+

+    n = s->hash_size;

+    p = &s->head[n] - 16;

+    do {

+        __m256i value, result;

+

+        value = _mm256_loadu_si256((__m256i *)p);

+        result= _mm256_subs_epu16(value, ymm_wsize);

+        _mm256_storeu_si256((__m256i *)p, result);

+        p -= 16;

+        n -= 16;

+    } while (n > 0);

+

+#ifndef FASTEST

+    n = wsize;

+    p = &s->prev[n] - 16;

+    do {

+        __m256i value, result;

+

+        value = _mm256_loadu_si256((__m256i *)p);

+        result= _mm256_subs_epu16(value, ymm_wsize);

+        _mm256_storeu_si256((__m256i *)p, result);

+

+        p -= 16;

+        n -= 16;

+    } while (n > 0);

+#endif

+}

--- a/deflate.c

+++ b/deflate.c

@@ -90,6 +90,9 @@

 local int deflateStateCheck      OF((z_streamp strm));

 local void slide_hash     OF((deflate_state *s));

+local void slide_hash_c     OF((deflate_state *s));

+extern void slide_hash_sse     (deflate_state *s);

+extern void slide_hash_avx2     (deflate_state *s);

 local void fill_window    OF((deflate_state *s));

 local block_state deflate_stored OF((deflate_state *s, int flush));

 local block_state deflate_fast   OF((deflate_state *s, int flush));

@@ -212,7 +215,7 @@

  * bit values at the expense of memory usage). We slide even when level == 0 to

  * keep the hash table consistent if we switch back to level > 0 later.

  */

-local void slide_hash(s)

+local void slide_hash_c(s)

     deflate_state *s;

 {

     unsigned n, m;

@@ -238,6 +241,13 @@

 #endif

 }

+local void slide_hash(deflate_state *s) {

+    #ifdef AVX2_SLIDE

+        slide_hash_avx2(s);

+    #endif

+        slide_hash_sse(s);

+}

+

 /* ========================================================================= */

 int ZEXPORT deflateInit_(strm, level, version, stream_size)

     z_streamp strm;