diff --git a/CMakeLists.txt b/CMakeLists.txt
index d266b7a..55557ee 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -22,4 +22,4 @@ if("${CMAKE_BUILD_TYPE}" MATCHES "Release")
     endif()
 endif()
 
-add_library("Aaru.Checksums.Native" SHARED adler32.h adler32.c crc16.h crc16.c crc16_ccitt.h crc16_ccitt.c crc32.c crc32.h crc64.c crc64.h fletcher16.h fletcher16.c fletcher32.h fletcher32.c library.h spamsum.c spamsum.h crc32_clmul.c crc64_clmul.c simd.c simd.h adler32_ssse3.c)
+add_library("Aaru.Checksums.Native" SHARED adler32.h adler32.c crc16.h crc16.c crc16_ccitt.h crc16_ccitt.c crc32.c crc32.h crc64.c crc64.h fletcher16.h fletcher16.c fletcher32.h fletcher32.c library.h spamsum.c spamsum.h crc32_clmul.c crc64_clmul.c simd.c simd.h adler32_ssse3.c adler32_avx2.c)
diff --git a/adler32.c b/adler32.c
index 19b314c..1ad47b5 100644
--- a/adler32.c
+++ b/adler32.c
@@ -51,6 +51,13 @@ AARU_EXPORT int AARU_CALL adler32_update(adler32_ctx* ctx, const uint8_t* data,
 
 #if defined(__x86_64__) || defined(__amd64) || defined(_M_AMD64) || defined(_M_X64) || defined(__I386__) ||            \
     defined(__i386__) || defined(__THW_INTEL) || defined(_M_IX86)
+    if(have_avx2())
+    {
+        adler32_avx2(&ctx->sum1, &ctx->sum2, data, len);
+
+        return 0;
+    }
+
     if(have_ssse3())
     {
         adler32_ssse3(&ctx->sum1, &ctx->sum2, data, len);
diff --git a/adler32.h b/adler32.h
index 92bda16..ad8f84f 100644
--- a/adler32.h
+++ b/adler32.h
@@ -38,6 +38,7 @@ AARU_EXPORT void AARU_CALL         adler32_free(adler32_ctx* ctx);
     defined(__i386__) || defined(__THW_INTEL) || defined(_M_IX86)
 
 void adler32_ssse3(uint16_t* sum1, uint16_t* sum2, const unsigned char* buf, size_t len);
+void adler32_avx2(uint16_t* sum1, uint16_t* sum2, const unsigned char* buf, size_t len);
 
 #endif
 
diff --git a/adler32_avx2.c b/adler32_avx2.c
new file mode 100644
index 0000000..7075e52
--- /dev/null
+++ b/adler32_avx2.c
@@ -0,0 +1,150 @@
+//
+// Created by claunia on 28/9/21.
+//
+
+#include <immintrin.h>
+#include <stdint.h>
+
+#include "library.h"
+#include "adler32.h"
+#include "simd.h"
+
+AVX2 void adler32_avx2(uint16_t* sum1, uint16_t* sum2, const unsigned char* buf, size_t len)
+{
+    uint32_t s1 = *sum1;
+    uint32_t s2 = *sum2;
+
+    /*
+     * Process the data in blocks.
+     */
+    const unsigned BLOCK_SIZE = 1 << 5;
+    size_t         blocks     = len / BLOCK_SIZE;
+    len -= blocks * BLOCK_SIZE;
+
+    while(blocks)
+    {
+        unsigned n = NMAX / BLOCK_SIZE; /* The NMAX constraint. */
+
+        if(n > blocks) n = (unsigned)blocks;
+        blocks -= n;
+
+        const __m256i tap  = _mm256_set_epi8(1,
+                                            2,
+                                            3,
+                                            4,
+                                            5,
+                                            6,
+                                            7,
+                                            8,
+                                            9,
+                                            10,
+                                            11,
+                                            12,
+                                            13,
+                                            14,
+                                            15,
+                                            16,
+                                            17,
+                                            18,
+                                            19,
+                                            20,
+                                            21,
+                                            22,
+                                            23,
+                                            24,
+                                            25,
+                                            26,
+                                            27,
+                                            28,
+                                            29,
+                                            30,
+                                            31,
+                                            32);
+        const __m256i zero = _mm256_setzero_si256();
+        const __m256i ones = _mm256_set1_epi16(1);
+
+        /*
+         * Process n blocks of data. At most NMAX data bytes can be
+         * processed before s2 must be reduced modulo BASE.
+         */
+        __m256i v_ps = _mm256_set_epi32(0, 0, 0, 0, 0, 0, 0, (s1 * n));
+        __m256i v_s2 = _mm256_set_epi32(0, 0, 0, 0, 0, 0, 0, s2);
+        __m256i v_s1 = _mm256_setzero_si256();
+        do {
+            /*
+             * Load 32 input bytes.
+             */
+            const __m256i bytes = _mm256_lddqu_si256((__m256i*)(buf));
+
+            /*
+             * Add previous block byte sum to v_ps.
+             */
+            v_ps = _mm256_add_epi32(v_ps, v_s1);
+            /*
+             * Horizontally add the bytes for s1, multiply-adds the
+             * bytes by [ 32, 31, 30, ... ] for s2.
+             */
+            v_s1              = _mm256_add_epi32(v_s1, _mm256_sad_epu8(bytes, zero));
+            const __m256i mad = _mm256_maddubs_epi16(bytes, tap);
+            v_s2              = _mm256_add_epi32(v_s2, _mm256_madd_epi16(mad, ones));
+
+            buf += BLOCK_SIZE;
+        } while(--n);
+
+        __m128i sum = _mm_add_epi32(_mm256_castsi256_si128(v_s1), _mm256_extracti128_si256(v_s1, 1));
+        __m128i hi  = _mm_unpackhi_epi64(sum, sum);
+        sum         = _mm_add_epi32(hi, sum);
+        hi          = _mm_shuffle_epi32(sum, 177);
+        sum         = _mm_add_epi32(sum, hi);
+        s1 += _mm_cvtsi128_si32(sum);
+
+        v_s2 = _mm256_add_epi32(v_s2, _mm256_slli_epi32(v_ps, 5));
+        sum  = _mm_add_epi32(_mm256_castsi256_si128(v_s2), _mm256_extracti128_si256(v_s2, 1));
+        hi   = _mm_unpackhi_epi64(sum, sum);
+        sum  = _mm_add_epi32(hi, sum);
+        hi   = _mm_shuffle_epi32(sum, 177);
+        sum  = _mm_add_epi32(sum, hi);
+        s2   = _mm_cvtsi128_si32(sum);
+
+        /*
+         * Reduce.
+         */
+        s1 %= ADLER_MODULE;
+        s2 %= ADLER_MODULE;
+    }
+
+    /*
+     * Handle leftover data.
+     */
+    if(len)
+    {
+        if(len >= 16)
+        {
+            s2 += (s1 += *buf++);
+            s2 += (s1 += *buf++);
+            s2 += (s1 += *buf++);
+            s2 += (s1 += *buf++);
+            s2 += (s1 += *buf++);
+            s2 += (s1 += *buf++);
+            s2 += (s1 += *buf++);
+            s2 += (s1 += *buf++);
+            s2 += (s1 += *buf++);
+            s2 += (s1 += *buf++);
+            s2 += (s1 += *buf++);
+            s2 += (s1 += *buf++);
+            s2 += (s1 += *buf++);
+            s2 += (s1 += *buf++);
+            s2 += (s1 += *buf++);
+            s2 += (s1 += *buf++);
+            len -= 16;
+        }
+        while(len--) { s2 += (s1 += *buf++); }
+        if(s1 >= ADLER_MODULE) s1 -= ADLER_MODULE;
+        s2 %= ADLER_MODULE;
+    }
+    /*
+     * Return the recombined sums.
+     */
+    *sum1 = s1 & 0xFFFF;
+    *sum2 = s2 & 0xFFFF;
+}
diff --git a/simd.c b/simd.c
index ae42f48..91fd687 100644
--- a/simd.c
+++ b/simd.c
@@ -35,6 +35,29 @@ static void cpuid(int info, unsigned* eax, unsigned* ebx, unsigned* ecx, unsigne
 #endif
 }
 
+static void cpuidex(int info, int count, unsigned* eax, unsigned* ebx, unsigned* ecx, unsigned* edx)
+{
+#ifdef _MSC_VER
+    unsigned int registers[4];
+    __cpuidex(registers, info, count);
+    *eax = registers[0];
+    *ebx = registers[1];
+    *ecx = registers[2];
+    *edx = registers[3];
+#else
+    /* GCC, clang */
+    unsigned int _eax;
+    unsigned int _ebx;
+    unsigned int _ecx;
+    unsigned int _edx;
+    __cpuid_count(info, count, _eax, _ebx, _ecx, _edx);
+    *eax = _eax;
+    *ebx = _ebx;
+    *ecx = _ecx;
+    *edx = _edx;
+#endif
+}
+
 int have_clmul(void)
 {
     unsigned eax, ebx, ecx, edx;
@@ -56,4 +79,11 @@ int have_ssse3(void)
     return ecx & 0x200;
 }
 
+int have_avx2(void)
+{
+    unsigned eax, ebx, ecx, edx;
+    cpuidex(7 /* extended feature bits */, 0, &eax, &ebx, &ecx, &edx);
+
+    return ebx & 0x20;
+}
 #endif
\ No newline at end of file
diff --git a/simd.h b/simd.h
index 301ec6a..a302eb7 100644
--- a/simd.h
+++ b/simd.h
@@ -1,6 +1,9 @@
 #if defined(__x86_64__) || defined(__amd64) || defined(_M_AMD64) || defined(_M_X64) || defined(__I386__) ||            \
     defined(__i386__) || defined(__THW_INTEL) || defined(_M_IX86)
 
+#define AVX2 __attribute__((target("avx2")))
+
 int have_clmul(void);
 int have_ssse3(void);
+int have_avx2(void);
 #endif