Merge LZX compression updates

[wimlib] / include / wimlib / matchfinder_avx2.h

diff --git a/include/wimlib/matchfinder_avx2.h b/include/wimlib/matchfinder_avx2.h
index fe98b636c9a0d27c5dd7dfab35786b34447be2cf..bdf10d212515aaa1dbcc7a794c5bf1c7b5e5ff05 100644 (file)
--- a/include/wimlib/matchfinder_avx2.h
+++ b/include/wimlib/matchfinder_avx2.h
@@ -2,6 +2,12 @@
  * matchfinder_avx2.h
  *
  * Matchfinding routines optimized for Intel AVX2 (Advanced Vector Extensions).
  * matchfinder_avx2.h
  *
  * Matchfinding routines optimized for Intel AVX2 (Advanced Vector Extensions).

+ *
+ * Author:     Eric Biggers
+ * Year:       2014, 2015
+ *
+ * The author dedicates this file to the public domain.
+ * You can do whatever you want with this file.

  */
 
 #include <immintrin.h>
  */
 
 #include <immintrin.h>


@@ -16,9 +22,9 @@ matchfinder_init_avx2(pos_t *data, size_t size)
                return false;
 
        if (sizeof(pos_t) == 2)
                return false;
 
        if (sizeof(pos_t) == 2)

-               v = _mm256_set1_epi16(MATCHFINDER_INITVAL);
+               v = _mm256_set1_epi16((u16)MATCHFINDER_NULL);

        else if (sizeof(pos_t) == 4)
        else if (sizeof(pos_t) == 4)

-               v = _mm256_set1_epi32(MATCHFINDER_INITVAL);
+               v = _mm256_set1_epi32((u32)MATCHFINDER_NULL);

        else
                return false;
 
        else
                return false;
 


@@ -33,32 +39,3 @@ matchfinder_init_avx2(pos_t *data, size_t size)
        } while (--n);
        return true;
 }
        } while (--n);
        return true;
 }

-
-static inline bool
-matchfinder_rebase_avx2(pos_t *data, size_t size)
-{
-       __m256i v, *p;
-       size_t n;
-
-       if ((size % sizeof(__m256i) * 4 != 0))
-               return false;
-
-       if (sizeof(pos_t) == 2)
-               v = _mm256_set1_epi16((pos_t)-MATCHFINDER_WINDOW_SIZE);
-       else if (sizeof(pos_t) == 4)
-               v = _mm256_set1_epi32((pos_t)-MATCHFINDER_WINDOW_SIZE);
-       else
-               return false;
-
-       p = (__m256i *)data;
-       n = size / (sizeof(__m256i) * 4);
-       do {
-               /* PADDSW: Add Packed Signed Integers With Signed Saturation  */
-               p[0] = _mm256_adds_epi16(p[0], v);
-               p[1] = _mm256_adds_epi16(p[1], v);
-               p[2] = _mm256_adds_epi16(p[2], v);
-               p[3] = _mm256_adds_epi16(p[3], v);
-               p += 4;
-       } while (--n);
-       return true;
-}