Version 1.0.1: Improved cache coherence for ARMv8 architecture.

IlyaGrebnov · Jun 21, 2022 · b15c658 · b15c658
1 parent 09afef5
commit b15c658
Show file tree

Hide file tree

Showing 7 changed files with 200 additions and 183 deletions.
diff --git a/CHANGES b/CHANGES
@@ -1,2 +1,6 @@
-* 2022-06-12 : Version 1.0.0
-  * Initial public release of the esa-matchfinder.
+Changes in 1.0.1 (June 19, 2022)
+- Improved cache coherence for ARMv8 architecture.
+
+Changes in 1.0.0 (June 12, 2022)
+- Initial public release of the esa-matchfinder.
+
diff --git a/README.md b/README.md
@@ -29,6 +29,8 @@ The esa-matchfinder finds all distance optimal matches (between min_match_length
 The esa-matchfinder released under the [Apache License Version 2.0](LICENSE "Apache license") and is considered suitable for production use. However, no warranty or fitness for a particular purpose is expressed or implied.
 
 ## Changes
+* June 19, 2022 (1.0.1)
+  * Improved cache coherence for ARMv8 architecture.
 * June 12, 2022 (1.0.0)
   * Initial public release of the esa-matchfinder.
 

diff --git a/VERSION b/VERSION
@@ -1 +1 @@
-1.0.0
+1.0.1
diff --git a/esa_matchfinder.c b/esa_matchfinder.c
@@ -127,20 +127,20 @@ typedef struct ESA_MF_CONTEXT
 #endif
 
 #if defined(ESA_MF_HAS_BUILTIN_PREFECTCH)
-    #define esa_matchfinder_prefetchr(address) __builtin_prefetch((const void *)(address), 0, 0)
-    #define esa_matchfinder_prefetchw(address) __builtin_prefetch((const void *)(address), 1, 0)
+    #define esa_matchfinder_prefetchr(address) __builtin_prefetch((const void *)(address), 0, 3)
+    #define esa_matchfinder_prefetchw(address) __builtin_prefetch((const void *)(address), 1, 3)
 #elif defined (_M_IX86) || defined (_M_AMD64)
     #include <intrin.h>
-    #define esa_matchfinder_prefetchr(address) _mm_prefetch((const void *)(address), _MM_HINT_NTA)
+    #define esa_matchfinder_prefetchr(address) _mm_prefetch((const void *)(address), _MM_HINT_T0)
     #define esa_matchfinder_prefetchw(address) _m_prefetchw((const void *)(address))
 #elif defined (_M_ARM)
     #include <intrin.h>
     #define esa_matchfinder_prefetchr(address) __prefetch((const void *)(address))
     #define esa_matchfinder_prefetchw(address) __prefetchw((const void *)(address))
 #elif defined (_M_ARM64)
     #include <intrin.h>
-    #define esa_matchfinder_prefetchr(address) __prefetch2((const void *)(address), 1)
-    #define esa_matchfinder_prefetchw(address) __prefetch2((const void *)(address), 17)
+    #define esa_matchfinder_prefetchr(address) __prefetch2((const void *)(address), 0)
+    #define esa_matchfinder_prefetchw(address) __prefetch2((const void *)(address), 16)
 #else
     #error Your compiler, configuration or platform is not supported.
 #endif
@@ -384,7 +384,9 @@ static ptrdiff_t esa_matchfinder_build_interval_tree
     for (ptrdiff_t i = omp_block_start + omp_block_size - 1; i >= omp_block_start; i -= 1)
     {
         esa_matchfinder_prefetchr(&sa_parent_link[i - 2 * prefetch_distance]);
+
         esa_matchfinder_prefetchw(&plcp_leaf_link[sa_parent_link[i - prefetch_distance]]);
+        esa_matchfinder_prefetchw(&sa_parent_link[next_interval_index - prefetch_distance]);
 
         uint64_t next_pos                   =  sa_parent_link[i];
         uint64_t next_lcp                   =  (uint64_t)plcp_leaf_link[next_pos] - min_match_length;

diff --git a/libsais/CHANGES b/libsais/CHANGES
@@ -1,3 +1,6 @@
+Changes in 2.7.1 (June 19, 2022)
+- Improved cache coherence for ARMv8 architecture.
+
 Changes in 2.7.0 (April 12, 2022)
 - Support for longest common prefix array (LCP) construction.
 

diff --git a/libsais/VERSION b/libsais/VERSION
@@ -1 +1 @@
-2.7.0
+2.7.1
diff --git a/libsais/libsais.c b/libsais/libsais.c