Merge pull request #143 from VectorCamp/develop

Prepare for new release 5.4.9

CMakeLists.txt

@@ -4,7 +4,7 @@ project (vectorscan C CXX)
 
 set (HS_MAJOR_VERSION 5)
 set (HS_MINOR_VERSION 4)
-set (HS_PATCH_VERSION 8)
+set (HS_PATCH_VERSION 9)
 set (HS_VERSION ${HS_MAJOR_VERSION}.${HS_MINOR_VERSION}.${HS_PATCH_VERSION})
 
 set(CMAKE_MODULE_PATH ${PROJECT_SOURCE_DIR}/cmake)
@@ -544,7 +544,6 @@ endif()
 set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${ARCH_C_FLAGS}")
 set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${ARCH_CXX_FLAGS}")
 
-add_subdirectory(util)
 add_subdirectory(doc/dev-reference)
 
 # PCRE check, we have a fixed requirement for PCRE to use Chimera
@@ -588,7 +587,7 @@ set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${EXTRA_C_FLAGS}")
 set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${EXTRA_CXX_FLAGS}")
 endif()
 
-set(RAGEL_C_FLAGS "-Wno-unused")
+set(RAGEL_C_FLAGS "-Wno-unused -funsigned-char")
 
 set_source_files_properties(
     ${CMAKE_BINARY_DIR}/src/parser/Parser.cpp
@@ -604,6 +603,8 @@ set_source_files_properties(
 
 ragelmaker(src/parser/control_verbs.rl)
 
+add_subdirectory(util)
+
 SET(hs_HEADERS
     src/hs.h
     src/hs_common.h

cmake/ragel.cmake

@@ -7,7 +7,7 @@ function(ragelmaker src_rl)
     add_custom_command(
         OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/${src_dir}/${src_file}.cpp
         COMMAND ${CMAKE_COMMAND} -E make_directory ${CMAKE_CURRENT_BINARY_DIR}/${src_dir}
-        COMMAND ${RAGEL} ${CMAKE_CURRENT_SOURCE_DIR}/${src_rl} -o ${rl_out}
+        COMMAND ${RAGEL} ${CMAKE_CURRENT_SOURCE_DIR}/${src_rl} -o ${rl_out} -G0
         DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/${src_rl}
         )
     add_custom_target(ragel_${src_file} DEPENDS ${rl_out})

src/fdr/teddy_runtime_common.h

@@ -348,7 +348,7 @@ static really_inline
 m512 vectoredLoad512(m512 *p_mask, const u8 *ptr, const size_t start_offset,
                      const u8 *lo, const u8 *hi, const u8 *hbuf, size_t hlen,
                      const u32 nMasks) {
-    m512 val;
+    m512 val = zeroes512();
 
     uintptr_t copy_start;
     uintptr_t copy_len;

src/hs.h

@@ -43,7 +43,7 @@
 
 #define HS_MAJOR      5
 #define HS_MINOR      4
-#define HS_PATCH      0
+#define HS_PATCH      9
 
 #include "hs_compile.h"
 #include "hs_runtime.h"

src/parser/Parser.rl

@@ -272,6 +272,7 @@ unichar readUtf8CodePoint4c(const char *s) {
 
 %%{
     machine regex;
+    alphtype unsigned char;
 
     action throwUnsupportedEscape {
         ostringstream str;

src/parser/control_verbs.rl

@@ -54,6 +54,7 @@ const char *read_control_verbs(const char *ptr, const char *end, size_t start,
 
     %%{
         machine ControlVerbs;
+        alphtype unsigned char;
 
         # Verbs that we recognise but do not support.
         unhandledVerbs = '(*' (

src/util/arch/common/simd_utils.h

@@ -254,7 +254,7 @@ static really_inline m256 loadu2x128(const void *ptr) {
 // aligned store
 static really_inline void store256(void *ptr, m256 a) {
     assert(ISALIGNED_N(ptr, alignof(m256)));
-    ptr = assume_aligned(ptr, 16);
+    ptr = vectorscan_assume_aligned(ptr, 16);
     *(m256 *)ptr = a;
 }
 
@@ -486,7 +486,7 @@ static really_inline m384 load384(const void *ptr) {
 // aligned store
 static really_inline void store384(void *ptr, m384 a) {
     assert(ISALIGNED_16(ptr));
-    ptr = assume_aligned(ptr, 16);
+    ptr = vectorscan_assume_aligned(ptr, 16);
     *(m384 *)ptr = a;
 }
 

src/util/arch/x86/simd_utils.h

@@ -278,14 +278,14 @@ static really_inline m128 andnot128(m128 a, m128 b) {
 // aligned load
 static really_inline m128 load128(const void *ptr) {
     assert(ISALIGNED_N(ptr, alignof(m128)));
-    ptr = assume_aligned(ptr, 16);
+    ptr = vectorscan_assume_aligned(ptr, 16);
     return _mm_load_si128((const m128 *)ptr);
 }
 
 // aligned store
 static really_inline void store128(void *ptr, m128 a) {
     assert(ISALIGNED_N(ptr, alignof(m128)));
-    ptr = assume_aligned(ptr, 16);
+    ptr = vectorscan_assume_aligned(ptr, 16);
     *(m128 *)ptr = a;
 }
 

src/util/simd_utils.h

@@ -40,17 +40,17 @@
 // it's available. Note that we need to handle C or C++ compilation.
 #ifdef __cplusplus
 #  ifdef HAVE_CXX_BUILTIN_ASSUME_ALIGNED
-#    define assume_aligned(x, y) __builtin_assume_aligned((x), (y))
+#    define vectorscan_assume_aligned(x, y) __builtin_assume_aligned((x), (y))
 #  endif
 #else
 #  ifdef HAVE_CC_BUILTIN_ASSUME_ALIGNED
-#    define assume_aligned(x, y) __builtin_assume_aligned((x), (y))
+#    define vectorscan_assume_aligned(x, y) __builtin_assume_aligned((x), (y))
 #  endif
 #endif
 
 // Fallback to identity case.
-#ifndef assume_aligned
-#define assume_aligned(x, y) (x)
+#ifndef vectorscan_assume_aligned
+#define vectorscan_assume_aligned(x, y) (x)
 #endif
 
 #ifdef __cplusplus

src/util/supervector/arch/arm/impl.cpp

@@ -518,7 +518,7 @@ template <>
 really_inline SuperVector<16> SuperVector<16>::load(void const *ptr)
 {
     assert(ISALIGNED_N(ptr, alignof(SuperVector::size)));
-    ptr = assume_aligned(ptr, SuperVector::size);
+    ptr = vectorscan_assume_aligned(ptr, SuperVector::size);
     return {vld1q_s32((const int32_t *)ptr)};
 }
 

src/util/supervector/arch/x86/impl.cpp

@@ -515,7 +515,7 @@ template <>
 really_inline SuperVector<16> SuperVector<16>::load(void const *ptr)
 {
     assert(ISALIGNED_N(ptr, alignof(SuperVector::size)));
-    ptr = assume_aligned(ptr, SuperVector::size);
+    ptr = vectorscan_assume_aligned(ptr, SuperVector::size);
     return _mm_load_si128((const m128 *)ptr);
 }
 
@@ -1119,7 +1119,7 @@ template <>
 really_inline SuperVector<32> SuperVector<32>::load(void const *ptr)
 {
     assert(ISALIGNED_N(ptr, alignof(SuperVector::size)));
-    ptr = assume_aligned(ptr, SuperVector::size);
+    ptr = vectorscan_assume_aligned(ptr, SuperVector::size);
     return {_mm256_load_si256((const m256 *)ptr)};
 }
 
@@ -1769,7 +1769,7 @@ template <>
 really_inline SuperVector<64> SuperVector<64>::load(void const *ptr)
 {
     assert(ISALIGNED_N(ptr, alignof(SuperVector::size)));
-    ptr = assume_aligned(ptr, SuperVector::size);
+    ptr = vectorscan_assume_aligned(ptr, SuperVector::size);
     return {_mm512_load_si512((const m512 *)ptr)};
 }
 

src/util/supervector/supervector.hpp

@@ -76,17 +76,17 @@ using Z_TYPE = u32;
 // it's available. Note that we need to handle C or C++ compilation.
 #ifdef __cplusplus
 #  ifdef HAVE_CXX_BUILTIN_ASSUME_ALIGNED
-#    define assume_aligned(x, y) __builtin_assume_aligned((x), (y))
+#    define vectorscan_assume_aligned(x, y) __builtin_assume_aligned((x), (y))
 #  endif
 #else
 #  ifdef HAVE_CC_BUILTIN_ASSUME_ALIGNED
-#    define assume_aligned(x, y) __builtin_assume_aligned((x), (y))
+#    define vectorscan_assume_aligned(x, y) __builtin_assume_aligned((x), (y))
 #  endif
 #endif
 
 // Fallback to identity case.
-#ifndef assume_aligned
-#define assume_aligned(x, y) (x)
+#ifndef vectorscan_assume_aligned
+#define vectorscan_assume_aligned(x, y) (x)
 #endif
 
 template <uint16_t SIZE>

tools/hscollider/ColliderCorporaParser.rl

@@ -57,6 +57,7 @@ char unhex(const char *start, UNUSED const char *end) {
 
 %%{
     machine FileCorporaParser;
+    alphtype unsigned char;
 
     action accumulateNum {
         num = (num * 10) + (fc - '0');

util/CMakeLists.txt

@@ -6,8 +6,10 @@ set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${EXTRA_CXX_FLAGS} ${HS_CXX_FLAGS}")
 include_directories(${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_BINARY_DIR}
     ${PROJECT_SOURCE_DIR})
 
+message("RAGEL_C_FLAGS" ${RAGEL_C_FLAGS})
+
 set_source_files_properties(
-    ${CMAKE_BINARY_DIR}/tools/ExpressionParser.cpp
+    ${CMAKE_BINARY_DIR}/util/ExpressionParser.cpp
     PROPERTIES
     COMPILE_FLAGS "${RAGEL_C_FLAGS}")
 

util/ExpressionParser.rl

@@ -55,6 +55,7 @@ enum ParamKey {
 
 %%{
     machine ExpressionParser;
+    alphtype unsigned char;
 
     action accumulateNum {
         num = (num * 10) + (fc - '0');