build avx, sse4 versions of dfloor dceil floor

Author: isuruf

runtime/libpgmath/lib/common/CMakeLists.txt

@@ -391,6 +391,23 @@ if(${LIBPGMATH_SYSTEM_PROCESSOR} MATCHES "x86_64" AND NOT ${LIBPGMATH_WITH_GENER
   endif()
   libmath_add_object_library("${TARGET_NAME}.c" "${FLAGS}" "${DEFINITIONS}" "${TARGET_NAME}_build")
   add_dependencies("${TARGET_NAME}_build" ${TARGET_NAME})
+
+  foreach (FILENAME dceil dfloor floor)
+    foreach (VARIANT sse4 avx)
+      set(TARGET_NAME "${FILENAME}_${VARIANT}")
+      add_custom_command(OUTPUT ${TARGET_NAME}.c DEPENDS ${CMAKE_CURRENT_SOURCE_DIR}/${FILENAME}.c PRE_BUILD
+        COMMAND ${CMAKE_COMMAND} -E copy ${CMAKE_CURRENT_SOURCE_DIR}/${FILENAME}.c ${TARGET_NAME}.c)
+      add_custom_target(${TARGET_NAME} ALL DEPENDS "${TARGET_NAME}.c")
+      if (${VARIANT} STREQUAL "sse4")
+        set_property(SOURCE ${TARGET_NAME}.c APPEND_STRING PROPERTY COMPILE_FLAGS "-msse4.1 -DBUILD_SSE4")
+      elseif (${VARIANT} STREQUAL "avx")
+        set_property(SOURCE ${TARGET_NAME}.c APPEND_STRING PROPERTY COMPILE_FLAGS "-mavx -DBUILD_AVX")
+      endif()
+      libmath_add_object_library("${TARGET_NAME}.c" "${FLAGS}" "${DEFINITIONS}" "${TARGET_NAME}_build")
+      add_dependencies("${TARGET_NAME}_build" ${TARGET_NAME})
+    endforeach()
+  endforeach()
+
 elseif(${LIBPGMATH_SYSTEM_PROCESSOR} MATCHES "ppc64le")
   # Generate mth_128mask.c
   set(TARGET_NAME "mth_128mask")

runtime/libpgmath/lib/common/dceil.c

@@ -10,13 +10,13 @@
 #include    <immintrin.h>
 #endif
 
-#if     defined(__AVX__)
+#if     defined(BUILD_AVX)
 double
 __mth_i_dceil_avx(double x)
 {
   return _mm_cvtsd_f64(_mm_ceil_sd(_mm_set1_pd(x), _mm_set1_pd(x)));
 }
-#elif   defined(__SSE4_1__)
+#elif   defined(BUILD_SSE4)
 double
 __mth_i_dceil_sse(double x)
 {

runtime/libpgmath/lib/common/dfloor.c

@@ -10,13 +10,13 @@
 #include    <immintrin.h>
 #endif
 
-#if     defined(__AVX__)
+#if     defined(BUILD_AVX)
 double
 __mth_i_dfloor_avx(double x)
 {
   return _mm_cvtsd_f64(_mm_floor_sd(_mm_set1_pd(x), _mm_set1_pd(x)));
 }
-#elif   defined(__SSE4_1__)
+#elif   defined(BUILD_SSE4)
 double
 __mth_i_dfloor_sse(double x)
 {

runtime/libpgmath/lib/common/floor.c

@@ -10,13 +10,13 @@
 #include    <immintrin.h>
 #endif
 
-#if     defined(__AVX__)
+#if     defined(BUILD_AVX)
 float
 __mth_i_floor_avx(float x)
 {
   return _mm_cvtss_f32(_mm_floor_ss(_mm_set1_ps(x), _mm_set1_ps(x)));
 }
-#elif   defined(__SSE4_1__)
+#elif   defined(BUILD_SSE4)
 float
 __mth_i_floor_sse(float x)
 {

runtime/libpgmath/lib/x86_64/math_tables/mth_ceildefs.h

@@ -7,6 +7,8 @@
 
 MTHINTRIN(ceil  , ss   , any        ,  __mth_i_ceil         ,  __mth_i_ceil         , __mth_i_ceil          ,__math_dispatch_error)
 MTHINTRIN(ceil  , ds   , any        ,  __mth_i_dceil        , __mth_i_dceil         , __mth_i_dceil         ,__math_dispatch_error)
+MTHINTRIN(ceil  , ds   , sse4       ,  __mth_i_dceil_sse    , __mth_i_dceil_sse     , __mth_i_dceil_sse     ,__math_dispatch_error)
+MTHINTRIN(ceil  , ds   , avx        ,  __mth_i_dceil_avx    , __mth_i_dceil_avx     , __mth_i_dceil_avx     ,__math_dispatch_error)
 MTHINTRIN(ceil  , sv4  , any        ,  __gs_ceil_4_f        ,  __gs_ceil_4_r        , __gs_ceil_4_p         ,__math_dispatch_error)
 MTHINTRIN(ceil  , dv2  , any        ,  __gd_ceil_2_f        ,  __gd_ceil_2_r        , __gd_ceil_2_p         ,__math_dispatch_error)
 MTHINTRIN(ceil  , sv8  , any        ,  __gs_ceil_8_f        ,  __gs_ceil_8_r        , __gs_ceil_8_p         ,__math_dispatch_error)

runtime/libpgmath/lib/x86_64/math_tables/mth_floordefs.h

@@ -7,6 +7,10 @@
 
 MTHINTRIN(floor  , ss   , any        ,  __mth_i_floor         ,  __mth_i_floor         , __mth_i_floor          ,__math_dispatch_error)
 MTHINTRIN(floor  , ds   , any        ,  __mth_i_dfloor        ,  __mth_i_dfloor        , __mth_i_dfloor         ,__math_dispatch_error)
+MTHINTRIN(floor  , ss   , sse4       ,  __mth_i_floor_sse     ,  __mth_i_floor_sse     , __mth_i_floor_sse      ,__math_dispatch_error)
+MTHINTRIN(floor  , ds   , sse4       ,  __mth_i_dfloor_sse    ,  __mth_i_dfloor_sse    , __mth_i_dfloor_sse     ,__math_dispatch_error)
+MTHINTRIN(floor  , ss   , avx        ,  __mth_i_floor_avx     ,  __mth_i_floor_avx     , __mth_i_floor_avx      ,__math_dispatch_error)
+MTHINTRIN(floor  , ds   , avx        ,  __mth_i_dfloor_avx    ,  __mth_i_dfloor_avx    , __mth_i_dfloor_avx     ,__math_dispatch_error)
 MTHINTRIN(floor  , sv4  , any        ,  __gs_floor_4_f        ,  __gs_floor_4_r        , __gs_floor_4_p         ,__math_dispatch_error)
 MTHINTRIN(floor  , dv2  , any        ,  __gd_floor_2_f        ,  __gd_floor_2_r        , __gd_floor_2_p         ,__math_dispatch_error)
 MTHINTRIN(floor  , sv8  , any        ,  __gs_floor_8_f        ,  __gs_floor_8_r        , __gs_floor_8_p         ,__math_dispatch_error)