[C++] Add RISC-V Vector (RVV) support and dynamic dispatch

.gitignore

@@ -59,6 +59,7 @@ python/doc/
 
 .vscode
 .idea/
+.icodemate/
 .pytest_cache/
 pkgs
 docker_cache

cpp/cmake_modules/DefineOptions.cmake

@@ -171,6 +171,10 @@ takes precedence over ccache if a storage backend is configured" ON)
                        "SVE128" # fixed size SVE
                        "SVE256" # "
                        "SVE512" # "
+                       "RVV" # size agnostic RVV
+                       "RVV128" # fixed size RVV
+                       "RVV256" # "
+                       "RVV512" # "
                        "DEFAULT")
 
   define_option_string(ARROW_RUNTIME_SIMD_LEVEL
@@ -183,6 +187,9 @@ takes precedence over ccache if a storage backend is configured" ON)
                        "SVE128" # fixed size SVE
                        "SVE256" # "
                        "SVE512" # "
+                       "RVV128" # fixed size RVV
+                       "RVV256" # "
+                       "RVV512" # "
                        "MAX")
 
   define_option(ARROW_ALTIVEC "Build with Altivec if compiler has support" ON)

cpp/cmake_modules/SetupCxxFlags.cmake

@@ -167,6 +167,24 @@ elseif(ARROW_CPU_FLAG STREQUAL "aarch64")
   if(ARROW_SIMD_LEVEL STREQUAL "DEFAULT")
     set(ARROW_SIMD_LEVEL "NEON")
   endif()
+elseif(ARROW_CPU_FLAG STREQUAL "riscv64")
+  set(ARROW_RVV_MARCH "rv64gcv")
+  check_cxx_compiler_flag("-march=${ARROW_RVV_MARCH}" CXX_SUPPORTS_RVV)
+  if(CXX_SUPPORTS_RVV AND ARROW_RUNTIME_SIMD_LEVEL MATCHES "^(RVV128|RVV256|RVV512|MAX)$")
+    set(ARROW_HAVE_RUNTIME_RVV128 ON)
+    add_definitions(-DARROW_HAVE_RUNTIME_RVV128)
+  endif()
+  if(CXX_SUPPORTS_RVV AND ARROW_RUNTIME_SIMD_LEVEL MATCHES "^(RVV256|RVV512|MAX)$")
+    set(ARROW_HAVE_RUNTIME_RVV256 ON)
+    add_definitions(-DARROW_HAVE_RUNTIME_RVV256)
+  endif()
+  if(CXX_SUPPORTS_RVV AND ARROW_RUNTIME_SIMD_LEVEL MATCHES "^(RVV512|MAX)$")
+    set(ARROW_HAVE_RUNTIME_RVV512 ON)
+    add_definitions(-DARROW_HAVE_RUNTIME_RVV512)
+  endif()
+  if(ARROW_SIMD_LEVEL STREQUAL "DEFAULT")
+    set(ARROW_SIMD_LEVEL "NONE")
+  endif()
 endif()
 
 # Support C11
@@ -547,6 +565,24 @@ if(ARROW_CPU_FLAG STREQUAL "aarch64")
     message(WARNING "ARROW_SIMD_LEVEL=${ARROW_SIMD_LEVEL} not supported by Arm.")
   endif()
 endif()
+if(ARROW_CPU_FLAG STREQUAL "riscv64")
+  if(ARROW_SIMD_LEVEL MATCHES "RVV")
+    if(NOT CXX_SUPPORTS_RVV)
+      message(FATAL_ERROR "RVV required but compiler doesn't support it.")
+    endif()
+    set(CXX_COMMON_FLAGS "${CXX_COMMON_FLAGS} -march=${ARROW_RVV_MARCH}")
+    string(REGEX MATCH "[0-9]+" RVV_VECTOR_BITS ${ARROW_SIMD_LEVEL})
+    if(RVV_VECTOR_BITS)
+      set(ARROW_HAVE_RVV${RVV_VECTOR_BITS} ON)
+      add_definitions(-DARROW_HAVE_RVV${RVV_VECTOR_BITS})
+    else()
+      set(ARROW_HAVE_RVV ON)
+      add_definitions(-DARROW_HAVE_RVV)
+    endif()
+  elseif(NOT ARROW_SIMD_LEVEL STREQUAL "NONE")
+    message(WARNING "ARROW_SIMD_LEVEL=${ARROW_SIMD_LEVEL} not supported by riscv64.")
+  endif()
+endif()
 
 if(NOT WIN32 AND NOT APPLE)
   if(ARROW_USE_MOLD)

cpp/src/arrow/util/cpu_info.cc

@@ -98,6 +98,14 @@ void OsRetrieveCpuInfo(int64_t* hardware_flags, CpuInfo::Vendor* vendor) {
   *hardware_flags |= (sve && sve_size == 32) ? CpuInfo::SVE256 : 0;
   *hardware_flags |= (sve && sve_size == 64) ? CpuInfo::SVE512 : 0;
 
+  // RVV and length
+  const bool rvv = cpu.rvv();
+  const auto rvv_size = cpu.rvv_size_bytes();
+  *hardware_flags |= rvv ? CpuInfo::RVV : 0;
+  *hardware_flags |= (rvv && rvv_size == 16) ? CpuInfo::RVV128 : 0;
+  *hardware_flags |= (rvv && rvv_size == 32) ? CpuInfo::RVV256 : 0;
+  *hardware_flags |= (rvv && rvv_size == 64) ? CpuInfo::RVV512 : 0;
+
   // x86 only
   switch (cpu.known_manufacturer()) {
     case (xsimd::x86_manufacturer::intel):

cpp/src/arrow/util/cpu_info.h

@@ -57,6 +57,12 @@ class ARROW_EXPORT CpuInfo {
   static constexpr int64_t SVE256 = (1LL << 34);
   static constexpr int64_t SVE512 = (1LL << 35);
 
+  /// RISC-V features
+  static constexpr int64_t RVV = (1LL << 37);
+  static constexpr int64_t RVV128 = (1LL << 38);
+  static constexpr int64_t RVV256 = (1LL << 39);
+  static constexpr int64_t RVV512 = (1LL << 40);
+
   /// Cache enums for L1 (data), L2 and L3
   enum class CacheLevel { L1 = 0, L2, L3, Last = L3 };
 

cpp/src/arrow/util/dispatch_internal.h

@@ -38,6 +38,9 @@ enum class DispatchLevel : int {
   SVE128,
   SVE256,
   SVE512,
+  RVV128,
+  RVV256,
+  RVV512,
   MAX
 };
 
@@ -221,6 +224,36 @@ constexpr DynamicDispatchTarget<Func> BestDispatchTarget(
 #  define ARROW_DISPATCH_TARGET_SVE512(func)
 #endif
 
+#if defined(ARROW_HAVE_RVV128) || defined(ARROW_HAVE_RUNTIME_RVV128)
+#  define ARROW_DISPATCH_TARGET_RVV128(func)      \
+    ::arrow::internal::DynamicDispatchTarget{     \
+        ::arrow::internal::DispatchLevel::RVV128, \
+        (func),                                   \
+    },
+#else
+#  define ARROW_DISPATCH_TARGET_RVV128(func)
+#endif
+
+#if defined(ARROW_HAVE_RVV256) || defined(ARROW_HAVE_RUNTIME_RVV256)
+#  define ARROW_DISPATCH_TARGET_RVV256(func)      \
+    ::arrow::internal::DynamicDispatchTarget{     \
+        ::arrow::internal::DispatchLevel::RVV256, \
+        (func),                                   \
+    },
+#else
+#  define ARROW_DISPATCH_TARGET_RVV256(func)
+#endif
+
+#if defined(ARROW_HAVE_RVV512) || defined(ARROW_HAVE_RUNTIME_RVV512)
+#  define ARROW_DISPATCH_TARGET_RVV512(func)      \
+    ::arrow::internal::DynamicDispatchTarget{     \
+        ::arrow::internal::DispatchLevel::RVV512, \
+        (func),                                   \
+    },
+#else
+#  define ARROW_DISPATCH_TARGET_RVV512(func)
+#endif
+
 /// A concept to specify how dynamic dispatch should be handled.
 ///
 /// A requirement is that the list of available targets must be compile time
@@ -352,6 +385,12 @@ class DynamicDispatch {
         return cpu_info->IsSupported(CpuInfo::SVE256);
       case DispatchLevel::SVE512:
         return cpu_info->IsSupported(CpuInfo::SVE512);
+      case DispatchLevel::RVV128:
+        return cpu_info->IsSupported(CpuInfo::RVV128);
+      case DispatchLevel::RVV256:
+        return cpu_info->IsSupported(CpuInfo::RVV256);
+      case DispatchLevel::RVV512:
+        return cpu_info->IsSupported(CpuInfo::RVV512);
       default:
         return false;
     }

cpp/src/arrow/util/io_util_test.cc

@@ -1109,6 +1109,14 @@ TEST(CpuInfo, Basic) {
   ASSERT_EQ(ci->hardware_flags(), 0);
   ci_rw->EnableFeature(original_hardware_flags, true);
   ASSERT_EQ(ci->hardware_flags(), original_hardware_flags);
+
+#if defined(__riscv) && __riscv_xlen == 64
+  if (ci->IsDetected(CpuInfo::RVV)) {
+    EXPECT_NE(ci->hardware_flags() & CpuInfo::RVV, 0);
+    EXPECT_NE(
+        ci->hardware_flags() & (CpuInfo::RVV128 | CpuInfo::RVV256 | CpuInfo::RVV512), 0);
+  }
+#endif
 }
 
 TEST(Memory, TotalMemory) {