Implement AOT support for RISCV (#649)

Enable RISCV AOT support, the supported ABIs are LP64 and LP64D for riscv64, ILP32 and ILP32D for riscv32.
For wamrc:
    use --target=riscv64/riscv32 to specify the target arch of output AOT file,
    use --target-abi=lp64d/lp64/ilp32d/ilp32 to specify the target ABI,
    if --target-abi isn't specified, by default lp64d is used for riscv64, and ilp32d is used for riscv32.

Signed-off-by: Huang Qi <[email protected]>
Co-authored-by: wenyongh <[email protected]>

Author: no1wudi

build-scripts/config_common.cmake

@@ -51,7 +51,8 @@ if (CMAKE_BUILD_TYPE STREQUAL "Debug")
 endif ()
 
 if (CMAKE_SIZEOF_VOID_P EQUAL 8)
-  if (WAMR_BUILD_TARGET STREQUAL "X86_64" OR WAMR_BUILD_TARGET STREQUAL "AMD_64" OR WAMR_BUILD_TARGET MATCHES "AARCH64.*" OR WAMR_BUILD_TARGET MATCHES "RISCV64.*")
+  if (WAMR_BUILD_TARGET STREQUAL "X86_64" OR WAMR_BUILD_TARGET STREQUAL "AMD_64"
+      OR WAMR_BUILD_TARGET MATCHES "AARCH64.*" OR WAMR_BUILD_TARGET MATCHES "RISCV64.*")
     if (NOT WAMR_BUILD_PLATFORM STREQUAL "windows")
       # Add -fPIC flag if build as 64-bit
       set (CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fPIC")
@@ -186,8 +187,12 @@ else ()
   add_definitions (-DWASM_DISABLE_HW_BOUND_CHECK=0)
 endif ()
 if (WAMR_BUILD_SIMD EQUAL 1)
-  add_definitions (-DWASM_ENABLE_SIMD=1)
-  message ("     SIMD enabled")
+  if (NOT WAMR_BUILD_TARGET MATCHES "RISCV64.*")
+    add_definitions (-DWASM_ENABLE_SIMD=1)
+    message ("     SIMD enabled")
+  else ()
+    message ("     SIMD disabled due to not supported on target RISCV64")
+  endif ()
 endif ()
 if (WAMR_BUILD_MEMORY_PROFILING EQUAL 1)
   add_definitions (-DWASM_ENABLE_MEMORY_PROFILING=1)

build-scripts/runtime_lib.cmake

@@ -52,11 +52,6 @@ if (WAMR_BUILD_INTERP EQUAL 1 OR WAMR_BUILD_JIT EQUAL 1)
     include (${IWASM_DIR}/interpreter/iwasm_interp.cmake)
 endif ()
 
-if (WAMR_BUILD_TARGET MATCHES "RISCV.*" AND WAMR_BUILD_AOT EQUAL 1)
-    set (WAMR_BUILD_AOT 0)
-    message ("-- WAMR AOT disabled as it isn't supported by riscv currently")
-endif ()
-
 if (WAMR_BUILD_AOT EQUAL 1)
     include (${IWASM_DIR}/aot/iwasm_aot.cmake)
     if (WAMR_BUILD_JIT EQUAL 1)

core/app-mgr/app-manager/module_wasm_app.c

@@ -1395,7 +1395,8 @@ wasm_app_module_on_install_request_byte_arrive(uint8 ch,
                 if (section->section_type == AOT_SECTION_TYPE_TEXT) {
                     int map_prot =
                         MMAP_PROT_READ | MMAP_PROT_WRITE | MMAP_PROT_EXEC;
-#if defined(BUILD_TARGET_X86_64) || defined(BUILD_TARGET_AMD_64)
+#if defined(BUILD_TARGET_X86_64) || defined(BUILD_TARGET_AMD_64) \
+    || defined(BUILD_TARGET_RISCV64_LP64D) || defined(BUILD_TARGET_RISCV64_LP64)
                     /* aot code and data in x86_64 must be in range 0 to 2G due to
                        relocation for R_X86_64_32/32S/PC32 */
                     int map_flags = MMAP_MAP_32BIT;

core/iwasm/aot/aot_loader.c

@@ -182,6 +182,7 @@ GET_U64_FROM_ADDR(uint32 *addr)
 #define E_MACHINE_MIPS_X   51       /* Stanford MIPS-X */
 #define E_MACHINE_X86_64   62       /* AMD x86-64 architecture */
 #define E_MACHINE_XTENSA   94       /* Tensilica Xtensa Architecture */
+#define E_MACHINE_RISCV    243      /* RISC-V 32/64 */
 #define E_MACHINE_WIN_X86_64 0x8664 /* Windowx x86-64 architecture */
 
 /* Legal values for e_version */
@@ -257,6 +258,9 @@ get_aot_file_target(AOTTargetInfo *target_info,
         case E_MACHINE_XTENSA:
             machine_type = "xtensa";
             break;
+        case E_MACHINE_RISCV:
+            machine_type = "riscv";
+            break;
         default:
             set_error_buf_v(error_buf, error_buf_size,
                             "unknown machine type %d",
@@ -1030,7 +1034,8 @@ load_object_data_sections(const uint8 **p_buf, const uint8 *buf_end,
     /* Create each data section */
     for (i = 0; i < module->data_section_count; i++) {
         int map_prot = MMAP_PROT_READ | MMAP_PROT_WRITE;
-#if defined(BUILD_TARGET_X86_64) || defined(BUILD_TARGET_AMD_64)
+#if defined(BUILD_TARGET_X86_64) || defined(BUILD_TARGET_AMD_64) \
+    || defined(BUILD_TARGET_RISCV64_LP64D) || defined(BUILD_TARGET_RISCV64_LP64)
         /* aot code and data in x86_64 must be in range 0 to 2G due to
            relocation for R_X86_64_32/32S/PC32 */
         int map_flags = MMAP_MAP_32BIT;
@@ -1501,6 +1506,7 @@ do_text_relocation(AOTModule *module,
             symbol_addr = module->code;
         }
         else if (!strcmp(symbol, ".data")
+                 || !strcmp(symbol, ".sdata")
                  || !strcmp(symbol, ".rdata")
                  || !strcmp(symbol, ".rodata")
                  /* ".rodata.cst4/8/16/.." */
@@ -2235,7 +2241,8 @@ create_sections(const uint8 *buf, uint32 size,
                 if (section_size > 0) {
                     int map_prot = MMAP_PROT_READ | MMAP_PROT_WRITE
                                    | MMAP_PROT_EXEC;
-#if defined(BUILD_TARGET_X86_64) || defined(BUILD_TARGET_AMD_64)
+#if defined(BUILD_TARGET_X86_64) || defined(BUILD_TARGET_AMD_64) \
+    || defined(BUILD_TARGET_RISCV64_LP64D) || defined(BUILD_TARGET_RISCV64_LP64)
                     /* aot code and data in x86_64 must be in range 0 to 2G due to
                        relocation for R_X86_64_32/32S/PC32 */
                     int map_flags = MMAP_MAP_32BIT;

core/iwasm/aot/aot_runtime.c

@@ -2776,12 +2776,12 @@ aot_table_copy(AOTModuleInstance *module_inst,
     /* if src_offset >= dst_offset, copy from front to back */
     /* if src_offset < dst_offset, copy from back to front */
     /* merge all together */
-    bh_memcpy_s((uint8 *)(dst_tbl_inst) + offsetof(AOTTableInstance, data)
-                  + dst_offset * sizeof(uint32),
-                (dst_tbl_inst->cur_size - dst_offset) * sizeof(uint32),
-                (uint8 *)(src_tbl_inst) + offsetof(AOTTableInstance, data)
-                  + src_offset * sizeof(uint32),
-                length * sizeof(uint32));
+    bh_memmove_s((uint8 *)(dst_tbl_inst) + offsetof(AOTTableInstance, data)
+                   + dst_offset * sizeof(uint32),
+                 (dst_tbl_inst->cur_size - dst_offset) * sizeof(uint32),
+                 (uint8 *)(src_tbl_inst) + offsetof(AOTTableInstance, data)
+                   + src_offset * sizeof(uint32),
+                 length * sizeof(uint32));
 }
 
 void