Merge pull request #61 from nature-lang/feature/arm64_debug

feat: arm64 test pass

Author: weiwenhao

CMakeLists.txt

@@ -28,7 +28,8 @@ if (CMAKE_BUILD_TYPE MATCHES Debug)
     #    add_compile_definitions(NATURE_DEBUG)
     add_compile_definitions(DEBUG_LIR)
     add_compile_definitions(DEBUG_ASM)
-    add_compile_definitions(DEBUG_PARSER)
+    # add_compile_definitions(DEBUG_PARSER)
+#    add_compile_definitions(DEBUG_INTERVAL)
     #    add_compile_definitions(DEBUG_INTERVAL)
     #    add_compile_definitions(DEBUG_MACHO)
     add_compile_definitions(ASSERT_ERROR)

cmake/linux-amd64-toolchain.cmake

@@ -1,5 +1,11 @@
-set(CMAKE_C_COMPILER x86_64-linux-musl-gcc)
-set(CMAKE_CXX_COMPILER x86_64-linux-musl-g++)
+find_program(CMAKE_C_COMPILER
+        NAMES
+        x86_64-linux-musl-gcc
+        musl-gcc
+        cc
+        REQUIRED
+)
+
 set(CMAKE_EXE_LINKER_FLAGS "-static")
 
 set(CMAKE_SYSTEM_NAME linux)

cmake/linux-arm64-toolchain.cmake

@@ -1,5 +1,11 @@
-set(CMAKE_C_COMPILER aarch64-linux-musl-gcc)
-set(CMAKE_CXX_COMPILER aarch64-linux-musl-g++)
+# 按优先级查找可用的编译器
+find_program(CMAKE_C_COMPILER 
+    NAMES 
+        aarch64-linux-musl-gcc 
+        musl-gcc 
+        cc
+    REQUIRED
+)
 set(CMAKE_EXE_LINKER_FLAGS "-static")
 #set(CMAKE_FIND_LIBRARY_SUFFIXES "")
 

include/uv.h

@@ -260,7 +260,9 @@ typedef struct uv_metrics_s uv_metrics_t;
 
 typedef enum {
   UV_LOOP_BLOCK_SIGNAL = 0,
-  UV_METRICS_IDLE_TIME
+  UV_METRICS_IDLE_TIME,
+  UV_LOOP_USE_IO_URING_SQPOLL
+#define UV_LOOP_USE_IO_URING_SQPOLL UV_LOOP_USE_IO_URING_SQPOLL
 } uv_loop_option;
 
 typedef enum {
@@ -604,7 +606,18 @@ UV_EXTERN int uv_tcp_simultaneous_accepts(uv_tcp_t* handle, int enable);
 
 enum uv_tcp_flags {
   /* Used with uv_tcp_bind, when an IPv6 address is used. */
-  UV_TCP_IPV6ONLY = 1
+  UV_TCP_IPV6ONLY = 1,
+
+  /* Enable SO_REUSEPORT socket option when binding the handle.
+   * This allows completely duplicate bindings by multiple processes
+   * or threads if they all set SO_REUSEPORT before binding the port.
+   * Incoming connections are distributed across the participating
+   * listener sockets.
+   *
+   * This flag is available only on Linux 3.9+, DragonFlyBSD 3.6+,
+   * FreeBSD 12.0+, Solaris 11.4, and AIX 7.2.5+ for now.
+   */
+  UV_TCP_REUSEPORT = 2,
 };
 
 UV_EXTERN int uv_tcp_bind(uv_tcp_t* handle,
@@ -645,10 +658,13 @@ enum uv_udp_flags {
   UV_UDP_PARTIAL = 2,
   /*
    * Indicates if SO_REUSEADDR will be set when binding the handle.
-   * This sets the SO_REUSEPORT socket flag on the BSDs and OS X. On other
-   * Unix platforms, it sets the SO_REUSEADDR flag.  What that means is that
-   * multiple threads or processes can bind to the same address without error
-   * (provided they all set the flag) but only the last one to bind will receive
+   * This sets the SO_REUSEPORT socket flag on the BSDs (except for
+   * DragonFlyBSD), OS X, and other platforms where SO_REUSEPORTs don't
+   * have the capability of load balancing, as the opposite of what
+   * UV_UDP_REUSEPORT would do. On other Unix platforms, it sets the
+   * SO_REUSEADDR flag. What that means is that multiple threads or
+   * processes can bind to the same address without error (provided
+   * they all set the flag) but only the last one to bind will receive
    * any traffic, in effect "stealing" the port from the previous listener.
    */
   UV_UDP_REUSEADDR = 4,
@@ -671,6 +687,18 @@ enum uv_udp_flags {
    * This flag is no-op on platforms other than Linux.
    */
   UV_UDP_LINUX_RECVERR = 32,
+  /*
+   * Indicates if SO_REUSEPORT will be set when binding the handle.
+   * This sets the SO_REUSEPORT socket option on supported platforms.
+   * Unlike UV_UDP_REUSEADDR, this flag will make multiple threads or
+   * processes that are binding to the same address and port "share"
+   * the port, which means incoming datagrams are distributed across
+   * the receiving sockets among threads or processes.
+   *
+   * This flag is available only on Linux 3.9+, DragonFlyBSD 3.6+,
+   * FreeBSD 12.0+, Solaris 11.4, and AIX 7.2.5+ for now.
+   */
+  UV_UDP_REUSEPORT = 64,
   /*
    * Indicates that recvmmsg should be used, if available.
    */
@@ -1106,7 +1134,14 @@ enum uv_process_flags {
    * option is only meaningful on Windows systems. On Unix it is silently
    * ignored.
    */
-  UV_PROCESS_WINDOWS_HIDE_GUI = (1 << 6)
+  UV_PROCESS_WINDOWS_HIDE_GUI = (1 << 6),
+  /*
+   * On Windows, if the path to the program to execute, specified in
+   * uv_process_options_t's file field, has a directory component,
+   * search for the exact file name before trying variants with
+   * extensions like '.exe' or '.cmd'.
+   */
+  UV_PROCESS_WINDOWS_FILE_PATH_EXACT_NAME = (1 << 7)
 };
 
 /*
@@ -1284,6 +1319,17 @@ UV_EXTERN uv_pid_t uv_os_getppid(void);
 UV_EXTERN int uv_os_getpriority(uv_pid_t pid, int* priority);
 UV_EXTERN int uv_os_setpriority(uv_pid_t pid, int priority);
 
+enum {
+  UV_THREAD_PRIORITY_HIGHEST = 2,
+  UV_THREAD_PRIORITY_ABOVE_NORMAL = 1,
+  UV_THREAD_PRIORITY_NORMAL = 0,
+  UV_THREAD_PRIORITY_BELOW_NORMAL = -1,
+  UV_THREAD_PRIORITY_LOWEST = -2,
+};
+
+UV_EXTERN int uv_thread_getpriority(uv_thread_t tid, int* priority);
+UV_EXTERN int uv_thread_setpriority(uv_thread_t tid, int priority);
+
 UV_EXTERN unsigned int uv_available_parallelism(void);
 UV_EXTERN int uv_cpu_info(uv_cpu_info_t** cpu_infos, int* count);
 UV_EXTERN void uv_free_cpu_info(uv_cpu_info_t* cpu_infos, int count);
@@ -1885,6 +1931,18 @@ struct uv_loop_s {
 UV_EXTERN void* uv_loop_get_data(const uv_loop_t*);
 UV_EXTERN void uv_loop_set_data(uv_loop_t*, void* data);
 
+/* Unicode utilities needed for dealing with Windows. */
+UV_EXTERN size_t uv_utf16_length_as_wtf8(const uint16_t* utf16,
+                                         ssize_t utf16_len);
+UV_EXTERN int uv_utf16_to_wtf8(const uint16_t* utf16,
+                               ssize_t utf16_len,
+                               char** wtf8_ptr,
+                               size_t* wtf8_len_ptr);
+UV_EXTERN ssize_t uv_wtf8_length_as_utf16(const char* wtf8);
+UV_EXTERN void uv_wtf8_to_utf16(const char* wtf8,
+                                uint16_t* utf16,
+                                size_t utf16_len);
+
 /* Don't export the private CPP symbols. */
 #undef UV_HANDLE_TYPE_PRIVATE
 #undef UV_REQ_TYPE_PRIVATE

include/uv/darwin.h

@@ -328,7 +328,10 @@ typedef struct {
 
 #define UV_TIMER_PRIVATE_FIELDS                                               \
   uv_timer_cb timer_cb;                                                       \
-  void* heap_node[3];                                                         \
+  union {                                                                     \
+    void* heap[3];                                                            \
+    struct uv__queue queue;                                                   \
+  } node;                                                                     \
   uint64_t timeout;                                                           \
   uint64_t repeat;                                                            \
   uint64_t start_id;

include/uv/unix.h

@@ -31,10 +31,10 @@
  */
 
 #define UV_VERSION_MAJOR 1
-#define UV_VERSION_MINOR 46
-#define UV_VERSION_PATCH 1
-#define UV_VERSION_IS_RELEASE 0
-#define UV_VERSION_SUFFIX "dev"
+#define UV_VERSION_MINOR 49
+#define UV_VERSION_PATCH 2
+#define UV_VERSION_IS_RELEASE 1
+#define UV_VERSION_SUFFIX ""
 
 #define UV_VERSION_HEX  ((UV_VERSION_MAJOR << 16) | \
                          (UV_VERSION_MINOR <<  8) | \

include/uv/version.h

@@ -11,6 +11,9 @@ include(../cmake/cross.cmake)
 
 set(CMAKE_C_STANDARD 11)
 
+set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -static")
+set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -static")
+
 add_compile_definitions(RUNTIME)
 message("add compile def RUNTIME")
 

lib/linux_arm64/crt1.o

@@ -23,11 +23,9 @@
 static uint64_t share_stack_base = MMAP_SHARE_STACK_BASE;
 
 void aco_runtime_test(void) {
-#ifdef __x86_64__
+#if  defined(__x86_64__) || defined(__aarch64__)
     assert(sizeof(void *) == 8 && "require 'sizeof(void*) == 8'");
     assert(sizeof(__uint128_t) == 16 && "require 'sizeof(__uint128_t) == 16'");
-#elif defined(__aarch64__)
-    assert(sizeof(void *) == 8 && "require 'sizeof(void*) == 8'");
 #else
 #error "platform no support yet"
 #endif
@@ -169,12 +167,6 @@ uv_key_t aco_gtls_co;
 
 uv_key_t aco_gtls_last_word_fp;
 
-// static __thread void* aco_gtls_fpucw_mxcsr[2];
-// #elif  __x86_64__
-// static __thread void *aco_gtls_fpucw_mxcsr[1];
-// #else
-// #error "platform no support yet"
-// #endif
 uv_key_t aco_gtls_fpucw_mxcsr;
 pthread_mutex_t share_stack_lock;
 
@@ -281,7 +273,14 @@ void *aco_share_stack_init(aco_share_stack_t *p, size_t sz) {
     uintptr_t u_p = (uintptr_t) (p->sz - (sizeof(void *) << 1) + (uintptr_t) p->ptr);
     u_p = (u_p >> 4) << 4;
     p->align_highptr = (void *) u_p;
-    p->align_retptr = (void *) (u_p - sizeof(void *));
+
+#ifdef __aarch64__
+    // aarch64 hardware-enforces 16 byte stack alignment
+    p->align_retptr  = (void*)(u_p - 16);
+#else
+    p->align_retptr  = (void*)(u_p - sizeof(void*));
+#endif
+
     *((void **) (p->align_retptr)) = (void *) (aco_funcp_protector_asm);
     assert(p->sz > (16 + (sizeof(void *) << 1) + sizeof(void *)));
     p->align_limit = p->sz - 16 - (sizeof(void *) << 1);
@@ -314,10 +313,8 @@ void aco_create_init(aco_t *aco, aco_t *main_co, aco_share_stack_t *share_stack,
         aco->reg[ACO_REG_IDX_SP] = aco->share_stack->align_retptr;
         aco->reg[ACO_REG_IDX_FPU] = uv_key_get(&aco_gtls_fpucw_mxcsr);
 #elif defined(__aarch64__)
-        aco->reg[ACO_REG_IDX_RETADDR] = (void *) fp;
+        aco->reg[ACO_REG_IDX_RETADDR] = (void*)fp;
         aco->reg[ACO_REG_IDX_SP] = aco->share_stack->align_retptr;
-        aco->reg[ACO_REG_IDX_BP] = aco->share_stack->align_retptr;
-        // ARM64 不需要保存 FPU 状态,因为它使用独立的 NEON/FP 寄存器
 #else
 #error "platform no support yet"
 #endif

lib/linux_arm64/libc.a

@@ -44,10 +44,10 @@
 #define ACO_REG_IDX_BP 7
 #define ACO_REG_IDX_FPU 8
 #elif defined(__aarch64__)
-#define ACO_REG_IDX_RETADDR 0
-#define ACO_REG_IDX_SP 1
-#define ACO_REG_IDX_BP 2
-#define ACO_REG_IDX_FPU 3
+#define ACO_REG_IDX_RETADDR 13
+#define ACO_REG_IDX_SP 14
+#define ACO_REG_IDX_BP 12
+#define ACO_REG_IDX_FPU 15
 #else
 #error "platform no support yet"
 #endif
@@ -75,7 +75,7 @@ typedef struct {
     void *ptr;
     size_t sz;
     void *align_highptr;
-    void *align_retptr; // 最大值
+    void *align_retptr;
     size_t align_validsz;
     size_t align_limit;
     aco_t *owner;
@@ -93,7 +93,14 @@ typedef struct {
 typedef void (*aco_cofuncp_t)(void);
 
 struct aco_s {
+#ifdef __X86_64__
     void *reg[9]; // amd64 试用
+#elif __aarch64__
+    void* reg[16];
+#else
+#error "platform no support yet"
+#endif
+
 
     aco_t *main_co;
     void *arg;