Shared heap enhancements for Interpreter and AOT (#4400)

Propose two enhancements:

- Shared heap created from preallocated memory buffer: The user can create a shared heap from a pre-allocated buffer and see that memory region as one large chunk; there's no need to dynamically manage it(malloc/free). The user needs to make sure the native address and size of that memory region are valid.
- Introduce shared heap chain: The user can create a shared heap chain, from the wasm app point of view, it's still a continuous memory region in wasm app's point of view while in the native it can consist of multiple shared heaps (each of which is a continuous memory region). For example, one 500MB shared heap 1 and one 500 MB shared heap 2 form a chain, in Wasm's point of view, it's one 1GB shared heap.

After these enhancements, the data sharing between wasm apps, and between hosts can be more efficient and flexible. Admittedly shared heap management can be more complex for users, but it's similar to the zero-overhead principle. No overhead will be imposed for the users who don't use the shared heap enhancement or don't use the shared heap at all.

core/iwasm/interpreter/wasm_interp_classic.c

@@ -46,28 +46,6 @@ typedef float64 CellType_F64;
 #define get_linear_mem_size() GET_LINEAR_MEMORY_SIZE(memory)
 #endif
 
-#if WASM_ENABLE_SHARED_HEAP != 0
-#if WASM_ENABLE_MULTI_MEMORY != 0
-/* Only enable shared heap for the default memory */
-#define is_default_memory (memidx == 0)
-#else
-#define is_default_memory true
-#endif
-#define app_addr_in_shared_heap(app_addr, bytes)                             \
-    (shared_heap && is_default_memory && (app_addr) >= shared_heap_start_off \
-     && (app_addr) <= shared_heap_end_off - bytes + 1)
-
-#define shared_heap_addr_app_to_native(app_addr, native_addr) \
-    native_addr = shared_heap_base_addr + ((app_addr)-shared_heap_start_off)
-
-#define CHECK_SHARED_HEAP_OVERFLOW(app_addr, bytes, native_addr) \
-    if (app_addr_in_shared_heap(app_addr, bytes))                \
-        shared_heap_addr_app_to_native(app_addr, native_addr);   \
-    else
-#else
-#define CHECK_SHARED_HEAP_OVERFLOW(app_addr, bytes, native_addr)
-#endif
-
 #if WASM_ENABLE_MEMORY64 == 0
 
 #if (!defined(OS_ENABLE_HW_BOUND_CHECK) \
@@ -1670,22 +1648,6 @@ wasm_interp_call_func_bytecode(WASMModuleInstance *module,
     if (memory)
         is_memory64 = memory->is_memory64;
 #endif
-#if WASM_ENABLE_SHARED_HEAP != 0
-    WASMSharedHeap *shared_heap = module->e->shared_heap;
-    uint8 *shared_heap_base_addr = shared_heap ? shared_heap->base_addr : NULL;
-#if WASM_ENABLE_MEMORY64 != 0
-    uint64 shared_heap_start_off =
-        shared_heap ? (is_memory64 ? shared_heap->start_off_mem64
-                                   : shared_heap->start_off_mem32)
-                    : 0;
-    uint64 shared_heap_end_off =
-        shared_heap ? (is_memory64 ? UINT64_MAX : UINT32_MAX) : 0;
-#else
-    uint64 shared_heap_start_off =
-        shared_heap ? shared_heap->start_off_mem32 : 0;
-    uint64 shared_heap_end_off = shared_heap ? UINT32_MAX : 0;
-#endif
-#endif /* end of WASM_ENABLE_SHARED_HEAP != 0 */
 #if WASM_ENABLE_MULTI_MEMORY != 0
     uint32 memidx = 0;
     uint32 memidx_cached = (uint32)-1;

core/iwasm/interpreter/wasm_interp_fast.c

@@ -41,22 +41,6 @@ typedef float64 CellType_F64;
 #define get_linear_mem_size() GET_LINEAR_MEMORY_SIZE(memory)
 #endif
 
-#if WASM_ENABLE_SHARED_HEAP != 0
-#define app_addr_in_shared_heap(app_addr, bytes)        \
-    (shared_heap && (app_addr) >= shared_heap_start_off \
-     && (app_addr) <= shared_heap_end_off - bytes + 1)
-
-#define shared_heap_addr_app_to_native(app_addr, native_addr) \
-    native_addr = shared_heap_base_addr + ((app_addr)-shared_heap_start_off)
-
-#define CHECK_SHARED_HEAP_OVERFLOW(app_addr, bytes, native_addr) \
-    if (app_addr_in_shared_heap(app_addr, bytes))                \
-        shared_heap_addr_app_to_native(app_addr, native_addr);   \
-    else
-#else
-#define CHECK_SHARED_HEAP_OVERFLOW(app_addr, bytes, native_addr)
-#endif
-
 #if !defined(OS_ENABLE_HW_BOUND_CHECK) \
     || WASM_CPU_SUPPORTS_UNALIGNED_ADDR_ACCESS == 0
 #define CHECK_MEMORY_OVERFLOW(bytes)                                           \
@@ -1590,21 +1574,13 @@ wasm_interp_call_func_bytecode(WASMModuleInstance *module,
     bool is_return_call = false;
 #endif
 #if WASM_ENABLE_SHARED_HEAP != 0
-    WASMSharedHeap *shared_heap = module->e ? module->e->shared_heap : NULL;
-    uint8 *shared_heap_base_addr = shared_heap ? shared_heap->base_addr : NULL;
-    /*
-#if WASM_ENABLE_MEMORY64 != 0
-    uint64 shared_heap_start_off =
-        shared_heap ? (is_memory64 ? shared_heap->start_off_mem64
-                                   : shared_heap->start_off_mem32)
-                    : 0;
-    uint64 shared_heap_end_off =
-        shared_heap ? (is_memory64 ? UINT64_MAX : UINT32_MAX) : 0;
-#else
-    */ /* TODO: uncomment the code when memory64 is enabled for fast-interp */
-    uint64 shared_heap_start_off =
-        shared_heap ? shared_heap->start_off_mem32 : 0;
-    uint64 shared_heap_end_off = shared_heap ? UINT32_MAX : 0;
+    /* TODO: currently flowing two variables are only dummy for shared heap
+     * boundary check, need to be updated when multi-memory or memory64
+     * proposals are to be implemented */
+    bool is_memory64 = false;
+    uint32 memidx = 0;
+    (void)is_memory64;
+    (void)memidx;
 /* #endif */
 #endif /* end of WASM_ENABLE_SHARED_HEAP != 0 */
 

core/iwasm/interpreter/wasm_runtime.c

@@ -2818,6 +2818,7 @@ wasm_instantiate(WASMModule *module, WASMModuleInstance *parent,
 #else
     module_inst->e->shared_heap_start_off.u32[0] = UINT32_MAX;
 #endif
+    module_inst->e->shared_heap = NULL;
 #endif
 
 #if WASM_ENABLE_GC != 0

core/iwasm/interpreter/wasm_runtime.h

@@ -93,12 +93,21 @@ typedef union {
 } MemBound;
 
 typedef struct WASMSharedHeap {
-    struct WASMSharedHeap *next;
-    void *heap_handle;
-    uint8 *base_addr;
+    /* The global shared heap list maintained in runtime, used for runtime
+     * destroy */
+    DefPointer(struct WASMSharedHeap *, next);
+    /* The logical shared heap chain the shared heap in */
+    DefPointer(struct WASMSharedHeap *, chain_next);
+    /* Will be null if shared heap is created from pre allocated memory chunk
+     * and don't need to dynamic malloc and free */
+    DefPointer(void *, heap_handle);
+    DefPointer(uint8 *, base_addr);
     uint64 size;
     uint64 start_off_mem64;
     uint64 start_off_mem32;
+    /* The number of wasm apps it attached to, for a shared heap chain, only the
+     * list head need to maintain the valid attached_count */
+    uint8 attached_count;
 } WASMSharedHeap;
 
 struct WASMMemoryInstance {
@@ -364,8 +373,6 @@ typedef struct WASMModuleInstanceExtra {
 #endif
 
 #if WASM_ENABLE_SHARED_HEAP != 0
-    WASMSharedHeap *shared_heap;
-#if WASM_ENABLE_JIT != 0
     /*
      * Adjusted shared heap based addr to simple the calculation
      * in the aot code. The value is:
@@ -373,7 +380,8 @@ typedef struct WASMModuleInstanceExtra {
      */
     uint8 *shared_heap_base_addr_adj;
     MemBound shared_heap_start_off;
-#endif
+    MemBound shared_heap_end_off;
+    WASMSharedHeap *shared_heap;
 #endif
 
 #if WASM_ENABLE_DEBUG_INTERP != 0                         \