解析PHP8底层内核源码-数组(四)
本篇文章给大家介绍《解析PHP8底层内核源码-数组(四)》。有一定的参考价值,有需要的朋友可以参考一下,希望对大家有所帮助。
相关文章推荐:《解析PHP8底层内核源码-数组(一)》《解析PHP8底层内核源码-数组(二)》《解析PHP8底层内核源码-数组(三)》
在Runningprocess 里已经知道代码需要经过词法分析 语法分析 编译 执行 四大步骤
- 《深入解析PHP底层之Running process》
PHP 8会在编译阶段(将AST抽象语法树编译成opcode时)就创建一个数组常量。这个数组常量和数字常量、字符串常量一样,是在编译阶段就确定并分配内存的。因此数组的初始化发生在编译阶段。
PHP的数组初始化代码 部分如下
//如果开启zend_debug #if !ZEND_DEBUG && defined(HAVE_BUILTIN_CONSTANT_P) # define zend_new_array(size) \ (__builtin_constant_p(size) ? \ ((((uint32_t)(size)) <= HT_MIN_SIZE) ? \ _zend_new_array_0() \ //走 _zend_new_array_0 : \ _zend_new_array((size)) \ ) \ : \ _zend_new_array((size)) \ ) #else //没有开启 也就是一般模式 走 _zend_new_array # define zend_new_array(size) \ _zend_new_array(size) #endif ZEND_API void ZEND_FASTCALL _zend_hash_init(HashTable *ht, uint32_t nSize, dtor_func_t pDestructor, zend_bool persistent) { _zend_hash_init_int(ht, nSize, pDestructor, persistent); } ZEND_API HashTable* ZEND_FASTCALL _zend_new_array_0(void) { //分配内存空间 HashTable *ht = emalloc(sizeof(HashTable)); //初始化 _zend_hash_init_int(ht, HT_MIN_SIZE, ZVAL_PTR_DTOR, 0); return ht; } //初始化方法 static zend_always_inline void _zend_hash_init_int(HashTable *ht, uint32_t nSize, dtor_func_t pDestructor, zend_bool persistent) { GC_SET_REFCOUNT(ht, 1); GC_TYPE_INFO(ht) = GC_ARRAY | (persistent ? ((GC_PERSISTENT|GC_NOT_COLLECTABLE) << GC_FLAGS_SHIFT) : 0); HT_FLAGS(ht) = HASH_FLAG_UNINITIALIZED; ht->nTableMask = HT_MIN_MASK; HT_SET_DATA_ADDR(ht, &uninitialized_bucket); ht->nNumUsed = 0; ht->nNumOfElements = 0; ht->nInternalPointer = 0; ht->nNextFreeElement = ZEND_LONG_MIN; ht->pDestructor = pDestructor; ht->nTableSize = zend_hash_check_size(nSize); } //初始化 bucket 也就是 ardata ZEND_API void ZEND_FASTCALL zend_hash_real_init(HashTable *ht, zend_bool packed) { IS_CONSISTENT(ht); HT_ASSERT_RC1(ht); //调用 zend_hash_real_init_ex方法 zend_hash_real_init_ex(ht, packed); } //zend_hash_real_init_ex方法 static zend_always_inline void zend_hash_real_init_ex(HashTable *ht, bool packed) { HT_ASSERT_RC1(ht); ZEND_ASSERT(HT_FLAGS(ht) & HASH_FLAG_UNINITIALIZED); if (packed) { //如果是packed_array zend_hash_real_init_packed_ex(ht); } else { //如果是 hash_array zend_hash_real_init_mixed_ex(ht); } } //paced_array 初始化bucket 的代码 static zend_always_inline void zend_hash_real_init_packed_ex(HashTable *ht) { void *data; if (UNEXPECTED(GC_FLAGS(ht) & IS_ARRAY_PERSISTENT)) { data = pemalloc(HT_SIZE_EX(ht->nTableSize, HT_MIN_MASK), 1); } else if (EXPECTED(ht->nTableSize == HT_MIN_SIZE)) { data = emalloc(HT_SIZE_EX(HT_MIN_SIZE, HT_MIN_MASK)); } else { data = emalloc(HT_SIZE_EX(ht->nTableSize, HT_MIN_MASK)); } HT_SET_DATA_ADDR(ht, data); /* Don't overwrite iterator count. */ ht->u.v.flags = HASH_FLAG_PACKED | HASH_FLAG_STATIC_KEYS; HT_HASH_RESET_PACKED(ht); } //hash_array 初始化bucket的代码 static zend_always_inline void zend_hash_real_init_mixed_ex(HashTable *ht) { void *data; uint32_t nSize = ht->nTableSize; if (UNEXPECTED(GC_FLAGS(ht) & IS_ARRAY_PERSISTENT)) { data = pemalloc(HT_SIZE_EX(nSize, HT_SIZE_TO_MASK(nSize)), 1); } else if (EXPECTED(nSize == HT_MIN_SIZE)) { data = emalloc(HT_SIZE_EX(HT_MIN_SIZE, HT_SIZE_TO_MASK(HT_MIN_SIZE))); ht->nTableMask = HT_SIZE_TO_MASK(HT_MIN_SIZE); HT_SET_DATA_ADDR(ht, data); /* Don't overwrite iterator count. */ ht->u.v.flags = HASH_FLAG_STATIC_KEYS; #ifdef __SSE2__ do { __m128i xmm0 = _mm_setzero_si128(); xmm0 = _mm_cmpeq_epi8(xmm0, xmm0); _mm_storeu_si128((__m128i*)&HT_HASH_EX(data, 0), xmm0); _mm_storeu_si128((__m128i*)&HT_HASH_EX(data, 4), xmm0); _mm_storeu_si128((__m128i*)&HT_HASH_EX(data, 8), xmm0); _mm_storeu_si128((__m128i*)&HT_HASH_EX(data, 12), xmm0); } while (0); #elif defined(__aarch64__) do { int32x4_t t = vdupq_n_s32(-1); vst1q_s32((int32_t*)&HT_HASH_EX(data, 0), t); vst1q_s32((int32_t*)&HT_HASH_EX(data, 4), t); vst1q_s32((int32_t*)&HT_HASH_EX(data, 8), t); vst1q_s32((int32_t*)&HT_HASH_EX(data, 12), t); } while (0); #else HT_HASH_EX(data, 0) = -1; HT_HASH_EX(data, 1) = -1; HT_HASH_EX(data, 2) = -1; HT_HASH_EX(data, 3) = -1; HT_HASH_EX(data, 4) = -1; HT_HASH_EX(data, 5) = -1; HT_HASH_EX(data, 6) = -1; HT_HASH_EX(data, 7) = -1; HT_HASH_EX(data, 8) = -1; HT_HASH_EX(data, 9) = -1; HT_HASH_EX(data, 10) = -1; HT_HASH_EX(data, 11) = -1; HT_HASH_EX(data, 12) = -1; HT_HASH_EX(data, 13) = -1; HT_HASH_EX(data, 14) = -1; HT_HASH_EX(data, 15) = -1; #endif return; } else { data = emalloc(HT_SIZE_EX(nSize, HT_SIZE_TO_MASK(nSize))); } ht->nTableMask = HT_SIZE_TO_MASK(nSize); HT_SET_DATA_ADDR(ht, data); HT_FLAGS(ht) = HASH_FLAG_STATIC_KEYS; HT_HASH_RESET(ht); } //数组赋值和更新值 static zend_always_inline zval *_zend_hash_index_add_or_update_i(HashTable *ht, zend_ulong h, zval *pData, uint32_t flag) { uint32_t nIndex; uint32_t idx; Bucket *p; IS_CONSISTENT(ht); HT_ASSERT_RC1(ht); if ((flag & HASH_ADD_NEXT) && h == ZEND_LONG_MIN) { h = 0; } if (HT_FLAGS(ht) & HASH_FLAG_PACKED) { if (h < ht->nNumUsed) { p = ht->arData + h; if (Z_TYPE(p->val) != IS_UNDEF) { replace: if (flag & HASH_ADD) { return NULL; } if (ht->pDestructor) { ht->pDestructor(&p->val); } ZVAL_COPY_VALUE(&p->val, pData); return &p->val; } else { /* we have to keep the order :( */ goto convert_to_hash; } } else if (EXPECTED(h < ht->nTableSize)) { add_to_packed: p = ht->arData + h; /* incremental initialization of empty Buckets */ if ((flag & (HASH_ADD_NEW|HASH_ADD_NEXT)) != (HASH_ADD_NEW|HASH_ADD_NEXT)) { if (h > ht->nNumUsed) { Bucket *q = ht->arData + ht->nNumUsed; while (q != p) { ZVAL_UNDEF(&q->val); q++; } } } ht->nNextFreeElement = ht->nNumUsed = h + 1; goto add; } else if ((h >> 1) < ht->nTableSize && (ht->nTableSize >> 1) < ht->nNumOfElements) { zend_hash_packed_grow(ht); goto add_to_packed; } else { if (ht->nNumUsed >= ht->nTableSize) { ht->nTableSize += ht->nTableSize; } convert_to_hash: zend_hash_packed_to_hash(ht); } } else if (HT_FLAGS(ht) & HASH_FLAG_UNINITIALIZED) { if (h < ht->nTableSize) { zend_hash_real_init_packed_ex(ht); goto add_to_packed; } zend_hash_real_init_mixed(ht); } else { if ((flag & HASH_ADD_NEW) == 0 || ZEND_DEBUG) { p = zend_hash_index_find_bucket(ht, h); if (p) { ZEND_ASSERT((flag & HASH_ADD_NEW) == 0); goto replace; } } ZEND_HASH_IF_FULL_DO_RESIZE(ht);/* If the Hash table is full, resize it */ } idx = ht->nNumUsed++; nIndex = h | ht->nTableMask; p = ht->arData + idx; Z_NEXT(p->val) = HT_HASH(ht, nIndex); HT_HASH(ht, nIndex) = HT_IDX_TO_HASH(idx); if ((zend_long)h >= ht->nNextFreeElement) { ht->nNextFreeElement = (zend_long)h < ZEND_LONG_MAX ? h + 1 : ZEND_LONG_MAX; } add: ht->nNumOfElements++; p->h = h; p->key = NULL; ZVAL_COPY_VALUE(&p->val, pData); return &p->val; }
_zend_hash_init_int 流程图如下
在PHP 8中,数组的初始化其实是分两步的。
第1步:分配HashTable结构体内存
第2步: 初始化HashTable结构体各个字段
第3步:分配bucket数组内存,修改一些字段值。
对于第3步,并不是每次都进行。比如像“$a = array()”这种写法,由于数组为空,PHP 不会额外申请bucket数组内存。而对于“$a = array(1, 2, 3)”这种写法,由于数组非空,因此PHP 需要执行第3步 分配bucket数组内存,修改一些字段值。
▏本文经原作者PHP崔雪峰同意,发布在php中文网,原文地址:https://zhuanlan.zhihu.com/p/361006441
以上就是解析PHP8底层内核源码-数组(四)的详细内容,更多请关注其它相关文章!