环形缓冲区的基本概念
环形缓冲区的基本概念来自Wikipedia:环形缓冲器
圆形缓冲区(circular buffer),也称作圆形队列(circular queue),循环缓冲区(cyclic buffer),环形缓冲区(ring buffer),是一种用于表示一个固定尺寸、头尾相连的缓冲区的数据结构,适合缓存数据流。
圆形缓冲区的一个有用特性是:当一个数据元素被用掉后,其余数据元素不需要移动其存储位置。相反,一个非圆形缓冲区(例如一个普通的队列)在用掉一个数据元素后,其余数据元素需要向前搬移。换句话说,圆形缓冲区适合实现先进先出缓冲区,而非圆形缓冲区适合后进先出缓冲区。
圆形缓冲区适合于事先明确了缓冲区的最大容量的情形。扩展一个圆形缓冲区的容量,需要搬移其中的数据。因此一个缓冲区如果需要经常调整其容量,用链表实现更为合适。
写操作覆盖圆形缓冲区中未被处理的数据在某些情况下是允许的。特别是在多媒体处理时。例如,音频的生产者可以覆盖掉声卡尚未来得及处理的音频数据。
圆形缓冲区的实现
环形缓冲区的特点:
1. 环形缓冲区的大小是固定的,在整个过程中只申请一次内存。
2. 是一种先进先出的缓冲区。
3. 写操作时,当要写入的size大于available write size时有两种处理方式:覆盖、不覆盖(哈哈,废话),依情况选择。
我只是代码的搬运工,代码来自webrtc的ring_buffer.h,具体位置是webrtc/common_audio/ring_buffer.h
。读写操作是非线程安全的,所以需要调用者自己去保证线程安全。写操作遇到要写入的大小大于可用大小时选择的方式是:部分写入的策略。所以在使用的时候不顺心可以改改改!
- 对外提供的API代码:
是一个完全独立的功能,不依赖任何webrtc的头文件,可以脱离webrtc在任意地方使用。考虑到并非每一个人都能访问googlesource,so直接列出源码。官方-ring_buffer.h
#ifndef WEBRTC_COMMON_AUDIO_RING_BUFFER_H_
#define WEBRTC_COMMON_AUDIO_RING_BUFFER_H_
#ifdef __cplusplus
extern "C" {
#endif
#include <stddef.h> // size_t
typedef struct RingBuffer RingBuffer;
// Creates and initializes the buffer. Returns NULL on failure.
RingBuffer* WebRtc_CreateBuffer(size_t element_count, size_t element_size);
void WebRtc_InitBuffer(RingBuffer* handle);
void WebRtc_FreeBuffer(void* handle);
// Reads data from the buffer. The |data_ptr| will point to the address where
// it is located. If all |element_count| data are feasible to read without
// buffer wrap around |data_ptr| will point to the location in the buffer.
// Otherwise, the data will be copied to |data| (memory allocation done by the
// user) and |data_ptr| points to the address of |data|. |data_ptr| is only
// guaranteed to be valid until the next call to WebRtc_WriteBuffer().
//
// To force a copying to |data|, pass a NULL |data_ptr|.
//
// Returns number of elements read.
size_t WebRtc_ReadBuffer(RingBuffer* handle,
void** data_ptr,
void* data,
size_t element_count);
// Writes |data| to buffer and returns the number of elements written.
size_t WebRtc_WriteBuffer(RingBuffer* handle, const void* data,
size_t element_count);
// Moves the buffer read position and returns the number of elements moved.
// Positive |element_count| moves the read position towards the write position,
// that is, flushing the buffer. Negative |element_count| moves the read
// position away from the the write position, that is, stuffing the buffer.
// Returns number of elements moved.
int WebRtc_MoveReadPtr(RingBuffer* handle, int element_count);
// Returns number of available elements to read.
size_t WebRtc_available_read(const RingBuffer* handle);
// Returns number of available elements for write.
size_t WebRtc_available_write(const RingBuffer* handle);
#ifdef __cplusplus
}
#endif
#endif // WEBRTC_COMMON_AUDIO_RING_BUFFER_H_
- 实现代码:
只依赖标准库,足够的说明,Good。官方-ring_buffer.c
#include "ring_buffer.h"
#include <stddef.h> // size_t
#include <stdlib.h>
#include <string.h>
enum Wrap {
SAME_WRAP,
DIFF_WRAP
};
struct RingBuffer {
size_t read_pos;
size_t write_pos;
size_t element_count;
size_t element_size;
enum Wrap rw_wrap;
char* data;
};
// Get address of region(s) from which we can read data.
// If the region is contiguous, |data_ptr_bytes_2| will be zero.
// If non-contiguous, |data_ptr_bytes_2| will be the size in bytes of the second
// region. Returns room available to be read or |element_count|, whichever is
// smaller.
static size_t GetBufferReadRegions(RingBuffer* buf,
size_t element_count,
void** data_ptr_1,
size_t* data_ptr_bytes_1,
void** data_ptr_2,
size_t* data_ptr_bytes_2) {
const size_t readable_elements = WebRtc_available_read(buf);
const size_t read_elements = (readable_elements < element_count ?
readable_elements : element_count);
const size_t margin = buf->element_count - buf->read_pos;
// Check to see if read is not contiguous.
if (read_elements > margin) {
// Write data in two blocks that wrap the buffer.
*data_ptr_1 = buf->data + buf->read_pos * buf->element_size;
*data_ptr_bytes_1 = margin * buf->element_size;
*data_ptr_2 = buf->data;
*data_ptr_bytes_2 = (read_elements - margin) * buf->element_size;
} else {
*data_ptr_1 = buf->data + buf->read_pos * buf->element_size;
*data_ptr_bytes_1 = read_elements * buf->element_size;
*data_ptr_2 = NULL;
*data_ptr_bytes_2 = 0;
}
return read_elements;
}
RingBuffer* WebRtc_CreateBuffer(size_t element_count, size_t element_size) {
RingBuffer* self = NULL;
if (element_count == 0 || element_size == 0) {
return NULL;
}
self = malloc(sizeof(RingBuffer));
if (!self) {
return NULL;
}
self->data = malloc(element_count * element_size);
if (!self->data) {
free(self);
self = NULL;
return NULL;
}
self->element_count = element_count;
self->element_size = element_size;
WebRtc_InitBuffer(self);
return self;
}
void WebRtc_InitBuffer(RingBuffer* self) {
self->read_pos = 0;
self->write_pos = 0;
self->rw_wrap = SAME_WRAP;
// Initialize buffer to zeros
memset(self->data, 0, self->element_count * self->element_size);
}
void WebRtc_FreeBuffer(void* handle) {
RingBuffer* self = (RingBuffer*)handle;
if (!self) {
return;
}
free(self->data);
free(self);
}
size_t WebRtc_ReadBuffer(RingBuffer* self,
void** data_ptr,
void* data,
size_t element_count) {
if (self == NULL) {
return 0;
}
if (data == NULL) {
return 0;
}
{
void* buf_ptr_1 = NULL;
void* buf_ptr_2 = NULL;
size_t buf_ptr_bytes_1 = 0;
size_t buf_ptr_bytes_2 = 0;
const size_t read_count = GetBufferReadRegions(self,
element_count,
&buf_ptr_1,
&buf_ptr_bytes_1,
&buf_ptr_2,
&buf_ptr_bytes_2);
if (buf_ptr_bytes_2 > 0) {
// We have a wrap around when reading the buffer. Copy the buffer data to
// |data| and point to it.
memcpy(data, buf_ptr_1, buf_ptr_bytes_1);
memcpy(((char*) data) + buf_ptr_bytes_1, buf_ptr_2, buf_ptr_bytes_2);
buf_ptr_1 = data;
} else if (!data_ptr) {
// No wrap, but a memcpy was requested.
memcpy(data, buf_ptr_1, buf_ptr_bytes_1);
}
if (data_ptr) {
// |buf_ptr_1| == |data| in the case of a wrap.
*data_ptr = buf_ptr_1;
}
// Update read position
WebRtc_MoveReadPtr(self, (int) read_count);
return read_count;
}
}
size_t WebRtc_WriteBuffer(RingBuffer* self,
const void* data,
size_t element_count) {
if (!self) {
return 0;
}
if (!data) {
return 0;
}
{
const size_t free_elements = WebRtc_available_write(self);
const size_t write_elements = (free_elements < element_count ? free_elements
: element_count);
size_t n = write_elements;
const size_t margin = self->element_count - self->write_pos;
if (write_elements > margin) {
// Buffer wrap around when writing.
memcpy(self->data + self->write_pos * self->element_size,
data, margin * self->element_size);
self->write_pos = 0;
n -= margin;
self->rw_wrap = DIFF_WRAP;
}
memcpy(self->data + self->write_pos * self->element_size,
((const char*) data) + ((write_elements - n) * self->element_size),
n * self->element_size);
self->write_pos += n;
return write_elements;
}
}
int WebRtc_MoveReadPtr(RingBuffer* self, int element_count) {
if (!self) {
return 0;
}
{
// We need to be able to take care of negative changes, hence use "int"
// instead of "size_t".
const int free_elements = (int) WebRtc_available_write(self);
const int readable_elements = (int) WebRtc_available_read(self);
int read_pos = (int) self->read_pos;
if (element_count > readable_elements) {
element_count = readable_elements;
}
if (element_count < -free_elements) {
element_count = -free_elements;
}
read_pos += element_count;
if (read_pos > (int) self->element_count) {
// Buffer wrap around. Restart read position and wrap indicator.
read_pos -= (int) self->element_count;
self->rw_wrap = SAME_WRAP;
}
if (read_pos < 0) {
// Buffer wrap around. Restart read position and wrap indicator.
read_pos += (int) self->element_count;
self->rw_wrap = DIFF_WRAP;
}
self->read_pos = (size_t) read_pos;
return element_count;
}
}
size_t WebRtc_available_read(const RingBuffer* self) {
if (!self) {
return 0;
}
if (self->rw_wrap == SAME_WRAP) {
return self->write_pos - self->read_pos;
} else {
return self->element_count - self->read_pos + self->write_pos;
}
}
size_t WebRtc_available_write(const RingBuffer* self) {
if (!self) {
return 0;
}
return self->element_count - WebRtc_available_read(self);
}
- 功能测试
Google的代码文档少,但是必定会有测试代码,一般是Gtest的单元测试,很专业。
ring_buffer_unittest.cc