dispatchThreadgroups的参数不正确会导致段错误吗？

Question

我已经使用金属着色器实现了插值功能。有时，执行代码时会报告段错误：

[commandEncoder dispatchThreadgroups:threadGroups threadsPerThreadgroup:threadGroupCounts];

我已经检查了我的程序的其他部分，似乎这个错误是由dispatchThreadgroups方法引起的。我不确定不正确的threadGroups和threadGroupCounts是否会导致错误。所有代码都列出如下：

void metal_v_texture( void *metal_context, int16_t *dst, uint8_t* _src, int _srcstride, int height, int mx, int my, int width )
{
#define LOCAL_BUF_SIZE 5184 //(64+8)*(64+8)
    int x, y;
    pixel        *src       = (pixel*)_src;
    ptrdiff_t     srcstride = _srcstride / sizeof(pixel);
    int16_t out_buf[LOCAL_BUF_SIZE];
    int16_t *pout = out_buf;
    uint8_t local_src[LOCAL_BUF_SIZE];
    MetalContext *mc = metal_context;

    memset( out_buf,   0, sizeof(int16_t)*LOCAL_BUF_SIZE );
    memset( local_src, 0, sizeof(uint8_t)*LOCAL_BUF_SIZE);

    //copy data to the local_src_buffer
    int s = _srcstride;
    //pcsrc has been checked, it did not lead to segment fault 
    uint8_t *pcsrc = _src - 4*s - 4;
    uint8_t *pcdst = local_src;
    for( int i = 0; i < height+8; i++ )
    {
       memcpy( pcdst, pcsrc, sizeof(uint8_t)*(width+8));
       pcsrc += s;
       pcdst += 72;
    }


    id<MTLDevice> device = mc->metal_device;
    id<MTLCommandQueue> commandQueue = mc->metal_commandqueue;
    id<MTLComputePipelineState> cpipeline = mc->metal_cps_put_hevc_qpel_v;
    assert( device );
    assert( commandQueue );
    assert( cpipeline );

    //in_texture
    int w = 72/4, h = height+4;
    MTLTextureDescriptor *in_desc = [MTLTextureDescriptor texture2DDescriptorWithPixelFormat:MTLPixelFormatRGBA8Uint width:w height:h mipmapped:NO];
    id<MTLTexture> in_texture = [device newTextureWithDescriptor:in_desc];
    assert( in_texture );

    MTLRegion texture_region = MTLRegionMake2D( 0, 0, w, h );
    [in_texture  replaceRegion:texture_region mipmapLevel:0 withBytes:local_src bytesPerRow:72];    

    //out_texture
    MTLTextureDescriptor *out_desc = [MTLTextureDescriptor texture2DDescriptorWithPixelFormat:MTLPixelFormatRGBA16Uint width:w height:h mipmapped:NO];
    out_desc.usage = MTLTextureUsageShaderWrite;
    id<MTLTexture> texture = [device newTextureWithDescriptor:out_desc];
    assert( texture );

    MTLSize threadGroupCounts = MTLSizeMake(1, 1, 1);
    MTLSize threadGroups = MTLSizeMake([texture width]  / threadGroupCounts.width,
    [texture height] / threadGroupCounts.height,
        1);

    id<MTLCommandBuffer> commandBuffer = [commandQueue commandBuffer];
    id<MTLComputeCommandEncoder> commandEncoder = [commandBuffer computeCommandEncoder];
    assert( commandBuffer );
    assert( commandEncoder );

    [commandEncoder setComputePipelineState:cpipeline];
    [commandEncoder setTexture:texture    atIndex:0];
    [commandEncoder setTexture:in_texture atIndex:1];

    //after this printf fuction, segment fault happed
    printf("%s %d commandEncoder: %x commandBuffer: %x in_texture: %x device: %x commandQueue: %x pipeline: %x\n", __FILE__, __LINE__, commandEncoder, commandBuffer, in_texture, device, commandQueue, cpipeline );
    [commandEncoder dispatchThreadgroups:threadGroups threadsPerThreadgroup:threadGroupCounts];
    [commandEncoder endEncoding];

    // synchronize texture from gpu to host mem
    id<MTLBlitCommandEncoder> blitEncoder = [commandBuffer blitCommandEncoder];
    [blitEncoder synchronizeTexture:texture slice:0 level:0];
    [blitEncoder endEncoding];


    [commandBuffer commit];
    [commandBuffer waitUntilCompleted];


    // Get the data back
    MTLRegion region = MTLRegionMake2D(0, 0, [texture width], [texture height]);
    [texture getBytes:out_buf bytesPerRow: 72*2 fromRegion:region mipmapLevel:0];

    pout = out_buf+4*72+4;
    int16_t *pdst = dst;
    for( int j = 0; j < height; j++ )
    {
        memcpy( pdst, pout, sizeof(int16_t)*width );
        pdst += MAX_PB_SIZE;
        pout += 72;
    }


   //release
   //[in_desc release];
   //[out_desc release];
   [in_texture release];
   [texture release];
   [commandBuffer release];
   [commandEncoder release];
   [blitEncoder release];        
}

设备，命令队列和computepipestat在外部创建并传递给此函数。已经检查了prsrc的使用，它没有导致段错误。

uint8_t *pcsrc = _src - 4*s - 4;

commandEncoder，commandBuffer，in_texture，device，commandQueue和cpipeline的地址有效，如段故障的打印信息所示

hmetal.m 548 commandEncoder: ea73af30 commandBuffer: ea529440 in_texture: eb0d7200 device: ec014800 commandQueue: ea70dc50 pipeline: ea710d70
Segmentation fault: 11

感谢您查看我的问题。