Macos Mac OSx上的OpenCL内核错误
我写了一些OpenCL代码，它在LINUX上运行良好，但在MacOSX上出现错误，所以失败了。有人能帮我确定为什么会发生这种情况吗。内核代码显示在错误之后。我的内核使用double，因此在顶部有相应的pragma。但我不知道为什么错误显示浮点数据类型：Macos Mac OSx上的OpenCL内核错误,macos,debugging,opencl,gpgpu,Macos,Debugging,Opencl,Gpgpu,我写了一些OpenCL代码，它在LINUX上运行良好，但在MacOSX上出现错误，所以失败了。有人能帮我确定为什么会发生这种情况吗。内核代码显示在错误之后。我的内核使用double，因此在顶部有相应的pragma。但我不知道为什么错误显示浮点数据类型： inline float8 __OVERLOAD__ _name(float8 x) { return _default_name(x); } \ ^ /System/Library/Framewor
inline float8 __OVERLOAD__ _name(float8 x) { return _default_name(x); } \
^
/System/Library/Frameworks/OpenCL.framework/Versions/A/lib/clang/3.2/include/cl_kernel.h:4606:30: note: candidate function
__CLFN_FD_1FD_FAST_RELAX(__fast_relax_log, native_log, __cl_log);
^
/System/Library/Frameworks/OpenCL.framework/Versions/A/lib/clang/3.2/include/cl_kernel.h:421:29:
note: expanded from macro '__CLFN_FD_1FD_FAST_RELAX'
inline float16 __OVERLOAD__ _name(float16 x){ return _default_name(x); }
^
<program source>:206:19: error: call to '__fast_relax_log' is ambiguous
det_zkinin + log((2.0) * 3.14));
^~~~~~~~~~~~~~~~~
/System/Library/Frameworks/OpenCL.framework/Versions/A/lib/clang/3.2/include/cl_kernel.h:4608:22:
note: expanded from macro 'log'
#define log(__x) __fast_relax_log(__x)
^~~~~~~~~~~~~~~~
/System/Library/Frameworks/OpenCL.framework/Versions/A/lib/clang/3.2/include/cl_kernel.h:4606:30:
note: candidate function
__CLFN_FD_1FD_FAST_RELAX(__fast_relax_log, native_log, __cl_log);
^
/System/Library/Frameworks/OpenCL.framework/Versions/A/lib/clang/3.2/include/cl_kernel.h:416:27:
note: expanded from macro '__CLFN_FD_1FD_FAST_RELAX'
inline float __OVERLOAD__ _name(float x) { return _default_name(x); } \
^
/System/Library/Frameworks/OpenCL.framework/Versions/A/lib/clang/3.2/include/cl_kernel.h:4606:30
note: candidate function
__CLFN_FD_1FD_FAST_RELAX(__fast_relax_log, native_log, __cl_log);
^
^
inlinefloat8\uuuuu重载\uuuuuu名称（float8x）{return\u default\u名称（x）；}\
^
/System/Library/Frameworks/OpenCL.framework/Versions/A/lib/clang/3.2/include/cl_kernel.h:4606:30:注意：候选函数
__CLFN\u FD\u 1FD\u FAST\u RELAX（\uuuu FAST\u RELAX\u log、native\u log和\uu cl\u log）；
^
/System/Library/Frameworks/OpenCL.framework/Versions/A/lib/clang/3.2/include/cl_kernel.h:421:29：
注意：从宏“\uuu CLFN\u FD\u 1FD\u FAST\u RELAX”展开
内联float16 uuu重载uuuu名称（float16 x）{return u default u名称（x）；}
^
：206:19:错误：对“\u快速\u松弛\u日志”的调用不明确
det_zkinin+log（（2.0）*3.14））；
^~~~~~~~~~~~~~~~~
/System/Library/Frameworks/OpenCL.framework/Versions/A/lib/clang/3.2/include/cl_kernel.h:4608:22：
注意：从宏“log”展开
#定义日志（\uuuux）\uuuuuu快速\uu松弛\u日志（\uuuux）
^~~~~~~~~~~~~~~~
/System/Library/Frameworks/OpenCL.framework/Versions/A/lib/clang/3.2/include/cl_kernel.h:4606:30：
注：候选函数
__CLFN\u FD\u 1FD\u FAST\u RELAX（\uuuu FAST\u RELAX\u log、native\u log和\uu cl\u log）；
^
/System/Library/Frameworks/OpenCL.framework/Versions/A/lib/clang/3.2/include/cl_kernel.h:416:27：
注意：从宏“\uuu CLFN\u FD\u 1FD\u FAST\u RELAX”展开
内联浮点uuu重载uuuu名称（float x）{返回u默认值u名称（x）}\
^
/System/Library/Frameworks/OpenCL.framework/Versions/A/lib/clang/3.2/include/cl_kernel.h:4606:30
注：候选函数
__CLFN\u FD\u 1FD\u FAST\u RELAX（\uuuu FAST\u RELAX\u log、native\u log和\uu cl\u log）；
^
^
#pragma OPENCL扩展cl_khr_fp64:启用
__内核无效ckf_kernel2（int-dimx，int-aligned_-dimx，
int numockf，int aligned_ckf，
国际热核实验堆，
双ε，
__全球双*名单，
__全球双*ZR列表，
__全球双*rlists，
__全局双*初始状态，
__全局双*初始变量，
__环球双星，
__全局双*共变矩阵，
__全球双*累积，
__全局双*临时变量，
__全球双*x_k_f，
__全球双*z_k_j，
__全局双*交叉矩阵，
__全球双*z_k_f，
__全局双*矩阵，
__全球双*zk_差异，
__全局双*减少增益矩阵，
__全球双*llk
)
{
int ckf_id=get_global_id（0）；
if（ckf_id=dimx）？-1:1；
sing_j[（i*dimx*2+idx）*对齐的_ckf+ckf_id]=临时变量[（i*dimx+列）*对齐的_ckf+ckf_id]*ε*模式+初始状态[i*对齐的_ckf+ckf_id]；
}
}
z_k_f[ckf_id]=0；
innCovMatrixSum[ckf_id]=0；
对于（intidx=0；idx#pragma OPENCL EXTENSION cl_khr_fp64: enable
__kernel void ckf_kernel2(int dimx, int aligned_dimx,
int numOfCKF, int aligned_ckf,
int iter,
double epsilon,
__global double * yrlists,
__global double * zrlists,
__global double * rlists,
__global double * init_state,
__global double * init_var,
__global double * sing_j,
__global double * covMatrixSum,
__global double * cummulative,
__global double * temp_var,
__global double * x_k_f,
__global double * z_k_j,
__global double * crossCovMatrixSum,
__global double * z_k_f,
__global double * innCovMatrixSum,
__global double * zk_diff,
__global double * reduce_gain_matrix,
__global double * llk
)
{
int ckf_id = get_global_id(0);
if( ckf_id < numOfCKF){
for (int i = 0 ; i < dimx ; i++)
{
for (int idx = 0; idx < dimx * 2 ; idx++)
{
int column = idx % dimx;
int mode = (idx >= dimx) ? -1 : 1;
sing_j[(i * dimx * 2 + idx) * aligned_ckf + ckf_id] = temp_var[(i * dimx + column) * aligned_ckf + ckf_id] * epsilon * mode + init_state[i * aligned_ckf + ckf_id];
}
}
z_k_f[ckf_id] = 0;
innCovMatrixSum[ckf_id] = 0;
for (int idx = 0; idx < dimx * 2 ; idx++)
{
z_k_j[idx * aligned_ckf + ckf_id] = 0;
for (int i = 0 ; i < dimx ; i++)
z_k_j[idx * aligned_ckf + ckf_id] += sing_j[(i * dimx * 2 + idx) * aligned_ckf + ckf_id] * zrlists[iter * aligned_dimx + i ];
z_k_f[ckf_id] += z_k_j[idx* aligned_ckf + ckf_id] ;
innCovMatrixSum[ckf_id] += z_k_j[idx* aligned_ckf + ckf_id] * z_k_j[idx* aligned_ckf + ckf_id];
}
z_k_f[ckf_id] = z_k_f[ckf_id] / (dimx * 2);
innCovMatrixSum[ckf_id] = innCovMatrixSum[ckf_id] / (dimx * 2);
innCovMatrixSum[ckf_id] = (innCovMatrixSum[ckf_id] - z_k_f[ckf_id] *z_k_f[ckf_id]) + rlists[ckf_id];
// calcualte crossCovMatrixSum
for (int idx = 0; idx < dimx; idx ++)
{
crossCovMatrixSum[idx * aligned_ckf + ckf_id] = 0;
for (int i = 0 ; i < 2 * dimx ; i++)
{
crossCovMatrixSum[idx * aligned_ckf + ckf_id] += sing_j[(idx * dimx*2 + i) * aligned_ckf + ckf_id ] * z_k_j[i* aligned_ckf + ckf_id];
}
crossCovMatrixSum[idx * aligned_ckf + ckf_id] = crossCovMatrixSum[idx * aligned_ckf + ckf_id]/ (dimx * 2);
crossCovMatrixSum[idx * aligned_ckf + ckf_id] = crossCovMatrixSum[idx * aligned_ckf + ckf_id] - x_k_f[idx* aligned_ckf + ckf_id] * z_k_f[ckf_id];
}
// calculate zk_diff
int z_check = (int)yrlists[iter];
if (z_check == -1)
zk_diff[ckf_id] = 0;
else
zk_diff[ckf_id] = yrlists[iter] - z_k_f[ckf_id];
// calculate reduce_gain_matrix and (reduce_state_matrix <==> init_state);
for (int idx = 0 ; idx < dimx; idx++)
{
reduce_gain_matrix[idx * aligned_ckf + ckf_id] = (crossCovMatrixSum[idx * aligned_ckf + ckf_id] / innCovMatrixSum[ckf_id]);
init_state[idx * aligned_ckf + ckf_id] = reduce_gain_matrix[idx * aligned_ckf + ckf_id] * zk_diff[ckf_id] + x_k_f[idx* aligned_ckf + ckf_id];
}
for (int idx = 0 ; idx < dimx; idx++)
{
init_var[idx * aligned_ckf + ckf_id ] = covMatrixSum[(idx * dimx + idx) * aligned_ckf + ckf_id] -
reduce_gain_matrix[idx * aligned_ckf + ckf_id] * innCovMatrixSum[ckf_id] *
reduce_gain_matrix[idx * aligned_ckf + ckf_id];
}
double det_zkinin = zk_diff[ckf_id] * zk_diff[ckf_id] * (1.0f /innCovMatrixSum[ckf_id]);
if (innCovMatrixSum[ckf_id] <= 0)
llk[ckf_id] = 0;
else
llk[ckf_id] = 0.5 * ((log(innCovMatrixSum[ckf_id])) +
det_zkinin + log((2.0) * 3.14));
cummulative[ckf_id] += llk[ckf_id];
}
}
cl_device_fp_config cfg;
clGetDeviceInfo(device, CL_DEVICE_DOUBLE_FP_CONFIG, sizeof(cfg), &cfg, NULL);
printf("Double FP config = %llu\n", cfg);