Skip to content

D
dlib
Commit ccb148b4 authored by Davis King's avatar Davis King
Browse files
Options

Cleaned up cuda error handling code

parent dbbce825
Hide whitespace changes
Inline Side-by-side
Showing with 145 additions and 108 deletions
dlib/dnn/cublas_dlibapi.cpp
View file @ ccb148b4
......@@ -9,27 +9,36 @@
#include <cublas_v2.h>
namespace dlib
static const char* cublas_get_error_string(cublasStatus_t s)
{
namespace cuda
switch(s)
{
case CUBLAS_STATUS_NOT_INITIALIZED:
return "CUDA Runtime API initialization failed.";
case CUBLAS_STATUS_ALLOC_FAILED:
return "CUDA Resources could not be allocated.";
default:
return "A call to cuBLAS failed";
}
}
// ----------------------------------------------------------------------------------------
// Check the return value of a call to the cuBLAS runtime for an error condition.
#define CHECK_CUBLAS(call) \
{ \
const cublasStatus_t error = call; \
if (error != CUBLAS_STATUS_SUCCESS) \
{ \
std::ostringstream sout; \
sout << "Error while calling " << #call << " in file " << __FILE__ << ":" << __LINE__ << ". ";\
sout << "code: " << error << ", reason: " << cublas_get_error_string(error);\
throw dlib::cublas_error(sout.str()); \
} \
}
// TODO, make into a macro that prints more information like the line number, etc.
static void check(cublasStatus_t s)
{
switch(s)
{
case CUBLAS_STATUS_SUCCESS: return;
case CUBLAS_STATUS_NOT_INITIALIZED:
throw cublas_error("CUDA Runtime API initialization failed.");
case CUBLAS_STATUS_ALLOC_FAILED:
throw cublas_error("CUDA Resources could not be allocated.");
default:
throw cublas_error("A call to cuBLAS failed");
}
}
namespace dlib
{
namespace cuda
{
// -----------------------------------------------------------------------------------
......@@ -42,7 +51,7 @@ namespace dlib
cublas_context()
{
check(cublasCreate(&handle));
CHECK_CUBLAS(cublasCreate(&handle));
}
~cublas_context()
{
......@@ -117,7 +126,7 @@ namespace dlib
}
const int k = trans_rhs ? rhs_nc : rhs_nr;
check(cublasSgemm(context(),
CHECK_CUBLAS(cublasSgemm(context(),
transb,
transa,
dest_nc, dest_nr, k,
......
dlib/dnn/cublas_dlibapi.h
View file @ ccb148b4
......@@ -6,20 +6,13 @@
#ifdef DLIB_USE_CUDA
#include "tensor.h"
#include "../error.h"
#include "cuda_errors.h"
namespace dlib
{
namespace cuda
{
// -----------------------------------------------------------------------------------
struct cublas_error : public error
{
cublas_error(const std::string& message): error(message) {}
};
// -----------------------------------------------------------------------------------
void gemm (
......
dlib/dnn/cuda_errors.h
View file @ ccb148b4
......@@ -30,6 +30,28 @@ namespace dlib
cudnn_error(const std::string& message): cuda_error(message) {}
};
struct curand_error : public cuda_error
{
/*!
WHAT THIS OBJECT REPRESENTS
This is the exception thrown if any calls to the NVIDIA cuRAND library
returns an error.
!*/
curand_error(const std::string& message): cuda_error(message) {}
};
struct cublas_error : public cuda_error
{
/*!
WHAT THIS OBJECT REPRESENTS
This is the exception thrown if any calls to the NVIDIA cuBLAS library
returns an error.
!*/
cublas_error(const std::string& message): cuda_error(message) {}
};
}
......
dlib/dnn/cudnn_dlibapi.cpp
View file @ ccb148b4
......@@ -12,6 +12,34 @@
#include <string>
#include "cuda_utils.h"
static const char* cudnn_get_error_string(cudnnStatus_t s)
{
switch(s)
{
case CUDNN_STATUS_NOT_INITIALIZED:
return "CUDA Runtime API initialization failed.";
case CUDNN_STATUS_ALLOC_FAILED:
return "CUDA Resources could not be allocated.";
case CUDNN_STATUS_BAD_PARAM:
return "CUDNN_STATUS_BAD_PARAM";
default:
return "A call to cuDNN failed";
}
}
// Check the return value of a call to the cuDNN runtime for an error condition.
#define CHECK_CUDNN(call) \
{ \
const cudnnStatus_t error = call; \
if (error != CUDNN_STATUS_SUCCESS) \
{ \
std::ostringstream sout; \
sout << "Error while calling " << #call << " in file " << __FILE__ << ":" << __LINE__ << ". ";\
sout << "code: " << error << ", reason: " << cudnn_get_error_string(error);\
throw dlib::cudnn_error(sout.str()); \
} \
}
namespace dlib
{
......@@ -19,23 +47,6 @@ namespace dlib
namespace cuda
{
// TODO, make into a macro that prints more information like the line number, etc.
static void check(cudnnStatus_t s)
{
switch(s)
{
case CUDNN_STATUS_SUCCESS: return;
case CUDNN_STATUS_NOT_INITIALIZED:
throw cudnn_error("CUDA Runtime API initialization failed.");
case CUDNN_STATUS_ALLOC_FAILED:
throw cudnn_error("CUDA Resources could not be allocated.");
case CUDNN_STATUS_BAD_PARAM:
throw cudnn_error("CUDNN_STATUS_BAD_PARAM");
default:
throw cudnn_error("A call to cuDNN failed: " + std::string(cudnnGetErrorString(s)));
}
}
// ------------------------------------------------------------------------------------
static cudnnTensorDescriptor_t descriptor(const tensor& t)
......@@ -58,7 +69,7 @@ namespace dlib
cudnn_context()
{
check(cudnnCreate(&handle));
CHECK_CUDNN(cudnnCreate(&handle));
}
~cudnn_context()
......@@ -112,10 +123,10 @@ namespace dlib
else
{
cudnnTensorDescriptor_t h;
check(cudnnCreateTensorDescriptor(&h));
CHECK_CUDNN(cudnnCreateTensorDescriptor(&h));
handle = h;
check(cudnnSetTensor4dDescriptor((cudnnTensorDescriptor_t)handle,
CHECK_CUDNN(cudnnSetTensor4dDescriptor((cudnnTensorDescriptor_t)handle,
CUDNN_TENSOR_NCHW,
CUDNN_DATA_FLOAT,
n,
......@@ -137,7 +148,7 @@ namespace dlib
{
int nStride, cStride, hStride, wStride;
cudnnDataType_t datatype;
check(cudnnGetTensor4dDescriptor((cudnnTensorDescriptor_t)handle,
CHECK_CUDNN(cudnnGetTensor4dDescriptor((cudnnTensorDescriptor_t)handle,
&datatype,
&n,
&k,
......@@ -172,7 +183,7 @@ namespace dlib
(dest.nc()==src.nc() || src.nc()==1) &&
(dest.k()==src.k() || src.k()==1), "");
check(cudnnAddTensor_v3(context(),
CHECK_CUDNN(cudnnAddTensor_v3(context(),
&alpha,
descriptor(src),
src.device(),
......@@ -188,7 +199,7 @@ namespace dlib
{
if (t.size() == 0)
return;
check(cudnnSetTensor(context(),
CHECK_CUDNN(cudnnSetTensor(context(),
descriptor(t),
t.device(),
&value));
......@@ -201,7 +212,7 @@ namespace dlib
{
if (t.size() == 0)
return;
check(cudnnScaleTensor(context(),
CHECK_CUDNN(cudnnScaleTensor(context(),
descriptor(t),
t.device(),
&value));
......@@ -222,7 +233,7 @@ namespace dlib
const float alpha = 1;
const float beta = 0;
check(cudnnConvolutionBackwardBias(context(),
CHECK_CUDNN(cudnnConvolutionBackwardBias(context(),
&alpha,
descriptor(gradient_input),
gradient_input.device(),
......@@ -304,16 +315,16 @@ namespace dlib
stride_y = stride_y_;
stride_x = stride_x_;
check(cudnnCreateFilterDescriptor((cudnnFilterDescriptor_t*)&filter_handle));
check(cudnnSetFilter4dDescriptor((cudnnFilterDescriptor_t)filter_handle,
CHECK_CUDNN(cudnnCreateFilterDescriptor((cudnnFilterDescriptor_t*)&filter_handle));
CHECK_CUDNN(cudnnSetFilter4dDescriptor((cudnnFilterDescriptor_t)filter_handle,
CUDNN_DATA_FLOAT,
filters.num_samples(),
filters.k(),
filters.nr(),
filters.nc()));
check(cudnnCreateConvolutionDescriptor((cudnnConvolutionDescriptor_t*)&conv_handle));
check(cudnnSetConvolution2dDescriptor((cudnnConvolutionDescriptor_t)conv_handle,
CHECK_CUDNN(cudnnCreateConvolutionDescriptor((cudnnConvolutionDescriptor_t*)&conv_handle));
CHECK_CUDNN(cudnnSetConvolution2dDescriptor((cudnnConvolutionDescriptor_t)conv_handle,
filters.nr()/2, // vertical padding
filters.nc()/2, // horizontal padding
stride_y,
......@@ -321,7 +332,7 @@ namespace dlib
1, 1, // must be 1,1
CUDNN_CONVOLUTION)); // could also be CUDNN_CROSS_CORRELATION
check(cudnnGetConvolution2dForwardOutputDim(
CHECK_CUDNN(cudnnGetConvolution2dForwardOutputDim(
(const cudnnConvolutionDescriptor_t)conv_handle,
descriptor(data),
(const cudnnFilterDescriptor_t)filter_handle,
......@@ -336,7 +347,7 @@ namespace dlib
// Pick which forward algorithm we will use and allocate the necessary
// workspace buffer.
cudnnConvolutionFwdAlgo_t forward_best_algo;
check(cudnnGetConvolutionForwardAlgorithm(
CHECK_CUDNN(cudnnGetConvolutionForwardAlgorithm(
context(),
descriptor(data),
(const cudnnFilterDescriptor_t)filter_handle,
......@@ -346,7 +357,7 @@ namespace dlib
std::numeric_limits<size_t>::max(),
&forward_best_algo));
forward_algo = forward_best_algo;
check(cudnnGetConvolutionForwardWorkspaceSize(
CHECK_CUDNN(cudnnGetConvolutionForwardWorkspaceSize(
context(),
descriptor(data),
(const cudnnFilterDescriptor_t)filter_handle,
......@@ -360,7 +371,7 @@ namespace dlib
// Pick which backward data algorithm we will use and allocate the
// necessary workspace buffer.
cudnnConvolutionBwdDataAlgo_t backward_data_best_algo;
check(cudnnGetConvolutionBackwardDataAlgorithm(
CHECK_CUDNN(cudnnGetConvolutionBackwardDataAlgorithm(
context(),
(const cudnnFilterDescriptor_t)filter_handle,
descriptor(dest_desc),
......@@ -370,7 +381,7 @@ namespace dlib
std::numeric_limits<size_t>::max(),
&backward_data_best_algo));
backward_data_algo = backward_data_best_algo;
check(cudnnGetConvolutionBackwardDataWorkspaceSize(
CHECK_CUDNN(cudnnGetConvolutionBackwardDataWorkspaceSize(
context(),
(const cudnnFilterDescriptor_t)filter_handle,
descriptor(dest_desc),
......@@ -384,7 +395,7 @@ namespace dlib
// Pick which backward filters algorithm we will use and allocate the
// necessary workspace buffer.
cudnnConvolutionBwdFilterAlgo_t backward_filters_best_algo;
check(cudnnGetConvolutionBackwardFilterAlgorithm(
CHECK_CUDNN(cudnnGetConvolutionBackwardFilterAlgorithm(
context(),
descriptor(data),
descriptor(dest_desc),
......@@ -394,7 +405,7 @@ namespace dlib
std::numeric_limits<size_t>::max(),
&backward_filters_best_algo));
backward_filters_algo = backward_filters_best_algo;
check(cudnnGetConvolutionBackwardFilterWorkspaceSize(
CHECK_CUDNN(cudnnGetConvolutionBackwardFilterWorkspaceSize(
context(),
descriptor(data),
descriptor(dest_desc),
......@@ -434,7 +445,7 @@ namespace dlib
const float alpha = 1;
const float beta = 0;
check(cudnnConvolutionForward(
CHECK_CUDNN(cudnnConvolutionForward(
context(),
&alpha,
descriptor(data),
......@@ -460,7 +471,7 @@ namespace dlib
const float beta = 1;
check(cudnnConvolutionBackwardData_v3(context(),
CHECK_CUDNN(cudnnConvolutionBackwardData_v3(context(),
&alpha,
(const cudnnFilterDescriptor_t)filter_handle,
filters.device(),
......@@ -484,7 +495,7 @@ namespace dlib
{
const float alpha = 1;
const float beta = 0;
check(cudnnConvolutionBackwardFilter_v3(context(),
CHECK_CUDNN(cudnnConvolutionBackwardFilter_v3(context(),
&alpha,
descriptor(data),
data.device(),
......@@ -535,10 +546,10 @@ namespace dlib
stride_x = stride_x_;
stride_y = stride_y_;
cudnnPoolingDescriptor_t poolingDesc;
check(cudnnCreatePoolingDescriptor(&poolingDesc));
CHECK_CUDNN(cudnnCreatePoolingDescriptor(&poolingDesc));
handle = poolingDesc;
check(cudnnSetPooling2dDescriptor(poolingDesc,
CHECK_CUDNN(cudnnSetPooling2dDescriptor(poolingDesc,
CUDNN_POOLING_MAX,
window_height,
window_width,
......@@ -559,7 +570,7 @@ namespace dlib
int outC;
int outH;
int outW;
check(cudnnGetPooling2dForwardOutputDim((const cudnnPoolingDescriptor_t)handle,
CHECK_CUDNN(cudnnGetPooling2dForwardOutputDim((const cudnnPoolingDescriptor_t)handle,
descriptor(src),
&outN,
&outC,
......@@ -574,7 +585,7 @@ namespace dlib
DLIB_CASSERT(dest.nr() == src.nr()/stride_y,"");
DLIB_CASSERT(dest.nc() == src.nc()/stride_x,"");
check(cudnnPoolingForward(context(),
CHECK_CUDNN(cudnnPoolingForward(context(),
(const cudnnPoolingDescriptor_t)handle,
&alpha,
descriptor(src),
......@@ -596,7 +607,7 @@ namespace dlib
const float alpha = 1;
const float beta = 0;
check(cudnnPoolingBackward(context(),
CHECK_CUDNN(cudnnPoolingBackward(context(),
(const cudnnPoolingDescriptor_t)handle,
&alpha,
descriptor(dest),
......@@ -625,7 +636,7 @@ namespace dlib
const float alpha = 1;
const float beta = 0;
check(cudnnSoftmaxForward(context(),
CHECK_CUDNN(cudnnSoftmaxForward(context(),
CUDNN_SOFTMAX_ACCURATE,
CUDNN_SOFTMAX_MODE_CHANNEL,
&alpha,
......@@ -651,7 +662,7 @@ namespace dlib
const float alpha = 1;
const float beta = 0;
check(cudnnSoftmaxBackward(context(),
CHECK_CUDNN(cudnnSoftmaxBackward(context(),
CUDNN_SOFTMAX_ACCURATE,
CUDNN_SOFTMAX_MODE_CHANNEL,
&alpha,
......@@ -678,7 +689,7 @@ namespace dlib
const float alpha = 1;
const float beta = 0;
check(cudnnActivationForward(context(),
CHECK_CUDNN(cudnnActivationForward(context(),
CUDNN_ACTIVATION_SIGMOID,
&alpha,
descriptor(src),
......@@ -702,7 +713,7 @@ namespace dlib
const float alpha = 1;
const float beta = 0;
check(cudnnActivationBackward(context(),
CHECK_CUDNN(cudnnActivationBackward(context(),
CUDNN_ACTIVATION_SIGMOID,
&alpha,
descriptor(dest),
......@@ -729,7 +740,7 @@ namespace dlib
const float alpha = 1;
const float beta = 0;
check(cudnnActivationForward(context(),
CHECK_CUDNN(cudnnActivationForward(context(),
CUDNN_ACTIVATION_RELU,
&alpha,
descriptor(src),
......@@ -753,7 +764,7 @@ namespace dlib
const float alpha = 1;
const float beta = 0;
check(cudnnActivationBackward(context(),
CHECK_CUDNN(cudnnActivationBackward(context(),
CUDNN_ACTIVATION_RELU,
&alpha,
descriptor(dest),
......@@ -780,7 +791,7 @@ namespace dlib
const float alpha = 1;
const float beta = 0;
check(cudnnActivationForward(context(),
CHECK_CUDNN(cudnnActivationForward(context(),
CUDNN_ACTIVATION_TANH,
&alpha,
descriptor(src),
......@@ -804,7 +815,7 @@ namespace dlib
const float alpha = 1;
const float beta = 0;
check(cudnnActivationBackward(context(),
CHECK_CUDNN(cudnnActivationBackward(context(),
CUDNN_ACTIVATION_TANH,
&alpha,
descriptor(dest),
......
dlib/dnn/curand_dlibapi.cpp
View file @ ccb148b4
......@@ -9,27 +9,36 @@
#include <curand.h>
#include "../string.h"
namespace dlib
static const char* curand_get_error_string(curandStatus_t s)
{
namespace cuda
switch(s)
{
case CURAND_STATUS_NOT_INITIALIZED:
return "CUDA Runtime API initialization failed.";
case CURAND_STATUS_LENGTH_NOT_MULTIPLE:
return "The requested length must be a multiple of two.";
default:
return "A call to cuRAND failed";
}
}
// ----------------------------------------------------------------------------------------
// Check the return value of a call to the cuDNN runtime for an error condition.
#define CHECK_CURAND(call) \
{ \
const curandStatus_t error = call; \
if (error != CURAND_STATUS_SUCCESS) \
{ \
std::ostringstream sout; \
sout << "Error while calling " << #call << " in file " << __FILE__ << ":" << __LINE__ << ". ";\
sout << "code: " << error << ", reason: " << curand_get_error_string(error);\
throw dlib::curand_error(sout.str()); \
} \
}
// TODO, make into a macro that prints more information like the line number, etc.
static void check(curandStatus_t s)
{
switch(s)
{
case CURAND_STATUS_SUCCESS: return;
case CURAND_STATUS_NOT_INITIALIZED:
throw curand_error("CUDA Runtime API initialization failed.");
case CURAND_STATUS_LENGTH_NOT_MULTIPLE:
throw curand_error("The requested length must be a multiple of two.");
default:
throw curand_error("A call to cuRAND failed: " + cast_to_string(s));
}
}
namespace dlib
{
namespace cuda
{
// ----------------------------------------------------------------------------------------
......@@ -39,10 +48,10 @@ namespace dlib
) : handle(nullptr)
{
curandGenerator_t gen;
check(curandCreateGenerator(&gen, CURAND_RNG_PSEUDO_DEFAULT));
CHECK_CURAND(curandCreateGenerator(&gen, CURAND_RNG_PSEUDO_DEFAULT));
handle = gen;
check(curandSetPseudoRandomGeneratorSeed(gen, seed));
CHECK_CURAND(curandSetPseudoRandomGeneratorSeed(gen, seed));
}
curand_generator::
......@@ -64,7 +73,7 @@ namespace dlib
if (data.size() == 0)
return;
check(curandGenerateNormal((curandGenerator_t)handle,
CHECK_CURAND(curandGenerateNormal((curandGenerator_t)handle,
data.device(),
data.size(),
mean,
......@@ -79,7 +88,7 @@ namespace dlib
if (data.size() == 0)
return;
check(curandGenerateUniform((curandGenerator_t)handle, data.device(), data.size()));
CHECK_CURAND(curandGenerateUniform((curandGenerator_t)handle, data.device(), data.size()));
}
// -----------------------------------------------------------------------------------
......
dlib/dnn/curand_dlibapi.h
View file @ ccb148b4
......@@ -6,7 +6,7 @@
#ifdef DLIB_USE_CUDA
#include "tensor.h"
#include "../error.h"
#include "cuda_errors.h"
namespace dlib
{
......@@ -15,13 +15,6 @@ namespace dlib
// -----------------------------------------------------------------------------------
struct curand_error : public error
{
curand_error(const std::string& message): error(message) {}
};
// ----------------------------------------------------------------------------------------
class curand_generator
{
public:
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment