void caffe_gpu_gemv(
const CBLAS_TRANSPOSE TransA,
const int m,
const int n,
const T alpha,
const T* A,
const int step_A,
const T* x,
const int step_x,
const T beta,
T* y,
const int step_y) {
clblasTranspose clTransA;
switch (TransA) {
case CblasNoTrans:
clTransA = clblasNoTrans;
break;
case CblasTrans:
clTransA = clblasTrans;
break;
case CblasConjTrans:
clTransA = clblasConjTrans;
break;
default:
LOG(ERROR)<< "unknown transpose mode.";
return;
}
#ifdef USE_CLGEMM
BOOL_CHECK(caffe::OpenCL::clgemv<T>(clTransA, m, n, alpha, A, step_A, x, step_x, beta, y, step_y));
#else
BOOL_CHECK(caffe::OpenCL::clBLASgemv<T>(clTransA, m, n, alpha, A, step_A, x, step_x, beta, y, step_y));
#endif
}
bool SA_fsup::GetMarkerFrequency(double &freqHz) {
char freqBuffer[200];
ZeroMemory(freqBuffer, 200);
BOOL_CHECK(Print("CALC:MARK1:X?;*WAI"));
BOOL_CHECK(Read(freqBuffer, 200));
freqHz = atof(freqBuffer);
return true;
}
bool SA_fsup::SetAverTrace(bool isEnable, int traceCount) {
if (isEnable) {
BOOL_CHECK(Print("AVER:COUN %d;*WAI", traceCount));
BOOL_CHECK(Print("AVER:STAT ON;*WAI"));
} else {
BOOL_CHECK(Print("AVER:STAT OFF;*WAI"));
}
return true;
}
void caffe_gpu_gemm(
const CBLAS_TRANSPOSE TransA,
const CBLAS_TRANSPOSE TransB,
const int M,
const int N,
const int K,
const T alpha,
const T* A,
const size_t idx_offset_A,
const T* B,
const size_t idx_offset_B,
const T beta,
T* C,
const size_t idx_offset_C,
cl_event* event) {
clblasTranspose clTransA;
switch (TransA) {
case CblasNoTrans:
clTransA = clblasNoTrans;
break;
case CblasTrans:
clTransA = clblasTrans;
break;
case CblasConjTrans:
clTransA = clblasConjTrans;
break;
default:
LOG(ERROR)<< "unknown transpose mode.";
return;
}
clblasTranspose clTransB;
switch (TransB) {
case CblasNoTrans:
clTransB = clblasNoTrans;
break;
case CblasTrans:
clTransB = clblasTrans;
break;
case CblasConjTrans:
clTransB = clblasConjTrans;
break;
default:
LOG(ERROR)<< "unknown transpose mode.";
return;
}
#ifdef USE_CLGEMM
BOOL_CHECK(caffe::OpenCL::clgemm<T>(clTransA, clTransB, M, N, K, alpha, A, idx_offset_A, B, idx_offset_B, beta, C, idx_offset_C, event));
#else
BOOL_CHECK(caffe::OpenCL::clBLASgemm<T>(clTransA, clTransB, M, N, K, alpha, A, idx_offset_A, B, idx_offset_B, beta, C, idx_offset_C));
#endif
}
bool SA_fsup::GetMarkerPwr(double &dbm) {
char dbmBuffer[200];
double freqHz;
ZeroMemory(dbmBuffer, 200);
BOOL_CHECK(Print("CALC:MARK1:Y?;*WAI"));
BOOL_CHECK(Read(dbmBuffer, 200));
BOOL_CHECK(GetMarkerFrequency(freqHz));
dbm = atof(dbmBuffer);
if (m_isEnableCableLoss) {
dbm -= m_cableLossData.GetData(freqHz, dbm);
}
return true;
}
bool SA_fsup::SetAverTraceAndGetData(int traceCount, int pointCount, double *data) {
int bufferSize = 18 * pointCount;
char *buffer = new char[bufferSize];
BOOL_CHECK(Print("SWE:POIN %d;*WAI", pointCount));
BOOL_CHECK(SetAverTrace(true, traceCount));
BOOL_CHECK(SweepOnce());
BOOL_CHECK(Print("FORM ASCii;*WAI"));
BOOL_CHECK(Print("TRAC? TRACE1;*WAI"));
BOOL_CHECK(Read(buffer, bufferSize));
int dataIdx = 0, lastIdx = 0;
for (int i = 0; i < bufferSize; i++) {
if (buffer[i] == ',' || i == bufferSize - 1) {
buffer[i] = 0;
data[dataIdx++] = atof(&buffer[lastIdx]);
lastIdx = i + 1;
i++;
}
}
delete[] buffer;
return true;
}
void caffe_gpu_axpy(const int N, const T alpha, const T* X, T* Y) {
BOOL_CHECK(caffe::OpenCL::clBLASaxpy<T>(N, alpha, X, 1, Y, 1));
}
void caffe_gpu_mul(const int n, const T* x, const T* y, T* z) {
BOOL_CHECK(caffe::OpenCL::clmul<T>(n, x, y, z));
}
void caffe_gpu_scale(const int n, const float alpha, const T* x, T* y) {
BOOL_CHECK(caffe::OpenCL::clBLASscal<T>(n, alpha, x, y));
}
void caffe_gpu_rng_bernoulli(const int n, const T1 p, T2* r) {
BOOL_CHECK(caffe::OpenCL::cl_caffe_gpu_rng_bernoulli(n, p, r));
}
void caffe_gpu_abs(const int n, const T* x, T* y) {
BOOL_CHECK(caffe::OpenCL::clabs<T>(n, x, y));
}
void caffe_gpu_sgnbit(const int n, const T* x, T* y) {
BOOL_CHECK(caffe::OpenCL::clsgnbit<T>(n, x, y));
}
void caffe_gpu_rng_gaussian(const int n, const T mu, const T sigma, T* r) {
BOOL_CHECK(caffe::OpenCL::cl_caffe_gpu_rng_gaussian(n, mu, sigma, r));
}
void caffe_gpu_add(const int n, const T* a, const T* b, T* y) {
BOOL_CHECK(caffe::OpenCL::cladd<T>(n, a, b, y));
}
void caffe_gpu_rng_uniform(const int n, const T a, const T b, T* r) {
BOOL_CHECK(caffe::OpenCL::cl_caffe_gpu_rng_uniform(n, a, b, r));
}
void caffe_gpu_powx(const int n, const T* a, const T alpha, T* y) {
BOOL_CHECK(caffe::OpenCL::clpowx<T>(n, a, alpha, y));
}
void caffe_gpu_add_scalar(const int N, const T alpha, T* Y) {
// NOLINT_NEXT_LINE(whitespace/operators)
//add_scalar_kernel<float><<<CAFFE_GET_BLOCKS(N), CAFFE_CUDA_NUM_THREADS>>>(N, alpha, Y);
BOOL_CHECK(caffe::OpenCL::cladd_scalar(N, alpha, Y));
}
void caffe_gpu_sign(const int n, const T* x, T* y) {
BOOL_CHECK(caffe::OpenCL::clsign<T>(n, x, y));
}
void caffe_gpu_asum(const int n, const T* x, T* y) {
BOOL_CHECK(caffe::OpenCL::clBLASasum<T>(n, x, y));
}
void caffe_gpu_memcpy(const size_t N, const void* X, void* Y, int type) {
if (X != Y) {
BOOL_CHECK(caffe::OpenCL::clMemcpy(Y, X, N, type)); // NOLINT(caffe/alt_fn)
}
}
void caffe_gpu_memcpy(const size_t N, const void* X, void* Y) {
if (X != Y) {
BOOL_CHECK(caffe::OpenCL::clMemcpy(Y, X, N, caffe::OpenCL::COPY_DEFAULT)); // NOLINT(caffe/alt_fn)
}
}
void caffe_gpu_exp(const int n, const T* a, T* y) {
BOOL_CHECK(caffe::OpenCL::clexp<T>(n, a, y));
}
bool SA_fsup::SweepOnce() {
BOOL_CHECK(Print("INIT1:CONT OFF;*WAI;"));
BOOL_CHECK(Print("INIT1:IMM;*WAI;"));
return true;
}
void caffe_gpu_scal(const int N, const T alpha, T *X) {
//CUBLAS_CHECK(cublasSscal(Caffe::cublas_handle(), N, &alpha, X, 1));
BOOL_CHECK(caffe::OpenCL::clBLASscal<T>(N, alpha, X, X));
}
void caffe_gpu_rng_uniform(const int n, unsigned int* r) {
BOOL_CHECK(caffe::OpenCL::cl_caffe_gpu_rng_uniform(n, r));
}
void caffe_gpu_dot(const int n, const T* x, const T* y, T* out) {
BOOL_CHECK(caffe::OpenCL::clBLASdot<T>(n, x, 1, y, 1, out));
}
