Mentions légales du service

Skip to content
Snippets Groups Projects
Commit fed57eb3 authored by hhakim's avatar hhakim
Browse files

Adjust GPU2 Faust::MatDense, Faust::MatSparse and Faust::Vect to support... 

Adjust GPU2 Faust::MatDense, Faust::MatSparse and Faust::Vect to support complex scalars (updating to gpu_mod@3920c368).
parent 64642111
Branches
Tags
No related merge requests found
Expand all Hide whitespace changes
Inline Side-by-side
Showing
with 262 additions and 255 deletions
CMakeLists.txt
+ 7
− 2
View file @ fed57eb3
...@@ -605,6 +605,7 @@ include(CMake/findExternalLibs.cmake) ...@@ -605,6 +605,7 @@ include(CMake/findExternalLibs.cmake)
######################################### #########################################
if(USE_GPU_MOD) if(USE_GPU_MOD)
# GPU2
find_package(GPU_MOD) #REQUIRED) find_package(GPU_MOD) #REQUIRED)
if(NOT GPU_MOD_INCLUDE_DIR) if(NOT GPU_MOD_INCLUDE_DIR)
set(GPU_MOD_INCLUDE_DIR ${PROJECT_SOURCE_DIR}/gpu_mod/src) # fallback to submodule directory in case find_package didn't work set(GPU_MOD_INCLUDE_DIR ${PROJECT_SOURCE_DIR}/gpu_mod/src) # fallback to submodule directory in case find_package didn't work
...@@ -620,9 +621,14 @@ if(USE_GPU_MOD) ...@@ -620,9 +621,14 @@ if(USE_GPU_MOD)
configure_file(${FAUST_SRC_LINEAR_OPERATOR_GPU2_DIR}/faust_Vect_gpu.cpp.in ${FAUST_SRC_LINEAR_OPERATOR_GPU2_DIR}/faust_Vect_gpu_double.cpp) configure_file(${FAUST_SRC_LINEAR_OPERATOR_GPU2_DIR}/faust_Vect_gpu.cpp.in ${FAUST_SRC_LINEAR_OPERATOR_GPU2_DIR}/faust_Vect_gpu_double.cpp)
#target_sources(${FAUST_TARGET} PUBLIC ${GPU_MOD_CPP_FILES}) #target_sources(${FAUST_TARGET} PUBLIC ${GPU_MOD_CPP_FILES})
#list(APPEND CPP_FILES ) #list(APPEND CPP_FILES )
set(FAUST_SCALAR_FOR_GM complex<double>) set(FAUST_SCALAR_FOR_GM std::complex<double>)
set(GM_SCALAR cuDoubleComplex) set(GM_SCALAR cuDoubleComplex)
set(GM_REINTERPRET_CAST_SCALAR double) set(GM_REINTERPRET_CAST_SCALAR double)
#configure_file(${FAUST_LINEAR_OPERATOR_CPU_SRC_DIR}/faust_gpu_mod_gen.hpp.in ${FAUST_LINEAR_OPERATOR_CPU_SRC_DIR}/faust_gpu_mod_complexdouble.hpp)
configure_file(${FAUST_SRC_LINEAR_OPERATOR_GPU2_DIR}/faust_MatDense_gpu.cpp.in ${FAUST_SRC_LINEAR_OPERATOR_GPU2_DIR}/faust_MatDense_gpu_complexdouble.cpp)
configure_file(${FAUST_SRC_LINEAR_OPERATOR_GPU2_DIR}/faust_MatSparse_gpu.cpp.in ${FAUST_SRC_LINEAR_OPERATOR_GPU2_DIR}/faust_MatSparse_gpu_complexdouble.cpp)
configure_file(${FAUST_SRC_LINEAR_OPERATOR_GPU2_DIR}/faust_Transform_gpu.cpp.in ${FAUST_SRC_LINEAR_OPERATOR_GPU2_DIR}/faust_Transform_gpu_complexdouble.cpp)
configure_file(${FAUST_SRC_LINEAR_OPERATOR_GPU2_DIR}/faust_Vect_gpu.cpp.in ${FAUST_SRC_LINEAR_OPERATOR_GPU2_DIR}/faust_Vect_gpu_complexdouble.cpp)
configure_file(${FAUST_LINEAR_OPERATOR_CPU_SRC_DIR}/faust_gpu_mod_gen.hpp.in ${FAUST_LINEAR_OPERATOR_CPU_SRC_DIR}/faust_gpu_mod_complexdouble.hpp) configure_file(${FAUST_LINEAR_OPERATOR_CPU_SRC_DIR}/faust_gpu_mod_gen.hpp.in ${FAUST_LINEAR_OPERATOR_CPU_SRC_DIR}/faust_gpu_mod_complexdouble.hpp)
# set install relative path of the lib (for each OS) # set install relative path of the lib (for each OS)
if(APPLE) if(APPLE)
...@@ -634,7 +640,6 @@ if(USE_GPU_MOD) ...@@ -634,7 +640,6 @@ if(USE_GPU_MOD)
endif() endif()
file(TO_NATIVE_PATH lib/${GPU_MOD_LIB} GPU_MOD_LIB_PATH) # used in setup.py file(TO_NATIVE_PATH lib/${GPU_MOD_LIB} GPU_MOD_LIB_PATH) # used in setup.py
file(TO_NATIVE_PATH ${CMAKE_INSTALL_PREFIX}/${GPU_MOD_LIB} GPU_MOD_LIB_INSTALL_PATH) # only for system packages (not for pip package) file(TO_NATIVE_PATH ${CMAKE_INSTALL_PREFIX}/${GPU_MOD_LIB} GPU_MOD_LIB_INSTALL_PATH) # only for system packages (not for pip package)
# GPU2
endif() endif()
......
gpu_mod @ 3920c368
Compare ed251cfa...3920c368
Subproject commit ed251cfad23484794e02500f1e413b86eb58eb85 Subproject commit 3920c368b6d21dd5b4f7d6b86535bfee96518357
misc/test/src/C++/test_matdense_gpu_mod.cpp.in
+ 71
− 68
View file @ fed57eb3
...@@ -8,10 +8,13 @@ ...@@ -8,10 +8,13 @@
#include "faust_cuda_device.h" #include "faust_cuda_device.h"
#include <cstdlib> #include <cstdlib>
#include <cmath> #include <cmath>
typedef @TEST_FPP@ FPP;
using namespace std; using namespace std;
using namespace Faust; using namespace Faust;
double calc_err(MatDense<double, GPU2> &gpu_mat, MatDense<double, Cpu> &cpu_mat) double calc_err(MatDense<FPP, GPU2> &gpu_mat, MatDense<FPP, Cpu> &cpu_mat)
{ {
auto gpu_mat_to_cpu = gpu_mat.tocpu(); auto gpu_mat_to_cpu = gpu_mat.tocpu();
auto err_diff = cpu_mat; auto err_diff = cpu_mat;
...@@ -26,10 +29,10 @@ void test_mul_gpu_dense() ...@@ -26,10 +29,10 @@ void test_mul_gpu_dense()
{ {
faust_unsigned_int nrows = 1024, ncols = 1024; faust_unsigned_int nrows = 1024, ncols = 1024;
faust_unsigned_int nrows2 = 1024, ncols2 = 1024; faust_unsigned_int nrows2 = 1024, ncols2 = 1024;
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData()); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData());
auto cpu_mat2 = Faust::MatDense<double,Cpu>::randMat(nrows2,ncols2); auto cpu_mat2 = Faust::MatDense<FPP,Cpu>::randMat(nrows2,ncols2);
MatDense<double,GPU2> gpu_mat2(nrows2, ncols2, cpu_mat2->getData()); MatDense<FPP,GPU2> gpu_mat2(nrows2, ncols2, cpu_mat2->getData());
// test MatDense<FPP,GPU2> * MatDense<FPP,GPU2> // test MatDense<FPP,GPU2> * MatDense<FPP,GPU2>
cout << "Mul. GPUDense*GPUDense in GPUDense" << endl; cout << "Mul. GPUDense*GPUDense in GPUDense" << endl;
gpu_mat1.multiply(gpu_mat2); gpu_mat1.multiply(gpu_mat2);
...@@ -50,7 +53,7 @@ void test_mul_gpu_dense() ...@@ -50,7 +53,7 @@ void test_mul_gpu_dense()
cout << "err mul.: " << err_diff.norm()/cpu_mat1_mat2_ref.norm() << endl; cout << "err mul.: " << err_diff.norm()/cpu_mat1_mat2_ref.norm() << endl;
cout << "OK" << endl; cout << "OK" << endl;
cout << "Mul. GPUDense*CPUSparse in CPUDense" << endl; cout << "Mul. GPUDense*CPUSparse in CPUDense" << endl;
Faust::MatSparse<double,Cpu> cpu_mat2_sparse(*cpu_mat2); Faust::MatSparse<FPP,Cpu> cpu_mat2_sparse(*cpu_mat2);
// cout << cpu_mat2_sparse.to_string(false, true) << endl; // cout << cpu_mat2_sparse.to_string(false, true) << endl;
// cout << cpu_mat2->to_string(false, true) << endl; // cout << cpu_mat2->to_string(false, true) << endl;
gpu_mat1.multiply(cpu_mat2_sparse, cpu_mat1_mat2_test); gpu_mat1.multiply(cpu_mat2_sparse, cpu_mat1_mat2_test);
...@@ -63,7 +66,7 @@ void test_mul_gpu_dense() ...@@ -63,7 +66,7 @@ void test_mul_gpu_dense()
cout << "err mul.: " << err_diff.norm()/cpu_mat1_mat2_ref.norm() << endl; cout << "err mul.: " << err_diff.norm()/cpu_mat1_mat2_ref.norm() << endl;
cout << "OK" << endl; cout << "OK" << endl;
cout << "Mul. GPUDense*CPUSparse in GPUDense" << endl; cout << "Mul. GPUDense*CPUSparse in GPUDense" << endl;
MatDense<double, GPU2> gpu_mat1_mat2_test(nrows, ncols2); MatDense<FPP, GPU2> gpu_mat1_mat2_test(nrows, ncols2);
gpu_mat1.multiply(cpu_mat2_sparse, gpu_mat1_mat2_test); gpu_mat1.multiply(cpu_mat2_sparse, gpu_mat1_mat2_test);
auto gpu_mat1_mat2_test_to_cpu = gpu_mat1_mat2_test.tocpu(); auto gpu_mat1_mat2_test_to_cpu = gpu_mat1_mat2_test.tocpu();
err_diff = gpu_mat1_mat2_test_to_cpu; err_diff = gpu_mat1_mat2_test_to_cpu;
...@@ -71,8 +74,8 @@ void test_mul_gpu_dense() ...@@ -71,8 +74,8 @@ void test_mul_gpu_dense()
cout << "err mul.: " << err_diff.norm()/cpu_mat1_mat2_ref.norm() << endl; cout << "err mul.: " << err_diff.norm()/cpu_mat1_mat2_ref.norm() << endl;
cout << "OK" << endl; cout << "OK" << endl;
cout << "MatDense*cpu_vec" << endl; cout << "MatDense*cpu_vec" << endl;
Faust::Vect<double, Cpu>* vec = Faust::Vect<double,Cpu>::rand(cpu_mat1->getNbCol()); Faust::Vect<FPP, Cpu>* vec = Faust::Vect<FPP,Cpu>::rand(cpu_mat1->getNbCol());
Faust::Vect<double, Cpu> vec_copy(*vec); Faust::Vect<FPP, Cpu> vec_copy(*vec);
cout << vec->norm() << endl; cout << vec->norm() << endl;
cout << vec_copy.norm() << endl; cout << vec_copy.norm() << endl;
gpu_mat1.multiply(*vec); gpu_mat1.multiply(*vec);
...@@ -94,10 +97,10 @@ void test_mul_eq_op() ...@@ -94,10 +97,10 @@ void test_mul_eq_op()
// convert the two cpu mats to gpu mats // convert the two cpu mats to gpu mats
faust_unsigned_int nrows = 1024, ncols = 1024; faust_unsigned_int nrows = 1024, ncols = 1024;
faust_unsigned_int nrows2 = 1024, ncols2 = 784; faust_unsigned_int nrows2 = 1024, ncols2 = 784;
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData()); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData());
auto cpu_mat2 = Faust::MatDense<double,Cpu>::randMat(nrows2,ncols2); auto cpu_mat2 = Faust::MatDense<FPP,Cpu>::randMat(nrows2,ncols2);
MatDense<double,GPU2> gpu_mat2(nrows2, ncols2, cpu_mat2->getData()); MatDense<FPP,GPU2> gpu_mat2(nrows2, ncols2, cpu_mat2->getData());
// test *= // test *=
*cpu_mat1 *= *cpu_mat2; *cpu_mat1 *= *cpu_mat2;
gpu_mat1 *= *cpu_mat2; gpu_mat1 *= *cpu_mat2;
...@@ -109,9 +112,9 @@ void test_mul_spm() ...@@ -109,9 +112,9 @@ void test_mul_spm()
{ {
cout << "test_mul_spm()" << endl; cout << "test_mul_spm()" << endl;
faust_unsigned_int nrows = 1024, ncols = 784; faust_unsigned_int nrows = 1024, ncols = 784;
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
Faust::MatSparse<double, Cpu> *cpu_mat2_sp = Faust::MatSparse<double, Cpu>::randMat(ncols, nrows, .2); Faust::MatSparse<FPP, Cpu> *cpu_mat2_sp = Faust::MatSparse<FPP, Cpu>::randMat(ncols, nrows, .2);
MatDense<double,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData()); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData());
gpu_mat1.multiplyLeft(*cpu_mat2_sp); gpu_mat1.multiplyLeft(*cpu_mat2_sp);
cpu_mat1->multiplyLeft(*cpu_mat2_sp); cpu_mat1->multiplyLeft(*cpu_mat2_sp);
// cout << gpu_mat1.tocpu().to_string(false, true) << endl; // cout << gpu_mat1.tocpu().to_string(false, true) << endl;
...@@ -125,8 +128,8 @@ void test_resize() ...@@ -125,8 +128,8 @@ void test_resize()
{ {
cout << "Test resize()" << endl; cout << "Test resize()" << endl;
faust_unsigned_int nrows = 1024, ncols = 784; faust_unsigned_int nrows = 1024, ncols = 784;
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData()); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData());
gpu_mat1.resize(45, 12); gpu_mat1.resize(45, 12);
cout << "OK" << endl; cout << "OK" << endl;
} }
...@@ -135,8 +138,8 @@ void test_setones() ...@@ -135,8 +138,8 @@ void test_setones()
{ {
cout << "Test setones()" << endl; cout << "Test setones()" << endl;
faust_unsigned_int nrows = 1024, ncols = 784; faust_unsigned_int nrows = 1024, ncols = 784;
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat1(nrows, ncols); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols);
gpu_mat1.setOnes(); gpu_mat1.setOnes();
auto gpu_mat1_to_cpu = gpu_mat1.tocpu(); auto gpu_mat1_to_cpu = gpu_mat1.tocpu();
// cout << gpu_mat1_to_cpu.to_string(false, true) << endl; // cout << gpu_mat1_to_cpu.to_string(false, true) << endl;
...@@ -151,8 +154,8 @@ void test_setzeros() ...@@ -151,8 +154,8 @@ void test_setzeros()
{ {
cout << "Test setzeros()" << endl; cout << "Test setzeros()" << endl;
faust_unsigned_int nrows = 1024, ncols = 784; faust_unsigned_int nrows = 1024, ncols = 784;
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat1(nrows, ncols); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols);
gpu_mat1.setZeros(); gpu_mat1.setZeros();
auto gpu_mat1_to_cpu = gpu_mat1.tocpu(); auto gpu_mat1_to_cpu = gpu_mat1.tocpu();
// cout << gpu_mat1_to_cpu.to_string(false, true) << endl; // cout << gpu_mat1_to_cpu.to_string(false, true) << endl;
...@@ -164,8 +167,8 @@ void test_seteyes() ...@@ -164,8 +167,8 @@ void test_seteyes()
{ {
cout << "Test seteyes()" << endl; cout << "Test seteyes()" << endl;
faust_unsigned_int nrows = 1024, ncols = 1024; faust_unsigned_int nrows = 1024, ncols = 1024;
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat1(nrows, ncols); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols);
gpu_mat1.setEyes(); gpu_mat1.setEyes();
auto gpu_mat1_to_cpu = gpu_mat1.tocpu(); auto gpu_mat1_to_cpu = gpu_mat1.tocpu();
cout << gpu_mat1_to_cpu.to_string(false, true) << endl; cout << gpu_mat1_to_cpu.to_string(false, true) << endl;
...@@ -180,8 +183,8 @@ void test_clone() ...@@ -180,8 +183,8 @@ void test_clone()
{ {
cout << "Test clone()" << endl; cout << "Test clone()" << endl;
faust_unsigned_int nrows = 1024, ncols = 1024; faust_unsigned_int nrows = 1024, ncols = 1024;
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData()); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData());
auto gpu_mat1_to_cpu = gpu_mat1.tocpu(); auto gpu_mat1_to_cpu = gpu_mat1.tocpu();
auto gpu_mat1_clone = gpu_mat1.clone(); auto gpu_mat1_clone = gpu_mat1.clone();
auto gpu_mat1_clone_to_cpu = gpu_mat1_clone->tocpu(); auto gpu_mat1_clone_to_cpu = gpu_mat1_clone->tocpu();
...@@ -198,8 +201,8 @@ void test_transpose() ...@@ -198,8 +201,8 @@ void test_transpose()
{ {
cout << "Test transpose" << endl; cout << "Test transpose" << endl;
faust_unsigned_int nrows = 1024, ncols = 1024; faust_unsigned_int nrows = 1024, ncols = 1024;
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData()); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData());
// cout << cpu_mat1->to_string(false, true) << endl; // cout << cpu_mat1->to_string(false, true) << endl;
cpu_mat1->transpose(); cpu_mat1->transpose();
gpu_mat1.transpose(); gpu_mat1.transpose();
...@@ -216,8 +219,8 @@ void test_adjoint() ...@@ -216,8 +219,8 @@ void test_adjoint()
{ {
cout << "Test adjoint" << endl; cout << "Test adjoint" << endl;
faust_unsigned_int nrows = 1024, ncols = 1024; faust_unsigned_int nrows = 1024, ncols = 1024;
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData()); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData());
// cout << cpu_mat1->to_string(false, true) << endl; // cout << cpu_mat1->to_string(false, true) << endl;
cpu_mat1->adjoint(); cpu_mat1->adjoint();
gpu_mat1.adjoint(); gpu_mat1.adjoint();
...@@ -234,8 +237,8 @@ void test_conjugate() ...@@ -234,8 +237,8 @@ void test_conjugate()
{ {
cout << "Test conjugate" << endl; cout << "Test conjugate" << endl;
faust_unsigned_int nrows = 1024, ncols = 1024; faust_unsigned_int nrows = 1024, ncols = 1024;
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData()); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData());
// cout << cpu_mat1->to_string(false, true) << endl; // cout << cpu_mat1->to_string(false, true) << endl;
cpu_mat1->conjugate(); cpu_mat1->conjugate();
gpu_mat1.conjugate(); gpu_mat1.conjugate();
...@@ -253,8 +256,8 @@ void test_norm() ...@@ -253,8 +256,8 @@ void test_norm()
cout << "test spectral norm" << endl; cout << "test spectral norm" << endl;
faust_unsigned_int nrows = 1024, ncols = 1024; faust_unsigned_int nrows = 1024, ncols = 1024;
int flag; int flag;
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData()); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData());
auto gpu_norm = gpu_mat1.spectralNorm(1000, (float)1e-6); auto gpu_norm = gpu_mat1.spectralNorm(1000, (float)1e-6);
auto cpu_norm = cpu_mat1->spectralNorm(1000, 1e-6, flag); auto cpu_norm = cpu_mat1->spectralNorm(1000, 1e-6, flag);
// cout << gpu_norm << endl; // cout << gpu_norm << endl;
...@@ -269,7 +272,7 @@ void test_norm() ...@@ -269,7 +272,7 @@ void test_norm()
cout << "test l1 norm" << endl; cout << "test l1 norm" << endl;
gpu_norm = gpu_mat1.normL1(); gpu_norm = gpu_mat1.normL1();
cpu_norm = cpu_mat1->normL1(); cpu_norm = cpu_mat1->normL1();
// cout << "gpu: " << gpu_norm << " cpu: " << cpu_norm << endl; cout << "gpu: " << gpu_norm << " cpu: " << cpu_norm << endl;
assert(abs(cpu_norm-gpu_norm) < 1e-6); assert(abs(cpu_norm-gpu_norm) < 1e-6);
cout << "OK" << endl; cout << "OK" << endl;
} }
...@@ -278,8 +281,8 @@ void test_normalize() ...@@ -278,8 +281,8 @@ void test_normalize()
{ {
cout << "test normalize" << endl; cout << "test normalize" << endl;
faust_unsigned_int nrows = 1024, ncols = 1024; faust_unsigned_int nrows = 1024, ncols = 1024;
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData()); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData());
cpu_mat1->normalize(); cpu_mat1->normalize();
gpu_mat1.normalize(); gpu_mat1.normalize();
auto gpu_mat1_to_cpu = gpu_mat1.tocpu(); auto gpu_mat1_to_cpu = gpu_mat1.tocpu();
...@@ -294,9 +297,9 @@ void test_scalar_mul() ...@@ -294,9 +297,9 @@ void test_scalar_mul()
{ {
cout << "test scalar mul" << endl; cout << "test scalar mul" << endl;
faust_unsigned_int nrows = 1024, ncols = 1024; faust_unsigned_int nrows = 1024, ncols = 1024;
auto lambda = 12.; auto lambda = FPP(12.);
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData()); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData());
cpu_mat1->scalarMultiply(lambda); cpu_mat1->scalarMultiply(lambda);
gpu_mat1.scalarMultiply(lambda); gpu_mat1.scalarMultiply(lambda);
auto gpu_mat1_to_cpu = gpu_mat1.tocpu(); auto gpu_mat1_to_cpu = gpu_mat1.tocpu();
...@@ -311,10 +314,10 @@ void test_abs() ...@@ -311,10 +314,10 @@ void test_abs()
cout << "test abs" << endl; cout << "test abs" << endl;
faust_unsigned_int nrows = 1024, ncols = 1024; faust_unsigned_int nrows = 1024, ncols = 1024;
auto lambda = 12.; auto lambda = 12.;
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
cpu_mat1->scalarMultiply(-1); cpu_mat1->scalarMultiply(FPP(-1));
cout << cpu_mat1->to_string(false, true) << endl; cout << cpu_mat1->to_string(false, true) << endl;
MatDense<double,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData()); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData());
cpu_mat1->abs(); cpu_mat1->abs();
gpu_mat1.abs(); gpu_mat1.abs();
auto gpu_mat1_to_cpu = gpu_mat1.tocpu(); auto gpu_mat1_to_cpu = gpu_mat1.tocpu();
...@@ -331,10 +334,10 @@ void test_add() ...@@ -331,10 +334,10 @@ void test_add()
cout << "test add" << endl; cout << "test add" << endl;
faust_unsigned_int nrows = 1024, ncols = 1024; faust_unsigned_int nrows = 1024, ncols = 1024;
auto lambda = 12.; auto lambda = 12.;
// test MatDense<double, GPU2> + MatDense<double, Cpu> // test MatDense<FPP, GPU2> + MatDense<FPP, Cpu>
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
auto cpu_mat2 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat2 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData()); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData());
cpu_mat1->add(*cpu_mat2); cpu_mat1->add(*cpu_mat2);
gpu_mat1.add(*cpu_mat2); gpu_mat1.add(*cpu_mat2);
auto gpu_mat1_to_cpu = gpu_mat1.tocpu(); auto gpu_mat1_to_cpu = gpu_mat1.tocpu();
...@@ -342,8 +345,8 @@ void test_add() ...@@ -342,8 +345,8 @@ void test_add()
err_diff -= gpu_mat1_to_cpu; err_diff -= gpu_mat1_to_cpu;
auto err = err_diff.norm()/cpu_mat1->norm(); auto err = err_diff.norm()/cpu_mat1->norm();
assert(err < 1e-6); assert(err < 1e-6);
// test MatDense<double, GPU2> + MatSparse<double, Cpu> // test MatDense<FPP, GPU2> + MatSparse<FPP, Cpu>
Faust::MatSparse<double, Cpu> *cpu_mat3_sp = Faust::MatSparse<double, Cpu>::randMat(nrows, ncols, .2); Faust::MatSparse<FPP, Cpu> *cpu_mat3_sp = Faust::MatSparse<FPP, Cpu>::randMat(nrows, ncols, .2);
// cpu_mat3_sp->Display(); // cpu_mat3_sp->Display();
// cout << cpu_mat3_sp->to_string(false, true) << endl; // cout << cpu_mat3_sp->to_string(false, true) << endl;
assert(calc_err(gpu_mat1, *cpu_mat1) < 1e-6); assert(calc_err(gpu_mat1, *cpu_mat1) < 1e-6);
...@@ -364,10 +367,10 @@ void test_sub() ...@@ -364,10 +367,10 @@ void test_sub()
cout << "test sub" << endl; cout << "test sub" << endl;
faust_unsigned_int nrows = 1024, ncols = 1024; faust_unsigned_int nrows = 1024, ncols = 1024;
auto lambda = 12.; auto lambda = 12.;
// test MatDense<double, GPU2> + MatDense<double, Cpu> // test MatDense<FPP, GPU2> + MatDense<FPP, Cpu>
auto cpu_mat1 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat1 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
auto cpu_mat2 = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat2 = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData()); MatDense<FPP,GPU2> gpu_mat1(nrows, ncols, cpu_mat1->getData());
cpu_mat1->sub(*cpu_mat2); cpu_mat1->sub(*cpu_mat2);
gpu_mat1.sub(*cpu_mat2); gpu_mat1.sub(*cpu_mat2);
auto gpu_mat1_to_cpu = gpu_mat1.tocpu(); auto gpu_mat1_to_cpu = gpu_mat1.tocpu();
...@@ -375,10 +378,10 @@ void test_sub() ...@@ -375,10 +378,10 @@ void test_sub()
err_diff -= gpu_mat1_to_cpu; err_diff -= gpu_mat1_to_cpu;
auto err = err_diff.norm()/cpu_mat1->norm(); auto err = err_diff.norm()/cpu_mat1->norm();
assert(err < 1e-6); assert(err < 1e-6);
// test MatDense<double, GPU2> + MatSparse<double, Cpu> // test MatDense<FPP, GPU2> + MatSparse<FPP, Cpu>
return; return;
//TODO: fix below test error //TODO: fix below test error
Faust::MatSparse<double, Cpu> *cpu_mat3_sp = Faust::MatSparse<double, Cpu>::randMat(nrows, ncols, .2); Faust::MatSparse<FPP, Cpu> *cpu_mat3_sp = Faust::MatSparse<FPP, Cpu>::randMat(nrows, ncols, .2);
// cpu_mat3_sp->Display(); // cpu_mat3_sp->Display();
// cout << cpu_mat3_sp->to_string(false, true) << endl; // cout << cpu_mat3_sp->to_string(false, true) << endl;
assert(calc_err(gpu_mat1, *cpu_mat1) < 1e-6); assert(calc_err(gpu_mat1, *cpu_mat1) < 1e-6);
...@@ -399,8 +402,8 @@ void test_get_device() ...@@ -399,8 +402,8 @@ void test_get_device()
cout << "test MatDense<FPP,GPU2>::getDevice" << endl; cout << "test MatDense<FPP,GPU2>::getDevice" << endl;
auto nrows = 5; auto nrows = 5;
auto ncols = 8; auto ncols = 8;
auto cpu_ds_mat = Faust::MatDense<double, Cpu>::randMat(nrows, ncols); auto cpu_ds_mat = Faust::MatDense<FPP, Cpu>::randMat(nrows, ncols);
Faust::MatDense<double, GPU2> gpu_ds_mat(nrows, ncols, cpu_ds_mat->getData()); Faust::MatDense<FPP, GPU2> gpu_ds_mat(nrows, ncols, cpu_ds_mat->getData());
auto dev_id = gpu_ds_mat.getDevice(); auto dev_id = gpu_ds_mat.getDevice();
cout << "device id: " << dev_id << endl; cout << "device id: " << dev_id << endl;
delete cpu_ds_mat; delete cpu_ds_mat;
...@@ -411,8 +414,8 @@ void test_move() ...@@ -411,8 +414,8 @@ void test_move()
cout << "test MatDense<FPP,GPU2>::move" << endl; cout << "test MatDense<FPP,GPU2>::move" << endl;
auto nrows = 5; auto nrows = 5;
auto ncols = 8; auto ncols = 8;
auto cpu_ds_mat = Faust::MatDense<double, Cpu>::randMat(nrows, ncols); auto cpu_ds_mat = Faust::MatDense<FPP, Cpu>::randMat(nrows, ncols);
Faust::MatDense<double, GPU2> gpu_ds_mat(nrows, ncols, cpu_ds_mat->getData()); Faust::MatDense<FPP, GPU2> gpu_ds_mat(nrows, ncols, cpu_ds_mat->getData());
auto ndevs = Faust::count_devices(); auto ndevs = Faust::count_devices();
cout << "device count: " << ndevs << endl; cout << "device count: " << ndevs << endl;
int32_t dest_dev = -1; int32_t dest_dev = -1;
...@@ -439,12 +442,12 @@ void test_mul_vec() ...@@ -439,12 +442,12 @@ void test_mul_vec()
auto size = 2048; auto size = 2048;
auto nrows = 1024; auto nrows = 1024;
auto ncols = size; auto ncols = size;
auto vec_mat = MatDense<double, Cpu>::randMat(size, 1); auto vec_mat = MatDense<FPP, Cpu>::randMat(size, 1);
Faust::Vect<double, Cpu> v_cpu(size, vec_mat->getData()); Faust::Vect<FPP, Cpu> v_cpu(size, vec_mat->getData());
Faust::Vect<double, GPU2> v_gpu(size, vec_mat->getData()); Faust::Vect<FPP, GPU2> v_gpu(size, vec_mat->getData());
Faust::Vect<double, GPU2> v_gpu_out(nrows); Faust::Vect<FPP, GPU2> v_gpu_out(nrows);
auto cpu_mat = Faust::MatDense<double,Cpu>::randMat(nrows,ncols); auto cpu_mat = Faust::MatDense<FPP,Cpu>::randMat(nrows,ncols);
MatDense<double,GPU2> gpu_mat(nrows, ncols, cpu_mat->getData()); MatDense<FPP,GPU2> gpu_mat(nrows, ncols, cpu_mat->getData());
gpu_mat.multiply(v_gpu, v_gpu_out); gpu_mat.multiply(v_gpu, v_gpu_out);
auto v_gpu_out_cpu = v_gpu_out.tocpu(); auto v_gpu_out_cpu = v_gpu_out.tocpu();
v_gpu_out_cpu.Display(); v_gpu_out_cpu.Display();
...@@ -459,9 +462,9 @@ void test_mul_vec() ...@@ -459,9 +462,9 @@ void test_mul_vec()
void test_gpu_matsparse2matdense() void test_gpu_matsparse2matdense()
{ {
cout << "test MatDense<FPP,GPU2>::MatDense<FPP,GPU2>(MatSparse<FPP,GPU2>&)" << endl; cout << "test MatDense<FPP,GPU2>::MatDense<FPP,GPU2>(MatSparse<FPP,GPU2>&)" << endl;
auto sM = Faust::MatSparse<double,Cpu>::randMat(22, 33, .2); auto sM = Faust::MatSparse<FPP,Cpu>::randMat(22, 33, .2);
Faust::MatSparse<double,GPU2> sM_gpu(*sM); Faust::MatSparse<FPP,GPU2> sM_gpu(*sM);
MatDense<double,GPU2> dM_gpu(sM_gpu); MatDense<FPP,GPU2> dM_gpu(sM_gpu);
cout << sM_gpu.norm() << " " << dM_gpu.norm() << endl; cout << sM_gpu.norm() << " " << dM_gpu.norm() << endl;
assert(abs(sM_gpu.norm()-dM_gpu.norm())< 1e-4); assert(abs(sM_gpu.norm()-dM_gpu.norm())< 1e-4);
cout << "OK" << endl; cout << "OK" << endl;
......
src/faust_linear_operator/GPU2/faust_MatDense_gpu.cpp.in
+ 163
− 168
View file @ fed57eb3
This diff is collapsed.
src/faust_linear_operator/GPU2/faust_MatDense_gpu.h
+ 1
− 1
View file @ fed57eb3
#ifndef __FAUST_MATDENSE_GPU2__ #ifndef __FAUST_MATDENSE_GPU2__
#define __FAUST_MATDENSE_GPU2__ #define __FAUST_MATDENSE_GPU2__
#ifdef USE_GPU_MOD #ifdef USE_GPU_MOD
#include <complex>
#include <cstdint> #include <cstdint>
#include "faust_MatDense.h" #include "faust_MatDense.h"
#include "faust_MatGeneric_gpu.h" #include "faust_MatGeneric_gpu.h"
#include "faust_Vect_gpu.h" #include "faust_Vect_gpu.h"
#include "faust_MatSparse_gpu.h" #include "faust_MatSparse_gpu.h"
#include "faust_gpu_mod_utils.h" #include "faust_gpu_mod_utils.h"
#include <cstdint>
namespace Faust namespace Faust
{ {
template <typename FPP> template <typename FPP>
......
src/faust_linear_operator/GPU2/faust_MatSparse_gpu.cpp.in
+ 5
− 5
View file @ fed57eb3
...@@ -46,7 +46,7 @@ namespace Faust ...@@ -46,7 +46,7 @@ namespace Faust
{ {
if(rowptr == nullptr || colinds == nullptr) if(rowptr == nullptr || colinds == nullptr)
throw std::runtime_error("All GPU buffers or none must be defined to instantiate a MatSparse<FPP,GPU2>."); throw std::runtime_error("All GPU buffers or none must be defined to instantiate a MatSparse<FPP,GPU2>.");
gpu_mat = spm_funcs->togpu_stream(nbRow, nbCol, nnz, rowptr, colinds, values, stream); gpu_mat = spm_funcs->togpu_stream(nbRow, nbCol, nnz, rowptr, colinds, (@GM_SCALAR@*) reinterpret_cast<const @GM_REINTERPRET_CAST_SCALAR@*>(values), stream);
} }
else if(!no_zero) else if(!no_zero)
{ // no values given, set the matrix as a zero matrix { // no values given, set the matrix as a zero matrix
...@@ -134,7 +134,7 @@ namespace Faust ...@@ -134,7 +134,7 @@ namespace Faust
{ {
auto spm_funcs = GPUModHandler::get_singleton()->spm_funcs((@FAUST_SCALAR_FOR_GM@)(0)); auto spm_funcs = GPUModHandler::get_singleton()->spm_funcs((@FAUST_SCALAR_FOR_GM@)(0));
if(gpu_mat != nullptr) if(gpu_mat != nullptr)
spm_funcs->mul_scalar(gpu_mat, reinterpret_cast<const @GM_REINTERPRET_CAST_SCALAR@*>(&scalar)); spm_funcs->mul_scalar(gpu_mat, (@GM_SCALAR@*) reinterpret_cast<const @GM_REINTERPRET_CAST_SCALAR@*>(&scalar));
} }
template<> template<>
...@@ -147,14 +147,14 @@ namespace Faust ...@@ -147,14 +147,14 @@ namespace Faust
void MatSparse<@FAUST_SCALAR_FOR_GM@,GPU2>::operator+=(const @FAUST_SCALAR_FOR_GM@& scalar) void MatSparse<@FAUST_SCALAR_FOR_GM@,GPU2>::operator+=(const @FAUST_SCALAR_FOR_GM@& scalar)
{ {
auto spm_funcs = GPUModHandler::get_singleton()->spm_funcs((@FAUST_SCALAR_FOR_GM@)(0)); auto spm_funcs = GPUModHandler::get_singleton()->spm_funcs((@FAUST_SCALAR_FOR_GM@)(0));
spm_funcs->add_scalar(gpu_mat, reinterpret_cast<const @GM_REINTERPRET_CAST_SCALAR@*>(&scalar)); spm_funcs->add_scalar(gpu_mat, (@GM_SCALAR@*) reinterpret_cast<const @GM_REINTERPRET_CAST_SCALAR@*>(&scalar));
} }
template<> template<>
void MatSparse<@FAUST_SCALAR_FOR_GM@,GPU2>::operator-=(const @FAUST_SCALAR_FOR_GM@& scalar) void MatSparse<@FAUST_SCALAR_FOR_GM@,GPU2>::operator-=(const @FAUST_SCALAR_FOR_GM@& scalar)
{ {
auto spm_funcs = GPUModHandler::get_singleton()->spm_funcs((@FAUST_SCALAR_FOR_GM@)(0)); auto spm_funcs = GPUModHandler::get_singleton()->spm_funcs((@FAUST_SCALAR_FOR_GM@)(0));
spm_funcs->sub_scalar(gpu_mat, reinterpret_cast<const @GM_REINTERPRET_CAST_SCALAR@*>(&scalar)); spm_funcs->sub_scalar(gpu_mat, (@GM_SCALAR@*) reinterpret_cast<const @GM_REINTERPRET_CAST_SCALAR@*>(&scalar));
} }
template<> template<>
...@@ -178,7 +178,7 @@ namespace Faust ...@@ -178,7 +178,7 @@ namespace Faust
*ncols = getNbCol(); *ncols = getNbCol();
if(nnz != nullptr) if(nnz != nullptr)
*nnz = getNonZeros(); *nnz = getNonZeros();
spm_funcs->tocpu(gpu_mat, row_ptr, col_ind, value_ptr); spm_funcs->tocpu(gpu_mat, row_ptr, col_ind, (@GM_SCALAR@*) reinterpret_cast<const @GM_REINTERPRET_CAST_SCALAR@*>(value_ptr));
} }
template<> template<>
......
src/faust_linear_operator/GPU2/faust_Vect_gpu.cpp.in
+ 14
− 10
View file @ fed57eb3
...@@ -47,7 +47,7 @@ namespace Faust ...@@ -47,7 +47,7 @@ namespace Faust
if(size() == v.size()) if(size() == v.size())
{ {
auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0)); auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0));
auto gpu_mat = dsm_funcs->togpu(this->getNbRow(), 1, const_cast<@FAUST_SCALAR_FOR_GM@*>(v.getData())); auto gpu_mat = dsm_funcs->togpu(this->getNbRow(), 1, (@GM_SCALAR@*) reinterpret_cast<const @GM_REINTERPRET_CAST_SCALAR@*>(v.getData()));
dsm_funcs->free(this->gpu_mat); dsm_funcs->free(this->gpu_mat);
this->gpu_mat = gpu_mat; this->gpu_mat = gpu_mat;
} }
...@@ -60,7 +60,7 @@ namespace Faust ...@@ -60,7 +60,7 @@ namespace Faust
{ {
auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0)); auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0));
FSFG _max_coeff; FSFG _max_coeff;
dsm_funcs->max(this->gpu_mat, &_max_coeff); dsm_funcs->max(this->gpu_mat, (@GM_SCALAR@*) reinterpret_cast<const @GM_REINTERPRET_CAST_SCALAR@*>(&_max_coeff));
return _max_coeff; return _max_coeff;
} }
...@@ -69,7 +69,7 @@ namespace Faust ...@@ -69,7 +69,7 @@ namespace Faust
{ {
auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0)); auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0));
FSFG _min_coeff; FSFG _min_coeff;
dsm_funcs->min(this->gpu_mat, &_min_coeff); dsm_funcs->min(this->gpu_mat, (@GM_SCALAR@*) reinterpret_cast<const @GM_REINTERPRET_CAST_SCALAR@*>(&_min_coeff));
return _min_coeff; return _min_coeff;
} }
...@@ -78,7 +78,7 @@ namespace Faust ...@@ -78,7 +78,7 @@ namespace Faust
{ {
FSFG alpha; FSFG alpha;
auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0)); auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0));
dsm_funcs->dot(this->gpu_mat, v.gpu_mat, &alpha); dsm_funcs->dot(this->gpu_mat, v.gpu_mat, (@GM_SCALAR@*) reinterpret_cast<const @GM_REINTERPRET_CAST_SCALAR@*>(&alpha));
return alpha; return alpha;
} }
...@@ -87,14 +87,18 @@ namespace Faust ...@@ -87,14 +87,18 @@ namespace Faust
{ {
FSFG s; FSFG s;
auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0)); auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0));
dsm_funcs->sum(this->gpu_mat, reinterpret_cast<@GM_REINTERPRET_CAST_SCALAR@*>(&s)); dsm_funcs->sum(this->gpu_mat, (@GM_SCALAR@*) reinterpret_cast<@GM_REINTERPRET_CAST_SCALAR@*>(&s));
return s; return s;
} }
template<> template<>
FSFG Vect<FSFG,GPU2>::mean() const FSFG Vect<FSFG,GPU2>::mean() const
{ {
return sum()/size(); // return sum()/size();
FSFG m;
auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0));
dsm_funcs->mean(this->gpu_mat, (@GM_SCALAR@*) reinterpret_cast<@GM_REINTERPRET_CAST_SCALAR@*>(&m));
return m;
} }
template<> template<>
...@@ -147,7 +151,7 @@ namespace Faust ...@@ -147,7 +151,7 @@ namespace Faust
void Vect<FSFG,GPU2>::setValues(const FSFG& val) void Vect<FSFG,GPU2>::setValues(const FSFG& val)
{ {
auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0)); auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0));
dsm_funcs->setval(this->gpu_mat, reinterpret_cast<@GM_REINTERPRET_CAST_SCALAR@*>(const_cast<FSFG*>(&val))); dsm_funcs->setval(this->gpu_mat, (@GM_SCALAR@*) reinterpret_cast<@GM_REINTERPRET_CAST_SCALAR@*>(const_cast<FSFG*>(&val)));
} }
template<> template<>
...@@ -156,7 +160,7 @@ namespace Faust ...@@ -156,7 +160,7 @@ namespace Faust
FSFG e; FSFG e;
auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0)); auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0));
dsm_funcs->mean_relerr(this->get_gpu_mat_ptr(), ref_vec.get_gpu_mat_ptr(), reinterpret_cast<@GM_REINTERPRET_CAST_SCALAR@*>(&e)); dsm_funcs->mean_relerr(this->get_gpu_mat_ptr(), ref_vec.get_gpu_mat_ptr(), (@GM_SCALAR@*) reinterpret_cast<@GM_REINTERPRET_CAST_SCALAR@*>(&e));
return e; return e;
} }
...@@ -165,7 +169,7 @@ namespace Faust ...@@ -165,7 +169,7 @@ namespace Faust
{ {
FSFG coeff; FSFG coeff;
auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0)); auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0));
dsm_funcs->coeff(this->gpu_mat, i, 0, reinterpret_cast<@GM_REINTERPRET_CAST_SCALAR@*>(&coeff)); dsm_funcs->coeff(this->gpu_mat, i, 0, (@GM_SCALAR@*) reinterpret_cast<@GM_REINTERPRET_CAST_SCALAR@*>(&coeff));
return coeff; return coeff;
} }
...@@ -174,6 +178,6 @@ namespace Faust ...@@ -174,6 +178,6 @@ namespace Faust
void Vect<FSFG,GPU2>::set_coeff(faust_unsigned_int i, const FSFG& val) void Vect<FSFG,GPU2>::set_coeff(faust_unsigned_int i, const FSFG& val)
{ {
auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0)); auto dsm_funcs = GPUModHandler::get_singleton()->dsm_funcs(@FAUST_SCALAR_FOR_GM@(0));
dsm_funcs->set_coeff(this->gpu_mat, i, 0, reinterpret_cast<const @GM_REINTERPRET_CAST_SCALAR@*>(&val)); dsm_funcs->set_coeff(this->gpu_mat, i, 0, (@GM_SCALAR@*) reinterpret_cast<const @GM_REINTERPRET_CAST_SCALAR@*>(&val));
} }
} }
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Please register or to comment