Klagarge/MSE-GPU-Cuda
1
0
Fork 0
Code Issues Pull Requests Releases Activity

feat(warmup): add solutions

Browse Source
This commit is contained in:
Rémi Heredero
2025-10-18 16:37:06 +02:00
parent f8ae49d666
commit f800b4f395
8 changed files with 91 additions and 60 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
Student_Cuda_Warmup/src/core/01_addArray/00_procedurale/addArrayProcedurale.cu
View File

@@ -9,6 +9,7 @@
#include "cudas.h"
#include "GM.h"
#include "Kernel.h"
#include "Hardware.h"
using std::cout;
using std::endl;
@@ -32,9 +33,34 @@ static __global__ void kaddArray(float* ptrGMV1 , float* ptrGMV2 , float* ptrGMW
* ptrW receptionne le resultat
* n nombre de case
*/
__host__ bool addArray_procedurale(float* ptrV1 , float* ptrV2 , float* ptrW , int n) // __host__ facultatif
__host__ void addArray_procedurale(float* ptrV1 , float* ptrV2 , float* ptrW , int n) // __host__ facultatif
{
// TODO addArray
size_t sizeVector = sizeof(float) * n;
float* ptrGMV1;
float* ptrGMV2;
float* ptrGMW;
GM::malloc(&ptrGMV1, sizeVector);
GM::malloc(&ptrGMV2, sizeVector);
GM::malloc(&ptrGMW, sizeVector);
GM::memcpyHToD(ptrGMV1, ptrV1, sizeVector);
GM::memcpyHToD(ptrGMV2, ptrV2, sizeVector);
const int MP = Hardware::getMPCount();
const int CORE_MP = Hardware::getCoreCountMP();
dim3 dg(MP, 3, 1);
dim3 db(CORE_MP, 2, 1); // <=1024
kaddArray<<<dg,db>>>( ptrGMV1 , ptrGMV2 , ptrGMW , n);
GM::memcpyDToH(ptrW, ptrGMW, sizeVector);
GM::free(ptrGMV1);
GM::free(ptrGMV2);
GM::free(ptrGMW);
}
/*--------------------------------------*\
@@ -50,8 +76,12 @@ __global__ void kaddArray(float* ptrGMV1 , float* ptrGMV2 , float* ptrGMW , int
const int TID = Thread2D::tid();
// pattern entrelacement
// TODO addArray
int s = TID;
while (s < n)
{
ptrGMW[s] = ptrGMV1[s] + ptrGMV2[s];
s += NB_THREAD;
}
}
/*----------------------------------------------------------------------*\

5
Student_Cuda_Warmup/src/core/01_addArray/00_procedurale/useAddArrayProcedurale.cpp
View File

@@ -37,16 +37,17 @@ bool exemple_addArray_procedurale()
ArrayTools::print(ptrV2, n);
cout << "--------------------------------------------------------------------" << endl;
ArrayTools::print(ptrW, n);
cout<<endl;
}
// Delete
{
delete ptrV1;
delete ptrV1;
delete ptrV2;
delete ptrW;
}
return isOk;
return true;
}
/*----------------------------------------------------------------------*\

19
Student_Cuda_Warmup/src/core/01_addArray/01_poo/01_pattern_entrelacement/device/addArray_device.cu
View File

@@ -11,23 +11,18 @@
/**
* output : void required, because kernel is asynchrone
*/
__global__ void addArray(float* ptrGMV1 , float* ptrGMV2 , float* ptrGMW , int n)
{
__global__ void addArray(float* ptrGMV1 , float* ptrGMV2 , float* ptrGMW , int n) {
const int NB_THREAD = Thread2D::nbThread();
const int TID = Thread2D::tid();
// Debug, facultatif
// if (TID == 0)
// {
// printf("Coucou from device tid = %d", TID);
// }
// pattern entrelacement
// TODO addArray
}
int s = TID;
while (s < n) {
ptrGMW[s] = ptrGMV1[s] + ptrGMV2[s];
s += NB_THREAD;
}
}
/*----------------------------------------------------------------------*\
|* End *|
\*---------------------------------------------------------------------*/

43
Student_Cuda_Warmup/src/core/01_addArray/01_poo/01_pattern_entrelacement/host/AddArray.cu
View File

@@ -12,7 +12,6 @@ using std::endl;
using std::to_string;
using std::string;
/*--------------------------------------*\
|* Imported *|
\*-------------------------------------*/
@@ -33,26 +32,24 @@ AddArray::AddArray(const Grid& grid , float* ptrV1 , float* ptrV2 , float* ptrW
ptrW(ptrW), //
n(n), //
dg(grid.dg), //
db(grid.db)
{
this->sizeVector = -1; // TODO addArray // octet
db(grid.db) {
this->sizeVector = sizeof(float) * n;
// MM (malloc Device)
{
GM::malloc(&ptrGMV1, sizeVector);
// TODO addArray
}
// MM (malloc Device)
{
GM::malloc(&ptrGMV1, sizeVector);
GM::malloc(&ptrGMV2, sizeVector);
GM::malloc(&ptrGMW, sizeVector);
}
}
AddArray::~AddArray()
{
AddArray::~AddArray() {
//MM (device free)
{
GM::free(ptrGMV1);
// TODO addArray
}
}
GM::free(ptrGMV2);
GM::free(ptrGMW);
}
/*--------------------------------------*\
|* Methode *|
@@ -61,23 +58,23 @@ AddArray::~AddArray()
/**
* override
*/
void AddArray::run()
{
void AddArray::run() {
// MM (copy Host->Device)
{
GM::memcpyHToD(ptrGMV1, ptrV1, sizeVector);
// TODO addArray
{
GM::memcpyHToD(ptrGMV1, ptrV1, sizeVector);
GM::memcpyHToD(ptrGMV2, ptrV2, sizeVector);
}
// TODO addArray // call kernel // assynchrone
addArray<<<dg,db>>>( ptrGMV1 , ptrGMV2 , ptrGMW , n);
//Kernel::synchronize();// inutile
// MM (Device -> Host)
{
// TODO addArray // MM barier de synchronisation implicite
GM::memcpyDToH(ptrW, ptrGMW, sizeVector);
}
}
}
/*----------------------------------------------------------------------*\
|* End *|

3
Student_Cuda_Warmup/src/core/01_addArray/01_poo/01_pattern_entrelacement/host/AddArray.h
View File

@@ -9,8 +9,7 @@
\*---------------------------------------------------------------------*/
class AddArray
{
class AddArray {
/*--------------------------------------*\
|* Constructor *|
\*-------------------------------------*/

13
Student_Cuda_Warmup/src/core/01_addArray/01_poo/01_pattern_entrelacement/useAddArrayObject.cpp
View File

@@ -45,12 +45,13 @@ bool exemple_addArray_object()
ArrayTools::print(ptrV2, n);
cout << "--------------------------------------------------------------------" << endl;
ArrayTools::print(ptrW, n);
cout<<endl;
}
// Delete
{
delete ptrV1;
delete ptrV1;
delete ptrV2;
delete ptrW;
}
@@ -69,16 +70,10 @@ static Grid createGrid()
const int MP = Hardware::getMPCount();
const int CORE_MP = Hardware::getCoreCountMP();
dim3 dg(1, 1, 1); // TODO addArray
dim3 db(1, 1, 1); // TODO addArray // produit <=1024
dim3 dg(MP, 2, 1);
dim3 db(CORE_MP, 3, 1); // produit <=1024
Grid grid(dg, db);
// to remove once coded
{
Couts::redln("aie aie aie, your best grid won t build itself");
assert(false);
}
return grid;
}

3
Student_Cuda_Warmup/src/core/01_addArray/01_poo/02_pattern_11/useAddArray11.cpp
View File

@@ -45,12 +45,13 @@ bool exemple_addArray_11()
ArrayTools::print(ptrV2, n);
cout << "--------------------------------------------------------------------" << endl;
ArrayTools::print(ptrW, n);
cout<<endl;
}
// Delete
{
delete ptrV1;
delete ptrV1;
delete ptrV2;
delete ptrW;
}

27
Student_Cuda_Warmup/src/main/main.cpp
View File

@@ -2,6 +2,7 @@
#include <stdlib.h>
#include "Couts.h"
#include "Hardware.h"
using std::cerr;
using std::cout;
@@ -29,17 +30,29 @@ static void scalar(bool& isOk);
int main(int argc , char** argv)
{
// Facultatif
{
Hardware::loadCudaDriverAll();
Hardware::printAllSimple();
const int DEVICE_ID = 0;
Hardware::setDevice(DEVICE_ID);
}
const int IS_VERBOSE = true;
bool isOk = true;
// run
{
bool isOk = true;
// Commenter ce dont vous n'avez pas besoin ci-dessous
scalar(isOk); // commenter dans la methode ci-dessous ce que vous ne voulez pas lancer
array(isOk); // commenter dans la methode ci-dessous ce que vous ne voulez pas lancer
// Commenter ce dont vous n'avez pas besoin ci-dessous
//scalar(isOk); // commenter dans la methode ci-dessous ce que vous ne voulez pas lancer
array(isOk); // commenter dans la methode ci-dessous ce que vous ne voulez pas lancer
Couts::statusln(isOk);
Couts::statusln(isOk);
return isOk ? EXIT_SUCCESS : EXIT_FAILURE;
return isOk ? EXIT_SUCCESS : EXIT_FAILURE;
}
}
/*-------------------------------------*\
@@ -56,7 +69,7 @@ static void array(bool& isOk)
{
isOk &= exemple_addArray_procedurale();
isOk &= exemple_addArray_object();
isOk &= exemple_addArray_11();
// isOk &= exemple_addArray_11();
}
/*----------------------------------------------------------------------*\