replica
/
Coyote
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

sw example pr.

This commit is contained in:
d-kor 2022-09-25 00:57:52 +02:00
parent 08479c64a1
commit 0c04e958a8
2 changed files with 71 additions and 62 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
.gitignore vendored
View File

@ -7,6 +7,7 @@ ip_dir/
xsim.dir/ xsim.dir/
tb_user/ tb_user/
apps/ apps/
rust/
.Xil/ .Xil/
csim/ csim/
.vscode/ .vscode/

132
sw/examples/pr_scheduling/main.cpp
View File

@ -14,11 +14,9 @@ using namespace std;
using namespace std::chrono; using namespace std::chrono;
constexpr auto const nRegions = 2; constexpr auto const nRegions = 2;
constexpr auto const nReps = 1; constexpr auto const defSize = 4 * 1024;
constexpr auto const defSize = 512; constexpr auto const nOper = 4;
constexpr auto const nOper = 3;
constexpr auto const nTasks = 10; constexpr auto const nTasks = 10;
constexpr auto const rsltLen = 4;
constexpr auto const randomOrder = true; constexpr auto const randomOrder = true;
@ -37,19 +35,18 @@ auto select_random(const S &s, size_t n) {
// Add + multiply // Add + multiply
constexpr auto const defAddmul = 2; constexpr auto const defAddmul = 2;
auto addmul = [](cThread *cthread, uint32_t size, uint32_t add, uint32_t mul) { auto addmul = [](cThread *cthread, uint32_t size, uint32_t add, uint32_t mul) {
// Allocate some memory // Allocate some memory
uint32_t *hMem = malloc(size); uint32_t *hMem = (uint32_t*) malloc(size);
uint32_t *rMem = malloc(size); uint32_t *rMem = (uint32_t*) malloc(size);
// Fill // Fill
for(int i = 0; i < size/4; i++) for(int i = 0; i < size/4; i++)
hMem[i] = defAddmul; hMem[i] = defAddmul;
// Prep // Prep
cthread->setCSR(add, 0); cthread->setCSR(mul, 0); // Addition
cthread->setCSR(mul, 1); cthread->setCSR(add, 1); // Multiplication
// Invoke // Invoke
cthread->invoke({CoyoteOper::TRANSFER, (void*)hMem, (void*)rMem, size, size}); cthread->invoke({CoyoteOper::TRANSFER, (void*)hMem, (void*)rMem, size, size});
@ -58,41 +55,53 @@ auto addmul = [](cThread *cthread, uint32_t size, uint32_t add, uint32_t mul) {
bool k = true; bool k = true;
for(int i = 0; i < size/4; i++) for(int i = 0; i < size/4; i++)
if(rMem[i] != defAddmul * mul + add) k = false; if(rMem[i] != defAddmul * mul + add) k = false;
if (!k) std::cout << "ADDMUL fail!" << std::endl; if (!k) std::cout << "ERR: Addmul failed!" << std::endl;
// Free
free((void*) hMem);
free((void*) rMem);
}; };
// Stream statistics // Stream statistics
auto minmaxsum = [](cThread *cthread, uint32_t *hmem, uint32_t *rmem, uint32_t size) { constexpr auto const defMin = 10;
constexpr auto const defMax = 20;
auto minmaxsum = [](cThread *cthread, uint32_t size) {
// Allocate some memory // Allocate some memory
uint32_t *hMem = malloc(size); uint32_t *hMem = (uint32_t*) malloc(size);
// Fill
uint32_t sum = 0;
for(int i = 0; i < size/4; i++) {
hMem[i] = i%2 ? defMin : defMax;
sum += hMem[i];
}
// Prep // Prep
cthread->setCSR(0x1, 0); cthread->setCSR(0x1, 0); // Start kernel
// Invoke // Invoke
cthread->invoke({CoyoteOper::READ, (void*)hmem, size, true, false}); cthread->invoke({CoyoteOper::READ, (void*)hMem, size, true, false});
while(cthread->getCSR(1) != 0x1) { nanosleep((const struct timespec[]){{0, 100L}}, NULL); } while(cthread->getCSR(1) != 0x1) { nanosleep((const struct timespec[]){{0, 100L}}, NULL); } // Poll for completion
// Stats // Check results
std::cout << "Min: " << cthread->getCSR(2) << std::endl; if((cthread->getCSR(2) != defMin) || (cthread->getCSR(2) != defMax) || (sum != cthread->getCSR(4)))
std::cout << "Max: " << cthread->getCSR(3) << std::endl; std::cout << "ERR: MinMaxSum failed!" << std::endl;
std::cout << "Sum: " << cthread->getCSR(4) << std::endl;
// Free
free((void*) hMem);
}; };
// Rotate // Rotation
constexpr auto const defRot = 2; constexpr auto const defRot = 0xefbeadde;
constexpr auto const expRot = 0xdeadbeef;
auto rotate = [](cThread *cthread, uint32_t *hmem, uint32_t *rmem, uint32_t size, uint32_t rot) { auto rotation = [](cThread *cthread, uint32_t size) {
// Allocate some memory // Allocate some memory
uint32_t *hMem = malloc(size); uint32_t *hMem = (uint32_t*) malloc(size);
uint32_t *rMem = malloc(size); uint32_t *rMem = (uint32_t*) malloc(size);
// Fill // Fill
for(int i = 0; i < size/4; i++) for(int i = 0; i < size/4; i++)
hMem[i] = defRot; hMem[i] = defRot;
// Prep
cthread->setCSR(rot, 0);
// Invoke // Invoke
cthread->invoke({CoyoteOper::TRANSFER, (void*)hMem, (void*)rMem, size, size}); cthread->invoke({CoyoteOper::TRANSFER, (void*)hMem, (void*)rMem, size, size});
@ -100,26 +109,38 @@ auto rotate = [](cThread *cthread, uint32_t *hmem, uint32_t *rmem, uint32_t size
// Check results // Check results
bool k = true; bool k = true;
for(int i = 0; i < size/4; i++) for(int i = 0; i < size/4; i++)
if(rMem[i] != defRot << rot) k = false; if(rMem[i] != expRot) k = false;
if (!k) std::cout << "ROTATE fail!" << std::endl; std::cout << "ERR: Rotate failed!" << std::endl;
// Free
free((void*) hMem);
free((void*) rMem);
}; };
// Testcount // Testcount
auto testcount = [](cThread *cthread, uint32_t *hmem, uint32_t *rmem, uint32_t size, uint32_t type, uint32_t cond) { auto testcount = [](cThread *cthread, uint32_t size, uint32_t type, uint32_t cond) {
// Allocate some memory // Allocate some memory
uint32_t *hMem = malloc(size); uint32_t *hMem = (uint32_t*) malloc(size);
// Fill
for(int i = 0; i < size/4; i++)
hMem[i] = i;
// Prep // Prep
cthread->setCSR(type, 2); cthread->setCSR(type, 2); // Type of comparison
cthread->setCSR(cond, 3); cthread->setCSR(cond, 3); // Predicate
cthread->setCSR(0x1, 0); cthread->setCSR(0x1, 0); // Start kernel
// Invoke // Invoke
cthread->invoke({CoyoteOper::TRANSFER, (void*)hmem, (void*)rmem, size, size, true, false}); cthread->invoke({CoyoteOper::READ, (void*)hMem, size, true, false});
while(cthread->getCSR(1) != 0x1) { nanosleep((const struct timespec[]){{0, 100L}}, NULL); } while(cthread->getCSR(1) != 0x1) { nanosleep((const struct timespec[]){{0, 100L}}, NULL); }
// Stats // Stats
rmem[0] = cthread->getCSR(4); // result if(cthread->getCSR(4) != size/4)
std::cout << "ERR: Testcount failed!" << std::endl;
// Free
free((void*) hMem);
}; };
@ -150,20 +171,20 @@ int main()
for(int j = 0; j < nTasks; j++) { for(int j = 0; j < nTasks; j++) {
switch (i) switch (i)
{ {
case 0: // addmul case 0: // Addmul
tasks[i].emplace_back(new cTask(i*nTasks + j, i, 1, addmul, 2, 3)); tasks[i].emplace_back(new cTask(i*nTasks + j, i, 1, addmul, defSize, 2, 3)); // TId, Oid, Priority, F, Args...
break; break;
case 1: // Minmaxsum
tasks[i].emplace_back(new cTask(i*nTasks + j, i, 1, minmaxsum, defSize)); // TId, Oid, Priority, F, Args...
break;
case 2: // Rotate
tasks[i].emplace_back(new cTask(i*nTasks + j, i, 1, rotation, defSize)); // TId, Oid, Priority, F, Args...
break;
case 3: // Testcount
tasks[i].emplace_back(new cTask(i*nTasks + j, i, 1, testcount, defSize, 0, 0)); // TId, Oid, Priority, F, Args...
default: default:
break; break;
} }
auto f = [=](cProc* cproc) {
std::cout << "This is an operator: " << i << ", task id: " << i*nTasks + j << std::endl;
sleep(1);
};
tasks[i].emplace_back(new cTask(i*nTasks + j, i, 1, f));
} }
} }
DBG1("All tasks created"); DBG1("All tasks created");
@ -204,20 +225,7 @@ int main()
high_resolution_clock::time_point begin = high_resolution_clock::now(); high_resolution_clock::time_point begin = high_resolution_clock::now();
while(carbiter.getCompletedCnt() != nOper * nTasks) nanosleep(&MSPAUSE, NULL); while(carbiter.getCompletedCnt() != nOper * nTasks) nanosleep(&MSPAUSE, NULL);
cout << "Arbiter tasks completed " << carbiter.getCompletedCnt() << endl;
int tcmpld0 = carbiter.getCompletedCnt();
cout << "COMPLETED COUNT: " << tcmpld0 << endl;
int32_t tid;
do {
tid = carbiter.getCompletedNext(0);
cout << "TID 0: " << tid << endl;
} while(tid != -1);
do {
tid = carbiter.getCompletedNext(1);
cout << "TID 1: " << tid << endl;
} while(tid != -1);
high_resolution_clock::time_point end = high_resolution_clock::now(); high_resolution_clock::time_point end = high_resolution_clock::now();
auto duration = duration_cast<microseconds>(end - begin).count(); auto duration = duration_cast<microseconds>(end - begin).count();