v4.5 dev update. (#3153)

This commit is contained in:
Junkai-Wu
2026-04-07 12:16:05 -04:00
committed by GitHub
parent 418d38a5de
commit a221da7ccf
265 changed files with 4913 additions and 1478 deletions
+5 -7
View File
@@ -2827,13 +2827,11 @@ struct TestbedImpl {
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}
@@ -2022,14 +2022,12 @@ struct TestbedImpl {
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}
+5 -5
View File
@@ -128,11 +128,11 @@ struct MultistageTestbed {
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
return false;
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}
+5 -5
View File
@@ -314,11 +314,11 @@ struct Testbed {
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
return false;
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}
+5 -5
View File
@@ -130,11 +130,11 @@ struct TestbedComplex : public Testbed<Gemm> {
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
return false;
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}
@@ -403,11 +403,11 @@ struct TestbedGemmWithBroadcast {
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
return false;
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}
@@ -377,11 +377,11 @@ struct TestbedGemmWithReduction {
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
return false;
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}
+5 -5
View File
@@ -131,11 +131,11 @@ struct InterleavedTestbed {
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
return false;
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}
@@ -139,11 +139,11 @@ public:
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
return false;
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}
@@ -300,10 +300,11 @@ struct TestbedRank2KUniversal {
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
return false;
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}
@@ -287,11 +287,11 @@ struct TestbedRank2KUniversal {
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
return false;
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}
+5 -5
View File
@@ -323,11 +323,11 @@ struct SparseTestbed {
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
return false;
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}
+5 -5
View File
@@ -87,11 +87,11 @@ struct TestbedSplitK : public Testbed<Gemm> {
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
return false;
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}
@@ -326,11 +326,11 @@ struct TestbedSymmUniversal {
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
return false;
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}
@@ -363,11 +363,11 @@ struct TestbedTrmmUniversal {
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
return false;
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}
+5 -5
View File
@@ -298,11 +298,11 @@ struct TestbedUniversal {
if (result != cudaSuccess) {
throw std::runtime_error("cudaGetDeviceProperties() failed");
}
if (properties.sharedMemPerBlockOptin < smem_size) {
return false;
}
if (properties.sharedMemPerBlockOptin < smem_size) {
printf("failed due to smem_size\n");
printf("hardware smem_size: %d, required smem_size: %d\n\n", int(properties.sharedMemPerBlockOptin), int(smem_size));
return false;
}
return true;
}