diff options
Diffstat (limited to 'test')
49 files changed, 923 insertions, 42 deletions
diff --git a/test/performance/routines/level1/xaxpy.cc b/test/performance/routines/level1/xaxpy.cc index 7ab15f28..b423bc3a 100644 --- a/test/performance/routines/level1/xaxpy.cc +++ b/test/performance/routines/level1/xaxpy.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXaxpy<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level1/xcopy.cc b/test/performance/routines/level1/xcopy.cc index 6277e8fb..c04c6c1c 100644 --- a/test/performance/routines/level1/xcopy.cc +++ b/test/performance/routines/level1/xcopy.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXcopy<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level1/xdot.cc b/test/performance/routines/level1/xdot.cc index 5aa76762..f4616464 100644 --- a/test/performance/routines/level1/xdot.cc +++ b/test/performance/routines/level1/xdot.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXdot<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level1/xdotc.cc b/test/performance/routines/level1/xdotc.cc index 81511085..5f36b80e 100644 --- a/test/performance/routines/level1/xdotc.cc +++ b/test/performance/routines/level1/xdotc.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kComplexSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kDouble: throw std::runtime_error("Unsupported precision mode"); diff --git a/test/performance/routines/level1/xdotu.cc b/test/performance/routines/level1/xdotu.cc index 888eede3..f19f751b 100644 --- a/test/performance/routines/level1/xdotu.cc +++ b/test/performance/routines/level1/xdotu.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kComplexSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kDouble: throw std::runtime_error("Unsupported precision mode"); diff --git a/test/performance/routines/level1/xscal.cc b/test/performance/routines/level1/xscal.cc index be49c066..bd38f43e 100644 --- a/test/performance/routines/level1/xscal.cc +++ b/test/performance/routines/level1/xscal.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXscal<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level1/xswap.cc b/test/performance/routines/level1/xswap.cc index 52fdc580..112641d3 100644 --- a/test/performance/routines/level1/xswap.cc +++ b/test/performance/routines/level1/xswap.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXswap<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xgbmv.cc b/test/performance/routines/level2/xgbmv.cc index 629e2182..b050184d 100644 --- a/test/performance/routines/level2/xgbmv.cc +++ b/test/performance/routines/level2/xgbmv.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXgbmv<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xgemv.cc b/test/performance/routines/level2/xgemv.cc index 2a1983de..51ab9a10 100644 --- a/test/performance/routines/level2/xgemv.cc +++ b/test/performance/routines/level2/xgemv.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXgemv<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xger.cc b/test/performance/routines/level2/xger.cc index 5fb0d91d..2d956346 100644 --- a/test/performance/routines/level2/xger.cc +++ b/test/performance/routines/level2/xger.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXger<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xgerc.cc b/test/performance/routines/level2/xgerc.cc index fd511e42..acd0fab7 100644 --- a/test/performance/routines/level2/xgerc.cc +++ b/test/performance/routines/level2/xgerc.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kComplexSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kDouble: throw std::runtime_error("Unsupported precision mode"); diff --git a/test/performance/routines/level2/xgeru.cc b/test/performance/routines/level2/xgeru.cc index 689ab2b1..a5973777 100644 --- a/test/performance/routines/level2/xgeru.cc +++ b/test/performance/routines/level2/xgeru.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kComplexSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kDouble: throw std::runtime_error("Unsupported precision mode"); diff --git a/test/performance/routines/level2/xhbmv.cc b/test/performance/routines/level2/xhbmv.cc index dabe6ec8..28b71045 100644 --- a/test/performance/routines/level2/xhbmv.cc +++ b/test/performance/routines/level2/xhbmv.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kComplexSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kDouble: throw std::runtime_error("Unsupported precision mode"); diff --git a/test/performance/routines/level2/xhemv.cc b/test/performance/routines/level2/xhemv.cc index 77447d76..622854a7 100644 --- a/test/performance/routines/level2/xhemv.cc +++ b/test/performance/routines/level2/xhemv.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kComplexSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kDouble: throw std::runtime_error("Unsupported precision mode"); diff --git a/test/performance/routines/level2/xher.cc b/test/performance/routines/level2/xher.cc index 4ef87e45..613d7766 100644 --- a/test/performance/routines/level2/xher.cc +++ b/test/performance/routines/level2/xher.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kComplexSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kDouble: throw std::runtime_error("Unsupported precision mode"); diff --git a/test/performance/routines/level2/xher2.cc b/test/performance/routines/level2/xher2.cc index 2d7e17ab..c335d3be 100644 --- a/test/performance/routines/level2/xher2.cc +++ b/test/performance/routines/level2/xher2.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kComplexSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kDouble: throw std::runtime_error("Unsupported precision mode"); diff --git a/test/performance/routines/level2/xhpmv.cc b/test/performance/routines/level2/xhpmv.cc index b9dd3f82..1e726569 100644 --- a/test/performance/routines/level2/xhpmv.cc +++ b/test/performance/routines/level2/xhpmv.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kComplexSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kDouble: throw std::runtime_error("Unsupported precision mode"); diff --git a/test/performance/routines/level2/xhpr.cc b/test/performance/routines/level2/xhpr.cc index f596682c..000b69af 100644 --- a/test/performance/routines/level2/xhpr.cc +++ b/test/performance/routines/level2/xhpr.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kComplexSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kDouble: throw std::runtime_error("Unsupported precision mode"); diff --git a/test/performance/routines/level2/xhpr2.cc b/test/performance/routines/level2/xhpr2.cc index 1c493226..19bafc46 100644 --- a/test/performance/routines/level2/xhpr2.cc +++ b/test/performance/routines/level2/xhpr2.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kComplexSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kDouble: throw std::runtime_error("Unsupported precision mode"); diff --git a/test/performance/routines/level2/xsbmv.cc b/test/performance/routines/level2/xsbmv.cc index febc6bfd..eabab3b7 100644 --- a/test/performance/routines/level2/xsbmv.cc +++ b/test/performance/routines/level2/xsbmv.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXsbmv<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xspmv.cc b/test/performance/routines/level2/xspmv.cc index 97c6b032..2a9ef925 100644 --- a/test/performance/routines/level2/xspmv.cc +++ b/test/performance/routines/level2/xspmv.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXspmv<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xspr.cc b/test/performance/routines/level2/xspr.cc index cc18d9b6..84331d74 100644 --- a/test/performance/routines/level2/xspr.cc +++ b/test/performance/routines/level2/xspr.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXspr<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xspr2.cc b/test/performance/routines/level2/xspr2.cc index 768452be..c42009a1 100644 --- a/test/performance/routines/level2/xspr2.cc +++ b/test/performance/routines/level2/xspr2.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXspr2<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xsymv.cc b/test/performance/routines/level2/xsymv.cc index 6748026f..3f72fe77 100644 --- a/test/performance/routines/level2/xsymv.cc +++ b/test/performance/routines/level2/xsymv.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXsymv<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xsyr.cc b/test/performance/routines/level2/xsyr.cc index 84510e5d..6b31d3a9 100644 --- a/test/performance/routines/level2/xsyr.cc +++ b/test/performance/routines/level2/xsyr.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXsyr<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xsyr2.cc b/test/performance/routines/level2/xsyr2.cc index b8c177d8..0ad59d2d 100644 --- a/test/performance/routines/level2/xsyr2.cc +++ b/test/performance/routines/level2/xsyr2.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXsyr2<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xtbmv.cc b/test/performance/routines/level2/xtbmv.cc index 1663dca0..a3297f34 100644 --- a/test/performance/routines/level2/xtbmv.cc +++ b/test/performance/routines/level2/xtbmv.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXtbmv<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xtbsv.cc b/test/performance/routines/level2/xtbsv.cc index e0cb9f2e..4dcd9a06 100644 --- a/test/performance/routines/level2/xtbsv.cc +++ b/test/performance/routines/level2/xtbsv.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXtbsv<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xtpmv.cc b/test/performance/routines/level2/xtpmv.cc index 407fdc8c..72477f2d 100644 --- a/test/performance/routines/level2/xtpmv.cc +++ b/test/performance/routines/level2/xtpmv.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXtpmv<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xtpsv.cc b/test/performance/routines/level2/xtpsv.cc index e402dc60..a3e3f7f1 100644 --- a/test/performance/routines/level2/xtpsv.cc +++ b/test/performance/routines/level2/xtpsv.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXtpsv<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xtrmv.cc b/test/performance/routines/level2/xtrmv.cc index c5563240..894a7952 100644 --- a/test/performance/routines/level2/xtrmv.cc +++ b/test/performance/routines/level2/xtrmv.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXtrmv<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level2/xtrsv.cc b/test/performance/routines/level2/xtrsv.cc index 136e2108..e8c65b0f 100644 --- a/test/performance/routines/level2/xtrsv.cc +++ b/test/performance/routines/level2/xtrsv.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXtrsv<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level3/xgemm.cc b/test/performance/routines/level3/xgemm.cc index 2082ceac..91897ee1 100644 --- a/test/performance/routines/level3/xgemm.cc +++ b/test/performance/routines/level3/xgemm.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXgemm<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level3/xhemm.cc b/test/performance/routines/level3/xhemm.cc index cc68e937..87650b9e 100644 --- a/test/performance/routines/level3/xhemm.cc +++ b/test/performance/routines/level3/xhemm.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kComplexSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kDouble: throw std::runtime_error("Unsupported precision mode"); diff --git a/test/performance/routines/level3/xher2k.cc b/test/performance/routines/level3/xher2k.cc index 70d76bed..06894816 100644 --- a/test/performance/routines/level3/xher2k.cc +++ b/test/performance/routines/level3/xher2k.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kComplexSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kDouble: throw std::runtime_error("Unsupported precision mode"); diff --git a/test/performance/routines/level3/xherk.cc b/test/performance/routines/level3/xherk.cc index b3b5dddf..d6f38fb2 100644 --- a/test/performance/routines/level3/xherk.cc +++ b/test/performance/routines/level3/xherk.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kComplexSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kDouble: throw std::runtime_error("Unsupported precision mode"); diff --git a/test/performance/routines/level3/xsymm.cc b/test/performance/routines/level3/xsymm.cc index f2292273..e0feadd1 100644 --- a/test/performance/routines/level3/xsymm.cc +++ b/test/performance/routines/level3/xsymm.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXsymm<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level3/xsyr2k.cc b/test/performance/routines/level3/xsyr2k.cc index 0c8f8f7c..4a82ddc4 100644 --- a/test/performance/routines/level3/xsyr2k.cc +++ b/test/performance/routines/level3/xsyr2k.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXsyr2k<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level3/xsyrk.cc b/test/performance/routines/level3/xsyrk.cc index ccd4511a..70f61322 100644 --- a/test/performance/routines/level3/xsyrk.cc +++ b/test/performance/routines/level3/xsyrk.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXsyrk<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level3/xtrmm.cc b/test/performance/routines/level3/xtrmm.cc index 8278d077..6f6041e4 100644 --- a/test/performance/routines/level3/xtrmm.cc +++ b/test/performance/routines/level3/xtrmm.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXtrmm<float>, float, float>(argc, argv); break; diff --git a/test/performance/routines/level3/xtrsm.cc b/test/performance/routines/level3/xtrsm.cc index 45f71c5e..76ef255a 100644 --- a/test/performance/routines/level3/xtrsm.cc +++ b/test/performance/routines/level3/xtrsm.cc @@ -18,7 +18,7 @@ using double2 = clblast::double2; // Main function (not within the clblast namespace) int main(int argc, char *argv[]) { - switch(clblast::GetPrecision(argc, argv)) { + switch(clblast::GetPrecision(argc, argv, clblast::Precision::kSingle)) { case clblast::Precision::kHalf: throw std::runtime_error("Unsupported precision mode"); case clblast::Precision::kSingle: clblast::RunClient<clblast::TestXtrsm<float>, float, float>(argc, argv); break; diff --git a/test/routines/level2/xgemv.h b/test/routines/level2/xgemv.h index 927bfaee..2924d498 100644 --- a/test/routines/level2/xgemv.h +++ b/test/routines/level2/xgemv.h @@ -35,7 +35,7 @@ class TestXgemv { // The list of arguments relevant for this routine static std::vector<std::string> GetOptions() { return {kArgM, kArgN, - kArgLayout, kArgATransp, + kArgLayout, kArgATransp, kArgALeadDim, kArgXInc, kArgYInc, kArgAOffset, kArgXOffset, kArgYOffset, kArgAlpha, kArgBeta}; diff --git a/test/routines/level2/xger.h b/test/routines/level2/xger.h new file mode 100644 index 00000000..98296e92 --- /dev/null +++ b/test/routines/level2/xger.h @@ -0,0 +1,131 @@ + +// ================================================================================================= +// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This +// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max- +// width of 100 characters per line. +// +// Author(s): +// Cedric Nugteren <www.cedricnugteren.nl> +// +// This file implements a class with static methods to describe the Xger routine. Examples of +// such 'descriptions' are how to calculate the size a of buffer or how to run the routine. These +// static methods are used by the correctness tester and the performance tester. +// +// ================================================================================================= + +#ifndef CLBLAST_TEST_ROUTINES_XGER_H_ +#define CLBLAST_TEST_ROUTINES_XGER_H_ + +#include <vector> +#include <string> + +#include "wrapper_clblas.h" + +namespace clblast { +// ================================================================================================= + +// See comment at top of file for a description of the class +template <typename T> +class TestXger { + public: + + // The BLAS level: 1, 2, or 3 + static size_t BLASLevel() { return 2; } + + // The list of arguments relevant for this routine + static std::vector<std::string> GetOptions() { + return {kArgM, kArgN, + kArgLayout, + kArgALeadDim, kArgXInc, kArgYInc, + kArgAOffset, kArgXOffset, kArgYOffset, + kArgAlpha}; + } + + // Describes how to obtain the sizes of the buffers + static size_t GetSizeX(const Arguments<T> &args) { + return args.m * args.x_inc + args.x_offset; + } + static size_t GetSizeY(const Arguments<T> &args) { + return args.n * args.y_inc + args.y_offset; + } + static size_t GetSizeA(const Arguments<T> &args) { + auto a_rotated = (args.layout == Layout::kRowMajor); + auto a_two = (a_rotated) ? args.m : args.n; + return a_two * args.a_ld + args.a_offset; + } + + // Describes how to set the sizes of all the buffers + static void SetSizes(Arguments<T> &args) { + args.a_size = GetSizeA(args); + args.x_size = GetSizeX(args); + args.y_size = GetSizeY(args); + } + + // Describes what the default values of the leading dimensions of the matrices are + static size_t DefaultLDA(const Arguments<T> &args) { return args.n; } + static size_t DefaultLDB(const Arguments<T> &) { return 1; } // N/A for this routine + static size_t DefaultLDC(const Arguments<T> &) { return 1; } // N/A for this routine + + // Describes which transpose options are relevant for this routine + using Transposes = std::vector<Transpose>; + static Transposes GetATransposes(const Transposes &) { return {}; } // N/A for this routine + static Transposes GetBTransposes(const Transposes &) { return {}; } // N/A for this routine + + // Describes how to run the CLBlast routine + static StatusCode RunRoutine(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) { + auto queue_plain = queue(); + auto event = cl_event{}; + auto status = Ger(args.layout, + args.m, args.n, args.alpha, + buffers.x_vec(), args.x_offset, args.x_inc, + buffers.y_vec(), args.y_offset, args.y_inc, + buffers.a_mat(), args.a_offset, args.a_ld, + &queue_plain, &event); + clWaitForEvents(1, &event); + return status; + } + + // Describes how to run the clBLAS routine (for correctness/performance comparison) + static StatusCode RunReference(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) { + auto queue_plain = queue(); + auto event = cl_event{}; + auto status = clblasXger(static_cast<clblasOrder>(args.layout), + args.m, args.n, args.alpha, + buffers.x_vec(), args.x_offset, args.x_inc, + buffers.y_vec(), args.y_offset, args.y_inc, + buffers.a_mat(), args.a_offset, args.a_ld, + 1, &queue_plain, 0, nullptr, &event); + clWaitForEvents(1, &event); + return static_cast<StatusCode>(status); + } + + // Describes how to download the results of the computation (more importantly: which buffer) + static std::vector<T> DownloadResult(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) { + std::vector<T> result(args.a_size, static_cast<T>(0)); + buffers.a_mat.Read(queue, args.a_size, result); + return result; + } + + // Describes how to compute the indices of the result buffer + static size_t ResultID1(const Arguments<T> &args) { return args.m; } + static size_t ResultID2(const Arguments<T> &args) { return args.n; } + static size_t GetResultIndex(const Arguments<T> &args, const size_t id1, const size_t id2) { + return (args.layout == Layout::kRowMajor) ? + id1*args.a_ld + id2 + args.a_offset: + id2*args.a_ld + id1 + args.a_offset; + } + + // Describes how to compute performance metrics + static size_t GetFlops(const Arguments<T> &args) { + return 2 * args.m * args.n; + } + static size_t GetBytes(const Arguments<T> &args) { + return (2*args.m*args.n + args.m + args.n) * sizeof(T); + } +}; + +// ================================================================================================= +} // namespace clblast + +// CLBLAST_TEST_ROUTINES_XGER_H_ +#endif diff --git a/test/routines/level2/xgerc.h b/test/routines/level2/xgerc.h new file mode 100644 index 00000000..77258d92 --- /dev/null +++ b/test/routines/level2/xgerc.h @@ -0,0 +1,131 @@ + +// ================================================================================================= +// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This +// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max- +// width of 100 characters per line. +// +// Author(s): +// Cedric Nugteren <www.cedricnugteren.nl> +// +// This file implements a class with static methods to describe the Xgerc routine. Examples of +// such 'descriptions' are how to calculate the size a of buffer or how to run the routine. These +// static methods are used by the correctness tester and the performance tester. +// +// ================================================================================================= + +#ifndef CLBLAST_TEST_ROUTINES_XGERC_H_ +#define CLBLAST_TEST_ROUTINES_XGERC_H_ + +#include <vector> +#include <string> + +#include "wrapper_clblas.h" + +namespace clblast { +// ================================================================================================= + +// See comment at top of file for a description of the class +template <typename T> +class TestXgerc { + public: + + // The BLAS level: 1, 2, or 3 + static size_t BLASLevel() { return 2; } + + // The list of arguments relevant for this routine + static std::vector<std::string> GetOptions() { + return {kArgM, kArgN, + kArgLayout, + kArgALeadDim, kArgXInc, kArgYInc, + kArgAOffset, kArgXOffset, kArgYOffset, + kArgAlpha}; + } + + // Describes how to obtain the sizes of the buffers + static size_t GetSizeX(const Arguments<T> &args) { + return args.m * args.x_inc + args.x_offset; + } + static size_t GetSizeY(const Arguments<T> &args) { + return args.n * args.y_inc + args.y_offset; + } + static size_t GetSizeA(const Arguments<T> &args) { + auto a_rotated = (args.layout == Layout::kRowMajor); + auto a_two = (a_rotated) ? args.m : args.n; + return a_two * args.a_ld + args.a_offset; + } + + // Describes how to set the sizes of all the buffers + static void SetSizes(Arguments<T> &args) { + args.a_size = GetSizeA(args); + args.x_size = GetSizeX(args); + args.y_size = GetSizeY(args); + } + + // Describes what the default values of the leading dimensions of the matrices are + static size_t DefaultLDA(const Arguments<T> &args) { return args.n; } + static size_t DefaultLDB(const Arguments<T> &) { return 1; } // N/A for this routine + static size_t DefaultLDC(const Arguments<T> &) { return 1; } // N/A for this routine + + // Describes which transpose options are relevant for this routine + using Transposes = std::vector<Transpose>; + static Transposes GetATransposes(const Transposes &) { return {}; } // N/A for this routine + static Transposes GetBTransposes(const Transposes &) { return {}; } // N/A for this routine + + // Describes how to run the CLBlast routine + static StatusCode RunRoutine(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) { + auto queue_plain = queue(); + auto event = cl_event{}; + auto status = Gerc(args.layout, + args.m, args.n, args.alpha, + buffers.x_vec(), args.x_offset, args.x_inc, + buffers.y_vec(), args.y_offset, args.y_inc, + buffers.a_mat(), args.a_offset, args.a_ld, + &queue_plain, &event); + clWaitForEvents(1, &event); + return status; + } + + // Describes how to run the clBLAS routine (for correctness/performance comparison) + static StatusCode RunReference(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) { + auto queue_plain = queue(); + auto event = cl_event{}; + auto status = clblasXgerc(static_cast<clblasOrder>(args.layout), + args.m, args.n, args.alpha, + buffers.x_vec(), args.x_offset, args.x_inc, + buffers.y_vec(), args.y_offset, args.y_inc, + buffers.a_mat(), args.a_offset, args.a_ld, + 1, &queue_plain, 0, nullptr, &event); + clWaitForEvents(1, &event); + return static_cast<StatusCode>(status); + } + + // Describes how to download the results of the computation (more importantly: which buffer) + static std::vector<T> DownloadResult(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) { + std::vector<T> result(args.a_size, static_cast<T>(0)); + buffers.a_mat.Read(queue, args.a_size, result); + return result; + } + + // Describes how to compute the indices of the result buffer + static size_t ResultID1(const Arguments<T> &args) { return args.m; } + static size_t ResultID2(const Arguments<T> &args) { return args.n; } + static size_t GetResultIndex(const Arguments<T> &args, const size_t id1, const size_t id2) { + return (args.layout == Layout::kRowMajor) ? + id1*args.a_ld + id2 + args.a_offset: + id2*args.a_ld + id1 + args.a_offset; + } + + // Describes how to compute performance metrics + static size_t GetFlops(const Arguments<T> &args) { + return 2 * args.m * args.n; + } + static size_t GetBytes(const Arguments<T> &args) { + return (2*args.m*args.n + args.m + args.n) * sizeof(T); + } +}; + +// ================================================================================================= +} // namespace clblast + +// CLBLAST_TEST_ROUTINES_XGERC_H_ +#endif diff --git a/test/routines/level2/xgeru.h b/test/routines/level2/xgeru.h new file mode 100644 index 00000000..e5f5f235 --- /dev/null +++ b/test/routines/level2/xgeru.h @@ -0,0 +1,131 @@ + +// ================================================================================================= +// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This +// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max- +// width of 100 characters per line. +// +// Author(s): +// Cedric Nugteren <www.cedricnugteren.nl> +// +// This file implements a class with static methods to describe the Xgeru routine. Examples of +// such 'descriptions' are how to calculate the size a of buffer or how to run the routine. These +// static methods are used by the correctness tester and the performance tester. +// +// ================================================================================================= + +#ifndef CLBLAST_TEST_ROUTINES_XGERU_H_ +#define CLBLAST_TEST_ROUTINES_XGERU_H_ + +#include <vector> +#include <string> + +#include "wrapper_clblas.h" + +namespace clblast { +// ================================================================================================= + +// See comment at top of file for a description of the class +template <typename T> +class TestXgeru { + public: + + // The BLAS level: 1, 2, or 3 + static size_t BLASLevel() { return 2; } + + // The list of arguments relevant for this routine + static std::vector<std::string> GetOptions() { + return {kArgM, kArgN, + kArgLayout, + kArgALeadDim, kArgXInc, kArgYInc, + kArgAOffset, kArgXOffset, kArgYOffset, + kArgAlpha}; + } + + // Describes how to obtain the sizes of the buffers + static size_t GetSizeX(const Arguments<T> &args) { + return args.m * args.x_inc + args.x_offset; + } + static size_t GetSizeY(const Arguments<T> &args) { + return args.n * args.y_inc + args.y_offset; + } + static size_t GetSizeA(const Arguments<T> &args) { + auto a_rotated = (args.layout == Layout::kRowMajor); + auto a_two = (a_rotated) ? args.m : args.n; + return a_two * args.a_ld + args.a_offset; + } + + // Describes how to set the sizes of all the buffers + static void SetSizes(Arguments<T> &args) { + args.a_size = GetSizeA(args); + args.x_size = GetSizeX(args); + args.y_size = GetSizeY(args); + } + + // Describes what the default values of the leading dimensions of the matrices are + static size_t DefaultLDA(const Arguments<T> &args) { return args.n; } + static size_t DefaultLDB(const Arguments<T> &) { return 1; } // N/A for this routine + static size_t DefaultLDC(const Arguments<T> &) { return 1; } // N/A for this routine + + // Describes which transpose options are relevant for this routine + using Transposes = std::vector<Transpose>; + static Transposes GetATransposes(const Transposes &) { return {}; } // N/A for this routine + static Transposes GetBTransposes(const Transposes &) { return {}; } // N/A for this routine + + // Describes how to run the CLBlast routine + static StatusCode RunRoutine(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) { + auto queue_plain = queue(); + auto event = cl_event{}; + auto status = Geru(args.layout, + args.m, args.n, args.alpha, + buffers.x_vec(), args.x_offset, args.x_inc, + buffers.y_vec(), args.y_offset, args.y_inc, + buffers.a_mat(), args.a_offset, args.a_ld, + &queue_plain, &event); + clWaitForEvents(1, &event); + return status; + } + + // Describes how to run the clBLAS routine (for correctness/performance comparison) + static StatusCode RunReference(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) { + auto queue_plain = queue(); + auto event = cl_event{}; + auto status = clblasXgeru(static_cast<clblasOrder>(args.layout), + args.m, args.n, args.alpha, + buffers.x_vec(), args.x_offset, args.x_inc, + buffers.y_vec(), args.y_offset, args.y_inc, + buffers.a_mat(), args.a_offset, args.a_ld, + 1, &queue_plain, 0, nullptr, &event); + clWaitForEvents(1, &event); + return static_cast<StatusCode>(status); + } + + // Describes how to download the results of the computation (more importantly: which buffer) + static std::vector<T> DownloadResult(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) { + std::vector<T> result(args.a_size, static_cast<T>(0)); + buffers.a_mat.Read(queue, args.a_size, result); + return result; + } + + // Describes how to compute the indices of the result buffer + static size_t ResultID1(const Arguments<T> &args) { return args.m; } + static size_t ResultID2(const Arguments<T> &args) { return args.n; } + static size_t GetResultIndex(const Arguments<T> &args, const size_t id1, const size_t id2) { + return (args.layout == Layout::kRowMajor) ? + id1*args.a_ld + id2 + args.a_offset: + id2*args.a_ld + id1 + args.a_offset; + } + + // Describes how to compute performance metrics + static size_t GetFlops(const Arguments<T> &args) { + return 2 * args.m * args.n; + } + static size_t GetBytes(const Arguments<T> &args) { + return (2*args.m*args.n + args.m + args.n) * sizeof(T); + } +}; + +// ================================================================================================= +} // namespace clblast + +// CLBLAST_TEST_ROUTINES_XGERU_H_ +#endif diff --git a/test/routines/level2/xher.h b/test/routines/level2/xher.h new file mode 100644 index 00000000..53c4200f --- /dev/null +++ b/test/routines/level2/xher.h @@ -0,0 +1,122 @@ + +// ================================================================================================= +// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This +// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max- +// width of 100 characters per line. +// +// Author(s): +// Cedric Nugteren <www.cedricnugteren.nl> +// +// This file implements a class with static methods to describe the Xher routine. Examples of +// such 'descriptions' are how to calculate the size a of buffer or how to run the routine. These +// static methods are used by the correctness tester and the performance tester. +// +// ================================================================================================= + +#ifndef CLBLAST_TEST_ROUTINES_XHER_H_ +#define CLBLAST_TEST_ROUTINES_XHER_H_ + +#include <vector> +#include <string> + +#include "wrapper_clblas.h" + +namespace clblast { +// ================================================================================================= + +// See comment at top of file for a description of the class +template <typename T, typename U> +class TestXher { + public: + + // The BLAS level: 1, 2, or 3 + static size_t BLASLevel() { return 2; } + + // The list of arguments relevant for this routine + static std::vector<std::string> GetOptions() { + return {kArgN, + kArgLayout, kArgTriangle, + kArgALeadDim, kArgXInc, + kArgAOffset, kArgXOffset, + kArgAlpha}; + } + + // Describes how to obtain the sizes of the buffers + static size_t GetSizeX(const Arguments<U> &args) { + return args.n * args.x_inc + args.x_offset; + } + static size_t GetSizeA(const Arguments<U> &args) { + return args.n * args.a_ld + args.a_offset; + } + + // Describes how to set the sizes of all the buffers + static void SetSizes(Arguments<U> &args) { + args.a_size = GetSizeA(args); + args.x_size = GetSizeX(args); + } + + // Describes what the default values of the leading dimensions of the matrices are + static size_t DefaultLDA(const Arguments<U> &args) { return args.n; } + static size_t DefaultLDB(const Arguments<U> &) { return 1; } // N/A for this routine + static size_t DefaultLDC(const Arguments<U> &) { return 1; } // N/A for this routine + + // Describes which transpose options are relevant for this routine + using Transposes = std::vector<Transpose>; + static Transposes GetATransposes(const Transposes &) { return {}; } // N/A for this routine + static Transposes GetBTransposes(const Transposes &) { return {}; } // N/A for this routine + + // Describes how to run the CLBlast routine + static StatusCode RunRoutine(const Arguments<U> &args, const Buffers<T> &buffers, Queue &queue) { + auto queue_plain = queue(); + auto event = cl_event{}; + auto status = Her(args.layout, args.triangle, + args.n, args.alpha, + buffers.x_vec(), args.x_offset, args.x_inc, + buffers.a_mat(), args.a_offset, args.a_ld, + &queue_plain, &event); + clWaitForEvents(1, &event); + return status; + } + + // Describes how to run the clBLAS routine (for correctness/performance comparison) + static StatusCode RunReference(const Arguments<U> &args, const Buffers<T> &buffers, Queue &queue) { + auto queue_plain = queue(); + auto event = cl_event{}; + auto status = clblasXher(static_cast<clblasOrder>(args.layout), + static_cast<clblasUplo>(args.triangle), + args.n, args.alpha, + buffers.x_vec(), args.x_offset, args.x_inc, + buffers.a_mat(), args.a_offset, args.a_ld, + 1, &queue_plain, 0, nullptr, &event); + clWaitForEvents(1, &event); + return static_cast<StatusCode>(status); + } + + // Describes how to download the results of the computation (more importantly: which buffer) + static std::vector<T> DownloadResult(const Arguments<U> &args, Buffers<T> &buffers, Queue &queue) { + std::vector<T> result(args.a_size, static_cast<T>(0)); + buffers.a_mat.Read(queue, args.a_size, result); + return result; + } + + // Describes how to compute the indices of the result buffer + static size_t ResultID1(const Arguments<U> &args) { return args.n; } + static size_t ResultID2(const Arguments<U> &args) { return args.n; } + static size_t GetResultIndex(const Arguments<U> &args, const size_t id1, const size_t id2) { + return id2*args.a_ld + id1 + args.a_offset; + } + + // Describes how to compute performance metrics + static size_t GetFlops(const Arguments<U> &args) { + return 3 * args.n * args.n; + } + static size_t GetBytes(const Arguments<U> &args) { + return (args.n*args.n + args.n) * sizeof(T); + } +}; + +// ================================================================================================= +} // namespace clblast + +// CLBLAST_TEST_ROUTINES_XHER_H_ +#endif diff --git a/test/routines/level2/xhpr.h b/test/routines/level2/xhpr.h new file mode 100644 index 00000000..03599ddc --- /dev/null +++ b/test/routines/level2/xhpr.h @@ -0,0 +1,122 @@ + +// ================================================================================================= +// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This +// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max- +// width of 100 characters per line. +// +// Author(s): +// Cedric Nugteren <www.cedricnugteren.nl> +// +// This file implements a class with static methods to describe the Xhpr routine. Examples of +// such 'descriptions' are how to calculate the size a of buffer or how to run the routine. These +// static methods are used by the correctness tester and the performance tester. +// +// ================================================================================================= + +#ifndef CLBLAST_TEST_ROUTINES_XHPR_H_ +#define CLBLAST_TEST_ROUTINES_XHPR_H_ + +#include <vector> +#include <string> + +#include "wrapper_clblas.h" + +namespace clblast { +// ================================================================================================= + +// See comment at top of file for a description of the class +template <typename T, typename U> +class TestXhpr { + public: + + // The BLAS level: 1, 2, or 3 + static size_t BLASLevel() { return 2; } + + // The list of arguments relevant for this routine + static std::vector<std::string> GetOptions() { + return {kArgN, + kArgLayout, kArgTriangle, + kArgXInc, + kArgAPOffset, kArgXOffset, + kArgAlpha}; + } + + // Describes how to obtain the sizes of the buffers + static size_t GetSizeX(const Arguments<U> &args) { + return args.n * args.x_inc + args.x_offset; + } + static size_t GetSizeAP(const Arguments<U> &args) { + return ((args.n*(args.n+1)) / 2) + args.ap_offset; + } + + // Describes how to set the sizes of all the buffers + static void SetSizes(Arguments<U> &args) { + args.ap_size = GetSizeAP(args); + args.x_size = GetSizeX(args); + } + + // Describes what the default values of the leading dimensions of the matrices are + static size_t DefaultLDA(const Arguments<U> &args) { return args.n; } + static size_t DefaultLDB(const Arguments<U> &) { return 1; } // N/A for this routine + static size_t DefaultLDC(const Arguments<U> &) { return 1; } // N/A for this routine + + // Describes which transpose options are relevant for this routine + using Transposes = std::vector<Transpose>; + static Transposes GetATransposes(const Transposes &) { return {}; } // N/A for this routine + static Transposes GetBTransposes(const Transposes &) { return {}; } // N/A for this routine + + // Describes how to run the CLBlast routine + static StatusCode RunRoutine(const Arguments<U> &args, const Buffers<T> &buffers, Queue &queue) { + auto queue_plain = queue(); + auto event = cl_event{}; + auto status = Hpr(args.layout, args.triangle, + args.n, args.alpha, + buffers.x_vec(), args.x_offset, args.x_inc, + buffers.ap_mat(), args.ap_offset, + &queue_plain, &event); + clWaitForEvents(1, &event); + return status; + } + + // Describes how to run the clBLAS routine (for correctness/performance comparison) + static StatusCode RunReference(const Arguments<U> &args, const Buffers<T> &buffers, Queue &queue) { + auto queue_plain = queue(); + auto event = cl_event{}; + auto status = clblasXhpr(static_cast<clblasOrder>(args.layout), + static_cast<clblasUplo>(args.triangle), + args.n, args.alpha, + buffers.x_vec(), args.x_offset, args.x_inc, + buffers.ap_mat(), args.ap_offset, + 1, &queue_plain, 0, nullptr, &event); + clWaitForEvents(1, &event); + return static_cast<StatusCode>(status); + } + + // Describes how to download the results of the computation (more importantly: which buffer) + static std::vector<T> DownloadResult(const Arguments<U> &args, Buffers<T> &buffers, Queue &queue) { + std::vector<T> result(args.ap_size, static_cast<T>(0)); + buffers.ap_mat.Read(queue, args.ap_size, result); + return result; + } + + // Describes how to compute the indices of the result buffer + static size_t ResultID1(const Arguments<U> &args) { return args.ap_size - args.ap_offset; } + static size_t ResultID2(const Arguments<U> &) { return 1; } // N/A for this routine + static size_t GetResultIndex(const Arguments<U> &args, const size_t id1, const size_t) { + return id1 + args.ap_offset; + } + + // Describes how to compute performance metrics + static size_t GetFlops(const Arguments<U> &args) { + return 3 * ((args.n*(args.n+1)) / 2); + } + static size_t GetBytes(const Arguments<U> &args) { + return ((args.n*(args.n+1)) + args.n) * sizeof(T); + } +}; + +// ================================================================================================= +} // namespace clblast + +// CLBLAST_TEST_ROUTINES_XHPR_H_ +#endif diff --git a/test/routines/level2/xspr.h b/test/routines/level2/xspr.h new file mode 100644 index 00000000..819b1ca8 --- /dev/null +++ b/test/routines/level2/xspr.h @@ -0,0 +1,122 @@ + +// ================================================================================================= +// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This +// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max- +// width of 100 characters per line. +// +// Author(s): +// Cedric Nugteren <www.cedricnugteren.nl> +// +// This file implements a class with static methods to describe the Xspr routine. Examples of +// such 'descriptions' are how to calculate the size a of buffer or how to run the routine. These +// static methods are used by the correctness tester and the performance tester. +// +// ================================================================================================= + +#ifndef CLBLAST_TEST_ROUTINES_XSPR_H_ +#define CLBLAST_TEST_ROUTINES_XSPR_H_ + +#include <vector> +#include <string> + +#include "wrapper_clblas.h" + +namespace clblast { +// ================================================================================================= + +// See comment at top of file for a description of the class +template <typename T> +class TestXspr { + public: + + // The BLAS level: 1, 2, or 3 + static size_t BLASLevel() { return 2; } + + // The list of arguments relevant for this routine + static std::vector<std::string> GetOptions() { + return {kArgN, + kArgLayout, kArgTriangle, + kArgXInc, + kArgAPOffset, kArgXOffset, + kArgAlpha}; + } + + // Describes how to obtain the sizes of the buffers + static size_t GetSizeX(const Arguments<T> &args) { + return args.n * args.x_inc + args.x_offset; + } + static size_t GetSizeAP(const Arguments<T> &args) { + return ((args.n*(args.n+1)) / 2) + args.ap_offset; + } + + // Describes how to set the sizes of all the buffers + static void SetSizes(Arguments<T> &args) { + args.ap_size = GetSizeAP(args); + args.x_size = GetSizeX(args); + } + + // Describes what the default values of the leading dimensions of the matrices are + static size_t DefaultLDA(const Arguments<T> &args) { return args.n; } + static size_t DefaultLDB(const Arguments<T> &) { return 1; } // N/A for this routine + static size_t DefaultLDC(const Arguments<T> &) { return 1; } // N/A for this routine + + // Describes which transpose options are relevant for this routine + using Transposes = std::vector<Transpose>; + static Transposes GetATransposes(const Transposes &) { return {}; } // N/A for this routine + static Transposes GetBTransposes(const Transposes &) { return {}; } // N/A for this routine + + // Describes how to run the CLBlast routine + static StatusCode RunRoutine(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) { + auto queue_plain = queue(); + auto event = cl_event{}; + auto status = Spr(args.layout, args.triangle, + args.n, args.alpha, + buffers.x_vec(), args.x_offset, args.x_inc, + buffers.ap_mat(), args.ap_offset, + &queue_plain, &event); + clWaitForEvents(1, &event); + return status; + } + + // Describes how to run the clBLAS routine (for correctness/performance comparison) + static StatusCode RunReference(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) { + auto queue_plain = queue(); + auto event = cl_event{}; + auto status = clblasXspr(static_cast<clblasOrder>(args.layout), + static_cast<clblasUplo>(args.triangle), + args.n, args.alpha, + buffers.x_vec(), args.x_offset, args.x_inc, + buffers.ap_mat(), args.ap_offset, + 1, &queue_plain, 0, nullptr, &event); + clWaitForEvents(1, &event); + return static_cast<StatusCode>(status); + } + + // Describes how to download the results of the computation (more importantly: which buffer) + static std::vector<T> DownloadResult(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) { + std::vector<T> result(args.ap_size, static_cast<T>(0)); + buffers.ap_mat.Read(queue, args.ap_size, result); + return result; + } + + // Describes how to compute the indices of the result buffer + static size_t ResultID1(const Arguments<T> &args) { return args.ap_size - args.ap_offset; } + static size_t ResultID2(const Arguments<T> &) { return 1; } // N/A for this routine + static size_t GetResultIndex(const Arguments<T> &args, const size_t id1, const size_t) { + return id1 + args.ap_offset; + } + + // Describes how to compute performance metrics + static size_t GetFlops(const Arguments<T> &args) { + return 3 * ((args.n*(args.n+1)) / 2); + } + static size_t GetBytes(const Arguments<T> &args) { + return ((args.n*(args.n+1)) + args.n) * sizeof(T); + } +}; + +// ================================================================================================= +} // namespace clblast + +// CLBLAST_TEST_ROUTINES_XSPR_H_ +#endif diff --git a/test/routines/level2/xsyr.h b/test/routines/level2/xsyr.h new file mode 100644 index 00000000..66b75c0c --- /dev/null +++ b/test/routines/level2/xsyr.h @@ -0,0 +1,122 @@ + +// ================================================================================================= +// This file is part of the CLBlast project. The project is licensed under Apache Version 2.0. This +// project loosely follows the Google C++ styleguide and uses a tab-size of two spaces and a max- +// width of 100 characters per line. +// +// Author(s): +// Cedric Nugteren <www.cedricnugteren.nl> +// +// This file implements a class with static methods to describe the Xsyr routine. Examples of +// such 'descriptions' are how to calculate the size a of buffer or how to run the routine. These +// static methods are used by the correctness tester and the performance tester. +// +// ================================================================================================= + +#ifndef CLBLAST_TEST_ROUTINES_XSYR_H_ +#define CLBLAST_TEST_ROUTINES_XSYR_H_ + +#include <vector> +#include <string> + +#include "wrapper_clblas.h" + +namespace clblast { +// ================================================================================================= + +// See comment at top of file for a description of the class +template <typename T> +class TestXsyr { + public: + + // The BLAS level: 1, 2, or 3 + static size_t BLASLevel() { return 2; } + + // The list of arguments relevant for this routine + static std::vector<std::string> GetOptions() { + return {kArgN, + kArgLayout, kArgTriangle, + kArgALeadDim, kArgXInc, + kArgAOffset, kArgXOffset, + kArgAlpha}; + } + + // Describes how to obtain the sizes of the buffers + static size_t GetSizeX(const Arguments<T> &args) { + return args.n * args.x_inc + args.x_offset; + } + static size_t GetSizeA(const Arguments<T> &args) { + return args.n * args.a_ld + args.a_offset; + } + + // Describes how to set the sizes of all the buffers + static void SetSizes(Arguments<T> &args) { + args.a_size = GetSizeA(args); + args.x_size = GetSizeX(args); + } + + // Describes what the default values of the leading dimensions of the matrices are + static size_t DefaultLDA(const Arguments<T> &args) { return args.n; } + static size_t DefaultLDB(const Arguments<T> &) { return 1; } // N/A for this routine + static size_t DefaultLDC(const Arguments<T> &) { return 1; } // N/A for this routine + + // Describes which transpose options are relevant for this routine + using Transposes = std::vector<Transpose>; + static Transposes GetATransposes(const Transposes &) { return {}; } // N/A for this routine + static Transposes GetBTransposes(const Transposes &) { return {}; } // N/A for this routine + + // Describes how to run the CLBlast routine + static StatusCode RunRoutine(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) { + auto queue_plain = queue(); + auto event = cl_event{}; + auto status = Syr(args.layout, args.triangle, + args.n, args.alpha, + buffers.x_vec(), args.x_offset, args.x_inc, + buffers.a_mat(), args.a_offset, args.a_ld, + &queue_plain, &event); + clWaitForEvents(1, &event); + return status; + } + + // Describes how to run the clBLAS routine (for correctness/performance comparison) + static StatusCode RunReference(const Arguments<T> &args, const Buffers<T> &buffers, Queue &queue) { + auto queue_plain = queue(); + auto event = cl_event{}; + auto status = clblasXsyr(static_cast<clblasOrder>(args.layout), + static_cast<clblasUplo>(args.triangle), + args.n, args.alpha, + buffers.x_vec(), args.x_offset, args.x_inc, + buffers.a_mat(), args.a_offset, args.a_ld, + 1, &queue_plain, 0, nullptr, &event); + clWaitForEvents(1, &event); + return static_cast<StatusCode>(status); + } + + // Describes how to download the results of the computation (more importantly: which buffer) + static std::vector<T> DownloadResult(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) { + std::vector<T> result(args.a_size, static_cast<T>(0)); + buffers.a_mat.Read(queue, args.a_size, result); + return result; + } + + // Describes how to compute the indices of the result buffer + static size_t ResultID1(const Arguments<T> &args) { return args.n; } + static size_t ResultID2(const Arguments<T> &args) { return args.n; } + static size_t GetResultIndex(const Arguments<T> &args, const size_t id1, const size_t id2) { + return id2*args.a_ld + id1 + args.a_offset; + } + + // Describes how to compute performance metrics + static size_t GetFlops(const Arguments<T> &args) { + return 3 * args.n * args.n; + } + static size_t GetBytes(const Arguments<T> &args) { + return (args.n*args.n + args.n) * sizeof(T); + } +}; + +// ================================================================================================= +} // namespace clblast + +// CLBLAST_TEST_ROUTINES_XSYR_H_ +#endif |