Moved the RunKernel and PadCopyTransposeMatrix functions out of the Routine class

8 files changed, 31 insertions, 31 deletions

diff --git a/src/routines/level3/xgemm.cc b/src/routines/level3/xgemm.cc
index 713bed8f..eced53ab 100644
--- a/src/routines/level3/xgemm.cc
+++ b/src/routines/level3/xgemm.cc
@@ -142,7 +142,7 @@ StatusCode Xgemm<T>::DoGemm(const Layout layout,
     // case nothing has to be done, these kernels can be skipped.
     if (!a_no_temp) {
       auto eventProcessA = Event();
-      status = PadCopyTransposeMatrix(eventProcessA.pointer(), emptyEventList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessA.pointer(), emptyEventList,
                                       a_one, a_two, a_ld, a_offset, a_buffer,
                                       m_ceiled, k_ceiled, m_ceiled, 0, a_temp,
                                       ConstantOne<T>(), program,
@@ -154,7 +154,7 @@ StatusCode Xgemm<T>::DoGemm(const Layout layout,
     // As above, but now for matrix B
     if (!b_no_temp) {
       auto eventProcessB = Event();
-      status = PadCopyTransposeMatrix(eventProcessB.pointer(), emptyEventList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessB.pointer(), emptyEventList,
                                       b_one, b_two, b_ld, b_offset, b_buffer,
                                       n_ceiled, k_ceiled, n_ceiled, 0, b_temp,
                                       ConstantOne<T>(), program,
@@ -166,7 +166,7 @@ StatusCode Xgemm<T>::DoGemm(const Layout layout,
     // As above, but now for matrix C. This is only necessary if C is used both as input and output.
     if (!c_no_temp && beta != static_cast<T>(0)) {
       auto eventProcessC = Event();
-      status = PadCopyTransposeMatrix(eventProcessC.pointer(), emptyEventList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessC.pointer(), emptyEventList,
                                       c_one, c_two, c_ld, c_offset, c_buffer,
                                       m_ceiled, n_ceiled, m_ceiled, 0, c_temp,
                                       ConstantOne<T>(), program,
@@ -199,13 +199,13 @@ StatusCode Xgemm<T>::DoGemm(const Layout layout,
       // Launches the kernel
       auto eventKernel = Event();
       auto eventPointer = (!c_no_temp) ? eventKernel.pointer() : event_;
-      status = RunKernel(kernel, global, local, eventPointer, eventWaitList);
+      status = RunKernel(kernel, queue_, device_, global, local, eventPointer, eventWaitList);
       if (ErrorIn(status)) { return status; }
 
       // Runs the post-processing kernel if needed
       if (!c_no_temp) {
         eventWaitList.push_back(eventKernel);
-        status = PadCopyTransposeMatrix(event_, eventWaitList,
+        status = PadCopyTransposeMatrix(queue_, device_, context_, db_, event_, eventWaitList,
                                         m_ceiled, n_ceiled, m_ceiled, 0, c_temp,
                                         c_one, c_two, c_ld, c_offset, c_buffer,
                                         ConstantOne<T>(), program,
diff --git a/src/routines/level3/xhemm.cc b/src/routines/level3/xhemm.cc
index a6e853e9..9791d7b4 100644
--- a/src/routines/level3/xhemm.cc
+++ b/src/routines/level3/xhemm.cc
@@ -80,7 +80,7 @@ StatusCode Xhemm<T>::DoHemm(const Layout layout, const Side side, const Triangle
                                         Ceil(CeilDiv(k, db_["PAD_WPTY"]), db_["PAD_DIMY"])};
       auto local = std::vector<size_t>{db_["PAD_DIMX"], db_["PAD_DIMY"]};
       auto kernelEvent = Event();
-      status = RunKernel(kernel, global, local, kernelEvent.pointer());
+      status = RunKernel(kernel, queue_, device_, global, local, kernelEvent.pointer());
       if (ErrorIn(status)) { return status; }
 
       // Synchronize now: 'DoGemm' does not accept a list of events to wait for
diff --git a/src/routines/level3/xher2k.cc b/src/routines/level3/xher2k.cc
index c891c202..43f7bb76 100644
--- a/src/routines/level3/xher2k.cc
+++ b/src/routines/level3/xher2k.cc
@@ -129,7 +129,7 @@ StatusCode Xher2k<T,U>::DoHer2k(const Layout layout, const Triangle triangle, co
     // case nothing has to be done, these kernels can be skipped.
     if (!a1_no_temp) {
       auto eventProcessA1 = Event();
-      status = PadCopyTransposeMatrix(eventProcessA1.pointer(), emptyEventList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessA1.pointer(), emptyEventList,
                                       ab_one, ab_two, a_ld, a_offset, a_buffer,
                                       n_ceiled, k_ceiled, n_ceiled, 0, a1_temp,
                                       ConstantOne<T>(), program,
@@ -139,7 +139,7 @@ StatusCode Xher2k<T,U>::DoHer2k(const Layout layout, const Triangle triangle, co
     }
     if (!a2_no_temp) {
       auto eventProcessA2 = Event();
-      status = PadCopyTransposeMatrix(eventProcessA2.pointer(), emptyEventList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessA2.pointer(), emptyEventList,
                                       ab_one, ab_two, a_ld, a_offset, a_buffer,
                                       n_ceiled, k_ceiled, n_ceiled, 0, a2_temp,
                                       ConstantOne<T>(), program,
@@ -149,7 +149,7 @@ StatusCode Xher2k<T,U>::DoHer2k(const Layout layout, const Triangle triangle, co
     }
     if (!b1_no_temp) {
       auto eventProcessB1 = Event();
-      status = PadCopyTransposeMatrix(eventProcessB1.pointer(), emptyEventList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessB1.pointer(), emptyEventList,
                                       ab_one, ab_two, b_ld, b_offset, b_buffer,
                                       n_ceiled, k_ceiled, n_ceiled, 0, b1_temp,
                                       ConstantOne<T>(), program,
@@ -159,7 +159,7 @@ StatusCode Xher2k<T,U>::DoHer2k(const Layout layout, const Triangle triangle, co
     }
     if (!b2_no_temp) {
       auto eventProcessB2 = Event();
-      status = PadCopyTransposeMatrix(eventProcessB2.pointer(), emptyEventList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessB2.pointer(), emptyEventList,
                                       ab_one, ab_two, b_ld, b_offset, b_buffer,
                                       n_ceiled, k_ceiled, n_ceiled, 0, b2_temp,
                                       ConstantOne<T>(), program,
@@ -171,7 +171,7 @@ StatusCode Xher2k<T,U>::DoHer2k(const Layout layout, const Triangle triangle, co
     // Furthermore, also creates a (possibly padded) copy of matrix C, since it is not allowed to
     // modify the other triangle.
     auto eventProcessC = Event();
-    status = PadCopyTransposeMatrix(eventProcessC.pointer(), emptyEventList,
+    status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessC.pointer(), emptyEventList,
                                     n, n, c_ld, c_offset, c_buffer,
                                     n_ceiled, n_ceiled, n_ceiled, 0, c_temp,
                                     ConstantOne<T>(), program,
@@ -201,7 +201,7 @@ StatusCode Xher2k<T,U>::DoHer2k(const Layout layout, const Triangle triangle, co
 
       // Launches the kernel
       auto eventKernel1 = Event();
-      status = RunKernel(kernel, global, local, eventKernel1.pointer(), eventWaitList);
+      status = RunKernel(kernel, queue_, device_, global, local, eventKernel1.pointer(), eventWaitList);
       if (ErrorIn(status)) { return status; }
       eventWaitList.push_back(eventKernel1);
 
@@ -217,14 +217,14 @@ StatusCode Xher2k<T,U>::DoHer2k(const Layout layout, const Triangle triangle, co
 
       // Runs the kernel again
       auto eventKernel2 = Event();
-      status = RunKernel(kernel, global, local, eventKernel2.pointer(), eventWaitList);
+      status = RunKernel(kernel, queue_, device_, global, local, eventKernel2.pointer(), eventWaitList);
       if (ErrorIn(status)) { return status; }
       eventWaitList.push_back(eventKernel2);
 
       // Runs the post-processing kernel
       auto upper = (triangle == Triangle::kUpper);
       auto lower = (triangle == Triangle::kLower);
-      status = PadCopyTransposeMatrix(event_, eventWaitList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, event_, eventWaitList,
                                       n_ceiled, n_ceiled, n_ceiled, 0, c_temp,
                                       n, n, c_ld, c_offset, c_buffer,
                                       ConstantOne<T>(), program,
diff --git a/src/routines/level3/xherk.cc b/src/routines/level3/xherk.cc
index 9d64af95..8ebcbfa8 100644
--- a/src/routines/level3/xherk.cc
+++ b/src/routines/level3/xherk.cc
@@ -121,7 +121,7 @@ StatusCode Xherk<T,U>::DoHerk(const Layout layout, const Triangle triangle, cons
     // case nothing has to be done, these kernels can be skipped. Two copies are created.
     if (!a_no_temp) {
       auto eventProcessA = Event();
-      status = PadCopyTransposeMatrix(eventProcessA.pointer(), emptyEventList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessA.pointer(), emptyEventList,
                                       a_one, a_two, a_ld, a_offset, a_buffer,
                                       n_ceiled, k_ceiled, n_ceiled, 0, a_temp,
                                       ConstantOne<T>(), program,
@@ -131,7 +131,7 @@ StatusCode Xherk<T,U>::DoHerk(const Layout layout, const Triangle triangle, cons
     }
     if (!b_no_temp) {
       auto eventProcessB = Event();
-      status = PadCopyTransposeMatrix(eventProcessB.pointer(), emptyEventList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessB.pointer(), emptyEventList,
                                       a_one, a_two, a_ld, a_offset, a_buffer,
                                       n_ceiled, k_ceiled, n_ceiled, 0, b_temp,
                                       ConstantOne<T>(), program,
@@ -143,7 +143,7 @@ StatusCode Xherk<T,U>::DoHerk(const Layout layout, const Triangle triangle, cons
     // Furthermore, also creates a (possibly padded) copy of matrix C, since it is not allowed to
     // modify the other triangle.
     auto eventProcessC = Event();
-    status = PadCopyTransposeMatrix(eventProcessC.pointer(), emptyEventList,
+    status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessC.pointer(), emptyEventList,
                                     n, n, c_ld, c_offset, c_buffer,
                                     n_ceiled, n_ceiled, n_ceiled, 0, c_temp,
                                     ConstantOne<T>(), program,
@@ -173,14 +173,14 @@ StatusCode Xherk<T,U>::DoHerk(const Layout layout, const Triangle triangle, cons
 
       // Launches the kernel
       auto eventKernel = Event();
-      status = RunKernel(kernel, global, local, eventKernel.pointer(), eventWaitList);
+      status = RunKernel(kernel, queue_, device_, global, local, eventKernel.pointer(), eventWaitList);
       if (ErrorIn(status)) { return status; }
       eventWaitList.push_back(eventKernel);
 
       // Runs the post-processing kernel
       auto upper = (triangle == Triangle::kUpper);
       auto lower = (triangle == Triangle::kLower);
-      status = PadCopyTransposeMatrix(event_, eventWaitList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, event_, eventWaitList,
                                       n_ceiled, n_ceiled, n_ceiled, 0, c_temp,
                                       n, n, c_ld, c_offset, c_buffer,
                                       ConstantOne<T>(), program,
diff --git a/src/routines/level3/xsymm.cc b/src/routines/level3/xsymm.cc
index 379e2908..650afbfc 100644
--- a/src/routines/level3/xsymm.cc
+++ b/src/routines/level3/xsymm.cc
@@ -80,7 +80,7 @@ StatusCode Xsymm<T>::DoSymm(const Layout layout, const Side side, const Triangle
                                         Ceil(CeilDiv(k, db_["PAD_WPTY"]), db_["PAD_DIMY"])};
       auto local = std::vector<size_t>{db_["PAD_DIMX"], db_["PAD_DIMY"]};
       auto kernelEvent = Event();
-      status = RunKernel(kernel, global, local, kernelEvent.pointer());
+      status = RunKernel(kernel, queue_, device_, global, local, kernelEvent.pointer());
       if (ErrorIn(status)) { return status; }
 
       // Synchronize now: 'DoGemm' does not accept a list of events to wait for
diff --git a/src/routines/level3/xsyr2k.cc b/src/routines/level3/xsyr2k.cc
index 886d1e16..4b436381 100644
--- a/src/routines/level3/xsyr2k.cc
+++ b/src/routines/level3/xsyr2k.cc
@@ -121,7 +121,7 @@ StatusCode Xsyr2k<T>::DoSyr2k(const Layout layout, const Triangle triangle, cons
     // case nothing has to be done, these kernels can be skipped.
     if (!a_no_temp) {
       auto eventProcessA = Event();
-      status = PadCopyTransposeMatrix(eventProcessA.pointer(), emptyEventList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessA.pointer(), emptyEventList,
                                       ab_one, ab_two, a_ld, a_offset, a_buffer,
                                       n_ceiled, k_ceiled, n_ceiled, 0, a_temp,
                                       ConstantOne<T>(), program,
@@ -131,7 +131,7 @@ StatusCode Xsyr2k<T>::DoSyr2k(const Layout layout, const Triangle triangle, cons
     }
     if (!b_no_temp) {
       auto eventProcessB = Event();
-      status = PadCopyTransposeMatrix(eventProcessB.pointer(), emptyEventList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessB.pointer(), emptyEventList,
                                       ab_one, ab_two, b_ld, b_offset, b_buffer,
                                       n_ceiled, k_ceiled, n_ceiled, 0, b_temp,
                                       ConstantOne<T>(), program,
@@ -143,7 +143,7 @@ StatusCode Xsyr2k<T>::DoSyr2k(const Layout layout, const Triangle triangle, cons
     // Furthermore, also creates a (possibly padded) copy of matrix C, since it is not allowed to
     // modify the other triangle.
     auto eventProcessC = Event();
-    status = PadCopyTransposeMatrix(eventProcessC.pointer(), emptyEventList,
+    status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessC.pointer(), emptyEventList,
                                     n, n, c_ld, c_offset, c_buffer,
                                     n_ceiled, n_ceiled, n_ceiled, 0, c_temp,
                                     ConstantOne<T>(), program,
@@ -173,7 +173,7 @@ StatusCode Xsyr2k<T>::DoSyr2k(const Layout layout, const Triangle triangle, cons
 
       // Launches the kernel
       auto eventKernel1 = Event();
-      status = RunKernel(kernel, global, local, eventKernel1.pointer(), eventWaitList);
+      status = RunKernel(kernel, queue_, device_, global, local, eventKernel1.pointer(), eventWaitList);
       if (ErrorIn(status)) { return status; }
       eventWaitList.push_back(eventKernel1);
 
@@ -186,14 +186,14 @@ StatusCode Xsyr2k<T>::DoSyr2k(const Layout layout, const Triangle triangle, cons
 
       // Runs the kernel again
       auto eventKernel2 = Event();
-      status = RunKernel(kernel, global, local, eventKernel2.pointer(), eventWaitList);
+      status = RunKernel(kernel, queue_, device_, global, local, eventKernel2.pointer(), eventWaitList);
       if (ErrorIn(status)) { return status; }
       eventWaitList.push_back(eventKernel2);
 
       // Runs the post-processing kernel
       auto upper = (triangle == Triangle::kUpper);
       auto lower = (triangle == Triangle::kLower);
-      status = PadCopyTransposeMatrix(event_, eventWaitList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, event_, eventWaitList,
                                       n_ceiled, n_ceiled, n_ceiled, 0, c_temp,
                                       n, n, c_ld, c_offset, c_buffer,
                                       ConstantOne<T>(), program,
diff --git a/src/routines/level3/xsyrk.cc b/src/routines/level3/xsyrk.cc
index 000347f3..5c49795b 100644
--- a/src/routines/level3/xsyrk.cc
+++ b/src/routines/level3/xsyrk.cc
@@ -114,7 +114,7 @@ StatusCode Xsyrk<T>::DoSyrk(const Layout layout, const Triangle triangle, const
     // case nothing has to be done, these kernels can be skipped.
     if (!a_no_temp) {
       auto eventProcessA = Event();
-      status = PadCopyTransposeMatrix(eventProcessA.pointer(), emptyEventList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessA.pointer(), emptyEventList,
                                       a_one, a_two, a_ld, a_offset, a_buffer,
                                       n_ceiled, k_ceiled, n_ceiled, 0, a_temp,
                                       ConstantOne<T>(), program,
@@ -126,7 +126,7 @@ StatusCode Xsyrk<T>::DoSyrk(const Layout layout, const Triangle triangle, const
     // Furthermore, also creates a (possibly padded) copy of matrix C, since it is not allowed to
     // modify the other triangle.
     auto eventProcessC = Event();
-    status = PadCopyTransposeMatrix(eventProcessC.pointer(), emptyEventList,
+    status = PadCopyTransposeMatrix(queue_, device_, context_, db_, eventProcessC.pointer(), emptyEventList,
                                     n, n, c_ld, c_offset, c_buffer,
                                     n_ceiled, n_ceiled, n_ceiled, 0, c_temp,
                                     ConstantOne<T>(), program,
@@ -156,14 +156,14 @@ StatusCode Xsyrk<T>::DoSyrk(const Layout layout, const Triangle triangle, const
 
       // Launches the kernel
       auto eventKernel = Event();
-      status = RunKernel(kernel, global, local, eventKernel.pointer(), eventWaitList);
+      status = RunKernel(kernel, queue_, device_, global, local, eventKernel.pointer(), eventWaitList);
       if (ErrorIn(status)) { return status; }
       eventWaitList.push_back(eventKernel);
 
       // Runs the post-processing kernel
       auto upper = (triangle == Triangle::kUpper);
       auto lower = (triangle == Triangle::kLower);
-      status = PadCopyTransposeMatrix(event_, eventWaitList,
+      status = PadCopyTransposeMatrix(queue_, device_, context_, db_, event_, eventWaitList,
                                       n_ceiled, n_ceiled, n_ceiled, 0, c_temp,
                                       n, n, c_ld, c_offset, c_buffer,
                                       ConstantOne<T>(), program,
diff --git a/src/routines/level3/xtrmm.cc b/src/routines/level3/xtrmm.cc
index c62305aa..98e0622d 100644
--- a/src/routines/level3/xtrmm.cc
+++ b/src/routines/level3/xtrmm.cc
@@ -83,7 +83,7 @@ StatusCode Xtrmm<T>::DoTrmm(const Layout layout, const Side side, const Triangle
                                         Ceil(CeilDiv(k, db_["PAD_WPTY"]), db_["PAD_DIMY"])};
       auto local = std::vector<size_t>{db_["PAD_DIMX"], db_["PAD_DIMY"]};
       auto kernelEvent = Event();
-      status = RunKernel(kernel, global, local, kernelEvent.pointer());
+      status = RunKernel(kernel, queue_, device_, global, local, kernelEvent.pointer());
       if (ErrorIn(status)) { return status; }
 
       // Synchronize now: 'DoGemm' does not accept a list of events to wait for