summaryrefslogtreecommitdiff
path: root/test/routines/level3
diff options
context:
space:
mode:
Diffstat (limited to 'test/routines/level3')
-rw-r--r--test/routines/level3/xgemm.hpp26
-rw-r--r--test/routines/level3/xhemm.hpp26
-rw-r--r--test/routines/level3/xher2k.hpp26
-rw-r--r--test/routines/level3/xherk.hpp20
-rw-r--r--test/routines/level3/xsymm.hpp26
-rw-r--r--test/routines/level3/xsyr2k.hpp26
-rw-r--r--test/routines/level3/xsyrk.hpp20
-rw-r--r--test/routines/level3/xtrmm.hpp20
-rw-r--r--test/routines/level3/xtrsm.hpp20
9 files changed, 105 insertions, 105 deletions
diff --git a/test/routines/level3/xgemm.hpp b/test/routines/level3/xgemm.hpp
index 1b12fb1c..5b220889 100644
--- a/test/routines/level3/xgemm.hpp
+++ b/test/routines/level3/xgemm.hpp
@@ -88,14 +88,14 @@ class TestXgemm {
std::vector<T>&, std::vector<T>&) {} // N/A for this routine
// Describes how to run the CLBlast routine
- static StatusCode RunRoutine(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunRoutine(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = Gemm(args.layout, args.a_transpose, args.b_transpose,
args.m, args.n, args.k, args.alpha,
- buffers.a_mat(), args.a_offset, args.a_ld,
- buffers.b_mat(), args.b_offset, args.b_ld, args.beta,
- buffers.c_mat(), args.c_offset, args.c_ld,
+ buffers[0].a_mat(), args.a_offset, args.a_ld,
+ buffers[0].b_mat(), args.b_offset, args.b_ld, args.beta,
+ buffers[0].c_mat(), args.c_offset, args.c_ld,
&queue_plain, &event);
if (status == StatusCode::kSuccess) { clWaitForEvents(1, &event); clReleaseEvent(event); }
return status;
@@ -103,16 +103,16 @@ class TestXgemm {
// Describes how to run the clBLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CLBLAS
- static StatusCode RunReference1(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference1(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = clblasXgemm(convertToCLBLAS(args.layout),
convertToCLBLAS(args.a_transpose),
convertToCLBLAS(args.b_transpose),
args.m, args.n, args.k, args.alpha,
- buffers.a_mat, args.a_offset, args.a_ld,
- buffers.b_mat, args.b_offset, args.b_ld, args.beta,
- buffers.c_mat, args.c_offset, args.c_ld,
+ buffers[0].a_mat, args.a_offset, args.a_ld,
+ buffers[0].b_mat, args.b_offset, args.b_ld, args.beta,
+ buffers[0].c_mat, args.c_offset, args.c_ld,
1, &queue_plain, 0, nullptr, &event);
clWaitForEvents(1, &event);
return static_cast<StatusCode>(status);
@@ -121,13 +121,13 @@ class TestXgemm {
// Describes how to run the CPU BLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CBLAS
- static StatusCode RunReference2(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference2(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
std::vector<T> a_mat_cpu(args.a_size, static_cast<T>(0));
std::vector<T> b_mat_cpu(args.b_size, static_cast<T>(0));
std::vector<T> c_mat_cpu(args.c_size, static_cast<T>(0));
- buffers.a_mat.Read(queue, args.a_size, a_mat_cpu);
- buffers.b_mat.Read(queue, args.b_size, b_mat_cpu);
- buffers.c_mat.Read(queue, args.c_size, c_mat_cpu);
+ buffers[0].a_mat.Read(queue, args.a_size, a_mat_cpu);
+ buffers[0].b_mat.Read(queue, args.b_size, b_mat_cpu);
+ buffers[0].c_mat.Read(queue, args.c_size, c_mat_cpu);
cblasXgemm(convertToCBLAS(args.layout),
convertToCBLAS(args.a_transpose),
convertToCBLAS(args.b_transpose),
@@ -135,7 +135,7 @@ class TestXgemm {
a_mat_cpu, args.a_offset, args.a_ld,
b_mat_cpu, args.b_offset, args.b_ld, args.beta,
c_mat_cpu, args.c_offset, args.c_ld);
- buffers.c_mat.Write(queue, args.c_size, c_mat_cpu);
+ buffers[0].c_mat.Write(queue, args.c_size, c_mat_cpu);
return StatusCode::kSuccess;
}
#endif
diff --git a/test/routines/level3/xhemm.hpp b/test/routines/level3/xhemm.hpp
index 76550b15..e6e8724f 100644
--- a/test/routines/level3/xhemm.hpp
+++ b/test/routines/level3/xhemm.hpp
@@ -88,14 +88,14 @@ class TestXhemm {
std::vector<T>&, std::vector<T>&) {} // N/A for this routine
// Describes how to run the CLBlast routine
- static StatusCode RunRoutine(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunRoutine(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = Hemm(args.layout, args.side, args.triangle,
args.m, args.n, args.alpha,
- buffers.a_mat(), args.a_offset, args.a_ld,
- buffers.b_mat(), args.b_offset, args.b_ld, args.beta,
- buffers.c_mat(), args.c_offset, args.c_ld,
+ buffers[0].a_mat(), args.a_offset, args.a_ld,
+ buffers[0].b_mat(), args.b_offset, args.b_ld, args.beta,
+ buffers[0].c_mat(), args.c_offset, args.c_ld,
&queue_plain, &event);
if (status == StatusCode::kSuccess) { clWaitForEvents(1, &event); clReleaseEvent(event); }
return status;
@@ -103,16 +103,16 @@ class TestXhemm {
// Describes how to run the clBLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CLBLAS
- static StatusCode RunReference1(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference1(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = clblasXhemm(convertToCLBLAS(args.layout),
convertToCLBLAS(args.side),
convertToCLBLAS(args.triangle),
args.m, args.n, args.alpha,
- buffers.a_mat, args.a_offset, args.a_ld,
- buffers.b_mat, args.b_offset, args.b_ld, args.beta,
- buffers.c_mat, args.c_offset, args.c_ld,
+ buffers[0].a_mat, args.a_offset, args.a_ld,
+ buffers[0].b_mat, args.b_offset, args.b_ld, args.beta,
+ buffers[0].c_mat, args.c_offset, args.c_ld,
1, &queue_plain, 0, nullptr, &event);
clWaitForEvents(1, &event);
return static_cast<StatusCode>(status);
@@ -121,13 +121,13 @@ class TestXhemm {
// Describes how to run the CPU BLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CBLAS
- static StatusCode RunReference2(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference2(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
std::vector<T> a_mat_cpu(args.a_size, static_cast<T>(0));
std::vector<T> b_mat_cpu(args.b_size, static_cast<T>(0));
std::vector<T> c_mat_cpu(args.c_size, static_cast<T>(0));
- buffers.a_mat.Read(queue, args.a_size, a_mat_cpu);
- buffers.b_mat.Read(queue, args.b_size, b_mat_cpu);
- buffers.c_mat.Read(queue, args.c_size, c_mat_cpu);
+ buffers[0].a_mat.Read(queue, args.a_size, a_mat_cpu);
+ buffers[0].b_mat.Read(queue, args.b_size, b_mat_cpu);
+ buffers[0].c_mat.Read(queue, args.c_size, c_mat_cpu);
cblasXhemm(convertToCBLAS(args.layout),
convertToCBLAS(args.side),
convertToCBLAS(args.triangle),
@@ -135,7 +135,7 @@ class TestXhemm {
a_mat_cpu, args.a_offset, args.a_ld,
b_mat_cpu, args.b_offset, args.b_ld, args.beta,
c_mat_cpu, args.c_offset, args.c_ld);
- buffers.c_mat.Write(queue, args.c_size, c_mat_cpu);
+ buffers[0].c_mat.Write(queue, args.c_size, c_mat_cpu);
return StatusCode::kSuccess;
}
#endif
diff --git a/test/routines/level3/xher2k.hpp b/test/routines/level3/xher2k.hpp
index 5ca3aac6..749eca11 100644
--- a/test/routines/level3/xher2k.hpp
+++ b/test/routines/level3/xher2k.hpp
@@ -86,15 +86,15 @@ class TestXher2k {
std::vector<T>&, std::vector<T>&) {} // N/A for this routine
// Describes how to run the CLBlast routine
- static StatusCode RunRoutine(const Arguments<U> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunRoutine(const Arguments<U> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto alpha2 = T{args.alpha, args.alpha};
auto status = Her2k(args.layout, args.triangle, args.a_transpose,
args.n, args.k, alpha2,
- buffers.a_mat(), args.a_offset, args.a_ld,
- buffers.b_mat(), args.b_offset, args.b_ld, args.beta,
- buffers.c_mat(), args.c_offset, args.c_ld,
+ buffers[0].a_mat(), args.a_offset, args.a_ld,
+ buffers[0].b_mat(), args.b_offset, args.b_ld, args.beta,
+ buffers[0].c_mat(), args.c_offset, args.c_ld,
&queue_plain, &event);
if (status == StatusCode::kSuccess) { clWaitForEvents(1, &event); clReleaseEvent(event); }
return status;
@@ -102,7 +102,7 @@ class TestXher2k {
// Describes how to run the clBLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CLBLAS
- static StatusCode RunReference1(const Arguments<U> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference1(const Arguments<U> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto alpha2 = T{args.alpha, args.alpha};
@@ -110,9 +110,9 @@ class TestXher2k {
convertToCLBLAS(args.triangle),
convertToCLBLAS(args.a_transpose),
args.n, args.k, alpha2,
- buffers.a_mat, args.a_offset, args.a_ld,
- buffers.b_mat, args.b_offset, args.b_ld, args.beta,
- buffers.c_mat, args.c_offset, args.c_ld,
+ buffers[0].a_mat, args.a_offset, args.a_ld,
+ buffers[0].b_mat, args.b_offset, args.b_ld, args.beta,
+ buffers[0].c_mat, args.c_offset, args.c_ld,
1, &queue_plain, 0, nullptr, &event);
clWaitForEvents(1, &event);
return static_cast<StatusCode>(status);
@@ -121,13 +121,13 @@ class TestXher2k {
// Describes how to run the CPU BLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CBLAS
- static StatusCode RunReference2(const Arguments<U> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference2(const Arguments<U> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
std::vector<T> a_mat_cpu(args.a_size, static_cast<T>(0));
std::vector<T> b_mat_cpu(args.b_size, static_cast<T>(0));
std::vector<T> c_mat_cpu(args.c_size, static_cast<T>(0));
- buffers.a_mat.Read(queue, args.a_size, a_mat_cpu);
- buffers.b_mat.Read(queue, args.b_size, b_mat_cpu);
- buffers.c_mat.Read(queue, args.c_size, c_mat_cpu);
+ buffers[0].a_mat.Read(queue, args.a_size, a_mat_cpu);
+ buffers[0].b_mat.Read(queue, args.b_size, b_mat_cpu);
+ buffers[0].c_mat.Read(queue, args.c_size, c_mat_cpu);
auto alpha2 = T{args.alpha, args.alpha};
cblasXher2k(convertToCBLAS(args.layout),
convertToCBLAS(args.triangle),
@@ -136,7 +136,7 @@ class TestXher2k {
a_mat_cpu, args.a_offset, args.a_ld,
b_mat_cpu, args.b_offset, args.b_ld, args.beta,
c_mat_cpu, args.c_offset, args.c_ld);
- buffers.c_mat.Write(queue, args.c_size, c_mat_cpu);
+ buffers[0].c_mat.Write(queue, args.c_size, c_mat_cpu);
return StatusCode::kSuccess;
}
#endif
diff --git a/test/routines/level3/xherk.hpp b/test/routines/level3/xherk.hpp
index e93d887a..e9193847 100644
--- a/test/routines/level3/xherk.hpp
+++ b/test/routines/level3/xherk.hpp
@@ -79,13 +79,13 @@ class TestXherk {
std::vector<T>&, std::vector<T>&) {} // N/A for this routine
// Describes how to run the CLBlast routine
- static StatusCode RunRoutine(const Arguments<U> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunRoutine(const Arguments<U> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = Herk(args.layout, args.triangle, args.a_transpose,
args.n, args.k, args.alpha,
- buffers.a_mat(), args.a_offset, args.a_ld, args.beta,
- buffers.c_mat(), args.c_offset, args.c_ld,
+ buffers[0].a_mat(), args.a_offset, args.a_ld, args.beta,
+ buffers[0].c_mat(), args.c_offset, args.c_ld,
&queue_plain, &event);
if (status == StatusCode::kSuccess) { clWaitForEvents(1, &event); clReleaseEvent(event); }
return status;
@@ -93,15 +93,15 @@ class TestXherk {
// Describes how to run the clBLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CLBLAS
- static StatusCode RunReference1(const Arguments<U> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference1(const Arguments<U> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = clblasXherk(convertToCLBLAS(args.layout),
convertToCLBLAS(args.triangle),
convertToCLBLAS(args.a_transpose),
args.n, args.k, args.alpha,
- buffers.a_mat, args.a_offset, args.a_ld, args.beta,
- buffers.c_mat, args.c_offset, args.c_ld,
+ buffers[0].a_mat, args.a_offset, args.a_ld, args.beta,
+ buffers[0].c_mat, args.c_offset, args.c_ld,
1, &queue_plain, 0, nullptr, &event);
clWaitForEvents(1, &event);
return static_cast<StatusCode>(status);
@@ -110,18 +110,18 @@ class TestXherk {
// Describes how to run the CPU BLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CBLAS
- static StatusCode RunReference2(const Arguments<U> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference2(const Arguments<U> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
std::vector<T> a_mat_cpu(args.a_size, static_cast<T>(0));
std::vector<T> c_mat_cpu(args.c_size, static_cast<T>(0));
- buffers.a_mat.Read(queue, args.a_size, a_mat_cpu);
- buffers.c_mat.Read(queue, args.c_size, c_mat_cpu);
+ buffers[0].a_mat.Read(queue, args.a_size, a_mat_cpu);
+ buffers[0].c_mat.Read(queue, args.c_size, c_mat_cpu);
cblasXherk(convertToCBLAS(args.layout),
convertToCBLAS(args.triangle),
convertToCBLAS(args.a_transpose),
args.n, args.k, args.alpha,
a_mat_cpu, args.a_offset, args.a_ld, args.beta,
c_mat_cpu, args.c_offset, args.c_ld);
- buffers.c_mat.Write(queue, args.c_size, c_mat_cpu);
+ buffers[0].c_mat.Write(queue, args.c_size, c_mat_cpu);
return StatusCode::kSuccess;
}
#endif
diff --git a/test/routines/level3/xsymm.hpp b/test/routines/level3/xsymm.hpp
index 9d127e26..bcd74fda 100644
--- a/test/routines/level3/xsymm.hpp
+++ b/test/routines/level3/xsymm.hpp
@@ -88,14 +88,14 @@ class TestXsymm {
std::vector<T>&, std::vector<T>&) {} // N/A for this routine
// Describes how to run the CLBlast routine
- static StatusCode RunRoutine(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunRoutine(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = Symm(args.layout, args.side, args.triangle,
args.m, args.n, args.alpha,
- buffers.a_mat(), args.a_offset, args.a_ld,
- buffers.b_mat(), args.b_offset, args.b_ld, args.beta,
- buffers.c_mat(), args.c_offset, args.c_ld,
+ buffers[0].a_mat(), args.a_offset, args.a_ld,
+ buffers[0].b_mat(), args.b_offset, args.b_ld, args.beta,
+ buffers[0].c_mat(), args.c_offset, args.c_ld,
&queue_plain, &event);
if (status == StatusCode::kSuccess) { clWaitForEvents(1, &event); clReleaseEvent(event); }
return status;
@@ -103,16 +103,16 @@ class TestXsymm {
// Describes how to run the clBLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CLBLAS
- static StatusCode RunReference1(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference1(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = clblasXsymm(convertToCLBLAS(args.layout),
convertToCLBLAS(args.side),
convertToCLBLAS(args.triangle),
args.m, args.n, args.alpha,
- buffers.a_mat, args.a_offset, args.a_ld,
- buffers.b_mat, args.b_offset, args.b_ld, args.beta,
- buffers.c_mat, args.c_offset, args.c_ld,
+ buffers[0].a_mat, args.a_offset, args.a_ld,
+ buffers[0].b_mat, args.b_offset, args.b_ld, args.beta,
+ buffers[0].c_mat, args.c_offset, args.c_ld,
1, &queue_plain, 0, nullptr, &event);
clWaitForEvents(1, &event);
return static_cast<StatusCode>(status);
@@ -121,13 +121,13 @@ class TestXsymm {
// Describes how to run the CPU BLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CBLAS
- static StatusCode RunReference2(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference2(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
std::vector<T> a_mat_cpu(args.a_size, static_cast<T>(0));
std::vector<T> b_mat_cpu(args.b_size, static_cast<T>(0));
std::vector<T> c_mat_cpu(args.c_size, static_cast<T>(0));
- buffers.a_mat.Read(queue, args.a_size, a_mat_cpu);
- buffers.b_mat.Read(queue, args.b_size, b_mat_cpu);
- buffers.c_mat.Read(queue, args.c_size, c_mat_cpu);
+ buffers[0].a_mat.Read(queue, args.a_size, a_mat_cpu);
+ buffers[0].b_mat.Read(queue, args.b_size, b_mat_cpu);
+ buffers[0].c_mat.Read(queue, args.c_size, c_mat_cpu);
cblasXsymm(convertToCBLAS(args.layout),
convertToCBLAS(args.side),
convertToCBLAS(args.triangle),
@@ -135,7 +135,7 @@ class TestXsymm {
a_mat_cpu, args.a_offset, args.a_ld,
b_mat_cpu, args.b_offset, args.b_ld, args.beta,
c_mat_cpu, args.c_offset, args.c_ld);
- buffers.c_mat.Write(queue, args.c_size, c_mat_cpu);
+ buffers[0].c_mat.Write(queue, args.c_size, c_mat_cpu);
return StatusCode::kSuccess;
}
#endif
diff --git a/test/routines/level3/xsyr2k.hpp b/test/routines/level3/xsyr2k.hpp
index d1bdac56..c722e0cf 100644
--- a/test/routines/level3/xsyr2k.hpp
+++ b/test/routines/level3/xsyr2k.hpp
@@ -86,14 +86,14 @@ class TestXsyr2k {
std::vector<T>&, std::vector<T>&) {} // N/A for this routine
// Describes how to run the CLBlast routine
- static StatusCode RunRoutine(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunRoutine(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = Syr2k(args.layout, args.triangle, args.a_transpose,
args.n, args.k, args.alpha,
- buffers.a_mat(), args.a_offset, args.a_ld,
- buffers.b_mat(), args.b_offset, args.b_ld, args.beta,
- buffers.c_mat(), args.c_offset, args.c_ld,
+ buffers[0].a_mat(), args.a_offset, args.a_ld,
+ buffers[0].b_mat(), args.b_offset, args.b_ld, args.beta,
+ buffers[0].c_mat(), args.c_offset, args.c_ld,
&queue_plain, &event);
if (status == StatusCode::kSuccess) { clWaitForEvents(1, &event); clReleaseEvent(event); }
return status;
@@ -101,16 +101,16 @@ class TestXsyr2k {
// Describes how to run the clBLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CLBLAS
- static StatusCode RunReference1(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference1(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = clblasXsyr2k(convertToCLBLAS(args.layout),
convertToCLBLAS(args.triangle),
convertToCLBLAS(args.a_transpose),
args.n, args.k, args.alpha,
- buffers.a_mat, args.a_offset, args.a_ld,
- buffers.b_mat, args.b_offset, args.b_ld, args.beta,
- buffers.c_mat, args.c_offset, args.c_ld,
+ buffers[0].a_mat, args.a_offset, args.a_ld,
+ buffers[0].b_mat, args.b_offset, args.b_ld, args.beta,
+ buffers[0].c_mat, args.c_offset, args.c_ld,
1, &queue_plain, 0, nullptr, &event);
clWaitForEvents(1, &event);
return static_cast<StatusCode>(status);
@@ -119,13 +119,13 @@ class TestXsyr2k {
// Describes how to run the CPU BLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CBLAS
- static StatusCode RunReference2(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference2(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
std::vector<T> a_mat_cpu(args.a_size, static_cast<T>(0));
std::vector<T> b_mat_cpu(args.b_size, static_cast<T>(0));
std::vector<T> c_mat_cpu(args.c_size, static_cast<T>(0));
- buffers.a_mat.Read(queue, args.a_size, a_mat_cpu);
- buffers.b_mat.Read(queue, args.b_size, b_mat_cpu);
- buffers.c_mat.Read(queue, args.c_size, c_mat_cpu);
+ buffers[0].a_mat.Read(queue, args.a_size, a_mat_cpu);
+ buffers[0].b_mat.Read(queue, args.b_size, b_mat_cpu);
+ buffers[0].c_mat.Read(queue, args.c_size, c_mat_cpu);
cblasXsyr2k(convertToCBLAS(args.layout),
convertToCBLAS(args.triangle),
convertToCBLAS(args.a_transpose),
@@ -133,7 +133,7 @@ class TestXsyr2k {
a_mat_cpu, args.a_offset, args.a_ld,
b_mat_cpu, args.b_offset, args.b_ld, args.beta,
c_mat_cpu, args.c_offset, args.c_ld);
- buffers.c_mat.Write(queue, args.c_size, c_mat_cpu);
+ buffers[0].c_mat.Write(queue, args.c_size, c_mat_cpu);
return StatusCode::kSuccess;
}
#endif
diff --git a/test/routines/level3/xsyrk.hpp b/test/routines/level3/xsyrk.hpp
index 1330924e..7d5c2039 100644
--- a/test/routines/level3/xsyrk.hpp
+++ b/test/routines/level3/xsyrk.hpp
@@ -79,13 +79,13 @@ class TestXsyrk {
std::vector<T>&, std::vector<T>&) {} // N/A for this routine
// Describes how to run the CLBlast routine
- static StatusCode RunRoutine(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunRoutine(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = Syrk(args.layout, args.triangle, args.a_transpose,
args.n, args.k, args.alpha,
- buffers.a_mat(), args.a_offset, args.a_ld, args.beta,
- buffers.c_mat(), args.c_offset, args.c_ld,
+ buffers[0].a_mat(), args.a_offset, args.a_ld, args.beta,
+ buffers[0].c_mat(), args.c_offset, args.c_ld,
&queue_plain, &event);
if (status == StatusCode::kSuccess) { clWaitForEvents(1, &event); clReleaseEvent(event); }
return status;
@@ -93,15 +93,15 @@ class TestXsyrk {
// Describes how to run the clBLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CLBLAS
- static StatusCode RunReference1(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference1(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = clblasXsyrk(convertToCLBLAS(args.layout),
convertToCLBLAS(args.triangle),
convertToCLBLAS(args.a_transpose),
args.n, args.k, args.alpha,
- buffers.a_mat, args.a_offset, args.a_ld, args.beta,
- buffers.c_mat, args.c_offset, args.c_ld,
+ buffers[0].a_mat, args.a_offset, args.a_ld, args.beta,
+ buffers[0].c_mat, args.c_offset, args.c_ld,
1, &queue_plain, 0, nullptr, &event);
clWaitForEvents(1, &event);
return static_cast<StatusCode>(status);
@@ -110,18 +110,18 @@ class TestXsyrk {
// Describes how to run the CPU BLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CBLAS
- static StatusCode RunReference2(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference2(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
std::vector<T> a_mat_cpu(args.a_size, static_cast<T>(0));
std::vector<T> c_mat_cpu(args.c_size, static_cast<T>(0));
- buffers.a_mat.Read(queue, args.a_size, a_mat_cpu);
- buffers.c_mat.Read(queue, args.c_size, c_mat_cpu);
+ buffers[0].a_mat.Read(queue, args.a_size, a_mat_cpu);
+ buffers[0].c_mat.Read(queue, args.c_size, c_mat_cpu);
cblasXsyrk(convertToCBLAS(args.layout),
convertToCBLAS(args.triangle),
convertToCBLAS(args.a_transpose),
args.n, args.k, args.alpha,
a_mat_cpu, args.a_offset, args.a_ld, args.beta,
c_mat_cpu, args.c_offset, args.c_ld);
- buffers.c_mat.Write(queue, args.c_size, c_mat_cpu);
+ buffers[0].c_mat.Write(queue, args.c_size, c_mat_cpu);
return StatusCode::kSuccess;
}
#endif
diff --git a/test/routines/level3/xtrmm.hpp b/test/routines/level3/xtrmm.hpp
index 7c5bd842..50cca6f8 100644
--- a/test/routines/level3/xtrmm.hpp
+++ b/test/routines/level3/xtrmm.hpp
@@ -79,13 +79,13 @@ class TestXtrmm {
std::vector<T>&, std::vector<T>&) {} // N/A for this routine
// Describes how to run the CLBlast routine
- static StatusCode RunRoutine(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunRoutine(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = Trmm(args.layout, args.side, args.triangle, args.a_transpose, args.diagonal,
args.m, args.n, args.alpha,
- buffers.a_mat(), args.a_offset, args.a_ld,
- buffers.b_mat(), args.b_offset, args.b_ld,
+ buffers[0].a_mat(), args.a_offset, args.a_ld,
+ buffers[0].b_mat(), args.b_offset, args.b_ld,
&queue_plain, &event);
if (status == StatusCode::kSuccess) { clWaitForEvents(1, &event); clReleaseEvent(event); }
return status;
@@ -93,7 +93,7 @@ class TestXtrmm {
// Describes how to run the clBLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CLBLAS
- static StatusCode RunReference1(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference1(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = clblasXtrmm(convertToCLBLAS(args.layout),
@@ -102,8 +102,8 @@ class TestXtrmm {
convertToCLBLAS(args.a_transpose),
convertToCLBLAS(args.diagonal),
args.m, args.n, args.alpha,
- buffers.a_mat, args.a_offset, args.a_ld,
- buffers.b_mat, args.b_offset, args.b_ld,
+ buffers[0].a_mat, args.a_offset, args.a_ld,
+ buffers[0].b_mat, args.b_offset, args.b_ld,
1, &queue_plain, 0, nullptr, &event);
clWaitForEvents(1, &event);
return static_cast<StatusCode>(status);
@@ -112,11 +112,11 @@ class TestXtrmm {
// Describes how to run the CPU BLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CBLAS
- static StatusCode RunReference2(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference2(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
std::vector<T> a_mat_cpu(args.a_size, static_cast<T>(0));
std::vector<T> b_mat_cpu(args.b_size, static_cast<T>(0));
- buffers.a_mat.Read(queue, args.a_size, a_mat_cpu);
- buffers.b_mat.Read(queue, args.b_size, b_mat_cpu);
+ buffers[0].a_mat.Read(queue, args.a_size, a_mat_cpu);
+ buffers[0].b_mat.Read(queue, args.b_size, b_mat_cpu);
cblasXtrmm(convertToCBLAS(args.layout),
convertToCBLAS(args.side),
convertToCBLAS(args.triangle),
@@ -125,7 +125,7 @@ class TestXtrmm {
args.m, args.n, args.alpha,
a_mat_cpu, args.a_offset, args.a_ld,
b_mat_cpu, args.b_offset, args.b_ld);
- buffers.b_mat.Write(queue, args.b_size, b_mat_cpu);
+ buffers[0].b_mat.Write(queue, args.b_size, b_mat_cpu);
return StatusCode::kSuccess;
}
#endif
diff --git a/test/routines/level3/xtrsm.hpp b/test/routines/level3/xtrsm.hpp
index a70ef03f..91f91d0b 100644
--- a/test/routines/level3/xtrsm.hpp
+++ b/test/routines/level3/xtrsm.hpp
@@ -91,13 +91,13 @@ class TestXtrsm {
}
// Describes how to run the CLBlast routine
- static StatusCode RunRoutine(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunRoutine(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = Trsm(args.layout, args.side, args.triangle, args.a_transpose, args.diagonal,
args.m, args.n, args.alpha,
- buffers.a_mat(), args.a_offset, args.a_ld,
- buffers.b_mat(), args.b_offset, args.b_ld,
+ buffers[0].a_mat(), args.a_offset, args.a_ld,
+ buffers[0].b_mat(), args.b_offset, args.b_ld,
&queue_plain, &event);
if (status == StatusCode::kSuccess) { clWaitForEvents(1, &event); clReleaseEvent(event); }
return status;
@@ -105,7 +105,7 @@ class TestXtrsm {
// Describes how to run the clBLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CLBLAS
- static StatusCode RunReference1(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference1(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
auto queue_plain = queue();
auto event = cl_event{};
auto status = clblasXtrsm(convertToCLBLAS(args.layout),
@@ -114,8 +114,8 @@ class TestXtrsm {
convertToCLBLAS(args.a_transpose),
convertToCLBLAS(args.diagonal),
args.m, args.n, args.alpha,
- buffers.a_mat, args.a_offset, args.a_ld,
- buffers.b_mat, args.b_offset, args.b_ld,
+ buffers[0].a_mat, args.a_offset, args.a_ld,
+ buffers[0].b_mat, args.b_offset, args.b_ld,
1, &queue_plain, 0, nullptr, &event);
clWaitForEvents(1, &event);
return static_cast<StatusCode>(status);
@@ -124,11 +124,11 @@ class TestXtrsm {
// Describes how to run the CPU BLAS routine (for correctness/performance comparison)
#ifdef CLBLAST_REF_CBLAS
- static StatusCode RunReference2(const Arguments<T> &args, Buffers<T> &buffers, Queue &queue) {
+ static StatusCode RunReference2(const Arguments<T> &args, std::vector<Buffers<T>> &buffers, Queue &queue) {
std::vector<T> a_mat_cpu(args.a_size, static_cast<T>(0));
std::vector<T> b_mat_cpu(args.b_size, static_cast<T>(0));
- buffers.a_mat.Read(queue, args.a_size, a_mat_cpu);
- buffers.b_mat.Read(queue, args.b_size, b_mat_cpu);
+ buffers[0].a_mat.Read(queue, args.a_size, a_mat_cpu);
+ buffers[0].b_mat.Read(queue, args.b_size, b_mat_cpu);
cblasXtrsm(convertToCBLAS(args.layout),
convertToCBLAS(args.side),
convertToCBLAS(args.triangle),
@@ -137,7 +137,7 @@ class TestXtrsm {
args.m, args.n, args.alpha,
a_mat_cpu, args.a_offset, args.a_ld,
b_mat_cpu, args.b_offset, args.b_ld);
- buffers.b_mat.Write(queue, args.b_size, b_mat_cpu);
+ buffers[0].b_mat.Write(queue, args.b_size, b_mat_cpu);
return StatusCode::kSuccess;
}
#endif
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-05 18:41:00 +0000