summaryrefslogtreecommitdiff
path: root/src/routines
diff options
context:
space:
mode:
Diffstat (limited to 'src/routines')
-rw-r--r--src/routines/common.hpp8
-rw-r--r--src/routines/level1/xaxpy.cpp8
-rw-r--r--src/routines/level2/xgemv.cpp10
-rw-r--r--src/routines/level2/xger.cpp6
-rw-r--r--src/routines/level2/xher.cpp6
-rw-r--r--src/routines/level2/xher2.cpp6
-rw-r--r--src/routines/level3/xgemm.cpp10
-rw-r--r--src/routines/level3/xher2k.cpp16
-rw-r--r--src/routines/level3/xherk.cpp10
-rw-r--r--src/routines/level3/xsyr2k.cpp13
-rw-r--r--src/routines/level3/xsyrk.cpp10
11 files changed, 24 insertions, 79 deletions
diff --git a/src/routines/common.hpp b/src/routines/common.hpp
index c99cd39d..e624a2b1 100644
--- a/src/routines/common.hpp
+++ b/src/routines/common.hpp
@@ -88,10 +88,6 @@ StatusCode PadCopyTransposeMatrix(Queue &queue, const Device &device, const Cont
}
}
- // Upload the scalar argument as a constant buffer to the device (needed for half-precision)
- auto alpha_buffer = Buffer<T>(context, 1);
- alpha_buffer.Write(queue, 1, &alpha);
-
// Retrieves the kernel from the compiled binary
try {
auto kernel = Kernel(program, kernel_name);
@@ -101,7 +97,7 @@ StatusCode PadCopyTransposeMatrix(Queue &queue, const Device &device, const Cont
kernel.SetArgument(0, static_cast<int>(src_ld));
kernel.SetArgument(1, src());
kernel.SetArgument(2, dest());
- kernel.SetArgument(3, alpha_buffer());
+ kernel.SetArgument(3, GetRealArg(alpha));
}
else {
kernel.SetArgument(0, static_cast<int>(src_one));
@@ -114,7 +110,7 @@ StatusCode PadCopyTransposeMatrix(Queue &queue, const Device &device, const Cont
kernel.SetArgument(7, static_cast<int>(dest_ld));
kernel.SetArgument(8, static_cast<int>(dest_offset));
kernel.SetArgument(9, dest());
- kernel.SetArgument(10, alpha_buffer());
+ kernel.SetArgument(10, GetRealArg(alpha));
if (do_pad) {
kernel.SetArgument(11, static_cast<int>(do_conjugate));
}
diff --git a/src/routines/level1/xaxpy.cpp b/src/routines/level1/xaxpy.cpp
index 5b6c9e77..3445e2b5 100644
--- a/src/routines/level1/xaxpy.cpp
+++ b/src/routines/level1/xaxpy.cpp
@@ -59,20 +59,16 @@ StatusCode Xaxpy<T>::DoAxpy(const size_t n, const T alpha,
const auto program = GetProgramFromCache(context_, PrecisionValue<T>(), routine_name_);
auto kernel = Kernel(program, kernel_name);
- // Upload the scalar argument as a constant buffer to the device (needed for half-precision)
- auto alpha_buffer = Buffer<T>(context_, 1);
- alpha_buffer.Write(queue_, 1, &alpha);
-
// Sets the kernel arguments
if (use_fast_kernel) {
kernel.SetArgument(0, static_cast<int>(n));
- kernel.SetArgument(1, alpha_buffer());
+ kernel.SetArgument(1, GetRealArg(alpha));
kernel.SetArgument(2, x_buffer());
kernel.SetArgument(3, y_buffer());
}
else {
kernel.SetArgument(0, static_cast<int>(n));
- kernel.SetArgument(1, alpha_buffer());
+ kernel.SetArgument(1, GetRealArg(alpha));
kernel.SetArgument(2, x_buffer());
kernel.SetArgument(3, static_cast<int>(x_offset));
kernel.SetArgument(4, static_cast<int>(x_inc));
diff --git a/src/routines/level2/xgemv.cpp b/src/routines/level2/xgemv.cpp
index 21fb397c..2842ef07 100644
--- a/src/routines/level2/xgemv.cpp
+++ b/src/routines/level2/xgemv.cpp
@@ -126,12 +126,6 @@ StatusCode Xgemv<T>::MatVec(const Layout layout, const Transpose a_transpose,
local_size = db_["WGS3"];
}
- // Upload the scalar arguments as constant buffers to the device (needed for half-precision)
- auto alpha_buffer = Buffer<T>(context_, 1);
- auto beta_buffer = Buffer<T>(context_, 1);
- alpha_buffer.Write(queue_, 1, &alpha);
- beta_buffer.Write(queue_, 1, &beta);
-
// Retrieves the Xgemv kernel from the compiled binary
try {
const auto program = GetProgramFromCache(context_, PrecisionValue<T>(), routine_name_);
@@ -140,8 +134,8 @@ StatusCode Xgemv<T>::MatVec(const Layout layout, const Transpose a_transpose,
// Sets the kernel arguments
kernel.SetArgument(0, static_cast<int>(m_real));
kernel.SetArgument(1, static_cast<int>(n_real));
- kernel.SetArgument(2, alpha_buffer());
- kernel.SetArgument(3, beta_buffer());
+ kernel.SetArgument(2, GetRealArg(alpha));
+ kernel.SetArgument(3, GetRealArg(beta));
kernel.SetArgument(4, static_cast<int>(a_rotated));
kernel.SetArgument(5, a_buffer());
kernel.SetArgument(6, static_cast<int>(a_offset));
diff --git a/src/routines/level2/xger.cpp b/src/routines/level2/xger.cpp
index 353047d2..29cffe0c 100644
--- a/src/routines/level2/xger.cpp
+++ b/src/routines/level2/xger.cpp
@@ -56,10 +56,6 @@ StatusCode Xger<T>::DoGer(const Layout layout,
status = TestVectorY(n, y_buffer, y_offset, y_inc);
if (ErrorIn(status)) { return status; }
- // Upload the scalar argument as a constant buffer to the device (needed for half-precision)
- auto alpha_buffer = Buffer<T>(context_, 1);
- alpha_buffer.Write(queue_, 1, &alpha);
-
// Retrieves the kernel from the compiled binary
try {
const auto program = GetProgramFromCache(context_, PrecisionValue<T>(), routine_name_);
@@ -68,7 +64,7 @@ StatusCode Xger<T>::DoGer(const Layout layout,
// Sets the kernel arguments
kernel.SetArgument(0, static_cast<int>(a_one));
kernel.SetArgument(1, static_cast<int>(a_two));
- kernel.SetArgument(2, alpha_buffer());
+ kernel.SetArgument(2, GetRealArg(alpha));
kernel.SetArgument(3, x_buffer());
kernel.SetArgument(4, static_cast<int>(x_offset));
kernel.SetArgument(5, static_cast<int>(x_inc));
diff --git a/src/routines/level2/xher.cpp b/src/routines/level2/xher.cpp
index ed8ba9e9..6dd95938 100644
--- a/src/routines/level2/xher.cpp
+++ b/src/routines/level2/xher.cpp
@@ -70,10 +70,6 @@ StatusCode Xher<T,U>::DoHer(const Layout layout, const Triangle triangle,
// Creates a matching version of alpha
const auto matching_alpha = GetAlpha(alpha);
- // Upload the scalar argument as a constant buffer to the device (needed for half-precision)
- auto alpha_buffer = Buffer<T>(context_, 1);
- alpha_buffer.Write(queue_, 1, &matching_alpha);
-
// Retrieves the kernel from the compiled binary
try {
const auto program = GetProgramFromCache(context_, PrecisionValue<T>(), routine_name_);
@@ -81,7 +77,7 @@ StatusCode Xher<T,U>::DoHer(const Layout layout, const Triangle triangle,
// Sets the kernel arguments
kernel.SetArgument(0, static_cast<int>(n));
- kernel.SetArgument(1, alpha_buffer());
+ kernel.SetArgument(1, GetRealArg(matching_alpha));
kernel.SetArgument(2, x_buffer());
kernel.SetArgument(3, static_cast<int>(x_offset));
kernel.SetArgument(4, static_cast<int>(x_inc));
diff --git a/src/routines/level2/xher2.cpp b/src/routines/level2/xher2.cpp
index 50572cea..3d57a9b9 100644
--- a/src/routines/level2/xher2.cpp
+++ b/src/routines/level2/xher2.cpp
@@ -58,10 +58,6 @@ StatusCode Xher2<T>::DoHer2(const Layout layout, const Triangle triangle,
status = TestVectorY(n, y_buffer, y_offset, y_inc);
if (ErrorIn(status)) { return status; }
- // Upload the scalar argument as a constant buffer to the device (needed for half-precision)
- auto alpha_buffer = Buffer<T>(context_, 1);
- alpha_buffer.Write(queue_, 1, &alpha);
-
// Retrieves the kernel from the compiled binary
try {
const auto program = GetProgramFromCache(context_, PrecisionValue<T>(), routine_name_);
@@ -69,7 +65,7 @@ StatusCode Xher2<T>::DoHer2(const Layout layout, const Triangle triangle,
// Sets the kernel arguments
kernel.SetArgument(0, static_cast<int>(n));
- kernel.SetArgument(1, alpha_buffer());
+ kernel.SetArgument(1, GetRealArg(alpha));
kernel.SetArgument(2, x_buffer());
kernel.SetArgument(3, static_cast<int>(x_offset));
kernel.SetArgument(4, static_cast<int>(x_inc));
diff --git a/src/routines/level3/xgemm.cpp b/src/routines/level3/xgemm.cpp
index 9ea5559c..97e8db7e 100644
--- a/src/routines/level3/xgemm.cpp
+++ b/src/routines/level3/xgemm.cpp
@@ -118,12 +118,6 @@ StatusCode Xgemm<T>::DoGemm(const Layout layout,
const auto b_temp = (b_no_temp) ? b_buffer : Buffer<T>(context_, k_ceiled*n_ceiled);
const auto c_temp = (c_no_temp) ? c_buffer : Buffer<T>(context_, m_ceiled*n_ceiled);
- // Upload the scalar arguments as constant buffers to the device (needed for half-precision)
- auto alpha_buffer = Buffer<T>(context_, 1);
- auto beta_buffer = Buffer<T>(context_, 1);
- alpha_buffer.Write(queue_, 1, &alpha);
- beta_buffer.Write(queue_, 1, &beta);
-
// Events of all kernels (including pre/post processing kernels)
auto eventWaitList = std::vector<Event>();
auto emptyEventList = std::vector<Event>();
@@ -174,8 +168,8 @@ StatusCode Xgemm<T>::DoGemm(const Layout layout,
kernel.SetArgument(0, static_cast<int>(m_ceiled));
kernel.SetArgument(1, static_cast<int>(n_ceiled));
kernel.SetArgument(2, static_cast<int>(k_ceiled));
- kernel.SetArgument(3, alpha_buffer());
- kernel.SetArgument(4, beta_buffer());
+ kernel.SetArgument(3, GetRealArg(alpha));
+ kernel.SetArgument(4, GetRealArg(beta));
kernel.SetArgument(5, a_temp());
kernel.SetArgument(6, b_temp());
kernel.SetArgument(7, c_temp());
diff --git a/src/routines/level3/xher2k.cpp b/src/routines/level3/xher2k.cpp
index bd7a053e..65e2be55 100644
--- a/src/routines/level3/xher2k.cpp
+++ b/src/routines/level3/xher2k.cpp
@@ -107,12 +107,8 @@ StatusCode Xher2k<T,U>::DoHer2k(const Layout layout, const Triangle triangle, co
auto b2_temp = (b2_no_temp) ? b_buffer : Buffer<T>(context_, k_ceiled*n_ceiled);
auto c_temp = Buffer<T>(context_, n_ceiled*n_ceiled);
- // Upload the scalar arguments as constant buffers to the device (needed for half-precision)
+ // Convert the arguments to complex versions
auto complex_beta = T{beta, static_cast<U>(0.0)};
- auto alpha_buffer = Buffer<T>(context_, 1);
- auto beta_buffer = Buffer<T>(context_, 1);
- alpha_buffer.Write(queue_, 1, &alpha);
- beta_buffer.Write(queue_, 1, &complex_beta);
// Events of all kernels (including pre/post processing kernels)
auto eventWaitList = std::vector<Event>();
@@ -180,8 +176,8 @@ StatusCode Xher2k<T,U>::DoHer2k(const Layout layout, const Triangle triangle, co
// Sets the kernel arguments
kernel.SetArgument(0, static_cast<int>(n_ceiled));
kernel.SetArgument(1, static_cast<int>(k_ceiled));
- kernel.SetArgument(2, alpha_buffer());
- kernel.SetArgument(3, beta_buffer());
+ kernel.SetArgument(2, GetRealArg(alpha));
+ kernel.SetArgument(3, GetRealArg(complex_beta));
kernel.SetArgument(4, a1_temp());
kernel.SetArgument(5, b2_temp());
kernel.SetArgument(6, c_temp());
@@ -202,10 +198,8 @@ StatusCode Xher2k<T,U>::DoHer2k(const Layout layout, const Triangle triangle, co
// Swaps the arguments for matrices A and B, sets 'beta' to 1, and conjugate alpha
auto conjugate_alpha = T{alpha.real(), -alpha.imag()};
auto complex_one = T{static_cast<U>(1.0), static_cast<U>(0.0)};
- alpha_buffer.Write(queue_, 1, &conjugate_alpha);
- beta_buffer.Write(queue_, 1, &complex_one);
- kernel.SetArgument(2, alpha_buffer());
- kernel.SetArgument(3, beta_buffer());
+ kernel.SetArgument(2, GetRealArg(conjugate_alpha));
+ kernel.SetArgument(3, GetRealArg(complex_one));
kernel.SetArgument(4, b1_temp());
kernel.SetArgument(5, a2_temp());
diff --git a/src/routines/level3/xherk.cpp b/src/routines/level3/xherk.cpp
index 6ef7f21f..cc87e3e9 100644
--- a/src/routines/level3/xherk.cpp
+++ b/src/routines/level3/xherk.cpp
@@ -98,13 +98,9 @@ StatusCode Xherk<T,U>::DoHerk(const Layout layout, const Triangle triangle, cons
auto b_temp = (b_no_temp) ? a_buffer : Buffer<T>(context_, k_ceiled*n_ceiled);
auto c_temp = Buffer<T>(context_, n_ceiled*n_ceiled);
- // Upload the scalar arguments as constant buffers to the device (needed for half-precision)
+ // Convert the arguments to complex versions
auto complex_alpha = T{alpha, static_cast<U>(0.0)};
auto complex_beta = T{beta, static_cast<U>(0.0)};
- auto alpha_buffer = Buffer<T>(context_, 1);
- auto beta_buffer = Buffer<T>(context_, 1);
- alpha_buffer.Write(queue_, 1, &complex_alpha);
- beta_buffer.Write(queue_, 1, &complex_beta);
// Events of all kernels (including pre/post processing kernels)
auto eventWaitList = std::vector<Event>();
@@ -152,8 +148,8 @@ StatusCode Xherk<T,U>::DoHerk(const Layout layout, const Triangle triangle, cons
// Sets the kernel arguments
kernel.SetArgument(0, static_cast<int>(n_ceiled));
kernel.SetArgument(1, static_cast<int>(k_ceiled));
- kernel.SetArgument(2, alpha_buffer());
- kernel.SetArgument(3, beta_buffer());
+ kernel.SetArgument(2, GetRealArg(complex_alpha));
+ kernel.SetArgument(3, GetRealArg(complex_beta));
kernel.SetArgument(4, a_temp());
kernel.SetArgument(5, b_temp());
kernel.SetArgument(6, c_temp());
diff --git a/src/routines/level3/xsyr2k.cpp b/src/routines/level3/xsyr2k.cpp
index 424d4d2d..18a1eac7 100644
--- a/src/routines/level3/xsyr2k.cpp
+++ b/src/routines/level3/xsyr2k.cpp
@@ -97,12 +97,6 @@ StatusCode Xsyr2k<T>::DoSyr2k(const Layout layout, const Triangle triangle, cons
auto b_temp = (b_no_temp) ? b_buffer : Buffer<T>(context_, k_ceiled*n_ceiled);
auto c_temp = Buffer<T>(context_, n_ceiled*n_ceiled);
- // Upload the scalar arguments as constant buffers to the device (needed for half-precision)
- auto alpha_buffer = Buffer<T>(context_, 1);
- auto beta_buffer = Buffer<T>(context_, 1);
- alpha_buffer.Write(queue_, 1, &alpha);
- beta_buffer.Write(queue_, 1, &beta);
-
// Events of all kernels (including pre/post processing kernels)
auto eventWaitList = std::vector<Event>();
auto emptyEventList = std::vector<Event>();
@@ -149,8 +143,8 @@ StatusCode Xsyr2k<T>::DoSyr2k(const Layout layout, const Triangle triangle, cons
// Sets the kernel arguments
kernel.SetArgument(0, static_cast<int>(n_ceiled));
kernel.SetArgument(1, static_cast<int>(k_ceiled));
- kernel.SetArgument(2, alpha_buffer());
- kernel.SetArgument(3, beta_buffer());
+ kernel.SetArgument(2, GetRealArg(alpha));
+ kernel.SetArgument(3, GetRealArg(beta));
kernel.SetArgument(4, a_temp());
kernel.SetArgument(5, b_temp());
kernel.SetArgument(6, c_temp());
@@ -170,8 +164,7 @@ StatusCode Xsyr2k<T>::DoSyr2k(const Layout layout, const Triangle triangle, cons
// Swaps the arguments for matrices A and B, and sets 'beta' to 1
auto one = static_cast<T>(1);
- beta_buffer.Write(queue_, 1, &one);
- kernel.SetArgument(3, beta_buffer());
+ kernel.SetArgument(3, GetRealArg(one));
kernel.SetArgument(4, b_temp());
kernel.SetArgument(5, a_temp());
diff --git a/src/routines/level3/xsyrk.cpp b/src/routines/level3/xsyrk.cpp
index f56c232b..1992cec1 100644
--- a/src/routines/level3/xsyrk.cpp
+++ b/src/routines/level3/xsyrk.cpp
@@ -90,12 +90,6 @@ StatusCode Xsyrk<T>::DoSyrk(const Layout layout, const Triangle triangle, const
auto a_temp = (a_no_temp) ? a_buffer : Buffer<T>(context_, k_ceiled*n_ceiled);
auto c_temp = Buffer<T>(context_, n_ceiled*n_ceiled);
- // Upload the scalar arguments as constant buffers to the device (needed for half-precision)
- auto alpha_buffer = Buffer<T>(context_, 1);
- auto beta_buffer = Buffer<T>(context_, 1);
- alpha_buffer.Write(queue_, 1, &alpha);
- beta_buffer.Write(queue_, 1, &beta);
-
// Events of all kernels (including pre/post processing kernels)
auto eventWaitList = std::vector<Event>();
auto emptyEventList = std::vector<Event>();
@@ -132,8 +126,8 @@ StatusCode Xsyrk<T>::DoSyrk(const Layout layout, const Triangle triangle, const
// Sets the kernel arguments
kernel.SetArgument(0, static_cast<int>(n_ceiled));
kernel.SetArgument(1, static_cast<int>(k_ceiled));
- kernel.SetArgument(2, alpha_buffer());
- kernel.SetArgument(3, beta_buffer());
+ kernel.SetArgument(2, GetRealArg(alpha));
+ kernel.SetArgument(3, GetRealArg(beta));
kernel.SetArgument(4, a_temp());
kernel.SetArgument(5, a_temp());
kernel.SetArgument(6, c_temp());
generated by cgit v1.2.3 (git 2.39.1) at 2024-08-01 20:57:37 +0000