haraldwolff/CLBlast
1
0
Fork 0
mirror of https://github.com/CNugteren/CLBlast.git synced 2024-07-07 12:23:46 +02:00
Code Issues Packages Projects Releases Wiki Activity

Added TRMM and TRSM clBLAS wrappers

Browse source
This commit is contained in:
CNugteren 2015-06-30 07:19:46 +02:00
parent cbf2eef179
commit 14186af590
1 changed files with 152 additions and 32 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

184
test/wrapper_clblas.h
View file

@ -76,33 +76,33 @@ clblasStatus clblasXaxpy(
// Calls {clblasSgemv, clblasDgemv, clblasCgemv, clblasZgemv} with the arguments forwarded.
clblasStatus clblasXgemv(
clblasOrder layout, clblasTranspose tran_a, size_t m, size_t n, float alpha,
clblasOrder layout, clblasTranspose a_transpose, size_t m, size_t n, float alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem x_vec, size_t x_offset, size_t x_inc, float beta,
const cl_mem y_vec, size_t y_offset, size_t y_inc,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasSgemv(layout, tran_a, m, n, alpha,
return clblasSgemv(layout, a_transpose, m, n, alpha,
a_mat, a_offset, a_ld,
x_vec, x_offset, static_cast<int>(x_inc), beta,
y_vec, y_offset, static_cast<int>(y_inc),
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXgemv(
clblasOrder layout, clblasTranspose tran_a, size_t m, size_t n, double alpha,
clblasOrder layout, clblasTranspose a_transpose, size_t m, size_t n, double alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem x_vec, size_t x_offset, size_t x_inc, double beta,
const cl_mem y_vec, size_t y_offset, size_t y_inc,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasDgemv(layout, tran_a, m, n, alpha,
return clblasDgemv(layout, a_transpose, m, n, alpha,
a_mat, a_offset, a_ld,
x_vec, x_offset, static_cast<int>(x_inc), beta,
y_vec, y_offset, static_cast<int>(y_inc),
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXgemv(
clblasOrder layout, clblasTranspose tran_a, size_t m, size_t n, float2 alpha,
clblasOrder layout, clblasTranspose a_transpose, size_t m, size_t n, float2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem x_vec, size_t x_offset, size_t x_inc, float2 beta,
const cl_mem y_vec, size_t y_offset, size_t y_inc,
@ -110,14 +110,14 @@ clblasStatus clblasXgemv(
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_float2{{alpha.real(), alpha.imag()}};
auto cl_beta = cl_float2{{beta.real(), beta.imag()}};
return clblasCgemv(layout, tran_a, m, n, cl_alpha,
return clblasCgemv(layout, a_transpose, m, n, cl_alpha,
a_mat, a_offset, a_ld,
x_vec, x_offset, static_cast<int>(x_inc), cl_beta,
y_vec, y_offset, static_cast<int>(y_inc),
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXgemv(
clblasOrder layout, clblasTranspose tran_a, size_t m, size_t n, double2 alpha,
clblasOrder layout, clblasTranspose a_transpose, size_t m, size_t n, double2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem x_vec, size_t x_offset, size_t x_inc, double2 beta,
const cl_mem y_vec, size_t y_offset, size_t y_inc,
@ -125,7 +125,7 @@ clblasStatus clblasXgemv(
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_double2{{alpha.real(), alpha.imag()}};
auto cl_beta = cl_double2{{beta.real(), beta.imag()}};
return clblasZgemv(layout, tran_a, m, n, cl_alpha,
return clblasZgemv(layout, a_transpose, m, n, cl_alpha,
a_mat, a_offset, a_ld,
x_vec, x_offset, static_cast<int>(x_inc), cl_beta,
y_vec, y_offset, static_cast<int>(y_inc),
@ -137,14 +137,14 @@ clblasStatus clblasXgemv(
// This calls {clblasSgemm, clblasDgemm, clblasCgemm, clblasZgemm} with the arguments forwarded.
clblasStatus clblasXgemm(
clblasOrder layout, clblasTranspose tran_a, clblasTranspose tran_b,
clblasOrder layout, clblasTranspose a_transpose, clblasTranspose b_transpose,
size_t m, size_t n, size_t k, float alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, float beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasSgemm(layout, tran_a, tran_b,
return clblasSgemm(layout, a_transpose, b_transpose,
m, n, k, alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, beta,
@ -152,14 +152,14 @@ clblasStatus clblasXgemm(
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXgemm(
clblasOrder layout, clblasTranspose tran_a, clblasTranspose tran_b,
clblasOrder layout, clblasTranspose a_transpose, clblasTranspose b_transpose,
size_t m, size_t n, size_t k, double alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, double beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasDgemm(layout, tran_a, tran_b,
return clblasDgemm(layout, a_transpose, b_transpose,
m, n, k, alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, beta,
@ -167,7 +167,7 @@ clblasStatus clblasXgemm(
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXgemm(
clblasOrder layout, clblasTranspose tran_a, clblasTranspose tran_b,
clblasOrder layout, clblasTranspose a_transpose, clblasTranspose b_transpose,
size_t m, size_t n, size_t k, float2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, float2 beta,
@ -176,7 +176,7 @@ clblasStatus clblasXgemm(
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_float2{{alpha.real(), alpha.imag()}};
auto cl_beta = cl_float2{{beta.real(), beta.imag()}};
return clblasCgemm(layout, tran_a, tran_b,
return clblasCgemm(layout, a_transpose, b_transpose,
m, n, k, cl_alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, cl_beta,
@ -184,7 +184,7 @@ clblasStatus clblasXgemm(
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXgemm(
clblasOrder layout, clblasTranspose tran_a, clblasTranspose tran_b,
clblasOrder layout, clblasTranspose a_transpose, clblasTranspose b_transpose,
size_t m, size_t n, size_t k, double2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, double2 beta,
@ -193,7 +193,7 @@ clblasStatus clblasXgemm(
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_double2{{alpha.real(), alpha.imag()}};
auto cl_beta = cl_double2{{beta.real(), beta.imag()}};
return clblasZgemm(layout, tran_a, tran_b,
return clblasZgemm(layout, a_transpose, b_transpose,
m, n, k, cl_alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, cl_beta,
@ -269,33 +269,33 @@ clblasStatus clblasXsymm(
// This calls {clblasSsyrk, clblasDsyrk, clblasCsyrk, clblasZsyrk} with the arguments forwarded.
clblasStatus clblasXsyrk(
clblasOrder layout, clblasUplo triangle, clblasTranspose tran_a,
clblasOrder layout, clblasUplo triangle, clblasTranspose a_transpose,
size_t n, size_t k, float alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld, float beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasSsyrk(layout, triangle, tran_a,
return clblasSsyrk(layout, triangle, a_transpose,
n, k, alpha,
a_mat, a_offset, a_ld, beta,
c_mat, c_offset, c_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXsyrk(
clblasOrder layout, clblasUplo triangle, clblasTranspose tran_a,
clblasOrder layout, clblasUplo triangle, clblasTranspose a_transpose,
size_t n, size_t k, double alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld, double beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasDsyrk(layout, triangle, tran_a,
return clblasDsyrk(layout, triangle, a_transpose,
n, k, alpha,
a_mat, a_offset, a_ld, beta,
c_mat, c_offset, c_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXsyrk(
clblasOrder layout, clblasUplo triangle, clblasTranspose tran_a,
clblasOrder layout, clblasUplo triangle, clblasTranspose a_transpose,
size_t n, size_t k, float2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld, float2 beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
@ -303,14 +303,14 @@ clblasStatus clblasXsyrk(
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_float2{{alpha.real(), alpha.imag()}};
auto cl_beta = cl_float2{{beta.real(), beta.imag()}};
return clblasCsyrk(layout, triangle, tran_a,
return clblasCsyrk(layout, triangle, a_transpose,
n, k, cl_alpha,
a_mat, a_offset, a_ld, cl_beta,
c_mat, c_offset, c_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXsyrk(
clblasOrder layout, clblasUplo triangle, clblasTranspose tran_a,
clblasOrder layout, clblasUplo triangle, clblasTranspose a_transpose,
size_t n, size_t k, double2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld, double2 beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
@ -318,7 +318,7 @@ clblasStatus clblasXsyrk(
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_double2{{alpha.real(), alpha.imag()}};
auto cl_beta = cl_double2{{beta.real(), beta.imag()}};
return clblasZsyrk(layout, triangle, tran_a,
return clblasZsyrk(layout, triangle, a_transpose,
n, k, cl_alpha,
a_mat, a_offset, a_ld, cl_beta,
c_mat, c_offset, c_ld,
@ -327,14 +327,14 @@ clblasStatus clblasXsyrk(
// This calls {clblasSsyr2k, clblasDsyr2k, clblasCsyr2k, clblasZsyr2k} with the arguments forwarded.
clblasStatus clblasXsyr2k(
clblasOrder layout, clblasUplo triangle, clblasTranspose tran_ab,
clblasOrder layout, clblasUplo triangle, clblasTranspose ab_transpose,
size_t n, size_t k, float alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, float beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasSsyr2k(layout, triangle, tran_ab,
return clblasSsyr2k(layout, triangle, ab_transpose,
n, k, alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, beta,
@ -342,14 +342,14 @@ clblasStatus clblasXsyr2k(
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXsyr2k(
clblasOrder layout, clblasUplo triangle, clblasTranspose tran_ab,
clblasOrder layout, clblasUplo triangle, clblasTranspose ab_transpose,
size_t n, size_t k, double alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, double beta,
cl_mem c_mat, size_t c_offset, size_t c_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasDsyr2k(layout, triangle, tran_ab,
return clblasDsyr2k(layout, triangle, ab_transpose,
n, k, alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, beta,
@ -357,7 +357,7 @@ clblasStatus clblasXsyr2k(
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXsyr2k(
clblasOrder layout, clblasUplo triangle, clblasTranspose tran_ab,
clblasOrder layout, clblasUplo triangle, clblasTranspose ab_transpose,
size_t n, size_t k, float2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, float2 beta,
@ -366,7 +366,7 @@ clblasStatus clblasXsyr2k(
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_float2{{alpha.real(), alpha.imag()}};
auto cl_beta = cl_float2{{beta.real(), beta.imag()}};
return clblasCsyr2k(layout, triangle, tran_ab,
return clblasCsyr2k(layout, triangle, ab_transpose,
n, k, cl_alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, cl_beta,
@ -374,7 +374,7 @@ clblasStatus clblasXsyr2k(
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXsyr2k(
clblasOrder layout, clblasUplo triangle, clblasTranspose tran_ab,
clblasOrder layout, clblasUplo triangle, clblasTranspose ab_transpose,
size_t n, size_t k, double2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld, double2 beta,
@ -383,7 +383,7 @@ clblasStatus clblasXsyr2k(
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_double2{{alpha.real(), alpha.imag()}};
auto cl_beta = cl_double2{{beta.real(), beta.imag()}};
return clblasZsyr2k(layout, triangle, tran_ab,
return clblasZsyr2k(layout, triangle, ab_transpose,
n, k, cl_alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld, cl_beta,
@ -391,6 +391,126 @@ clblasStatus clblasXsyr2k(
num_queues, queues, num_wait_events, wait_events, events);
}
// This calls {clblasStrmm, clblasDtrmm, clblasCtrmm, clblasZtrmm} with the arguments forwarded.
clblasStatus clblasXtrmm(
clblasOrder layout, clblasSide side, clblasUplo triangle,
clblasTranspose a_transpose, clblasDiag diagonal,
size_t m, size_t n, float alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasStrmm(layout, side, triangle, a_transpose, diagonal,
m, n, alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXtrmm(
clblasOrder layout, clblasSide side, clblasUplo triangle,
clblasTranspose a_transpose, clblasDiag diagonal,
size_t m, size_t n, double alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasDtrmm(layout, side, triangle, a_transpose, diagonal,
m, n, alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXtrmm(
clblasOrder layout, clblasSide side, clblasUplo triangle,
clblasTranspose a_transpose, clblasDiag diagonal,
size_t m, size_t n, float2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_float2{{alpha.real(), alpha.imag()}};
return clblasCtrmm(layout, side, triangle, a_transpose, diagonal,
m, n, cl_alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXtrmm(
clblasOrder layout, clblasSide side, clblasUplo triangle,
clblasTranspose a_transpose, clblasDiag diagonal,
size_t m, size_t n, double2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_double2{{alpha.real(), alpha.imag()}};
return clblasZtrmm(layout, side, triangle, a_transpose, diagonal,
m, n, cl_alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
// This calls {clblasStrsm, clblasDtrsm, clblasCtrsm, clblasZtrsm} with the arguments forwarded.
clblasStatus clblasXtrsm(
clblasOrder layout, clblasSide side, clblasUplo triangle,
clblasTranspose a_transpose, clblasDiag diagonal,
size_t m, size_t n, float alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasStrsm(layout, side, triangle, a_transpose, diagonal,
m, n, alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXtrsm(
clblasOrder layout, clblasSide side, clblasUplo triangle,
clblasTranspose a_transpose, clblasDiag diagonal,
size_t m, size_t n, double alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
return clblasDtrsm(layout, side, triangle, a_transpose, diagonal,
m, n, alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXtrsm(
clblasOrder layout, clblasSide side, clblasUplo triangle,
clblasTranspose a_transpose, clblasDiag diagonal,
size_t m, size_t n, float2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_float2{{alpha.real(), alpha.imag()}};
return clblasCtrsm(layout, side, triangle, a_transpose, diagonal,
m, n, cl_alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
clblasStatus clblasXtrsm(
clblasOrder layout, clblasSide side, clblasUplo triangle,
clblasTranspose a_transpose, clblasDiag diagonal,
size_t m, size_t n, double2 alpha,
const cl_mem a_mat, size_t a_offset, size_t a_ld,
const cl_mem b_mat, size_t b_offset, size_t b_ld,
cl_uint num_queues, cl_command_queue *queues,
cl_uint num_wait_events, const cl_event *wait_events, cl_event *events) {
auto cl_alpha = cl_double2{{alpha.real(), alpha.imag()}};
return clblasZtrsm(layout, side, triangle, a_transpose, diagonal,
m, n, cl_alpha,
a_mat, a_offset, a_ld,
b_mat, b_offset, b_ld,
num_queues, queues, num_wait_events, wait_events, events);
}
// =================================================================================================
} // namespace clblast