aboutsummaryrefslogtreecommitdiff
path: root/src/core/gpu/cl/kernels/ClMulKernel.cpp
diff options
context:
space:
mode:
authorGiorgio Arena <giorgio.arena@arm.com>2021-06-28 11:00:27 +0100
committerGiorgio Arena <giorgio.arena@arm.com>2021-06-29 11:06:10 +0000
commit4a95bba6ca61ce99995ece6fd237b5498c9f322c (patch)
tree04ef2dee1cdc015ce14ce6b664d9a956a96c3a05 /src/core/gpu/cl/kernels/ClMulKernel.cpp
parent731fe667e3009bfbfee8b0eb74ecb68b291e4311 (diff)
downloadComputeLibrary-4a95bba6ca61ce99995ece6fd237b5498c9f322c.tar.gz
Set up the framework to choose the default LWS
Resolve COMPMID-4486 Signed-off-by: Giorgio Arena <giorgio.arena@arm.com> Change-Id: Ib38b7943bd776a6d75d1da163908724c49eae73d Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/5864 Comments-Addressed: Arm Jenkins <bsgcomp@arm.com> Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com> Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com> Tested-by: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/gpu/cl/kernels/ClMulKernel.cpp')
-rw-r--r--src/core/gpu/cl/kernels/ClMulKernel.cpp12
1 files changed, 11 insertions, 1 deletions
diff --git a/src/core/gpu/cl/kernels/ClMulKernel.cpp b/src/core/gpu/cl/kernels/ClMulKernel.cpp
index b8081bbacf..65f3bec099 100644
--- a/src/core/gpu/cl/kernels/ClMulKernel.cpp
+++ b/src/core/gpu/cl/kernels/ClMulKernel.cpp
@@ -92,6 +92,11 @@ Status validate_arguments(const ITensorInfo *src1, const ITensorInfo *src2, cons
}
} // namespace
+ClMulKernel::ClMulKernel()
+{
+ _type = CLKernelType::ELEMENTWISE;
+}
+
void ClMulKernel::configure(const CLCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, float scale,
ConvertPolicy overflow_policy, RoundingPolicy rounding_policy, const ActivationLayerInfo &act_info)
{
@@ -128,7 +133,7 @@ void ClMulKernel::configure(const CLCompileContext &compile_context, ITensorInfo
else
{
if(src1->element_size() == 4 || src2->element_size() == 4)
- {
+ {
// use 64 bit accumulator for 32-bit input
acc_type = "long";
}
@@ -316,6 +321,11 @@ Status validate_arguments_complex(const ITensorInfo *src1, const ITensorInfo *sr
}
} // namespace
+ClComplexMulKernel::ClComplexMulKernel()
+{
+ _type = CLKernelType::ELEMENTWISE;
+}
+
void ClComplexMulKernel::configure(const CLCompileContext &compile_context, ITensorInfo *src1, ITensorInfo *src2, ITensorInfo *dst, const ActivationLayerInfo &act_info)
{
ARM_COMPUTE_ERROR_ON_NULLPTR(src1, src2, dst);