Merge commit '75891161' into gemm_layernorm_welford

2732d06c · rocking · dd0255ba · 75891161 · 2732d06c · 2732d06c
Commit 2732d06c authored Sep 06, 2022 by rocking
14 changed files
--- a/include/ck/tensor_operation/operator_transform/transform_conv_fwd_to_gemm.hpp
+++ b/include/ck/tensor_operation/operator_transform/transform_conv_fwd_to_gemm.hpp
--- a/include/ck/utility/data_type.hpp
+++ b/include/ck/utility/data_type.hpp
@@ -1023,6 +1023,8 @@ struct NumericLimits
    {
        return std::numeric_limits<T>::quiet_NaN();
    }
+    __host__ __device__ static constexpr T Infinity() { return std::numeric_limits<T>::infinity(); }
 };
 template <>

--- a/include/ck/utility/reduction_operator.hpp
+++ b/include/ck/utility/reduction_operator.hpp
--- a/library/include/ck/library/reference_tensor_operation/cpu/reference_batched_gemm.hpp
+++ b/library/include/ck/library/reference_tensor_operation/cpu/reference_batched_gemm.hpp
--- a/library/include/ck/library/utility/convolution_host_tensor_descriptor_helper.hpp
+++ b/library/include/ck/library/utility/convolution_host_tensor_descriptor_helper.hpp
--- a/library/include/ck/library/utility/convolution_parameter.hpp
+++ b/library/include/ck/library/utility/convolution_parameter.hpp
--- a/library/include/ck/library/utility/host_tensor.hpp
+++ b/library/include/ck/library/utility/host_tensor.hpp
--- a/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f32_f32_f32_km_kn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f32_f32_f32_km_kn_mn_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f32_f32_f32_km_nk_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f32_f32_f32_km_nk_mn_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f32_f32_f32_mk_kn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f32_f32_f32_mk_kn_mn_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f32_f32_f32_mk_nk_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f32_f32_f32_mk_nk_mn_instance.cpp
--- a/profiler/include/profile_batched_gemm_softmax_gemm_impl.hpp
+++ b/profiler/include/profile_batched_gemm_softmax_gemm_impl.hpp
--- a/script/run_full_performance_tests.sh
+++ b/script/run_full_performance_tests.sh
--- a/test/gemm_split_k/gemm_split_k.cpp
+++ b/test/gemm_split_k/gemm_split_k.cpp