Merge branch 'develop' into lwpck-471

19a08d65 · Rostyslav Geyyer · GitHub · 2056491f · afdfef74 · 19a08d65
Unverified Commit 19a08d65 authored Feb 03, 2023 by Rostyslav Geyyer Committed by GitHub Feb 03, 2023
20 changed files
--- a/include/ck/tensor_operation/gpu/device/impl/device_multiple_reduce_threadwise.hpp
+++ b/include/ck/tensor_operation/gpu/device/impl/device_multiple_reduce_threadwise.hpp
--- a/include/ck/tensor_operation/gpu/device/impl/device_normalization_impl.hpp
+++ b/include/ck/tensor_operation/gpu/device/impl/device_normalization_impl.hpp
--- a/include/ck/tensor_operation/gpu/device/impl/device_reduce_multiblock.hpp
+++ b/include/ck/tensor_operation/gpu/device/impl/device_reduce_multiblock.hpp
--- a/include/ck/tensor_operation/gpu/device/impl/device_reduce_threadwise.hpp
+++ b/include/ck/tensor_operation/gpu/device/impl/device_reduce_threadwise.hpp
--- a/include/ck/tensor_operation/gpu/device/impl/device_softmax_impl.hpp
+++ b/include/ck/tensor_operation/gpu/device/impl/device_softmax_impl.hpp
--- a/include/ck/tensor_operation/gpu/device/impl/device_sparse_embedding3_forward_layernorm.hpp
+++ b/include/ck/tensor_operation/gpu/device/impl/device_sparse_embedding3_forward_layernorm.hpp
--- a/include/ck/tensor_operation/gpu/element/element_wise_operation.hpp
+++ b/include/ck/tensor_operation/gpu/element/element_wise_operation.hpp
--- a/include/ck/tensor_operation/gpu/grid/block_to_ctile_map.hpp
+++ b/include/ck/tensor_operation/gpu/grid/block_to_ctile_map.hpp
--- a/include/ck/tensor_operation/gpu/grid/gemm_layernorm/gridwise_gemm_multiple_d_welford_first_half_xdl_cshuffle.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gemm_layernorm/gridwise_gemm_multiple_d_welford_first_half_xdl_cshuffle.hpp
--- a/include/ck/tensor_operation/gpu/grid/gemm_layernorm/gridwise_welford_second_half_layernorm2d.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gemm_layernorm/gridwise_welford_second_half_layernorm2d.hpp
--- a/include/ck/tensor_operation/gpu/grid/gridwise_elementwise_layernorm_welford_variance.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_elementwise_layernorm_welford_variance.hpp
@@ -434,7 +434,7 @@ struct GridwiseElementwiseLayernormWelfordVariance_mk_to_mk
            });
            static_for<0, MThreadSliceSize, 1>{}([&](auto iM) {
-                auto divisor = 1 / __builtin_amdgcn_sqrtf(var_thread_buf(iM) + epsilon);
+                auto divisor = 1 / ck::math::sqrt(var_thread_buf(iM) + epsilon);
                static_for<0, XThreadBufferNumber, 1>{}([&](auto iK0) {
                    static_for<0, XSrcVectorSize, 1>{}([&](auto iK1) {
                        constexpr auto offset_m_k =

--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_waveletmodel.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_waveletmodel.hpp
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_wmma.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_wmma.hpp
--- a/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_waveletmodel_cshuffle.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdl_waveletmodel_cshuffle.hpp
--- a/include/ck/tensor_operation/gpu/grid/gridwise_normalization_welford_variance.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_normalization_welford_variance.hpp
--- a/include/ck/tensor_operation/gpu/grid/gridwise_sparse_embedding3_forward_layernorm.hpp
+++ b/include/ck/tensor_operation/gpu/grid/gridwise_sparse_embedding3_forward_layernorm.hpp
--- a/include/ck/tensor_operation/gpu/warp/wmma_gemm.hpp
+++ b/include/ck/tensor_operation/gpu/warp/wmma_gemm.hpp
--- a/include/ck/utility/amd_inline_asm.hpp
+++ b/include/ck/utility/amd_inline_asm.hpp
--- a/include/ck/utility/amd_wmma.hpp
+++ b/include/ck/utility/amd_wmma.hpp
--- a/include/ck/utility/math_v2.hpp
+++ b/include/ck/utility/math_v2.hpp