diff options
author | Yi Kong <yikong@google.com> | 2022-02-25 16:41:05 +0000 |
---|---|---|
committer | Automerger Merge Worker <android-build-automerger-merge-worker@system.gserviceaccount.com> | 2022-02-25 16:41:05 +0000 |
commit | bc0f5df265caa21a2120c22453655a7fcc941991 (patch) | |
tree | fb979fb4cf4f8052c8cc66b1ec9516d91fcd859b /unsupported/Eigen/CXX11/src/Tensor/TensorContractionBlocking.h | |
parent | 8fd413e275f78a4c240f1442ce5cf77c73a20a55 (diff) | |
parent | 7cb50013986f04dce5fac87bebf319bb8db37a36 (diff) | |
download | eigen-t_frc_odp_330442040.tar.gz |
Merge changes Iee153445,Iee274471 am: 79df15ea88 am: 10f298fc41 am: 7cb5001398t_frc_odp_330442040t_frc_odp_330442000t_frc_ase_330444010android-wear-13.0.0-gpl_r3android-wear-13.0.0-gpl_r2android-wear-13.0.0-gpl_r1android-vts-13.0_r8android-vts-13.0_r7android-vts-13.0_r6android-vts-13.0_r5android-vts-13.0_r4android-vts-13.0_r3android-vts-13.0_r2android-t-qpr3-beta-3-gplandroid-t-qpr3-beta-1-gplandroid-t-qpr2-beta-3-gplandroid-t-qpr2-beta-2-gplandroid-t-qpr1-beta-3-gplandroid-t-qpr1-beta-1-gplandroid-cts-13.0_r8android-cts-13.0_r7android-cts-13.0_r6android-cts-13.0_r5android-cts-13.0_r4android-cts-13.0_r3android-cts-13.0_r2android-13.0.0_r83android-13.0.0_r82android-13.0.0_r81android-13.0.0_r80android-13.0.0_r79android-13.0.0_r78android-13.0.0_r77android-13.0.0_r76android-13.0.0_r75android-13.0.0_r74android-13.0.0_r73android-13.0.0_r72android-13.0.0_r71android-13.0.0_r70android-13.0.0_r69android-13.0.0_r68android-13.0.0_r67android-13.0.0_r66android-13.0.0_r65android-13.0.0_r64android-13.0.0_r63android-13.0.0_r62android-13.0.0_r61android-13.0.0_r60android-13.0.0_r59android-13.0.0_r58android-13.0.0_r57android-13.0.0_r56android-13.0.0_r55android-13.0.0_r54android-13.0.0_r53android-13.0.0_r52android-13.0.0_r51android-13.0.0_r50android-13.0.0_r49android-13.0.0_r48android-13.0.0_r47android-13.0.0_r46android-13.0.0_r45android-13.0.0_r44android-13.0.0_r43android-13.0.0_r42android-13.0.0_r41android-13.0.0_r40android-13.0.0_r39android-13.0.0_r38android-13.0.0_r37android-13.0.0_r36android-13.0.0_r35android-13.0.0_r34android-13.0.0_r33android-13.0.0_r32android-13.0.0_r30android-13.0.0_r29android-13.0.0_r28android-13.0.0_r27android-13.0.0_r24android-13.0.0_r23android-13.0.0_r22android-13.0.0_r21android-13.0.0_r20android-13.0.0_r19android-13.0.0_r18android-13.0.0_r17android-13.0.0_r16aml_go_odp_330912000aml_go_ads_330915100aml_go_ads_330915000aml_go_ads_330913000android13-tests-releaseandroid13-tests-devandroid13-qpr3-s9-releaseandroid13-qpr3-s8-releaseandroid13-qpr3-s7-releaseandroid13-qpr3-s6-releaseandroid13-qpr3-s5-releaseandroid13-qpr3-s4-releaseandroid13-qpr3-s3-releaseandroid13-qpr3-s2-releaseandroid13-qpr3-s14-releaseandroid13-qpr3-s13-releaseandroid13-qpr3-s12-releaseandroid13-qpr3-s11-releaseandroid13-qpr3-s10-releaseandroid13-qpr3-s1-releaseandroid13-qpr3-releaseandroid13-qpr3-c-s8-releaseandroid13-qpr3-c-s7-releaseandroid13-qpr3-c-s6-releaseandroid13-qpr3-c-s5-releaseandroid13-qpr3-c-s4-releaseandroid13-qpr3-c-s3-releaseandroid13-qpr3-c-s2-releaseandroid13-qpr3-c-s12-releaseandroid13-qpr3-c-s11-releaseandroid13-qpr3-c-s10-releaseandroid13-qpr3-c-s1-releaseandroid13-qpr2-s9-releaseandroid13-qpr2-s8-releaseandroid13-qpr2-s7-releaseandroid13-qpr2-s6-releaseandroid13-qpr2-s5-releaseandroid13-qpr2-s3-releaseandroid13-qpr2-s2-releaseandroid13-qpr2-s12-releaseandroid13-qpr2-s11-releaseandroid13-qpr2-s10-releaseandroid13-qpr2-s1-releaseandroid13-qpr2-releaseandroid13-qpr2-b-s1-releaseandroid13-qpr1-s8-releaseandroid13-qpr1-s7-releaseandroid13-qpr1-s6-releaseandroid13-qpr1-s5-releaseandroid13-qpr1-s4-releaseandroid13-qpr1-s3-releaseandroid13-qpr1-s2-releaseandroid13-qpr1-s1-releaseandroid13-qpr1-releaseandroid13-mainline-go-adservices-releaseandroid13-frc-odp-releaseandroid13-devandroid13-d4-s2-releaseandroid13-d4-s1-releaseandroid13-d4-releaseandroid13-d3-s1-releaseandroid13-d2-releaseandroid-wear-13.0.0-gpl_r1
Original change: https://android-review.googlesource.com/c/platform/external/eigen/+/1999079
Change-Id: I4c76dc5ddc7fb0ae9fc42436f28bd8bf9de50a97
Diffstat (limited to 'unsupported/Eigen/CXX11/src/Tensor/TensorContractionBlocking.h')
-rw-r--r-- | unsupported/Eigen/CXX11/src/Tensor/TensorContractionBlocking.h | 39 |
1 files changed, 28 insertions, 11 deletions
diff --git a/unsupported/Eigen/CXX11/src/Tensor/TensorContractionBlocking.h b/unsupported/Eigen/CXX11/src/Tensor/TensorContractionBlocking.h index 5cf7b4f71..974feb0ad 100644 --- a/unsupported/Eigen/CXX11/src/Tensor/TensorContractionBlocking.h +++ b/unsupported/Eigen/CXX11/src/Tensor/TensorContractionBlocking.h @@ -21,14 +21,28 @@ enum { // Default Blocking Strategy -template <typename LhsMapper, typename RhsMapper, typename Index, int ShardingType=ShardByCol> +template<typename ResScalar, typename LhsScalar, typename RhsScalar, typename StorageIndex, int ShardingType = ShardByCol> class TensorContractionBlocking { public: - typedef typename LhsMapper::Scalar LhsScalar; - typedef typename RhsMapper::Scalar RhsScalar; + /* + adding EIGEN_DEVICE_FUNC unconditionally to 'TensorContractionBlocking' constructor in `TensorContractionBlocking.h` + requires adding EIGEN_DEVICE_FUNC to `computeProductBlockingSizes` in `GeneralBlockPanelKernel.h` + which in turn, requires adding EIGEN_DEVICE_FUNC to `evaluateProductBlockingSizesHeuristic` in `GeneralBlockPanelKernel.h` + which in turn, requires adding EIGEN_DEVICE_FUNC to `manage_caching_sizes` in `GeneralBlockPanelKernel.h` + (else HIPCC will error out) - EIGEN_DEVICE_FUNC TensorContractionBlocking(Index k, Index m, Index n, Index num_threads = 1) : + However adding EIGEN_DEVICE_FUNC to `manage_caching_sizes` in `GeneralBlockPanelKernel.h` + results in NVCC erroring out with the following error + + ../Eigen/src/Core/products/GeneralBlockPanelKernel.h(57): error #2901: + dynamic initialization is not supported for function-scope static variables within a __device__/__global__ function + */ + + #if !defined(EIGEN_HIPCC) + EIGEN_DEVICE_FUNC + #endif + TensorContractionBlocking(StorageIndex k, StorageIndex m, StorageIndex n, StorageIndex num_threads = 1) : kc_(k), mc_(m), nc_(n) { if (ShardingType == ShardByCol) { @@ -37,19 +51,22 @@ class TensorContractionBlocking { else { computeProductBlockingSizes<LhsScalar, RhsScalar, 1>(kc_, nc_, mc_, num_threads); } + + const int rhs_packet_size = internal::packet_traits<RhsScalar>::size; + kc_ = (rhs_packet_size <= 8 || kc_ <= rhs_packet_size) ? + kc_ : (kc_ / rhs_packet_size) * rhs_packet_size; } - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Index kc() const { return kc_; } - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Index mc() const { return mc_; } - EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE Index nc() const { return nc_; } + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE StorageIndex kc() const { return kc_; } + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE StorageIndex mc() const { return mc_; } + EIGEN_DEVICE_FUNC EIGEN_ALWAYS_INLINE StorageIndex nc() const { return nc_; } private: - Index kc_; - Index mc_; - Index nc_; + StorageIndex kc_; + StorageIndex mc_; + StorageIndex nc_; }; - } // end namespace internal } // end namespace Eigen |