en Copyright 2015 Java http://172.16.0.5:8080/history/llvm-project-15.0.7/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-float.ll#022bd92c [LowerMatrixMultiplication] Switch dummy values from undef to poison [NFC] List of files: /llvm-project-15.0.7/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-float.ll Sun, 03 Jul 2022 10:56:29 +0000 Nuno Lopes <nuno.lopes@tecnico.ulisboa.pt> http://172.16.0.5:8080/history/llvm-project-15.0.7/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-float.ll#7c4f45f8 Revert [LowerMatrixMultiplication] Switch dummy values from undef to poison [NFC]This reverts commits 47e6f98f84ac3 and 3e701bcd2a6aee2 List of files: /llvm-project-15.0.7/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-float.ll Fri, 01 Jul 2022 22:53:41 +0000 Nuno Lopes <nuno.lopes@tecnico.ulisboa.pt> http://172.16.0.5:8080/history/llvm-project-15.0.7/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-float.ll#47e6f98f [LowerMatrixMultiplication] Switch dummy values from undef to poison [NFC] List of files: /llvm-project-15.0.7/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-float.ll Fri, 01 Jul 2022 22:31:31 +0000 Nuno Lopes <nuno.lopes@tecnico.ulisboa.pt> http://172.16.0.5:8080/history/llvm-project-15.0.7/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-float.ll#dec9be85 [test][LowerMatrixIntrinsics] Use new PM RUN lines List of files: /llvm-project-15.0.7/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-float.ll Tue, 08 Mar 2022 21:37:22 +0000 Arthur Eubanks <aeubanks@google.com> http://172.16.0.5:8080/history/llvm-project-15.0.7/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-float.ll#9b296102 Use unary CreateShuffleVector if possibleAs mentioned in D93793, there are quite a few places where unary `IRBuilder::CreateShuffleVector(X, Mask)` can be usedinstead of `IRBuilder::CreateShuffleVector(X, Undef, Mask)`.Let's update them.Actually, it would have been more natural if the patches were made in this order:(1) let them use unary CreateShuffleVector first(2) update IRBuilder::CreateShuffleVector to use poison as a placeholder value (D93793)The order is swapped, but in terms of correctness it is still fine.Reviewed By: spatelDifferential Revision: https://reviews.llvm.org/D93923 List of files: /llvm-project-15.0.7/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-float.ll Tue, 29 Dec 2020 22:28:17 +0000 Juneyoung Lee <aqjune@gmail.com> http://172.16.0.5:8080/history/llvm-project-15.0.7/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-float.ll#526244b1 [Matrix] Add first set of matrix intrinsics and initial lowering pass.This is the first patch adding an initial set of matrix intrinsics and acorresponding lowering pass. This has been discussed on llvm-dev:http://lists.llvm.org/pipermail/llvm-dev/2019-October/136240.htmlThe first patch introduces four new intrinsics (transpose, multiply,columnwise load and store) and a LowerMatrixIntrinsics pass, thatlowers those intrinsics to vector operations.Matrixes are embedded in a 'flat' vector (e.g. a 4 x 4 float matrixembedded in a <16 x float> vector) and the intrinsics take the dimensioninformation as parameters. Those parameters need to be ConstantInt.For the memory layout, we initially assume column-major, but in the RFCwe also described how to extend the intrinsics to support row-major aswell.For the initial lowering, we split the input of the intrinsics into aset of column vectors, transform those column vectors and concatenatethe result columns to a flat result vector.This allows us to lower the intrinsics without any shape propagation, asmentioned in the RFC. In follow-up patches, we plan to submit thefollowing improvements: * Shape propagation to eliminate the embedding/splitting for each intrinsic. * Fused & tiled lowering of multiply and other operations. * Optimization remarks highlighting matrix expressions and costs. * Generate loops for operations on large matrixes. * More general block processing for operation on large vectors, exploiting shape information.We would like to add dedicated transpose, columnwise load and storeintrinsics, even though they are not strictly necessary. For example, wecould instead emit a large shufflevector instruction instead of thetranspose. But we expect that to (1) become unwieldy for larger matrixes (even for 16x16 matrixes, the resulting shufflevector masks would be huge), (2) risk instcombine making small changes, causing us to fail to detect the transpose, preventing better loweringsFor the load/store, we are additionally planning on exploiting theintrinsics for better alias analysis.Reviewers: anemet, Gerolf, reames, hfinkel, andrew.w.kaylor, efriedma, rengolinReviewed By: anemetDifferential Revision: https://reviews.llvm.org/D70456 List of files: /llvm-project-15.0.7/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-float.ll Thu, 12 Dec 2019 15:27:28 +0000 Florian Hahn <flo@fhahn.com>