mlir

mlir copied to clipboard

How to convert MLIR to SPIR-V?

9

Would you like to give me any information how to convert from MLIR to SPIR-V? I tried to use mlir-translate with serialize-spirv option, but I couldn't realize how to use...

stakemura

installation steps

1

@alextp @ewilderj @d0k @benvanik @flaub i have installed mlir but i have missed someting can u please help me ~/llvm-project/build$ cmake --build . --target check-mlir [0/1] Running the MLIR regression...

arunchannaraju18

three unexpected test failures:mlir-cuda-runner

4

``` [1732/1733] Running the MLIR regression tests FAIL: MLIR :: mlir-cuda-runner/all-reduce-op.mlir (362 of 364) ******************** TEST 'MLIR :: mlir-cuda-runner/all-reduce-op.mlir' FAILED ******************** Script: -- : 'RUN: at line 1'; mlir-cuda-runner /home/sun/llvm-project/llvm/projects/mlir/test/mlir-cuda-runner/all-reduce-op.mlir...

MingliSun

[spirv] Add a pass to convert a sequence of composite insert to one composite construct

6

Sometimes we will see the following code pattern ``` %0 = spv.undef : vector %1 = spv.CompositeInsert %val0 into %0 : vector %2 = spv.CompositeInsert %val1 into %1 : vector...

antiagainst

(**NOTE**: This is not updated anymore. Please see https://mlir.llvm.org/docs/Dialects/SPIR-V/ for supported features.) This is a tracking bug for completing the SPIR-V dialect. It will be periodically updated to provide the...

antiagainst

memref lowering: stack overflow due to alloca's for memref descriptors for calls inside a loop

1

When lowering MLIR to LLVM, since memrefs are lowered through llvm structs that hold the descriptor info, the alloca's for these structs can exhaust stack space when there are calls...

bondhugula

[spirv] [tracking] SPIR-V lowering

29

This is a tracking bug for implementing lowering towards the SPIR-V dialect. It will be periodically updated to provide the current status. Anybody interested in helping is very welcome to...

antiagainst

Custom lowering of MemRefs to LLVM

5

This PR introduces the infrastructure to provide a custom Std-to-LLVM lowering for MemRef type: * `MemRefDescriptor` class is turned into an abstract API that defines the methods needed to perform...

dcaballe

[LLVM] Implement missing LLVM IR instructions

11

The following LLVM IR instructions are currently not modeled by the LLVM dialect. Each of this is a great starter bug to implement a new Op in MLIR. - [...

ftynse

Fix the wrong computation of dynamic strides for lowering AllocOp to LLVM

2

The computation of **dynamic strides** in lowering AllocOp was wrong. Given a MemRefType with `sizes = [5, 10]`, it computed `strides = [5, 1]`, while the correct answer should be...

tungld