LLVMGPU

-extract-address-computation-gpu

Extract address computations from memory accesses

This pass is similar to extract-address-computation except it also supports memory accesses that are specific to GPUs.

-iree-amdgpu-emulate-narrow-type

Emulate narrow integer operations including amdgpu operations

-iree-convert-to-nvvm

Perform final conversion from builtin/GPU/HAL/standard dialect to LLVM and NVVM dialects

-iree-convert-to-rocdl

Perform final conversion from builtin/GPU/HAL/standard dialect to LLVM and ROCDL dialects

-iree-llvmgpu-assign-constant-ordinals

Assigns executable constant ordinals across all LLVMGPU variants.

-iree-llvmgpu-cast-address-space-function

Cast address space to generic in CallOp and FuncOp

-iree-llvmgpu-cast-type-to-fit-mma

Perform type extension/truncation over vector.contract types to target GPU MMA intrinsics

-iree-llvmgpu-configure-tensor-layouts

Pass to set layouts on tensors for later vector distribution

-iree-llvmgpu-link-executables

Links LLVMGPU HAL executables within the top-level program module.

Options

-target : Target backend name whose executables will be linked by this pass.

-iree-llvmgpu-lower-executable-target

Perform lowering of executable target using one of the IREE::HAL::DispatchLoweringPassPipeline

Options

-for-rocdl : Enable features only supported on ROCDL such as delaying lowering of subgroup reduce.

-iree-llvmgpu-pack-shared-memory-alloc

Pass pack shared memory allocation in order to reduce memory usage.

-iree-llvmgpu-prefetch-shared-memory

Rotate scf.for loops to prefetch shared memory with distance 1. This pass is only applicableto ROCDL targets because its effectiveness on non-AMD GPUs lacks testing and evaluation.

-iree-llvmgpu-select-lowering-strategy

Select a IREE::HAL::DispatchLoweringPassPipeline for lowering the target variant

-iree-llvmgpu-tensorcore-vectorization

Pass to convert linalg into Vector and transform it to a form that can be lowered to GPU MMA ops

-iree-llvmgpu-tile-and-distribute

Pass to tile and distribute linalg ops within a workgroup.

-iree-llvmgpu-vector-distribute

Pass to distribute vectorized functions.

-iree-llvmgpu-vector-lowering

Pass to lower Vector ops before conversion to LLVM.

-iree-llvmgpu-vector-to-gpu

Pass to convert vector to gpu.

-iree-test-llvmgpu-legalize-ops

Test pass for several legalization patterns.