[all-commits] [llvm/llvm-project] 8ec406: [VPlan] Implement unrolling as VPlan-to-VPlan tran...

[Florian Hahn via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 8ec406757cb92f352888a3d4092397f2b5a7d1d9
      https://github.com/llvm/llvm-project/commit/8ec406757cb92f352888a3d4092397f2b5a7d1d9
  Author: Florian Hahn <flo at fhahn.com>
  Date:   2024-09-21 (Sat, 21 Sep 2024)

  Changed paths:
    M llvm/lib/Transforms/Vectorize/CMakeLists.txt
    M llvm/lib/Transforms/Vectorize/LoopVectorizationPlanner.h
    M llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
    M llvm/lib/Transforms/Vectorize/VPlan.cpp
    M llvm/lib/Transforms/Vectorize/VPlan.h
    M llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
    M llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
    M llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
    M llvm/lib/Transforms/Vectorize/VPlanTransforms.h
    A llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp
    M llvm/lib/Transforms/Vectorize/VPlanUtils.cpp
    M llvm/lib/Transforms/Vectorize/VPlanUtils.h
    M llvm/test/Transforms/LoopVectorize/AArch64/arbitrary-induction-step.ll
    M llvm/test/Transforms/LoopVectorize/AArch64/deterministic-type-shrinkage.ll
    M llvm/test/Transforms/LoopVectorize/AArch64/divs-with-scalable-vfs.ll
    M llvm/test/Transforms/LoopVectorize/AArch64/force-target-instruction-cost.ll
    M llvm/test/Transforms/LoopVectorize/AArch64/reduction-recurrence-costs-sve.ll
    M llvm/test/Transforms/LoopVectorize/AArch64/store-costs-sve.ll
    M llvm/test/Transforms/LoopVectorize/AArch64/sve-inductions-unusual-types.ll
    M llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll
    M llvm/test/Transforms/LoopVectorize/RISCV/dead-ops-cost.ll
    M llvm/test/Transforms/LoopVectorize/RISCV/divrem.ll
    M llvm/test/Transforms/LoopVectorize/RISCV/interleaved-accesses.ll
    M llvm/test/Transforms/LoopVectorize/RISCV/riscv-vector-reverse.ll
    M llvm/test/Transforms/LoopVectorize/RISCV/uniform-load-store.ll
    M llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-interleave.ll
    M llvm/test/Transforms/LoopVectorize/X86/cost-model.ll
    M llvm/test/Transforms/LoopVectorize/X86/epilog-vectorization-inductions.ll
    M llvm/test/Transforms/LoopVectorize/X86/interleaving.ll
    M llvm/test/Transforms/LoopVectorize/X86/masked-store-cost.ll
    M llvm/test/Transforms/LoopVectorize/X86/pr47437.ll
    M llvm/test/Transforms/LoopVectorize/X86/uniform_mem_op.ll
    A llvm/test/Transforms/LoopVectorize/first-order-recurrence-interleave-only.ll
    M llvm/test/Transforms/LoopVectorize/first-order-recurrence.ll
    M llvm/test/Transforms/LoopVectorize/float-induction.ll
    M llvm/test/Transforms/LoopVectorize/induction.ll
    M llvm/test/Transforms/LoopVectorize/interleave-and-scalarize-only.ll
    M llvm/test/Transforms/LoopVectorize/pr45679-fold-tail-by-masking.ll
    M llvm/test/Transforms/LoopVectorize/reduction-inloop-uf4.ll
    M llvm/test/Transforms/LoopVectorize/vplan-predicate-switch.ll
    M llvm/test/Transforms/LoopVectorize/vplan-printing-before-execute.ll

  Log Message:
  -----------
  [VPlan] Implement unrolling as VPlan-to-VPlan transform. (#95842)

This patch implements explicit unrolling by UF  as VPlan transform. In
follow up patches this will allow simplifying VPTransform state (no need
to store unrolled parts) as well as recipe execution (no need to
generate code for multiple parts in an each recipe). It also allows for
more general optimziations (e.g. avoid generating code for recipes that
are uniform-across parts).

It also unifies the logic dealing with unrolled parts in a single place,
rather than spreading it out across multiple places (e.g. VPlan post
processing for header-phi recipes previously.)

In the initial implementation, a number of recipes still take the
unrolled part as additional, optional argument, if their execution
depends on the unrolled part.

The computation for start/step values for scalable inductions changed
slightly. Previously the step would be computed as scalar and then
splatted, now vscale gets splatted and multiplied by the step in a
vector mul.

This has been split off https://github.com/llvm/llvm-project/pull/94339
which also includes changes to simplify VPTransfomState and recipes'
::execute.

The current version mostly leaves existing ::execute untouched and
instead sets VPTransfomState::UF to 1.

A follow-up patch will clean up all references to VPTransformState::UF.

Another follow-up patch will simplify VPTransformState to only store a
single vector value per VPValue.

PR: https://github.com/llvm/llvm-project/pull/95842



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications