[PATCH] D49761: [AMDGPU] Use AssumptionCacheTracker in the divrem32 expansion

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec created this revision.
rampitec added reviewers: artem.tamazov, arsenm.
Herald added subscribers: t-tye, tpr, dstuttard, yaxunl, nhaehnle, wdng, kzhuravl.

This allows to use a short div/rem expansion given a kernel with __builtin_assume() call:

  kernel void divrem24_uint(global int *a, uint x) {
    uint id = get_local_id(0);
    __builtin_assume(x < 42);
    a[id / x] = 0;
  }


https://reviews.llvm.org/D49761

Files:
  lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp
  test/CodeGen/AMDGPU/divrem24-assume.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D49761.157137.patch
Type: text/x-patch
Size: 7485 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180724/34800bd6/attachment.bin>