[PATCH] D108607: [RISCV] Optimize (add (mul x, c0), c1)

[Nathan Chancellor via Phabricator via llvm-commits]

nathanchance added a comment.
Herald added a subscriber: achieveartificialintelligence.

This commit causes an infinite loop when compiling the Linux kernel:

  $ git bisect log
  # bad: [43d6991c2a4cc2ac374e68c029634f2b59ffdfdf] [IR] Look through bitcast in hasFnAttribute()
  # good: [6a2c2263fbca07a59b9f41181c1418df052e24d1] [X86] Improve i8 all-ones element insertion in pre-SSE4.1
  git bisect start '43d6991c2a4cc2ac374e68c029634f2b59ffdfdf' '6a2c2263fbca07a59b9f41181c1418df052e24d1'
  # good: [792101fff749191dfd4dadabe2ecd30a4d8cd973] [RISCV] Add test cases for missed opportunity to use vfmacc.vf. NFC
  git bisect good 792101fff749191dfd4dadabe2ecd30a4d8cd973
  # good: [20b58855e0cfc263d609e8bb59e692024ecb42aa] [CodeGen] SelectionDAGBuilder - Use const-ref iterator in for-range loops. NFCI.
  git bisect good 20b58855e0cfc263d609e8bb59e692024ecb42aa
  # bad: [08ef71ca92d9bd474ac4bbb5109e9ced19c77f0e] [InstCombine] move/add tests for trunc-of-lshr; NFC
  git bisect bad 08ef71ca92d9bd474ac4bbb5109e9ced19c77f0e
  # bad: [744ec74b305a5039dc74e2d043e1c136e06beac1] [NFC] `goto fail` has failed us in the past...
  git bisect bad 744ec74b305a5039dc74e2d043e1c136e06beac1
  # good: [b8e7f5320825812fc4b597ee5df3dccc53fe78bb] [AMDGPU][MC][GFX10] Enabled dlc for FLAT and GLOBAL atomics
  git bisect good b8e7f5320825812fc4b597ee5df3dccc53fe78bb
  # good: [ee31ad0ab5f7d443b0bd582582a3524dcf7f13f0] [clang-offload-bundler][docs][NFC] Add archive unbundling documentation
  git bisect good ee31ad0ab5f7d443b0bd582582a3524dcf7f13f0
  # good: [32b994bca66641cdac8586f25315daf349921ebc] [OpenCL] Defines helper function for OpenCL default address space
  git bisect good 32b994bca66641cdac8586f25315daf349921ebc
  # bad: [b3052013b43617defd777b1f3b4339540c4c07df] [RISCV] Optimize (add (mul x, c0), c1)
  git bisect bad b3052013b43617defd777b1f3b4339540c4c07df
  # first bad commit: [b3052013b43617defd777b1f3b4339540c4c07df] [RISCV] Optimize (add (mul x, c0), c1)
  
  $ cat evergreen.i
  int evergreen_dram_bandwidth_for_display_wm_0,
      evergreen_dram_bandwidth_for_display_disp_dram_allocation_0,
      evergreen_dram_bandwidth_for_display_bandwidth_0,
      evergreen_dram_bandwidth_for_display_a_0;
  void evergreen_dram_bandwidth_for_display() {
    evergreen_dram_bandwidth_for_display_a_0 = 1000 << 12;
    evergreen_dram_bandwidth_for_display_disp_dram_allocation_0 = 3 << 12;
    long tmp = evergreen_dram_bandwidth_for_display_disp_dram_allocation_0 << 13;
    tmp = tmp / evergreen_dram_bandwidth_for_display_a_0;
    evergreen_dram_bandwidth_for_display_bandwidth_0 =
        evergreen_dram_bandwidth_for_display_wm_0 * tmp + 2048;
  }
  
  $ timeout 10s clang -O2 --target=riscv64-linux-gnu -c -o /dev/null evergreen.i
  
  $ echo $?
  124

This is still reproducible at current ToT (3778c1cd6ef5a3286d5d49e842a2c65fffb8f3a6 <https://reviews.llvm.org/rG3778c1cd6ef5a3286d5d49e842a2c65fffb8f3a6>).


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D108607/new/

https://reviews.llvm.org/D108607