[PATCH] D42717: [JumpThreading] sync DT for LVI analysis (PR 36133)

[Brian Rzycki via Phabricator via llvm-commits]

brzycki added a comment.

Six runs of tramp3d-v4 with this patch applied show comperable variance and similar low/high bounds:

  $ cat baz.sh
  #!/bin/bash
  
  for i in 1 2 3 4 5 6; do
      for j in 1 2 3 4 5 6; do
          ~/git/me/llvm/litdiff \
              build.D42717_$i/results.json \
              build.D42717_$j/results.json \
              compile_time  | grep tramp3d
      done
  done
  brzycki at cc01 /work/brzycki/test-suite $ ./baz.sh
      494.1120 -> 487.4120     [  1.37%]  CTMark/tramp3d-v4/tramp3d-v4.test
      494.1120 -> 485.5440     [  1.76%]  CTMark/tramp3d-v4/tramp3d-v4.test
      494.1120 <- 494.9720     [  0.17%]  CTMark/tramp3d-v4/tramp3d-v4.test
      494.1120 <- 502.4760     [  1.69%]  CTMark/tramp3d-v4/tramp3d-v4.test
      494.1120 <- 536.5600     [  8.59%]  CTMark/tramp3d-v4/tramp3d-v4.test
      487.4120 <- 494.1120     [  1.37%]  CTMark/tramp3d-v4/tramp3d-v4.test
      487.4120 -> 485.5440     [  0.38%]  CTMark/tramp3d-v4/tramp3d-v4.test
      487.4120 <- 494.9720     [  1.55%]  CTMark/tramp3d-v4/tramp3d-v4.test
      487.4120 <- 502.4760     [  3.09%]  CTMark/tramp3d-v4/tramp3d-v4.test
      487.4120 <- 536.5600     [ 10.08%]  CTMark/tramp3d-v4/tramp3d-v4.test
      485.5440 <- 494.1120     [  1.76%]  CTMark/tramp3d-v4/tramp3d-v4.test
      485.5440 <- 487.4120     [  0.38%]  CTMark/tramp3d-v4/tramp3d-v4.test
      485.5440 <- 494.9720     [  1.94%]  CTMark/tramp3d-v4/tramp3d-v4.test
      485.5440 <- 502.4760     [  3.49%]  CTMark/tramp3d-v4/tramp3d-v4.test
      485.5440 <- 536.5600     [ 10.51%]  CTMark/tramp3d-v4/tramp3d-v4.test
      494.9720 -> 494.1120     [  0.17%]  CTMark/tramp3d-v4/tramp3d-v4.test
      494.9720 -> 487.4120     [  1.55%]  CTMark/tramp3d-v4/tramp3d-v4.test
      494.9720 -> 485.5440     [  1.94%]  CTMark/tramp3d-v4/tramp3d-v4.test
      494.9720 <- 502.4760     [  1.52%]  CTMark/tramp3d-v4/tramp3d-v4.test
      494.9720 <- 536.5600     [  8.40%]  CTMark/tramp3d-v4/tramp3d-v4.test
      502.4760 -> 494.1120     [  1.69%]  CTMark/tramp3d-v4/tramp3d-v4.test
      502.4760 -> 487.4120     [  3.09%]  CTMark/tramp3d-v4/tramp3d-v4.test
      502.4760 -> 485.5440     [  3.49%]  CTMark/tramp3d-v4/tramp3d-v4.test
      502.4760 -> 494.9720     [  1.52%]  CTMark/tramp3d-v4/tramp3d-v4.test
      502.4760 <- 536.5600     [  6.78%]  CTMark/tramp3d-v4/tramp3d-v4.test
      536.5600 -> 494.1120     [  8.59%]  CTMark/tramp3d-v4/tramp3d-v4.test
      536.5600 -> 487.4120     [ 10.08%]  CTMark/tramp3d-v4/tramp3d-v4.test
      536.5600 -> 485.5440     [ 10.51%]  CTMark/tramp3d-v4/tramp3d-v4.test
      536.5600 -> 494.9720     [  8.40%]  CTMark/tramp3d-v4/tramp3d-v4.test
      536.5600 -> 502.4760     [  6.78%]  CTMark/tramp3d-v4/tramp3d-v4.test

The best time was `485.544` and the worst time was `536.56`.

@dberlin I think this patch can be applied as-is to fix pr36133.


https://reviews.llvm.org/D42717