[PATCH] D74484: [AggressiveInstCombine] Add support for ICmp instr that feeds a select intsr's condition operand.

[Ayman Musa via Phabricator via llvm-commits]

aymanmus added a comment.

In D74484#1873745 <https://reviews.llvm.org/D74484#1873745>, @nikic wrote:

> I think this is missing a test-case where the select could be truncated if you ignored the icmp, but not if you require the icmp to also be truncated. That should show up as a regression in the test diffs.


Not sure I understand what you mean exactly.
Shrinking the cmp is not *required* in order to shrink the select. If there is a cmp instruction that can be shrank along with the select, we shrink it. If not, it does not affect the shrinking of the select.
Is the following test case relevant?

define dso_local i16 @cmp_shrink_select_not_cmp(i8 %a, i8 %b, i32 %c, i32 %d) {
; CHECK-LABEL: @cmp_shrink_select_not_cmp(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[CONV:%.*]] = sext i8 [[A:%.*]] to i16
; CHECK-NEXT:    [[CONV2:%.*]] = sext i8 [[B:%.*]] to i16
; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[C:%.*]], [[D:%.*]]
; CHECK-NEXT:    [[COND:%.*]] = select i1 [[CMP]], i16 [[CONV2]], i16 [[CONV]]
; CHECK-NEXT:    ret i16 [[COND]]
;
entry:

  %conv = sext i8 %a to i32
  %conv2 = sext i8 %b to i32
  %cmp = icmp slt i32 %c, %d
  %cond = select i1 %cmp, i32 %conv2, i32 %conv
  %conv4 = trunc i32 %cond to i16
  ret i16 %conv4

}


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D74484/new/

https://reviews.llvm.org/D74484