[InstCombine] Add tests cases related to demanded use bits for ashr

When trying to improve value tracking in
   #97693
some regressions was found due to a "weirdness" in simplify demanded
use bits for ashr. Normally an ashr is replaced by lshr when the
shifted in bits aren't demanded. Some years ago (see commit
22178dd) there was a test case motivating to keep the ashr
when any sign bit (besides the shifted in bits) was demanded. The
weird part about it is that the better we get at analysing known sign
bits, the less likely it is that we canonicalize from ashr to lshr.
That makes it hard to tune other combines to work based on the
canonicalization, as well as possibly resulting in unexpected
regressions when improving value tracking.

This patch adds a test case for which it would be better to
canonicalize ashr into lshr when possible. Worth mentioning is also
that reverting 22178dd doesn't seem to cause regressions
in any other lit tests (not even the one added in 22178dd).

Author: bjope

llvm/test/Transforms/InstCombine/ashr-demand.ll

@@ -52,3 +52,39 @@ define <2 x i32> @srem2_ashr_mask_vector_nonconstant(<2 x i32> %a0, <2 x i32> %a
   %mask = and <2 x i32> %ashr, <i32 2, i32 2>
   ret <2 x i32> %mask
 }
+
+
+; If it does not matter if we do ashr or lshr, then we canonicalize to lshr.
+
+define i16 @ashr_can_be_lshr(i32 %a) {
+; CHECK-LABEL: @ashr_can_be_lshr(
+; CHECK-NEXT:    [[ASHR:%.*]] = lshr exact i32 [[A:%.*]], 16
+; CHECK-NEXT:    [[TRUNC:%.*]] = trunc nuw i32 [[ASHR]] to i16
+; CHECK-NEXT:    ret i16 [[TRUNC]]
+;
+  %ashr = ashr exact i32 %a, 16
+  %trunc = trunc nsw i32 %ashr to i16
+  ret i16 %trunc
+}
+
+; Historically SimplifyDemandedUseBits skipped replacing ashr with lshr here
+; due to known sign bits analysis indicating that %ashr had more than 33 sign
+; bits. It does however seem weird not to always canonicalize to lshr when
+; possible, and in this case rewriting into lshr would trigger further
+; optimizations.
+define i32 @ashr_can_be_lshr_2(i32 %a) {
+; CHECK-LABEL: @ashr_can_be_lshr_2(
+; CHECK-NEXT:    [[TMP1:%.*]] = or i32 [[A:%.*]], 1056964608
+; CHECK-NEXT:    [[OR:%.*]] = zext i32 [[TMP1]] to i64
+; CHECK-NEXT:    [[SHL:%.*]] = shl i64 [[OR]], 34
+; CHECK-NEXT:    [[ASHR:%.*]] = ashr exact i64 [[SHL]], 32
+; CHECK-NEXT:    [[TRUNC:%.*]] = trunc nsw i64 [[ASHR]] to i32
+; CHECK-NEXT:    ret i32 [[TRUNC]]
+;
+  %ext = zext i32 %a to i64
+  %or = or i64 %ext, 4278190080
+  %shl = shl i64 %or, 34
+  %ashr = ashr exact i64 %shl, 32
+  %trunc = trunc nsw i64 %ashr to i32
+  ret i32 %trunc
+}