[llvm] [LV] Improve cost model for some replicating recipes (PR #192056)

[David Sherwood via llvm-commits]

================
@@ -380,22 +380,70 @@ exit:
 define void @sdiv_power_of_2_divisor_in_replicate_region(i32 %x, ptr %dst, i64 %n) {
 ; CHECK-LABEL: define void @sdiv_power_of_2_divisor_in_replicate_region(
 ; CHECK-SAME: i32 [[X:%.*]], ptr [[DST:%.*]], i64 [[N:%.*]]) {
-; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:  [[ENTRY:.*:]]
 ; CHECK-NEXT:    [[C_1:%.*]] = icmp sgt i64 [[N]], 0
 ; CHECK-NEXT:    [[C_2:%.*]] = icmp slt i64 [[N]], 9
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[C_1]])
 ; CHECK-NEXT:    call void @llvm.assume(i1 [[C_2]])
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
-; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N]], 3
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
+; CHECK:       [[VECTOR_BODY]]:
----------------
david-arm wrote:

It's possibly not profitable, but then I don't think that's because of this PR. I think it's clear from the IR output that the result of the sdiv is never inserted into a vector and so we shouldn't be artificially inflating the cost of the sdiv just to hide problems elsewhere.

I'm happy to look into whether there is an independent cost model issue and also I can test this locally as a simple benchmark to see whether it really is better or not. It's worth noting that this PR is a requirement for follow-on PRs that stop skipping computation for other opcodes such as fadd, etc. I will share one of these follow-on PRs, which should make it clearer why this PR is needed.

https://github.com/llvm/llvm-project/pull/192056