3097c60928
Currently it's very difficult to improve the cost model for tail-folded loops because as soon as you add a VPInstruction::computeCost function that adds the costs of instructions such as VPInstruction::ActiveLaneMask and VPInstruction::ExplicitVectorLength the assert in LoopVectorizationPlanner::computeBestVF fails for some tests. This is because the VF chosen by the legacy cost model doesn't match the vplan cost model. See PR #90191. This assert is currently making it difficult to improve the cost model. Hopefully we will be in a position to remove the assert soon, however in order to do that we have to fix up a whole bunch of tests that rely upon the legacy cost model output. I've tried my best to update these tests to use vplan output instead. There is still work needed for the VF=1 case because the vplan cost model is not printed out in this case. I've not attempted to fix those in this patch.
69 lines
2.6 KiB
LLVM
69 lines
2.6 KiB
LLVM
; RUN: opt < %s -passes=loop-vectorize -mattr=+sse4.2 -debug-only=loop-vectorize 2>&1 -S | FileCheck %s
|
|
; REQUIRES: asserts
|
|
; Make sure we use the right select kind when querying select costs.
|
|
|
|
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
|
|
target triple = "x86_64-apple-macosx10.8.0"
|
|
|
|
@a = common global [2048 x i32] zeroinitializer, align 16
|
|
@b = common global [2048 x i32] zeroinitializer, align 16
|
|
@c = common global [2048 x i32] zeroinitializer, align 16
|
|
|
|
; CHECK: Checking a loop in 'scalarselect'
|
|
define void @scalarselect(i1 %cond) {
|
|
br label %1
|
|
|
|
; <label>:1
|
|
%indvars.iv = phi i64 [ 0, %0 ], [ %indvars.iv.next, %1 ]
|
|
%2 = getelementptr inbounds [2048 x i32], ptr @b, i64 0, i64 %indvars.iv
|
|
%3 = load i32, ptr %2, align 4
|
|
%4 = getelementptr inbounds [2048 x i32], ptr @c, i64 0, i64 %indvars.iv
|
|
%5 = load i32, ptr %4, align 4
|
|
%6 = add nsw i32 %5, %3
|
|
%7 = getelementptr inbounds [2048 x i32], ptr @a, i64 0, i64 %indvars.iv
|
|
|
|
; CHECK: cost of 1 for VF 1 {{.*}} select i1 %cond, i32 %6, i32 0
|
|
; CHECK: Cost of 2 for VF 2: WIDEN-SELECT ir<%sel> = select ir<%cond>, ir<%6>, ir<0> (condition is loop invariant)
|
|
; CHECK: Cost of 2 for VF 4: WIDEN-SELECT ir<%sel> = select ir<%cond>, ir<%6>, ir<0> (condition is loop invariant)
|
|
|
|
%sel = select i1 %cond, i32 %6, i32 zeroinitializer
|
|
store i32 %sel, ptr %7, align 4
|
|
%indvars.iv.next = add i64 %indvars.iv, 1
|
|
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
|
|
%exitcond = icmp eq i32 %lftr.wideiv, 256
|
|
br i1 %exitcond, label %8, label %1
|
|
|
|
; <label>:8
|
|
ret void
|
|
}
|
|
|
|
; CHECK: Checking a loop in 'vectorselect'
|
|
define void @vectorselect(i1 %cond) {
|
|
br label %1
|
|
|
|
; <label>:1
|
|
%indvars.iv = phi i64 [ 0, %0 ], [ %indvars.iv.next, %1 ]
|
|
%2 = getelementptr inbounds [2048 x i32], ptr @b, i64 0, i64 %indvars.iv
|
|
%3 = load i32, ptr %2, align 4
|
|
%4 = getelementptr inbounds [2048 x i32], ptr @c, i64 0, i64 %indvars.iv
|
|
%5 = load i32, ptr %4, align 4
|
|
%6 = add nsw i32 %5, %3
|
|
%7 = getelementptr inbounds [2048 x i32], ptr @a, i64 0, i64 %indvars.iv
|
|
%8 = icmp ult i64 %indvars.iv, 8
|
|
|
|
; CHECK: cost of 1 for VF 1 {{.*}} select i1 %8, i32 %6, i32 0
|
|
; CHECK: Cost of 2 for VF 2: WIDEN-SELECT ir<%sel> = select ir<%8>, ir<%6>, ir<0>
|
|
; CHECK: Cost of 2 for VF 4: WIDEN-SELECT ir<%sel> = select ir<%8>, ir<%6>, ir<0>
|
|
|
|
%sel = select i1 %8, i32 %6, i32 zeroinitializer
|
|
store i32 %sel, ptr %7, align 4
|
|
%indvars.iv.next = add i64 %indvars.iv, 1
|
|
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
|
|
%exitcond = icmp eq i32 %lftr.wideiv, 256
|
|
br i1 %exitcond, label %9, label %1
|
|
|
|
; <label>:9
|
|
ret void
|
|
}
|
|
|