23c2f2e6b2
This patch marks the induction increment of the main induction variable of the vector loop as NUW when not folding the tail. If the tail is not folded, we know that End - Start >= Step (either statically or through the minimum iteration checks). We also know that both Start % Step == 0 and End % Step == 0. We exit the vector loop if %IV + %Step == %End. Hence we must exit the loop before %IV + %Step unsigned overflows and we can mark the induction increment as NUW. This should make SCEV return more precise bounds for the created vector loops, used by later optimizations, like late unrolling. At the moment quite a few tests still need to be updated, but before doing so I'd like to get initial feedback to make sure I am not missing anything. Note that this could probably be further improved by using information from the original IV. Attempt of modeling of the assumption in Alive2: https://alive2.llvm.org/ce/z/H_DL_g Part of a set of fixes required for PR50412. Reviewed By: mkazantsev Differential Revision: https://reviews.llvm.org/D103255
88 lines
3.5 KiB
LLVM
88 lines
3.5 KiB
LLVM
; REQUIRES: asserts
|
|
; RUN: opt < %s -loop-vectorize -force-vector-width=2 -force-vector-interleave=1 -instcombine -debug-only=loop-vectorize -disable-output -print-after=instcombine -enable-new-pm=0 2>&1 | FileCheck %s
|
|
; RUN: opt < %s -passes='function(loop-vectorize,instcombine)' -force-vector-width=2 -force-vector-interleave=1 -debug-only=loop-vectorize -disable-output -print-after=instcombine 2>&1 | FileCheck %s
|
|
|
|
target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
|
|
|
|
; CHECK-LABEL: phi_two_incoming_values
|
|
; CHECK: LV: Found an estimated cost of 1 for VF 2 For instruction: %i = phi i64 [ %i.next, %if.end ], [ 0, %entry ]
|
|
; CHECK: LV: Found an estimated cost of 1 for VF 2 For instruction: %tmp5 = phi i32 [ %tmp1, %for.body ], [ %tmp4, %if.then ]
|
|
; CHECK: vector.body:
|
|
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
|
|
; CHECK: [[WIDE_LOAD:%.*]] = load <2 x i32>, <2 x i32>* {{.*}}
|
|
; CHECK: [[TMP5:%.*]] = icmp sgt <2 x i32> [[WIDE_LOAD]], zeroinitializer
|
|
; CHECK-NEXT: [[TMP6:%.*]] = zext <2 x i1> [[TMP5]] to <2 x i32>
|
|
; CHECK-NEXT: [[PREDPHI:%.*]] = add <2 x i32> [[WIDE_LOAD]], [[TMP6]]
|
|
; CHECK: store <2 x i32> [[PREDPHI]], <2 x i32>* {{.*}}
|
|
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
|
|
;
|
|
define void @phi_two_incoming_values(i32* %a, i32* %b, i64 %n) {
|
|
entry:
|
|
br label %for.body
|
|
|
|
for.body:
|
|
%i = phi i64 [ %i.next, %if.end ], [ 0, %entry ]
|
|
%tmp0 = getelementptr inbounds i32, i32* %a, i64 %i
|
|
%tmp1 = load i32, i32* %tmp0, align 4
|
|
%tmp2 = getelementptr inbounds i32, i32* %b, i64 %i
|
|
%tmp3 = icmp sgt i32 %tmp1, 0
|
|
br i1 %tmp3, label %if.then, label %if.end
|
|
|
|
if.then:
|
|
%tmp4 = add i32 %tmp1, 1
|
|
br label %if.end
|
|
|
|
if.end:
|
|
%tmp5 = phi i32 [ %tmp1, %for.body ], [ %tmp4, %if.then ]
|
|
store i32 %tmp5, i32* %tmp2, align 4
|
|
%i.next = add i64 %i, 1
|
|
%cond = icmp eq i64 %i, %n
|
|
br i1 %cond, label %for.end, label %for.body
|
|
|
|
for.end:
|
|
ret void
|
|
}
|
|
|
|
; CHECK-LABEL: phi_three_incoming_values
|
|
; CHECK: LV: Found an estimated cost of 1 for VF 2 For instruction: %i = phi i64 [ %i.next, %if.end ], [ 0, %entry ]
|
|
; CHECK: LV: Found an estimated cost of 2 for VF 2 For instruction: %tmp8 = phi i32 [ 9, %for.body ], [ 3, %if.then ], [ %tmp7, %if.else ]
|
|
; CHECK: vector.body:
|
|
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
|
|
; CHECK: [[PREDPHI:%.*]] = select <2 x i1> {{.*}}, <2 x i32> <i32 3, i32 3>, <2 x i32> <i32 9, i32 9>
|
|
; CHECK: [[PREDPHI7:%.*]] = select <2 x i1> {{.*}}, <2 x i32> {{.*}}, <2 x i32> [[PREDPHI]]
|
|
; CHECK: store <2 x i32> [[PREDPHI7]], <2 x i32>* {{.*}}
|
|
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
|
|
;
|
|
define void @phi_three_incoming_values(i32* %a, i32* %b, i64 %n) {
|
|
entry:
|
|
br label %for.body
|
|
|
|
for.body:
|
|
%i = phi i64 [ %i.next, %if.end ], [ 0, %entry ]
|
|
%tmp0 = getelementptr inbounds i32, i32* %a, i64 %i
|
|
%tmp1 = load i32, i32* %tmp0, align 4
|
|
%tmp2 = getelementptr inbounds i32, i32* %b, i64 %i
|
|
%tmp3 = load i32, i32* %tmp2, align 4
|
|
%tmp4 = icmp sgt i32 %tmp1, %tmp3
|
|
br i1 %tmp4, label %if.then, label %if.end
|
|
|
|
if.then:
|
|
%tmp5 = icmp sgt i32 %tmp1, 19
|
|
br i1 %tmp5, label %if.end, label %if.else
|
|
|
|
if.else:
|
|
%tmp6 = icmp slt i32 %tmp3, 4
|
|
%tmp7 = select i1 %tmp6, i32 4, i32 5
|
|
br label %if.end
|
|
|
|
if.end:
|
|
%tmp8 = phi i32 [ 9, %for.body ], [ 3, %if.then ], [ %tmp7, %if.else ]
|
|
store i32 %tmp8, i32* %tmp0, align 4
|
|
%i.next = add i64 %i, 1
|
|
%cond = icmp eq i64 %i, %n
|
|
br i1 %cond, label %for.end, label %for.body
|
|
|
|
for.end:
|
|
ret void
|
|
}
|