268467869b
In rL316552, we ban intersection of unsigned latch range with signed range check and vice
versa, unless the entire range check iteration space is known positive. It was a correct
functional fix that saved us from dealing with ambiguous values, but it also appeared
to be a very restrictive limitation. In particular, in the following case:
loop:
%iv = phi i32 [ 0, %preheader ], [ %iv.next, %latch]
%iv.offset = add i32 %iv, 10
%rc = icmp slt i32 %iv.offset, %len
br i1 %rc, label %latch, label %deopt
latch:
%iv.next = add i32 %iv, 11
%cond = icmp i32 ult %iv.next, 100
br it %cond, label %loop, label %exit
Here, the unsigned iteration range is `[0, 100)`, and the safe range for range
check is `[-10, %len - 10)`. For unsigned iteration spaces, we use unsigned
min/max functions for range intersection. Given this, we wanted to avoid dealing
with `-10` because it is interpreted as a very big unsigned value. Semantically, range
check's safe range goes through unsigned border, so in fact it is two disjoint
ranges in IV's iteration space. Intersection of such ranges is not trivial, so we prohibited
this case saying that we are not allowed to intersect such ranges.
What semantics of this safe range actually means is that we can start from `-10` and go
up increasing the `%iv` by one until we reach `%len - 10` (for simplicity let's assume that
`%len - 10` is a reasonably big positive value).
In particular, this safe iteration space includes `0, 1, 2, ..., %len - 11`. So if we were able to return
safe iteration space `[0, %len - 10)`, we could safely intersect it with IV's iteration space. All
values in this range are non-negative, so using signed/unsigned min/max for them is unambiguous.
In this patch, we alter the algorithm of safe range calculation so that it returnes a subset of the
original safe space which is represented by one continuous range that does not go through wrap.
In order to reach this, we use modified SCEV substraction function. It can be imagined as a function
that substracts by `1` (or `-1`) as long as the further substraction does not cause a wrap in IV iteration
space. This allows us to perform IRCE in many situations when we deal with IV space and range check
of different types (in terms of signed/unsigned).
We apply this approach for both matching and not matching types of IV iteration space and the
range check. One implication of this is that now IRCE became smarter in detection of empty safe
ranges. For example, in this case:
loop:
%iv = phi i32 [ %begin, %preheader ], [ %iv.next, %latch]
%iv.offset = sub i32 %iv, 10
%rc = icmp ult i32 %iv.offset, %len
br i1 %rc, label %latch, label %deopt
latch:
%iv.next = add i32 %iv, 11
%cond = icmp i32 ult %iv.next, 100
br it %cond, label %loop, label %exit
If `%len` was less than 10 but SCEV failed to trivially prove that `%begin - 10 >u %len- 10`,
we could end up executing entire loop in safe preloop while the main loop was still generated,
but never executed. Now, cutting the ranges so that if both `begin - 10` and `%len - 10` overflow,
we have a trivially empty range of `[0, 0)`. This in some cases prevents us from meaningless optimization.
Differential Revision: https://reviews.llvm.org/D39954
llvm-svn: 318639
95 lines
3.8 KiB
LLVM
95 lines
3.8 KiB
LLVM
; RUN: opt -verify-loop-info -irce-print-changed-loops -irce -S < %s 2>&1 | FileCheck %s
|
|
|
|
; The test demonstrates that incorrect behavior of Clamp may lead to incorrect
|
|
; calculation of post-loop exit condition.
|
|
|
|
; CHECK-LABEL: irce: in function test_01: constrained Loop at depth 1 containing: %loop<header><exiting>,%in_bounds<exiting>,%not_zero<latch><exiting>
|
|
; CHECK-NOT: irce: in function test_02: constrained Loop
|
|
|
|
define void @test_01() {
|
|
|
|
; CHECK-LABEL: test_01
|
|
|
|
entry:
|
|
%indvars.iv.next467 = add nuw nsw i64 2, 1
|
|
%length.i167 = load i32, i32 addrspace(1)* undef, align 8
|
|
%tmp21 = zext i32 %length.i167 to i64
|
|
%tmp34 = load atomic i32, i32 addrspace(1)* undef unordered, align 4
|
|
%tmp35 = add i32 %tmp34, -9581
|
|
%tmp36 = icmp ugt i32 %length.i167, 1
|
|
br i1 %tmp36, label %preheader, label %exit
|
|
|
|
exit: ; preds = %in_bounds, %loop, %not_zero, %entry
|
|
ret void
|
|
|
|
preheader: ; preds = %entry
|
|
; CHECK: preheader:
|
|
; CHECK-NEXT: %length_gep.i146 = getelementptr inbounds i8, i8 addrspace(1)* undef, i64 8
|
|
; CHECK-NEXT: %length_gep_typed.i147 = bitcast i8 addrspace(1)* undef to i32 addrspace(1)*
|
|
; CHECK-NEXT: %tmp43 = icmp ult i64 %indvars.iv.next467, %tmp21
|
|
; CHECK-NEXT: [[C0:%[^ ]+]] = icmp ugt i64 %tmp21, 1
|
|
; CHECK-NEXT: %exit.mainloop.at = select i1 [[C0]], i64 %tmp21, i64 1
|
|
; CHECK-NEXT: [[C1:%[^ ]+]] = icmp ult i64 1, %exit.mainloop.at
|
|
; CHECK-NEXT: br i1 [[C1]], label %loop.preheader, label %main.pseudo.exit
|
|
|
|
%length_gep.i146 = getelementptr inbounds i8, i8 addrspace(1)* undef, i64 8
|
|
%length_gep_typed.i147 = bitcast i8 addrspace(1)* undef to i32 addrspace(1)*
|
|
%tmp43 = icmp ult i64 %indvars.iv.next467, %tmp21
|
|
br label %loop
|
|
|
|
not_zero: ; preds = %in_bounds
|
|
; CHECK: not_zero:
|
|
; CHECK: %tmp56 = icmp ult i64 %indvars.iv.next, %tmp21
|
|
; CHECK-NEXT: [[COND:%[^ ]+]] = icmp ult i64 %indvars.iv.next, %exit.mainloop.at
|
|
; CHECK-NEXT: br i1 [[COND]], label %loop, label %main.exit.selector
|
|
|
|
%tmp51 = trunc i64 %indvars.iv.next to i32
|
|
%tmp53 = mul i32 %tmp51, %tmp51
|
|
%tmp54 = add i32 %tmp53, -9582
|
|
%tmp55 = add i32 %tmp54, %tmp62
|
|
%tmp56 = icmp ult i64 %indvars.iv.next, %tmp21
|
|
br i1 %tmp56, label %loop, label %exit
|
|
|
|
loop: ; preds = %not_zero, %preheader
|
|
%tmp62 = phi i32 [ 1, %preheader ], [ %tmp55, %not_zero ]
|
|
%indvars.iv750 = phi i64 [ 1, %preheader ], [ %indvars.iv.next, %not_zero ]
|
|
%length.i148 = load i32, i32 addrspace(1)* %length_gep_typed.i147, align 8
|
|
%tmp68 = zext i32 %length.i148 to i64
|
|
%tmp97 = icmp ult i64 2, %tmp68
|
|
%or.cond = and i1 %tmp43, %tmp97
|
|
%tmp99 = icmp ult i64 %indvars.iv750, %tmp21
|
|
%or.cond1 = and i1 %or.cond, %tmp99
|
|
br i1 %or.cond1, label %in_bounds, label %exit
|
|
|
|
in_bounds: ; preds = %loop
|
|
%indvars.iv.next = add nuw nsw i64 %indvars.iv750, 3
|
|
%tmp107 = icmp ult i64 %indvars.iv.next, 2
|
|
br i1 %tmp107, label %not_zero, label %exit
|
|
}
|
|
|
|
define void @test_02() {
|
|
|
|
; Now IRCE is smart enough to understand that the safe range here is empty.
|
|
; Previously it executed the entire loop in safe preloop and never actually
|
|
; entered the main loop.
|
|
|
|
entry:
|
|
br label %loop
|
|
|
|
loop: ; preds = %in_bounds, %entry
|
|
%iv1 = phi i64 [ 3, %entry ], [ %iv1.next, %in_bounds ]
|
|
%iv2 = phi i64 [ 4294967295, %entry ], [ %iv2.next, %in_bounds ]
|
|
%iv2.offset = add i64 %iv2, 1
|
|
%rc = icmp ult i64 %iv2.offset, 400
|
|
br i1 %rc, label %in_bounds, label %bci_321
|
|
|
|
bci_321: ; preds = %in_bounds, %loop
|
|
ret void
|
|
|
|
in_bounds: ; preds = %loop
|
|
%iv1.next = add nuw nsw i64 %iv1, 2
|
|
%iv2.next = add nuw nsw i64 %iv2, 2
|
|
%cond = icmp ugt i64 %iv1, 204
|
|
br i1 %cond, label %bci_321, label %loop
|
|
}
|