mirror of
https://github.com/RPCS3/llvm.git
synced 2024-12-30 16:34:03 +00:00
160f9b9c10
Prior to this change, the loop vectorizer did not make use of the alias
analysis infrastructure. Instead, it performed memory dependence analysis using
ScalarEvolution-based linear dependence checks within equivalence classes
derived from the results of ValueTracking's GetUnderlyingObjects.
Unfortunately, this meant that:
1. The loop vectorizer had logic that essentially duplicated that in BasicAA
for aliasing based on identified objects.
2. The loop vectorizer could not partition the space of dependency checks
based on information only easily available from within AA (TBAA metadata is
currently the prime example).
This means, for example, regardless of whether -fno-strict-aliasing was
provided, the vectorizer would only vectorize this loop with a runtime
memory-overlap check:
void foo(int *a, float *b) {
for (int i = 0; i < 1600; ++i)
a[i] = b[i];
}
This is suboptimal because the TBAA metadata already provides the information
necessary to show that this check unnecessary. Of course, the vectorizer has a
limit on the number of such checks it will insert, so in practice, ignoring
TBAA means not vectorizing more-complicated loops that we should.
This change causes the vectorizer to use an AliasSetTracker to keep track of
the pointers in the loop. The resulting alias sets are then used to partition
the space of dependency checks, and potential runtime checks; this results in
more-efficient vectorizations.
When pointer locations are added to the AliasSetTracker, two things are done:
1. The location size is set to UnknownSize (otherwise you'd not catch
inter-iteration dependencies)
2. For instructions in blocks that would need to be predicated, TBAA is
removed (because the metadata might have a control dependency on the condition
being speculated).
For non-predicated blocks, you can leave the TBAA metadata. This is safe
because you can't have an iteration dependency on the TBAA metadata (if you
did, and you unrolled sufficiently, you'd end up with the same pointer value
used by two accesses that TBAA says should not alias, and that would yield
undefined behavior).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@213486 91177308-0d34-0410-b5e6-96231b3b80d8
236 lines
8.1 KiB
LLVM
236 lines
8.1 KiB
LLVM
; RUN: opt -S -march=r600 -mcpu=cayman -basicaa -loop-vectorize -force-vector-unroll=1 -force-vector-width=4 -dce -instcombine < %s | FileCheck %s
|
|
|
|
; Check vectorization that would ordinarily require a runtime bounds
|
|
; check on the pointers when mixing address spaces. For now we cannot
|
|
; assume address spaces do not alias, and we can't assume that
|
|
; different pointers are directly comparable.
|
|
;
|
|
; These all test this basic loop for different combinations of address
|
|
; spaces, and swapping in globals or adding noalias.
|
|
;
|
|
;void foo(int addrspace(N)* [noalias] a, int addrspace(M)* [noalias] b, int n)
|
|
;{
|
|
; for (int i = 0; i < n; ++i)
|
|
; {
|
|
; a[i] = 3 * b[i];
|
|
; }
|
|
;}
|
|
|
|
; Artificial datalayout
|
|
target datalayout = "e-p:32:32:32-p1:16:16:16-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-v16:16:16-v24:32:32-v32:32:32-v48:64:64-v64:64:64-v96:128:128-v128:128:128-v192:256:256-v256:256:256-v512:512:512-v1024:1024:1024-v2048:2048:2048-n32:64"
|
|
|
|
|
|
@g_as1 = common addrspace(1) global [1024 x i32] zeroinitializer, align 16
|
|
@q_as2 = common addrspace(2) global [1024 x i32] zeroinitializer, align 16
|
|
|
|
; Both parameters are unidentified objects with the same address
|
|
; space, so this should vectorize normally.
|
|
define void @foo(i32 addrspace(1)* %a, i32 addrspace(1)* %b, i32 %n) #0 {
|
|
; CHECK-LABEL: @foo(
|
|
; CHECK: <4 x i32>
|
|
; CHECK: ret
|
|
|
|
entry:
|
|
br label %for.cond
|
|
|
|
for.cond: ; preds = %for.body, %entry
|
|
%i.0 = phi i32 [ 0, %entry ], [ %inc, %for.body ]
|
|
%cmp = icmp slt i32 %i.0, %n
|
|
br i1 %cmp, label %for.body, label %for.end
|
|
|
|
for.body: ; preds = %for.cond
|
|
%idxprom = sext i32 %i.0 to i64
|
|
%arrayidx = getelementptr inbounds i32 addrspace(1)* %b, i64 %idxprom
|
|
%0 = load i32 addrspace(1)* %arrayidx, align 4
|
|
%mul = mul nsw i32 %0, 3
|
|
%idxprom1 = sext i32 %i.0 to i64
|
|
%arrayidx2 = getelementptr inbounds i32 addrspace(1)* %a, i64 %idxprom1
|
|
store i32 %mul, i32 addrspace(1)* %arrayidx2, align 4
|
|
%inc = add nsw i32 %i.0, 1
|
|
br label %for.cond
|
|
|
|
for.end: ; preds = %for.cond
|
|
ret void
|
|
}
|
|
|
|
; Parameters are unidentified and different address spaces, so cannot vectorize.
|
|
define void @bar0(i32* %a, i32 addrspace(1)* %b, i32 %n) #0 {
|
|
; CHECK-LABEL: @bar0(
|
|
; CHECK-NOT: <4 x i32>
|
|
; CHECK: ret
|
|
|
|
entry:
|
|
br label %for.cond
|
|
|
|
for.cond: ; preds = %for.body, %entry
|
|
%i.0 = phi i32 [ 0, %entry ], [ %inc, %for.body ]
|
|
%cmp = icmp slt i32 %i.0, %n
|
|
br i1 %cmp, label %for.body, label %for.end
|
|
|
|
for.body: ; preds = %for.cond
|
|
%idxprom = sext i32 %i.0 to i64
|
|
%arrayidx = getelementptr inbounds i32 addrspace(1)* %b, i64 %idxprom
|
|
%0 = load i32 addrspace(1)* %arrayidx, align 4
|
|
%mul = mul nsw i32 %0, 3
|
|
%idxprom1 = sext i32 %i.0 to i64
|
|
%arrayidx2 = getelementptr inbounds i32* %a, i64 %idxprom1
|
|
store i32 %mul, i32* %arrayidx2, align 4
|
|
%inc = add nsw i32 %i.0, 1
|
|
br label %for.cond
|
|
|
|
for.end: ; preds = %for.cond
|
|
ret void
|
|
}
|
|
|
|
; Swapped arguments should be the same
|
|
define void @bar1(i32 addrspace(1)* %a, i32* %b, i32 %n) #0 {
|
|
; CHECK-LABEL: @bar1(
|
|
; CHECK-NOT: <4 x i32>
|
|
; CHECK: ret
|
|
|
|
entry:
|
|
br label %for.cond
|
|
|
|
for.cond: ; preds = %for.body, %entry
|
|
%i.0 = phi i32 [ 0, %entry ], [ %inc, %for.body ]
|
|
%cmp = icmp slt i32 %i.0, %n
|
|
br i1 %cmp, label %for.body, label %for.end
|
|
|
|
for.body: ; preds = %for.cond
|
|
%idxprom = sext i32 %i.0 to i64
|
|
%arrayidx = getelementptr inbounds i32* %b, i64 %idxprom
|
|
%0 = load i32* %arrayidx, align 4
|
|
%mul = mul nsw i32 %0, 3
|
|
%idxprom1 = sext i32 %i.0 to i64
|
|
%arrayidx2 = getelementptr inbounds i32 addrspace(1)* %a, i64 %idxprom1
|
|
store i32 %mul, i32 addrspace(1)* %arrayidx2, align 4
|
|
%inc = add nsw i32 %i.0, 1
|
|
br label %for.cond
|
|
|
|
for.end: ; preds = %for.cond
|
|
ret void
|
|
}
|
|
|
|
; We should still be able to vectorize with noalias even if the
|
|
; address spaces are different.
|
|
define void @bar2(i32* noalias %a, i32 addrspace(1)* noalias %b, i32 %n) #0 {
|
|
; CHECK-LABEL: @bar2(
|
|
; CHECK: <4 x i32>
|
|
; CHECK: ret
|
|
|
|
entry:
|
|
br label %for.cond
|
|
|
|
for.cond: ; preds = %for.body, %entry
|
|
%i.0 = phi i32 [ 0, %entry ], [ %inc, %for.body ]
|
|
%cmp = icmp slt i32 %i.0, %n
|
|
br i1 %cmp, label %for.body, label %for.end
|
|
|
|
for.body: ; preds = %for.cond
|
|
%idxprom = sext i32 %i.0 to i64
|
|
%arrayidx = getelementptr inbounds i32 addrspace(1)* %b, i64 %idxprom
|
|
%0 = load i32 addrspace(1)* %arrayidx, align 4
|
|
%mul = mul nsw i32 %0, 3
|
|
%idxprom1 = sext i32 %i.0 to i64
|
|
%arrayidx2 = getelementptr inbounds i32* %a, i64 %idxprom1
|
|
store i32 %mul, i32* %arrayidx2, align 4
|
|
%inc = add nsw i32 %i.0, 1
|
|
br label %for.cond
|
|
|
|
for.end: ; preds = %for.cond
|
|
ret void
|
|
}
|
|
|
|
; Store to identified global with different address space. This isn't
|
|
; generally safe and shouldn't be vectorized.
|
|
define void @arst0(i32* %b, i32 %n) #0 {
|
|
; CHECK-LABEL: @arst0(
|
|
; CHECK-NOT: <4 x i32>
|
|
; CHECK: ret
|
|
|
|
entry:
|
|
br label %for.cond
|
|
|
|
for.cond: ; preds = %for.body, %entry
|
|
%i.0 = phi i32 [ 0, %entry ], [ %inc, %for.body ]
|
|
%cmp = icmp slt i32 %i.0, %n
|
|
br i1 %cmp, label %for.body, label %for.end
|
|
|
|
for.body: ; preds = %for.cond
|
|
%idxprom = sext i32 %i.0 to i64
|
|
%arrayidx = getelementptr inbounds i32* %b, i64 %idxprom
|
|
%0 = load i32* %arrayidx, align 4
|
|
%mul = mul nsw i32 %0, 3
|
|
%idxprom1 = sext i32 %i.0 to i64
|
|
%arrayidx2 = getelementptr inbounds [1024 x i32] addrspace(1)* @g_as1, i64 0, i64 %idxprom1
|
|
store i32 %mul, i32 addrspace(1)* %arrayidx2, align 4
|
|
%inc = add nsw i32 %i.0, 1
|
|
br label %for.cond
|
|
|
|
for.end: ; preds = %for.cond
|
|
ret void
|
|
}
|
|
|
|
|
|
; Load from identified global with different address space.
|
|
; This isn't generally safe and shouldn't be vectorized.
|
|
define void @arst1(i32* %b, i32 %n) #0 {
|
|
; CHECK-LABEL: @arst1(
|
|
; CHECK-NOT: <4 x i32>
|
|
; CHECK: ret
|
|
|
|
entry:
|
|
br label %for.cond
|
|
|
|
for.cond: ; preds = %for.body, %entry
|
|
%i.0 = phi i32 [ 0, %entry ], [ %inc, %for.body ]
|
|
%cmp = icmp slt i32 %i.0, %n
|
|
br i1 %cmp, label %for.body, label %for.end
|
|
|
|
for.body: ; preds = %for.cond
|
|
%idxprom = sext i32 %i.0 to i64
|
|
%arrayidx = getelementptr inbounds [1024 x i32] addrspace(1)* @g_as1, i64 0, i64 %idxprom
|
|
%0 = load i32 addrspace(1)* %arrayidx, align 4
|
|
%mul = mul nsw i32 %0, 3
|
|
%idxprom1 = sext i32 %i.0 to i64
|
|
%arrayidx2 = getelementptr inbounds i32* %b, i64 %idxprom1
|
|
store i32 %mul, i32* %arrayidx2, align 4
|
|
%inc = add nsw i32 %i.0, 1
|
|
br label %for.cond
|
|
|
|
for.end: ; preds = %for.cond
|
|
ret void
|
|
}
|
|
|
|
; Read and write to 2 identified globals in different address
|
|
; spaces. This should be vectorized.
|
|
define void @aoeu(i32 %n) #0 {
|
|
; CHECK-LABEL: @aoeu(
|
|
; CHECK: <4 x i32>
|
|
; CHECK: ret
|
|
|
|
entry:
|
|
br label %for.cond
|
|
|
|
for.cond: ; preds = %for.body, %entry
|
|
%i.0 = phi i32 [ 0, %entry ], [ %inc, %for.body ]
|
|
%cmp = icmp slt i32 %i.0, %n
|
|
br i1 %cmp, label %for.body, label %for.end
|
|
|
|
for.body: ; preds = %for.cond
|
|
%idxprom = sext i32 %i.0 to i64
|
|
%arrayidx = getelementptr inbounds [1024 x i32] addrspace(2)* @q_as2, i64 0, i64 %idxprom
|
|
%0 = load i32 addrspace(2)* %arrayidx, align 4
|
|
%mul = mul nsw i32 %0, 3
|
|
%idxprom1 = sext i32 %i.0 to i64
|
|
%arrayidx2 = getelementptr inbounds [1024 x i32] addrspace(1)* @g_as1, i64 0, i64 %idxprom1
|
|
store i32 %mul, i32 addrspace(1)* %arrayidx2, align 4
|
|
%inc = add nsw i32 %i.0, 1
|
|
br label %for.cond
|
|
|
|
for.end: ; preds = %for.cond
|
|
ret void
|
|
}
|
|
|
|
attributes #0 = { nounwind uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
|