09e3697f44
This patch allows invariant loads to be used in the SCoP description,
e.g., as loop bounds, conditions or in memory access functions.
First we collect "required invariant loads" during SCoP detection that
would otherwise make an expression we care about non-affine. To this
end a new level of abstraction was introduced before
SCEVValidator::isAffineExpr() namely ScopDetection::isAffine() and
ScopDetection::onlyValidRequiredInvariantLoads(). Here we can decide
if we want a load inside the region to be optimistically assumed
invariant or not. If we do, it will be marked as required and in the
SCoP generation we bail if it is actually not invariant. If we don't
it will be a non-affine expression as before. At the moment we
optimistically assume all "hoistable" (namely non-loop-carried) loads
to be invariant. This causes us to expand some SCoPs and dismiss them
later but it also allows us to detect a lot we would dismiss directly
if we would ask e.g., AliasAnalysis::canBasicBlockModify(). We also
allow potential aliases between optimistically assumed invariant loads
and other pointers as our runtime alias checks are sound in case the
loads are actually invariant. Together with the invariant checks this
combination allows to handle a lot more than LICM can.
The code generation of the invariant loads had to be extended as we
can now have dependences between parameters and invariant (hoisted)
loads as well as the other way around, e.g.,
test/Isl/CodeGen/invariant_load_parameters_cyclic_dependence.ll
First, it is important to note that we cannot have real cycles but
only dependences from a hoisted load to a parameter and from another
parameter to that hoisted load (and so on). To handle such cases we
materialize llvm::Values for parameters that are referred by a hoisted
load on demand and then materialize the remaining parameters. Second,
there are new kinds of dependences between hoisted loads caused by the
constraints on their execution. If a hoisted load is conditionally
executed it might depend on the value of another hoisted load. To deal
with such situations we sort them already in the ScopInfo such that
they can be generated in the order they are listed in the
Scop::InvariantAccesses list (see compareInvariantAccesses). The
dependences between hoisted loads caused by indirect accesses are
handled the same way as before.
llvm-svn: 249607
103 lines
4.7 KiB
LLVM
103 lines
4.7 KiB
LLVM
; RUN: opt %loadPolly -basicaa -polly-ast -analyze < %s | FileCheck %s
|
|
|
|
;#include <string.h>
|
|
;#include <stdio.h>
|
|
;#define N 1021
|
|
;
|
|
;int main () {
|
|
; int i;
|
|
; int A[N];
|
|
; int RED[1];
|
|
;
|
|
; memset(A, 0, sizeof(int) * N);
|
|
;
|
|
; A[0] = 1;
|
|
; A[1] = 1;
|
|
; RED[0] = 0;
|
|
;
|
|
; for (i = 2; i < N; i++) {
|
|
; A[i] = A[i-1] + A[i-2];
|
|
; RED[0] += A[i-2];
|
|
; }
|
|
;
|
|
; if (RED[0] != 382399368)
|
|
; return 1;
|
|
;}
|
|
|
|
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"
|
|
|
|
define i32 @main() nounwind {
|
|
entry:
|
|
%A = alloca [1021 x i32], align 4 ; <[1021 x i32]*> [#uses=6]
|
|
%RED = alloca [1 x i32], align 4 ; <[1 x i32]*> [#uses=3]
|
|
%arraydecay = getelementptr inbounds [1021 x i32], [1021 x i32]* %A, i32 0, i32 0 ; <i32*> [#uses=1]
|
|
%conv = bitcast i32* %arraydecay to i8* ; <i8*> [#uses=1]
|
|
call void @llvm.memset.p0i8.i64(i8* %conv, i8 0, i64 4084, i32 1, i1 false)
|
|
%arraydecay1 = getelementptr inbounds [1021 x i32], [1021 x i32]* %A, i32 0, i32 0 ; <i32*> [#uses=1]
|
|
%arrayidx = getelementptr inbounds i32, i32* %arraydecay1, i64 0 ; <i32*> [#uses=1]
|
|
store i32 1, i32* %arrayidx
|
|
%arraydecay2 = getelementptr inbounds [1021 x i32], [1021 x i32]* %A, i32 0, i32 0 ; <i32*> [#uses=1]
|
|
%arrayidx3 = getelementptr inbounds i32, i32* %arraydecay2, i64 1 ; <i32*> [#uses=1]
|
|
store i32 1, i32* %arrayidx3
|
|
%arraydecay4 = getelementptr inbounds [1 x i32], [1 x i32]* %RED, i32 0, i32 0 ; <i32*> [#uses=1]
|
|
%arrayidx5 = getelementptr inbounds i32, i32* %arraydecay4, i64 0 ; <i32*> [#uses=1]
|
|
store i32 0, i32* %arrayidx5
|
|
br label %for.cond
|
|
|
|
for.cond: ; preds = %for.inc, %entry
|
|
%indvar = phi i64 [ %indvar.next, %for.inc ], [ 0, %entry ] ; <i64> [#uses=5]
|
|
%arrayidx15 = getelementptr [1021 x i32], [1021 x i32]* %A, i64 0, i64 %indvar ; <i32*> [#uses=2]
|
|
%tmp = add i64 %indvar, 2 ; <i64> [#uses=1]
|
|
%arrayidx20 = getelementptr [1021 x i32], [1021 x i32]* %A, i64 0, i64 %tmp ; <i32*> [#uses=1]
|
|
%tmp1 = add i64 %indvar, 1 ; <i64> [#uses=1]
|
|
%arrayidx9 = getelementptr [1021 x i32], [1021 x i32]* %A, i64 0, i64 %tmp1 ; <i32*> [#uses=1]
|
|
%exitcond = icmp ne i64 %indvar, 1019 ; <i1> [#uses=1]
|
|
br i1 %exitcond, label %for.body, label %for.end
|
|
|
|
for.body: ; preds = %for.cond
|
|
%tmp10 = load i32, i32* %arrayidx9 ; <i32> [#uses=1]
|
|
%tmp16 = load i32, i32* %arrayidx15 ; <i32> [#uses=1]
|
|
%add = add nsw i32 %tmp10, %tmp16 ; <i32> [#uses=1]
|
|
store i32 %add, i32* %arrayidx20
|
|
%tmp26 = load i32, i32* %arrayidx15 ; <i32> [#uses=1]
|
|
%arraydecay27 = getelementptr inbounds [1 x i32], [1 x i32]* %RED, i32 0, i32 0 ; <i32*> [#uses=1]
|
|
%arrayidx28 = getelementptr inbounds i32, i32* %arraydecay27, i64 0 ; <i32*> [#uses=2]
|
|
%tmp29 = load i32, i32* %arrayidx28 ; <i32> [#uses=1]
|
|
%add30 = add nsw i32 %tmp29, %tmp26 ; <i32> [#uses=1]
|
|
store i32 %add30, i32* %arrayidx28
|
|
br label %for.inc
|
|
|
|
for.inc: ; preds = %for.body
|
|
%indvar.next = add i64 %indvar, 1 ; <i64> [#uses=1]
|
|
br label %for.cond
|
|
|
|
for.end: ; preds = %for.cond
|
|
%arraydecay32 = getelementptr inbounds [1 x i32], [1 x i32]* %RED, i32 0, i32 0 ; <i32*> [#uses=1]
|
|
%arrayidx33 = getelementptr inbounds i32, i32* %arraydecay32, i64 0 ; <i32*> [#uses=1]
|
|
%tmp34 = load i32, i32* %arrayidx33 ; <i32> [#uses=1]
|
|
%cmp35 = icmp ne i32 %tmp34, 382399368 ; <i1> [#uses=1]
|
|
br i1 %cmp35, label %if.then, label %if.end
|
|
|
|
if.then: ; preds = %for.end
|
|
br label %if.end
|
|
|
|
if.end: ; preds = %if.then, %for.end
|
|
%retval.0 = phi i32 [ 1, %if.then ], [ 0, %for.end ] ; <i32> [#uses=1]
|
|
ret i32 %retval.0
|
|
}
|
|
|
|
declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i32, i1) nounwind
|
|
|
|
; Negative test. At the moment we will optimistically assume RED[0] in the conditional after the
|
|
; loop might be invariant and expand the SCoP from the loop to include the conditional. However,
|
|
; during SCoP generation we will realize that RED[0] is in fact not invariant and bail.
|
|
;
|
|
; Possible solutions could be:
|
|
; - Do not optimistically assume it to be invariant (as before this commit), however we would loose
|
|
; a lot of invariant cases due to possible aliasing.
|
|
; - Reduce the size of the SCoP if an assumed invariant access is in fact not invariant instead of
|
|
; rejecting the whole region.
|
|
;
|
|
; CHECK-NOT: for (int c0 = 0; c0 <= 1018; c0 += 1)
|
|
; CHECK-NOT: Stmt_for_body(c0);
|