summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--polly/include/polly/Support/GICHelper.h8
-rw-r--r--polly/lib/Transform/DeLICM.cpp47
-rw-r--r--polly/test/DeLICM/reject_loadafterstore.ll67
-rw-r--r--polly/test/DeLICM/reject_outofquota.ll65
-rw-r--r--polly/test/DeLICM/reject_storeafterstore.ll67
-rw-r--r--polly/test/DeLICM/reject_storeinsubregion.ll79
-rw-r--r--polly/test/DeLICM/reject_unusualstore.ll71
7 files changed, 401 insertions, 3 deletions
diff --git a/polly/include/polly/Support/GICHelper.h b/polly/include/polly/Support/GICHelper.h
index c23513ea96c..4dc2b8f0ee8 100644
--- a/polly/include/polly/Support/GICHelper.h
+++ b/polly/include/polly/Support/GICHelper.h
@@ -15,6 +15,7 @@
#define POLLY_SUPPORT_GIC_HELPER_H
#include "llvm/ADT/APInt.h"
+#include "llvm/IR/DiagnosticInfo.h"
#include "llvm/Support/raw_ostream.h"
#include "isl/aff.h"
#include "isl/ctx.h"
@@ -319,6 +320,13 @@ llvm::raw_ostream &operator<<(llvm::raw_ostream &OS, const IslPtr<T> &Obj) {
return OS;
}
+template <typename T>
+llvm::DiagnosticInfoOptimizationBase &
+operator<<(llvm::DiagnosticInfoOptimizationBase &OS, const IslPtr<T> &Obj) {
+ OS << IslObjTraits<T>::to_str(Obj.keep());
+ return OS;
+}
+
/// Smart pointer to an ISL object, but does not release it when destroyed.
///
/// This is meant to be used as function parameter type. The caller guarantees
diff --git a/polly/lib/Transform/DeLICM.cpp b/polly/lib/Transform/DeLICM.cpp
index 3b064e2892d..45c91437c13 100644
--- a/polly/lib/Transform/DeLICM.cpp
+++ b/polly/lib/Transform/DeLICM.cpp
@@ -641,6 +641,17 @@ public:
}
};
+std::string printIntruction(Instruction *Instr, bool IsForDebug = false) {
+ std::string Result;
+ raw_string_ostream OS(Result);
+ Instr->print(OS, IsForDebug);
+ OS.flush();
+ size_t i = 0;
+ while (i < Result.size() && Result[i] == ' ')
+ i += 1;
+ return Result.substr(i);
+}
+
/// Base class for algorithms based on zones, like DeLICM.
class ZoneAlgorithm {
protected:
@@ -724,8 +735,15 @@ private:
if (MA->isRead()) {
// Reject load after store to same location.
- if (!isl_union_map_is_disjoint(Stores.keep(), AccRel.keep()))
+ if (!isl_union_map_is_disjoint(Stores.keep(), AccRel.keep())) {
+ OptimizationRemarkMissed R(DEBUG_TYPE, "LoadAfterStore",
+ MA->getAccessInstruction());
+ R << "load after store of same element in same statement";
+ R << " (previous stores: " << Stores;
+ R << ", loading: " << AccRel << ")";
+ S->getFunction().getContext().diagnose(R);
return false;
+ }
Loads = give(isl_union_map_union(Loads.take(), AccRel.take()));
@@ -734,18 +752,35 @@ private:
if (!isa<StoreInst>(MA->getAccessInstruction())) {
DEBUG(dbgs() << "WRITE that is not a StoreInst not supported\n");
+ OptimizationRemarkMissed R(DEBUG_TYPE, "UnusualStore",
+ MA->getAccessInstruction());
+ R << "encountered write that is not a StoreInst: "
+ << printIntruction(MA->getAccessInstruction());
+ S->getFunction().getContext().diagnose(R);
return false;
}
// In region statements the order is less clear, eg. the load and store
// might be in a boxed loop.
if (Stmt->isRegionStmt() &&
- !isl_union_map_is_disjoint(Loads.keep(), AccRel.keep()))
+ !isl_union_map_is_disjoint(Loads.keep(), AccRel.keep())) {
+ OptimizationRemarkMissed R(DEBUG_TYPE, "StoreInSubregion",
+ MA->getAccessInstruction());
+ R << "store is in a non-affine subregion";
+ S->getFunction().getContext().diagnose(R);
return false;
+ }
// Do not allow more than one store to the same location.
- if (!isl_union_map_is_disjoint(Stores.keep(), AccRel.keep()))
+ if (!isl_union_map_is_disjoint(Stores.keep(), AccRel.keep())) {
+ OptimizationRemarkMissed R(DEBUG_TYPE, "StoreAfterStore",
+ MA->getAccessInstruction());
+ R << "store after store of same element in same statement";
+ R << " (previous stores: " << Stores;
+ R << ", storing: " << AccRel << ")";
+ S->getFunction().getContext().diagnose(R);
return false;
+ }
Stores = give(isl_union_map_union(Stores.take(), AccRel.take()));
}
@@ -1537,6 +1572,12 @@ public:
if (isl_ctx_last_error(IslCtx.get()) == isl_error_quota) {
DeLICMOutOfQuota++;
DEBUG(dbgs() << "DeLICM analysis exceeded max_operations\n");
+ DebugLoc Begin, End;
+ getDebugLocations(getBBPairForRegion(&S->getRegion()), Begin, End);
+ OptimizationRemarkAnalysis R(DEBUG_TYPE, "OutOfQuota", Begin,
+ S->getEntry());
+ R << "maximal number of operations exceeded during zone analysis";
+ S->getFunction().getContext().diagnose(R);
}
DeLICMAnalyzed++;
diff --git a/polly/test/DeLICM/reject_loadafterstore.ll b/polly/test/DeLICM/reject_loadafterstore.ll
new file mode 100644
index 00000000000..9f1a05d5f3e
--- /dev/null
+++ b/polly/test/DeLICM/reject_loadafterstore.ll
@@ -0,0 +1,67 @@
+; RUN: opt %loadPolly -polly-delicm -analyze -pass-remarks-missed=polly-delicm < %s 2>&1 | FileCheck %s
+;
+; void func(double *A) {
+; for (int j = 0; j < 2; j += 1) { /* outer */
+; double phi = 0.0;
+; for (int i = 0; i < 4; i += 1) /* reduction */
+; phi += 4.2;
+; A[j] = phi;
+; (void)A[j];
+; }
+; }
+;
+define void @func(double* noalias nonnull %A) {
+entry:
+ br label %outer.preheader
+
+outer.preheader:
+ br label %outer.for
+
+outer.for:
+ %j = phi i32 [0, %outer.preheader], [%j.inc, %outer.inc]
+ %j.cmp = icmp slt i32 %j, 2
+ br i1 %j.cmp, label %reduction.preheader, label %outer.exit
+
+
+ reduction.preheader:
+ br label %reduction.for
+
+ reduction.for:
+ %i = phi i32 [0, %reduction.preheader], [%i.inc, %reduction.inc]
+ %phi = phi double [0.0, %reduction.preheader], [%add, %reduction.inc]
+ %i.cmp = icmp slt i32 %i, 4
+ br i1 %i.cmp, label %body, label %reduction.exit
+
+
+
+ body:
+ %add = fadd double %phi, 4.2
+ br label %reduction.inc
+
+
+
+ reduction.inc:
+ %i.inc = add nuw nsw i32 %i, 1
+ br label %reduction.for
+
+ reduction.exit:
+ %A_idx = getelementptr inbounds double, double* %A, i32 %j
+ store double %phi, double* %A_idx
+ %dummy = load double, double* %A_idx
+ br label %outer.inc
+
+
+
+outer.inc:
+ %j.inc = add nuw nsw i32 %j, 1
+ br label %outer.for
+
+outer.exit:
+ br label %return
+
+return:
+ ret void
+}
+
+
+; CHECK: load after store of same element in same statement
diff --git a/polly/test/DeLICM/reject_outofquota.ll b/polly/test/DeLICM/reject_outofquota.ll
new file mode 100644
index 00000000000..eac194354b8
--- /dev/null
+++ b/polly/test/DeLICM/reject_outofquota.ll
@@ -0,0 +1,65 @@
+; RUN: opt %loadPolly -polly-delicm -analyze -pass-remarks-analysis=polly-delicm -polly-delicm-max-ops=1 < %s 2>&1 | FileCheck %s
+;
+; void func(double *A) {
+; for (int j = 0; j < 2; j += 1) { /* outer */
+; double phi = 0.0;
+; for (int i = 0; i < 4; i += 1) /* reduction */
+; phi += 4.2;
+; A[j] = phi;
+; }
+; }
+;
+define void @func(double* noalias nonnull %A) {
+entry:
+ br label %outer.preheader
+
+outer.preheader:
+ br label %outer.for
+
+outer.for:
+ %j = phi i32 [0, %outer.preheader], [%j.inc, %outer.inc]
+ %j.cmp = icmp slt i32 %j, 2
+ br i1 %j.cmp, label %reduction.preheader, label %outer.exit
+
+
+ reduction.preheader:
+ br label %reduction.for
+
+ reduction.for:
+ %i = phi i32 [0, %reduction.preheader], [%i.inc, %reduction.inc]
+ %phi = phi double [0.0, %reduction.preheader], [%add, %reduction.inc]
+ %i.cmp = icmp slt i32 %i, 4
+ br i1 %i.cmp, label %body, label %reduction.exit
+
+
+
+ body:
+ %add = fadd double %phi, 4.2
+ br label %reduction.inc
+
+
+
+ reduction.inc:
+ %i.inc = add nuw nsw i32 %i, 1
+ br label %reduction.for
+
+ reduction.exit:
+ %A_idx = getelementptr inbounds double, double* %A, i32 %j
+ store double %phi, double* %A_idx
+ br label %outer.inc
+
+
+
+outer.inc:
+ %j.inc = add nuw nsw i32 %j, 1
+ br label %outer.for
+
+outer.exit:
+ br label %return
+
+return:
+ ret void
+}
+
+
+; CHECK: maximal number of operations exceeded during zone analysis
diff --git a/polly/test/DeLICM/reject_storeafterstore.ll b/polly/test/DeLICM/reject_storeafterstore.ll
new file mode 100644
index 00000000000..f3f8955a1c5
--- /dev/null
+++ b/polly/test/DeLICM/reject_storeafterstore.ll
@@ -0,0 +1,67 @@
+; RUN: opt %loadPolly -polly-delicm -analyze -pass-remarks-missed=polly-delicm < %s 2>&1 | FileCheck %s
+;
+; void func(double *A) {
+; for (int j = 0; j < 2; j += 1) { /* outer */
+; double phi = 0.0;
+; for (int i = 0; i < 4; i += 1) /* reduction */
+; phi += 4.2;
+; A[j] = 0.0;
+; A[j] = phi;
+; }
+; }
+;
+define void @func(double* noalias nonnull %A) {
+entry:
+ br label %outer.preheader
+
+outer.preheader:
+ br label %outer.for
+
+outer.for:
+ %j = phi i32 [0, %outer.preheader], [%j.inc, %outer.inc]
+ %j.cmp = icmp slt i32 %j, 2
+ br i1 %j.cmp, label %reduction.preheader, label %outer.exit
+
+
+ reduction.preheader:
+ br label %reduction.for
+
+ reduction.for:
+ %i = phi i32 [0, %reduction.preheader], [%i.inc, %reduction.inc]
+ %phi = phi double [0.0, %reduction.preheader], [%add, %reduction.inc]
+ %i.cmp = icmp slt i32 %i, 4
+ br i1 %i.cmp, label %body, label %reduction.exit
+
+
+
+ body:
+ %add = fadd double %phi, 4.2
+ br label %reduction.inc
+
+
+
+ reduction.inc:
+ %i.inc = add nuw nsw i32 %i, 1
+ br label %reduction.for
+
+ reduction.exit:
+ %A_idx = getelementptr inbounds double, double* %A, i32 %j
+ store double 0.0, double* %A_idx
+ store double %phi, double* %A_idx
+ br label %outer.inc
+
+
+
+outer.inc:
+ %j.inc = add nuw nsw i32 %j, 1
+ br label %outer.for
+
+outer.exit:
+ br label %return
+
+return:
+ ret void
+}
+
+
+; CHECK: store after store of same element in same statement
diff --git a/polly/test/DeLICM/reject_storeinsubregion.ll b/polly/test/DeLICM/reject_storeinsubregion.ll
new file mode 100644
index 00000000000..4a96ace3fdf
--- /dev/null
+++ b/polly/test/DeLICM/reject_storeinsubregion.ll
@@ -0,0 +1,79 @@
+; RUN: opt %loadPolly -polly-delicm -analyze -pass-remarks-missed=polly-delicm < %s 2>&1 | FileCheck %s
+;
+; void func(double *A) {
+; for (int j = 0; j < 2; j += 1) { /* outer */
+; double phi = 0.0;
+; for (int i = 0; i < 4; i += 1) /* reduction */
+; phi += 4.2;
+; if (phi > 42.0)
+; (void)A[j];
+; else
+; A[j] = 0.0;
+; A[j] = phi;
+; }
+; }
+;
+define void @func(double* noalias nonnull %A) {
+entry:
+ br label %outer.preheader
+
+outer.preheader:
+ br label %outer.for
+
+outer.for:
+ %j = phi i32 [0, %outer.preheader], [%j.inc, %outer.inc]
+ %j.cmp = icmp slt i32 %j, 2
+ br i1 %j.cmp, label %reduction.preheader, label %outer.exit
+
+
+ reduction.preheader:
+ br label %reduction.for
+
+ reduction.for:
+ %i = phi i32 [0, %reduction.preheader], [%i.inc, %reduction.inc]
+ %phi = phi double [0.0, %reduction.preheader], [%add, %reduction.inc]
+ %i.cmp = icmp slt i32 %i, 4
+ br i1 %i.cmp, label %body, label %reduction.exit
+
+
+
+ body:
+ %add = fadd double %phi, 4.2
+ %A_idxp = getelementptr inbounds double, double* %A, i32 %j
+ %add.cmp = fcmp ogt double %add, 42.0
+ br i1 %add.cmp , label %body_true, label %body_false
+
+ body_true:
+ %dummy = load double, double* %A_idxp
+ br label %reduction.inc
+
+ body_false:
+ store double 0.0, double* %A_idxp
+ br label %reduction.inc
+
+
+
+ reduction.inc:
+ %i.inc = add nuw nsw i32 %i, 1
+ br label %reduction.for
+
+ reduction.exit:
+ %A_idx = getelementptr inbounds double, double* %A, i32 %j
+ store double %phi, double* %A_idx
+ br label %outer.inc
+
+
+
+outer.inc:
+ %j.inc = add nuw nsw i32 %j, 1
+ br label %outer.for
+
+outer.exit:
+ br label %return
+
+return:
+ ret void
+}
+
+
+; CHECK: store is in a non-affine subregion
diff --git a/polly/test/DeLICM/reject_unusualstore.ll b/polly/test/DeLICM/reject_unusualstore.ll
new file mode 100644
index 00000000000..43db78b0763
--- /dev/null
+++ b/polly/test/DeLICM/reject_unusualstore.ll
@@ -0,0 +1,71 @@
+; RUN: opt %loadPolly -polly-delicm -analyze -pass-remarks-missed=polly-delicm < %s 2>&1 | FileCheck %s
+;
+; void func(double *A) {
+; for (int j = 0; j < 2; j += 1) { /* outer */
+; memset(A[j], 0, sizeof(double));
+; double phi = 0.0;
+; for (int i = 0; i < 4; i += 1) /* reduction */
+; phi += 4.2;
+; A[j] = phi;
+; }
+; }
+;
+
+declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i32, i1)
+
+define void @func(double* noalias nonnull %A) {
+entry:
+ br label %outer.preheader
+
+outer.preheader:
+ br label %outer.for
+
+outer.for:
+ %j = phi i32 [0, %outer.preheader], [%j.inc, %outer.inc]
+ %j.cmp = icmp slt i32 %j, 2
+ br i1 %j.cmp, label %reduction.preheader, label %outer.exit
+
+
+ reduction.preheader:
+ %A_idx = getelementptr inbounds double, double* %A, i32 %j
+ %tmp = bitcast double* %A_idx to i8*
+ call void @llvm.memset.p0i8.i64(i8* %tmp, i8 0, i64 8, i32 1, i1 false)
+ br label %reduction.for
+
+ reduction.for:
+ %i = phi i32 [0, %reduction.preheader], [%i.inc, %reduction.inc]
+ %phi = phi double [0.0, %reduction.preheader], [%add, %reduction.inc]
+ %i.cmp = icmp slt i32 %i, 4
+ br i1 %i.cmp, label %body, label %reduction.exit
+
+
+
+ body:
+ %add = fadd double %phi, 4.2
+ br label %reduction.inc
+
+
+
+ reduction.inc:
+ %i.inc = add nuw nsw i32 %i, 1
+ br label %reduction.for
+
+ reduction.exit:
+ store double %phi, double* %A_idx
+ br label %outer.inc
+
+
+
+outer.inc:
+ %j.inc = add nuw nsw i32 %j, 1
+ br label %outer.for
+
+outer.exit:
+ br label %return
+
+return:
+ ret void
+}
+
+
+; CHECK: encountered write that is not a StoreInst: call void @llvm.memset.p0i8.i64(i8* %tmp, i8 0, i64 8, i32 1, i1 false)
OpenPOWER on IntegriCloud