Index: include/polly/CodeGen/IslNodeBuilder.h =================================================================== --- include/polly/CodeGen/IslNodeBuilder.h +++ include/polly/CodeGen/IslNodeBuilder.h @@ -75,6 +75,12 @@ void addParameters(__isl_take isl_set *Context); + /// Materialize all fortran array parameters in the current scop. + /// + /// @returns False, iff a problem occurred and a fortran array was not + /// materialized. + bool materializeFortranArrayOutermostDimensionParameters(); + /// Generate code that evaluates @p Condition at run-time. /// /// This function is typically called to generate the LLVM-IR for the Index: include/polly/ScopInfo.h =================================================================== --- include/polly/ScopInfo.h +++ include/polly/ScopInfo.h @@ -263,6 +263,9 @@ /// with old sizes bool updateSizes(ArrayRef Sizes, bool CheckConsistency = true); + /// Make the ScopArrayInfo model a Fortran Array + void makeFortranArray(Value *FAD); + /// Destructor to free the isl id of the base pointer. ~ScopArrayInfo(); @@ -419,6 +422,10 @@ /// The scop this SAI object belongs to. Scop &S; + + /// If this arrays models a Fortran array, contains a pointer + /// to the Fortran array descriptor + Value *FAD; }; /// Represent memory accesses in statements. @@ -895,6 +902,10 @@ /// the dimension of the innermost loop containing the statement. __isl_give isl_set *getStride(__isl_take const isl_map *Schedule) const; + /// Get the FortranArrayDescriptor corresponding to this memory access if + /// it exists, and nullptr otherwise. + Value *getFortranArrayDescriptor() const { return this->FAD; }; + /// Is the stride of the access equal to a certain width? Schedule is a map /// from the statement to a schedule where the innermost dimension is the /// dimension of the innermost loop containing the statement. @@ -2035,6 +2046,10 @@ /// all memory accesses have been modeled and canonicalized. void assumeNoOutOfBounds(); + /// Mark Arrays that have memory accesses with FortranArrayDescriptor + /// as fortran arrays + void markFortranArrays(); + /// Finalize all access relations. /// /// When building up access relations, temporary access relations that Index: lib/Analysis/ScopInfo.cpp =================================================================== --- lib/Analysis/ScopInfo.cpp +++ lib/Analysis/ScopInfo.cpp @@ -240,7 +240,8 @@ ArrayRef Sizes, MemoryKind Kind, const DataLayout &DL, Scop *S, const char *BaseName) - : BasePtr(BasePtr), ElementType(ElementType), Kind(Kind), DL(DL), S(*S) { + : BasePtr(BasePtr), ElementType(ElementType), Kind(Kind), DL(DL), S(*S), + FAD(nullptr) { std::string BasePtrName = BaseName ? BaseName : getIslCompatibleName("MemRef", BasePtr, S->getNextArrayIdx(), @@ -311,6 +312,36 @@ } } +/// Make the ScopArrayInfo model a Fortran Array +void ScopArrayInfo::makeFortranArray(Value *FAD) { + assert(FAD != nullptr && "got invalid FortranArraydescriptor"); + if (this->FAD != nullptr) { + assert(this->FAD == FAD && + "receiving different array descriptors for same array"); + return; + } + + assert(DimensionSizesPw.size() > 0 && DimensionSizesPw[0] == nullptr); + assert(this->FAD == nullptr); + this->FAD = FAD; + + isl_space *space = isl_space_set_alloc(S.getIslCtx(), 1, 0); + + std::string param_name = this->getName(); + param_name += "_fortranarr_size"; + isl_id *id_for_pa = isl_id_alloc(S.getIslCtx(), param_name.c_str(), nullptr); + + space = isl_space_set_dim_id(space, isl_dim_param, 0, id_for_pa); + isl_basic_set *identity = isl_basic_set_universe(space); + isl_local_space *ls = isl_basic_set_get_local_space(identity); + isl_basic_set_free(identity); + + isl_aff *aff = isl_aff_var_on_domain(ls, isl_dim_param, 0); + isl_pw_aff *pa = isl_pw_aff_from_aff(aff); + + DimensionSizesPw[0] = pa; +} + bool ScopArrayInfo::updateSizes(ArrayRef NewSizes, bool CheckConsistency) { int SharedDims = std::min(NewSizes.size(), DimensionSizes.size()); @@ -367,7 +398,12 @@ void ScopArrayInfo::print(raw_ostream &OS, bool SizeAsPwAff) const { OS.indent(8) << *getElementType() << " " << getName(); unsigned u = 0; - if (getNumberOfDimensions() > 0 && !getDimensionSize(0)) { + // If this is a Fortran array, then we can print the outermost dimension + // as a pwAff even though there is no SCEV information. + const bool shouldPrintOutermostDim = SizeAsPwAff && FAD != nullptr; + + if (!shouldPrintOutermostDim && getNumberOfDimensions() > 0 && + !getDimensionSize(0)) { OS << "[*]"; u++; } @@ -2131,6 +2167,33 @@ } } +isl_set *getContextWithFortranArrays(__isl_keep isl_set *ScopContext, + Scop::array_range arrays) { + isl_set *FortranContext = isl_set_copy(ScopContext); + int ContextNParams = isl_set_dim(FortranContext, isl_dim_param); + for (auto arr : arrays) { + // is a fortran array + // HACK: actually need to check if it has a FAD, but for now this works + if (arr->getNumberOfDimensions() > 0) { + isl_pw_aff *pwaff = arr->getDimensionSizePw(0); + if (pwaff == nullptr) + continue; + isl_id *id_for_pa = isl_pw_aff_get_dim_id(pwaff, isl_dim_param, 0); + + assert(id_for_pa != nullptr); + + FortranContext = isl_set_add_dims(FortranContext, isl_dim_param, 1); + FortranContext = isl_set_set_dim_id(FortranContext, isl_dim_param, + ContextNParams, id_for_pa); + ContextNParams++; + + isl_pw_aff_free(pwaff); + } + } + + return FortranContext; +} + void Scop::realignParams() { if (PollyIgnoreParamBounds) return; @@ -2153,8 +2216,9 @@ for (ScopStmt &Stmt : *this) Stmt.realignParams(); - // Simplify the schedule according to the context too. - Schedule = isl_schedule_gist_domain_params(Schedule, getContext()); + isl_set *FortranArraysContext = + getContextWithFortranArrays(Context, arrays()); + Schedule = isl_schedule_gist_domain_params(Schedule, FortranArraysContext); } static __isl_give isl_set * @@ -2164,7 +2228,7 @@ // simplify the context with the constraints that are needed for anything to // be executed at all. However, if we have error blocks in the SCoP we already // assumed some parameter combinations cannot occur and removed them from the - // domains, thus we cannot use the remaining domain to simplify the + // domains, thus we cannot use the remaining domain to simplify thearr // assumptions. if (!S.hasErrorBlock()) { isl_set *DomainParameters = isl_union_set_params(S.getDomains()); @@ -3379,11 +3443,28 @@ return; } +void Scop::markFortranArrays() { + for (ScopStmt &Stmt : this->Stmts) { + for (MemoryAccess *memAccess : Stmt) { + Value *FAD = memAccess->getFortranArrayDescriptor(); + if (!FAD) + continue; + // HACK: const_cast ing to edit + ScopArrayInfo *SAI = + const_cast(memAccess->getLatestScopArrayInfo()); + assert(SAI != nullptr && "memory access into a fortran array does not " + "have an associated ScopArrayInfo"); + SAI->makeFortranArray(FAD); + } + } +} + void Scop::finalizeAccesses() { updateAccessDimensionality(); foldSizeConstantsToRight(); foldAccessRelations(); assumeNoOutOfBounds(); + markFortranArrays(); } Scop::~Scop() { Index: lib/CodeGen/IslNodeBuilder.cpp =================================================================== --- lib/CodeGen/IslNodeBuilder.cpp +++ lib/CodeGen/IslNodeBuilder.cpp @@ -995,6 +995,89 @@ return true; } +/// %"struct.array3_integer(kind=4)" = type { i8*, i64, i64, [3 x +/// %struct.descriptor_dimension] } %struct.descriptor_dimension = type { i64, +/// i64, i64 } +/// @__src_soil_MOD_arr = global %"struct.array3_integer(kind=4)" +/// zeroinitializer, align 32 +/// ... +/// %0 = load i64, i64* getelementptr inbounds +/// (%"struct.array3_integer(kind=4)", %"struct.array3_integer(kind=4)"* +/// @__src_soil_MOD_arr, i64 0, i32 3, i64 0, i32 2), align 8, !tbaa !0 %1 = +/// load i64, i64* getelementptr inbounds (%"struct.array3_integer(kind=4)", +/// %"struct.array3_integer(kind=4)"* @__src_soil_MOD_arr, i64 0, i32 3, i64 0, +/// i32 1), align 8, !tbaa !0 %2 = sub nsw i64 %0, %1 %3 = add nsw i64 %2, 1 +Value *buildFortranArrayDescriptorOutermostDimensionLoad( + Value *GlobalDescriptor, PollyIRBuilder &Builder, std::string ArrayName) { + assert(GlobalDescriptor != nullptr && "invalid global descriptor given"); + + Value *endIdx[4] = {Builder.getInt64(0), Builder.getInt32(3), + Builder.getInt64(0), Builder.getInt32(2)}; + Value *endPtr = Builder.CreateInBoundsGEP(GlobalDescriptor, endIdx, + ArrayName + "_end_ptr"); + Value *end = Builder.CreateLoad(endPtr, ArrayName + "_end"); + + Value *beginIdx[4] = {Builder.getInt64(0), Builder.getInt32(3), + Builder.getInt64(0), Builder.getInt32(1)}; + Value *beginPtr = Builder.CreateInBoundsGEP(GlobalDescriptor, beginIdx, + ArrayName + "_begin_ptr"); + Value *begin = Builder.CreateLoad(beginPtr, ArrayName + "_begin"); + + Value *size = + Builder.CreateNSWSub(end, begin, ArrayName + "_end_begin_delta"); + Type *endType = dyn_cast(end->getType()); + assert(endType != nullptr && "expected type of end to be integral"); + + size = Builder.CreateNSWAdd(end, + ConstantInt::get(endType, 1, /* signed = */ true), + ArrayName + "_size"); + + return size; +} + +bool IslNodeBuilder::materializeFortranArrayOutermostDimensionParameters() { + for (const ScopStmt &Stmt : S) { + for (const MemoryAccess *Access : Stmt) { + if (!Access->isArrayKind()) + continue; + + const ScopArrayInfo *Array = Access->getScopArrayInfo(); + if (!Array) + continue; + + if (Array->getNumberOfDimensions() == 0) + continue; + + Value *FAD = const_cast(Access->getFortranArrayDescriptor()); + if (FAD == nullptr) + continue; + + isl_pw_aff *parametric_pw_aff = Array->getDimensionSizePw(0); + assert(parametric_pw_aff != nullptr && "parameteric pw_aff corresponding " + "to outermost dimension does not " + "exist"); + + isl_id *Id = isl_pw_aff_get_dim_id(parametric_pw_aff, isl_dim_param, 0); + isl_pw_aff_free(parametric_pw_aff); + + assert(Id != nullptr && "pw_aff is not parametric"); + + if (IDToValue.count(Id)) { + isl_id_free(Id); + continue; + } + + Value *finalValue = buildFortranArrayDescriptorOutermostDimensionLoad( + dyn_cast(FAD), Builder, Array->getName()); + assert(finalValue != nullptr && "unable to build fortran array " + "descriptor load of outermost dimension"); + IDToValue[Id] = finalValue; + isl_id_free(Id); + } + } + return true; +} + /// Add the number of dimensions in @p BS to @p U. static isl_stat countTotalDims(__isl_take isl_basic_set *BS, void *U) { unsigned *NumTotalDim = static_cast(U); @@ -1313,6 +1396,12 @@ // Materialize values for the parameters of the SCoP. materializeParameters(); + // materialize the outermost dimension parameters for a fortran array. + // NOTE: materializeParameters() does not work since it looks through + // the SCEVs. We don't have a corresponding SCEV for the array size + // parameter + materializeFortranArrayOutermostDimensionParameters(); + // Generate values for the current loop iteration for all surrounding loops. // // We may also reference loops outside of the scop which do not contain the Index: lib/CodeGen/PPCGCodeGeneration.cpp =================================================================== --- lib/CodeGen/PPCGCodeGeneration.cpp +++ lib/CodeGen/PPCGCodeGeneration.cpp @@ -2163,9 +2163,18 @@ for (unsigned i = 1; i < NumDims; ++i) Extent = isl_set_lower_bound_si(Extent, isl_dim_set, i, 0); - for (unsigned i = 1; i < NumDims; ++i) { + for (unsigned i = 0; i < NumDims; ++i) { isl_pw_aff *PwAff = const_cast(Array->getDimensionSizePw(i)); + + // pwAff can be NULL for zero dimension. only in the case of a fortran + // array will we have a legitimate dimension + if (!PwAff) { + if (i != 0) + assert(false && "invalid dimension pwAff for nonzero dimension"); + continue; + } + isl_pw_aff *Val = isl_pw_aff_from_aff(isl_aff_var_on_domain( isl_local_space_from_space(Array->getSpace()), isl_dim_set, i)); PwAff = isl_pw_aff_add_dims(PwAff, isl_dim_in, Index: test/Isl/CodeGen/fortran_array_runtime_size_generation.ll =================================================================== --- /dev/null +++ test/Isl/CodeGen/fortran_array_runtime_size_generation.ll @@ -0,0 +1,100 @@ +; Check that the runtime size computation is generation for fortran arrays +; PPCG code generation backend: +; RUN: opt %loadPolly -S -polly-process-unprofitable -polly-allow-nonaffine \ +; RUN: -polly-ignore-aliasing -polly-invariant-load-hoisting \ +; RUN: -polly-canonicalize -polly-target=gpu -polly-acc-dump-kernel-ir \ +; RUN: -polly-acc-mincompute=0 -polly-detect-fortran-arrays -polly-codegen-ppcg \ +; RUN: < %s | FileCheck %s + +; Reguar code generation backend: +; RUN: opt %loadPolly -S -polly-process-unprofitable -polly-allow-nonaffine \ +; RUN: -polly-ignore-aliasing -polly-invariant-load-hoisting \ +; RUN: -polly-canonicalize -polly-detect-fortran-arrays -polly-codegen < %s | FileCheck %s + +; What the input fortran code should look like. NOTE: this is fake, the +; .ll file was hand-written. +; +; MODULE testmod +; USE data_parameters, ONLY : & +; IMPLICIT NONE +; +; INTEGER (KIND=iintegers), ALLOCATABLE, PRIVATE :: & +; arrin(:), arrout(:) +; CONTAINS +; +; SUBROUTINE test() +; INTEGER (KIND=iintegers) :: i +; +; DO i = 1, 100 +; arrout(i) = arrin(i) * arrin(i) +; END DO +; END SUBROUTINE test +; END MODULE testmod + +target datalayout = "e-p:64:64:64-S128-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i32:64:64-f16:16:16-f32:32:32-f64:64:64-f128:128:128-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64" +target triple = "x86_64-unknown-linux-gnu" + +module asm "\09.ident\09\22GCC: (GNU) 4.6.4 LLVM: 3.3.1\22" + +%"struct.array1_real(kind=8)" = type { i8*, i32, i32, [1 x %struct.descriptor_dimension] } +%struct.descriptor_dimension = type { i32, i32, i32 } + +@arrin = unnamed_addr global %"struct.array1_real(kind=8)" zeroinitializer, align 32 +@arrout = unnamed_addr global %"struct.array1_real(kind=8)" zeroinitializer, align 32 + +; Function Attrs: nounwind uwtable +define void @__src_soil_MOD_terra1() unnamed_addr #0 { +entry: + %rawmemin = load i8*, i8** getelementptr inbounds (%"struct.array1_real(kind=8)", %"struct.array1_real(kind=8)"* @arrin, i32 0, i32 0), align 32, !tbaa !5 + %typedmemin = bitcast i8* %rawmemin to i32* + + + %rawmemout = load i8*, i8** getelementptr inbounds (%"struct.array1_real(kind=8)", %"struct.array1_real(kind=8)"* @arrout, i32 0, i32 0), align 32, !tbaa !5 + %typedmemout = bitcast i8* %rawmemout to i32* + br label %for.cond + +for.cond: ; preds = %for.inc, %entry + %indvars.iv = phi i32 [ %indvars.iv.next, %for.inc ], [ 1, %entry ] + %exitcond = icmp ne i32 %indvars.iv, 100 + br i1 %exitcond, label %for.body, label %return + +for.body: ; preds = %"3", %"3.preheader" + ; %6 = add nsw i32 %indvars.iv, %0 + %inslot = getelementptr inbounds i32, i32* %typedmemin, i32 %indvars.iv + %inval = load i32, i32* %inslot, align 8 + + %outslot = getelementptr inbounds i32, i32* %typedmemout, i32 %indvars.iv + + %out = mul nsw i32 %inval, %inval + store i32 %out, i32* %outslot, align 8 + + br label %for.inc + +for.inc: + %indvars.iv.next = add i32 %indvars.iv, 1 + br label %for.cond + +return: ; preds = %"3", %entry + ret void +} + +attributes #0 = { nounwind uwtable } + +!0 = !{!1, !1, i32 0} +!1 = !{!"alias set 11: integer(kind=4)", !2} +!2 = distinct !{!2} +!3 = !{!4, !4, i32 0} +!4 = !{!"alias set 4: integer(kind=8)", !2} +!5 = !{!6, !6, i32 0} +!6 = !{!"alias set 3: void*", !2} +!7 = !{!8, !8, i32 0} +!8 = !{!"alias set 18: real(kind=8)", !2} + +; CHECK: %MemRef_rawmemin1_end = load i32, i32* getelementptr inbounds (%"struct.array1_real(kind=8)", %"struct.array1_real(kind=8)"* @arrin, i64 0, i32 3, i64 0, i32 2) +; CHECK-NEXT: %MemRef_rawmemin1_begin = load i32, i32* getelementptr inbounds (%"struct.array1_real(kind=8)", %"struct.array1_real(kind=8)"* @arrin, i64 0, i32 3, i64 0, i32 1) +; CHECK-NEXT: %MemRef_rawmemin1_end_begin_delta = sub nsw i32 %MemRef_rawmemin1_end, %MemRef_rawmemin1_begin +; CHECK-NEXT: %MemRef_rawmemin1_size = add nsw i32 %MemRef_rawmemin1_end, 1 +; CHECK-NEXT: %MemRef_rawmemout2_end = load i32, i32* getelementptr inbounds (%"struct.array1_real(kind=8)", %"struct.array1_real(kind=8)"* @arrout, i64 0, i32 3, i64 0, i32 2) +; CHECK-NEXT: %MemRef_rawmemout2_begin = load i32, i32* getelementptr inbounds (%"struct.array1_real(kind=8)", %"struct.array1_real(kind=8)"* @arrout, i64 0, i32 3, i64 0, i32 1) +; CHECK-NEXT: %MemRef_rawmemout2_end_begin_delta = sub nsw i32 %MemRef_rawmemout2_end, %MemRef_rawmemout2_begin +; CHECK-NEXT: %MemRef_rawmemout2_size = add nsw i32 %MemRef_rawmemout2_end, 1