| ; RUN: opt %loadPolly -analyze -polly-scops \ |
| ; RUN: -polly-detect-fortran-arrays \ |
| ; RUN: -polly-invariant-load-hoisting \ |
| ; RUN: -polly-use-llvm-names \ |
| ; RUN: -polly-stmt-granularity=bb \ |
| ; RUN: < %s | FileCheck %s --check-prefix=SCOP |
| |
| ; RUN: opt %loadPolly -S \ |
| ; RUN: -polly-detect-fortran-arrays \ |
| ; RUN: -polly-codegen-ppcg \ |
| ; RUN: -polly-invariant-load-hoisting \ |
| ; RUN: -polly-use-llvm-names \ |
| ; RUN: -polly-acc-fail-on-verify-module-failure \ |
| ; RUN: < %s | FileCheck %s --check-prefix=HOST-IR |
| |
| ; REQUIRES: pollyacc |
| |
| ; In Polly, we specifically add a parameter to represent the outermost dimension |
| ; size of fortran arrays. We do this because this information is statically |
| ; available from the fortran metadata generated by dragonegg. |
| ; However, we were only materializing these parameters (meaning, creating an |
| ; llvm::Value to back the isl_id) from *memory accesses*. This is wrong, |
| ; we should materialize parameters from *scop array info*. |
| |
| ; It is wrong because if there is a case where we detect 2 fortran arrays, |
| ; but only one of them is accessed, we may not materialize the other array's |
| ; dimensions at all. |
| |
| ; This test case checks that we do not fail if there is an array that does |
| ; not have an access (In this case, `memory`), we still generate the |
| ; parameter. |
| |
| ; Check that we detect the function as a Scop. |
| ; SCOP: Function: f |
| ; SCOP-NEXT: Region: %loop.prepare---%for.exit |
| ; SCOP-NEXT: Max Loop Depth: 1 |
| |
| ; Check that we detect fortran arrays. |
| ; SCOP: Arrays (Bounds as pw_affs) { |
| ; SCOP: double* MemRef_global_arr[*]; // Element size 8 |
| ; SCOP-NEXT: double MemRef_memory[ [MemRef_memory_fortranarr_size] -> { [] -> [(MemRef_memory_fortranarr_size)] } ]; [BasePtrOrigin: MemRef_global_arr] // Element size 8 |
| ; SCOP-NEXT: double MemRef_memory2[ [MemRef_memory2_fortranarr_size] -> { [] -> [(MemRef_memory2_fortranarr_size)] } ]; [BasePtrOrigin: MemRef_global_arr] // Element size 8 |
| ; SCOP-NEXT: } |
| |
| ; Check that we have writes *only* into memory2, not into memory. |
| ; SCOP: Statements { |
| ; SCOP: Stmt_for_body |
| ; SCOP: MustWriteAccess := [Reduction Type: NONE] [Fortran array descriptor: global_arr] [Scalar: 0] |
| ; SCOP-NEXT: [start_val, end_val, offset, MemRef_memory_fortranarr_size, MemRef_memory2_fortranarr_size] -> { Stmt_for_body[i0] -> MemRef_memory2[start_val + offset + i0] }; |
| ; SCOP-NEXT: ReadAccess := [Reduction Type: NONE] [Fortran array descriptor: global_arr] [Scalar: 0] |
| ; SCOP-NEXT: [start_val, end_val, offset, MemRef_memory_fortranarr_size, MemRef_memory2_fortranarr_size] -> { Stmt_for_body[i0] -> MemRef_memory2[start_val + offset + i0] }; |
| ; SCOP-NEXT: } |
| |
| ; Check that we materialize the sizes and send it across to the kernel. |
| ; HOST-IR: store i64 %MemRef_memory_size, i64* %polly_launch_0_param_4 |
| ; HOST-IR: store i64 %MemRef_memory2_size, i64* %polly_launch_0_param_5 |
| target datalayout = "e-p:64:64:64-S128-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-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*, i64, i64, [1 x %struct.descriptor_dimension] } |
| %struct.descriptor_dimension = type { i64, i64, i64 } |
| |
| @global_arr = external unnamed_addr global %"struct.array1_real(kind=8)", align 32 |
| |
| ; Function Attrs: nounwind uwtable |
| define void @f(i32* noalias %ipstart, i32* noalias %ipend) unnamed_addr #0 { |
| entry: |
| br label %loop.prepare |
| |
| |
| loop.prepare: ; preds = %"6", %"3.preheader" |
| %start.val = load i32, i32* %ipstart, align 4 |
| %end.val = load i32, i32* %ipend, align 4 |
| %should.loop = icmp sgt i32 %start.val, %end.val |
| br i1 %should.loop, label %for.exit, label %for.body |
| |
| |
| for.body: ; preds = %for.body, %"4.preheader" |
| %i = phi i32 [ %i.next, %for.body ], [ %start.val, %loop.prepare ] |
| %i.sext = sext i32 %i to i64 |
| %memory = load double*, double** bitcast (%"struct.array1_real(kind=8)"* @global_arr to double**), align 32 |
| %offset = load i64, i64* getelementptr inbounds (%"struct.array1_real(kind=8)", %"struct.array1_real(kind=8)"* @global_arr, i64 0, i32 1), align 8 |
| %idx = add i64 %offset, %i.sext |
| %slot = getelementptr double, double* %memory, i64 %idx |
| store double 1.0, double* %slot, align 8 |
| |
| %memory2 = load double*, double** bitcast (%"struct.array1_real(kind=8)"* @global_arr to double**), align 32 |
| %offset2 = load i64, i64* getelementptr inbounds (%"struct.array1_real(kind=8)", %"struct.array1_real(kind=8)"* @global_arr, i64 0, i32 1), align 8 |
| %idx2 = add i64 %offset2, %i.sext |
| %slot2 = getelementptr double, double* %memory2, i64 %idx2 |
| %val = load double, double* %slot2, align 8 |
| |
| %should.loopexit = icmp eq i32 %i, %end.val |
| %i.next = add i32 %i, 1 |
| br i1 %should.loopexit, label %for.exit, label %for.body |
| |
| for.exit: ; preds = %for.body |
| ret void |
| } |
| |
| attributes #0 = { nounwind uwtable } |