7a0f7dbf2d
Reapply of a22d1c2225543aa9ae7882f6b1a97ee7b2c95574. Using this PR for pre-merge CI. Instead of relying on any pass manager to schedule Polly's passes, add Polly's own pipeline manager which is seen as a monolithic pass in LLVM's pass manager. Polly's former passes are now phases of the new PhaseManager component. Relying on LLVM's pass manager (the legacy as well as the New Pass Manager) to manage Polly's phases never was a good fit that the PhaseManager resolves: * Polly passes were modifying analysis results, in particular RegionInfo and ScopInfo. This means that there was not just one unique and "definite" analysis result, the actual result depended on which analyses ran prior, and the pass manager was not allowed to throw away cached analyses or prior SCoP optimizations would have been forgotten. The LLVM pass manger's persistance of analysis results is not contractual but designed for caching. * Polly depends on a particular execution order of passes and regions (e.g. regression tests, invalidation of consecutive SCoPs). LLVM's pass manager does not guarantee any excecution order. * Polly does not completely preserve DominatorTree, RegionInfo, LoopInfo, or ScalarEvolution, but only as-needed for Polly's own uses. Because the ScopDetection object stores references to those analyses, it still had to lie to the pass manager that they would be preserved, or the pass manager would have released and recomputed the invalidated analysis objects that ScopDetection/ScopInfo was still referencing. To ensure that no non-Polly pass would see these not-completely-preserved analyses, all analyses still had to be thrown away after the ScopPassManager, respectively with a BarrierNoopPass in case of the LPM. * The NPM's PassInstrumentation wraps the IR unit into an `llvm::Any` object, but implementations such as PrintIRInstrumentation call llvm_unreachable on encountering an unknown IR unit, such as SCoPs, with no extension points to add support. Hence LLVM crashes when dumping IR between SCoP passes (such as `-print-before-changed` with Polly being active). The new PhaseManager uses some command line options that previously belonged to Polly's legacy passes, such as `-polly-print-detect` (so the option will continue to work). Hence the LPM support is incompatible with the new approach and support for it is removed.
184 lines
8.6 KiB
LLVM
184 lines
8.6 KiB
LLVM
; RUN: opt %loadNPMPolly '-passes=polly<no-default-opts>' -S -polly-codegen-add-debug-printing -polly-ignore-aliasing < %s | FileCheck %s
|
|
|
|
; #define N 10
|
|
; void foo(float A[restrict], double B[restrict], char C[restrict],
|
|
; int D[restrict], long E[restrict]) {
|
|
; for (long i = 0; i < N; i++)
|
|
; A[i] += B[i] + C[i] + D[i] + E[i];
|
|
; }
|
|
;
|
|
; int main() {
|
|
; float A[N];
|
|
; double B[N];
|
|
; char C[N];
|
|
; int D[N];
|
|
; long E[N];
|
|
;
|
|
; for (long i = 0; i < N; i++) {
|
|
; __sync_synchronize();
|
|
; A[i] = B[i] = C[i] = D[i] = E[i] = 42;
|
|
; }
|
|
;
|
|
; foo(A, B, C, D, E);
|
|
;
|
|
; return A[8];
|
|
; }
|
|
|
|
; CHECK: @0 = private unnamed_addr addrspace(4) constant [11 x i8] c"Load from \00"
|
|
; CHECK: @1 = private unnamed_addr addrspace(4) constant [3 x i8] c": \00"
|
|
; CHECK: @2 = private unnamed_addr addrspace(4) constant [2 x i8] c"\0A\00"
|
|
; CHECK: @3 = private unnamed_addr constant [12 x i8] c"%s%ld%s%f%s\00"
|
|
; CHECK: @4 = private unnamed_addr addrspace(4) constant [11 x i8] c"Load from \00"
|
|
; CHECK: @5 = private unnamed_addr addrspace(4) constant [3 x i8] c": \00"
|
|
; CHECK: @6 = private unnamed_addr addrspace(4) constant [2 x i8] c"\0A\00"
|
|
; CHECK: @7 = private unnamed_addr constant [13 x i8] c"%s%ld%s%ld%s\00"
|
|
; CHECK: @8 = private unnamed_addr addrspace(4) constant [11 x i8] c"Load from \00"
|
|
; CHECK: @9 = private unnamed_addr addrspace(4) constant [3 x i8] c": \00"
|
|
; CHECK: @10 = private unnamed_addr addrspace(4) constant [2 x i8] c"\0A\00"
|
|
; CHECK: @11 = private unnamed_addr constant [13 x i8] c"%s%ld%s%ld%s\00"
|
|
; CHECK: @12 = private unnamed_addr addrspace(4) constant [11 x i8] c"Load from \00"
|
|
; CHECK: @13 = private unnamed_addr addrspace(4) constant [3 x i8] c": \00"
|
|
; CHECK: @14 = private unnamed_addr addrspace(4) constant [2 x i8] c"\0A\00"
|
|
; CHECK: @15 = private unnamed_addr constant [13 x i8] c"%s%ld%s%ld%s\00"
|
|
; CHECK: @16 = private unnamed_addr addrspace(4) constant [11 x i8] c"Load from \00"
|
|
; CHECK: @17 = private unnamed_addr addrspace(4) constant [3 x i8] c": \00"
|
|
; CHECK: @18 = private unnamed_addr addrspace(4) constant [2 x i8] c"\0A\00"
|
|
; CHECK: @19 = private unnamed_addr constant [12 x i8] c"%s%ld%s%f%s\00"
|
|
; CHECK: @20 = private unnamed_addr addrspace(4) constant [11 x i8] c"Store to \00"
|
|
; CHECK: @21 = private unnamed_addr addrspace(4) constant [3 x i8] c": \00"
|
|
; CHECK: @22 = private unnamed_addr addrspace(4) constant [2 x i8] c"\0A\00"
|
|
; CHECK: @23 = private unnamed_addr constant [12 x i8] c"%s%ld%s%f%s\00"
|
|
|
|
; CHECK: %0 = shl nuw nsw i64 %polly.indvar, 3
|
|
; CHECK: %scevgep = getelementptr i8, ptr %B, i64 %0
|
|
; CHECK: %tmp3_p_scalar_ = load double, ptr %scevgep, align 8, !alias.scope !2, !noalias !5
|
|
; CHECK: %1 = ptrtoint ptr %scevgep to i64
|
|
; CHECK: %2 = call i32 (...) @printf(ptr @3, ptr addrspace(4) @0, i64 %1, ptr addrspace(4) @1, double %tmp3_p_scalar_, ptr addrspace(4) @2)
|
|
; CHECK: %3 = call i32 @fflush(ptr null)
|
|
; CHECK: %scevgep1 = getelementptr i8, ptr %C, i64 %polly.indvar
|
|
; CHECK: %tmp5_p_scalar_ = load i8, ptr %scevgep1, align 1, !alias.scope !10, !noalias !11
|
|
; CHECK: %4 = ptrtoint ptr %scevgep1 to i64
|
|
; CHECK: %5 = sext i8 %tmp5_p_scalar_ to i64
|
|
; CHECK: %6 = call i32 (...) @printf(ptr @7, ptr addrspace(4) @4, i64 %4, ptr addrspace(4) @5, i64 %5, ptr addrspace(4) @6)
|
|
; CHECK: %7 = call i32 @fflush(ptr null)
|
|
; CHECK: %p_tmp6 = sitofp i8 %tmp5_p_scalar_ to double
|
|
; CHECK: %p_tmp7 = fadd double %tmp3_p_scalar_, %p_tmp6
|
|
; CHECK: %8 = shl nuw nsw i64 %polly.indvar, 2
|
|
; CHECK: %scevgep2 = getelementptr i8, ptr %D, i64 %8
|
|
; CHECK: %tmp9_p_scalar_ = load i32, ptr %scevgep2, align 4, !alias.scope !12, !noalias !13
|
|
; CHECK: %9 = ptrtoint ptr %scevgep2 to i64
|
|
; CHECK: %10 = sext i32 %tmp9_p_scalar_ to i64
|
|
; CHECK: %11 = call i32 (...) @printf(ptr @11, ptr addrspace(4) @8, i64 %9, ptr addrspace(4) @9, i64 %10, ptr addrspace(4) @10)
|
|
; CHECK: %12 = call i32 @fflush(ptr null)
|
|
; CHECK: %p_tmp10 = sitofp i32 %tmp9_p_scalar_ to double
|
|
; CHECK: %p_tmp11 = fadd double %p_tmp7, %p_tmp10
|
|
; CHECK: %13 = shl nuw nsw i64 %polly.indvar, 3
|
|
; CHECK: %scevgep3 = getelementptr i8, ptr %E, i64 %13
|
|
; CHECK: %tmp13_p_scalar_ = load i64, ptr %scevgep3, align 8, !alias.scope !14, !noalias !15
|
|
; CHECK: %14 = ptrtoint ptr %scevgep3 to i64
|
|
; CHECK: %15 = call i32 (...) @printf(ptr @15, ptr addrspace(4) @12, i64 %14, ptr addrspace(4) @13, i64 %tmp13_p_scalar_, ptr addrspace(4) @14)
|
|
; CHECK: %16 = call i32 @fflush(ptr null)
|
|
; CHECK: %p_tmp14 = sitofp i64 %tmp13_p_scalar_ to double
|
|
; CHECK: %p_tmp15 = fadd double %p_tmp11, %p_tmp14
|
|
; CHECK: %17 = shl nuw nsw i64 %polly.indvar, 2
|
|
; CHECK: %scevgep4 = getelementptr i8, ptr %A, i64 %17
|
|
; CHECK: %tmp17_p_scalar_ = load float, ptr %scevgep4, align 4, !alias.scope !16, !noalias !17
|
|
; CHECK: %18 = ptrtoint ptr %scevgep4 to i64
|
|
; CHECK: %19 = fpext float %tmp17_p_scalar_ to double
|
|
; CHECK: %20 = call i32 (...) @printf(ptr @19, ptr addrspace(4) @16, i64 %18, ptr addrspace(4) @17, double %19, ptr addrspace(4) @18)
|
|
; CHECK: %21 = call i32 @fflush(ptr null)
|
|
; CHECK: %p_tmp18 = fpext float %tmp17_p_scalar_ to double
|
|
; CHECK: %p_tmp19 = fadd double %p_tmp18, %p_tmp15
|
|
; CHECK: %p_tmp20 = fptrunc double %p_tmp19 to float
|
|
; CHECK: %22 = ptrtoint ptr %scevgep4 to i64
|
|
; CHECK: %23 = fpext float %p_tmp20 to double
|
|
; CHECK: %24 = call i32 (...) @printf(ptr @23, ptr addrspace(4) @20, i64 %22, ptr addrspace(4) @21, double %23, ptr addrspace(4) @22)
|
|
; CHECK: %25 = call i32 @fflush(ptr null)
|
|
|
|
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
|
|
|
|
define void @foo(ptr noalias %A, ptr noalias %B, ptr noalias %C, ptr noalias %D, ptr noalias %E) {
|
|
bb:
|
|
br label %bb1
|
|
|
|
bb1: ; preds = %bb21, %bb
|
|
%i.0 = phi i64 [ 0, %bb ], [ %tmp22, %bb21 ]
|
|
%exitcond = icmp ne i64 %i.0, 10
|
|
br i1 %exitcond, label %bb2, label %bb23
|
|
|
|
bb2: ; preds = %bb1
|
|
%tmp = getelementptr inbounds double, ptr %B, i64 %i.0
|
|
%tmp3 = load double, ptr %tmp, align 8
|
|
%tmp4 = getelementptr inbounds i8, ptr %C, i64 %i.0
|
|
%tmp5 = load i8, ptr %tmp4, align 1
|
|
%tmp6 = sitofp i8 %tmp5 to double
|
|
%tmp7 = fadd double %tmp3, %tmp6
|
|
%tmp8 = getelementptr inbounds i32, ptr %D, i64 %i.0
|
|
%tmp9 = load i32, ptr %tmp8, align 4
|
|
%tmp10 = sitofp i32 %tmp9 to double
|
|
%tmp11 = fadd double %tmp7, %tmp10
|
|
%tmp12 = getelementptr inbounds i64, ptr %E, i64 %i.0
|
|
%tmp13 = load i64, ptr %tmp12, align 8
|
|
%tmp14 = sitofp i64 %tmp13 to double
|
|
%tmp15 = fadd double %tmp11, %tmp14
|
|
%tmp16 = getelementptr inbounds float, ptr %A, i64 %i.0
|
|
%tmp17 = load float, ptr %tmp16, align 4
|
|
%tmp18 = fpext float %tmp17 to double
|
|
%tmp19 = fadd double %tmp18, %tmp15
|
|
%tmp20 = fptrunc double %tmp19 to float
|
|
store float %tmp20, ptr %tmp16, align 4
|
|
br label %bb21
|
|
|
|
bb21: ; preds = %bb2
|
|
%tmp22 = add nsw i64 %i.0, 1
|
|
br label %bb1
|
|
|
|
bb23: ; preds = %bb1
|
|
ret void
|
|
}
|
|
|
|
define i32 @main() {
|
|
bb:
|
|
%A = alloca [10 x float], align 16
|
|
%B = alloca [10 x double], align 16
|
|
%C = alloca [10 x i8], align 1
|
|
%D = alloca [10 x i32], align 16
|
|
%E = alloca [10 x i64], align 16
|
|
br label %bb1
|
|
|
|
bb1: ; preds = %bb7, %bb
|
|
%i.0 = phi i64 [ 0, %bb ], [ %tmp8, %bb7 ]
|
|
%exitcond = icmp ne i64 %i.0, 10
|
|
br i1 %exitcond, label %bb2, label %bb9
|
|
|
|
bb2: ; preds = %bb1
|
|
fence seq_cst
|
|
%tmp = getelementptr inbounds [10 x i64], ptr %E, i64 0, i64 %i.0
|
|
store i64 42, ptr %tmp, align 8
|
|
%tmp3 = getelementptr inbounds [10 x i32], ptr %D, i64 0, i64 %i.0
|
|
store i32 42, ptr %tmp3, align 4
|
|
%tmp4 = getelementptr inbounds [10 x i8], ptr %C, i64 0, i64 %i.0
|
|
store i8 42, ptr %tmp4, align 1
|
|
%tmp5 = getelementptr inbounds [10 x double], ptr %B, i64 0, i64 %i.0
|
|
store double 4.200000e+01, ptr %tmp5, align 8
|
|
%tmp6 = getelementptr inbounds [10 x float], ptr %A, i64 0, i64 %i.0
|
|
store float 4.200000e+01, ptr %tmp6, align 4
|
|
br label %bb7
|
|
|
|
bb7: ; preds = %bb2
|
|
%tmp8 = add nsw i64 %i.0, 1
|
|
br label %bb1
|
|
|
|
bb9: ; preds = %bb1
|
|
%tmp10 = getelementptr inbounds [10 x float], ptr %A, i64 0, i64 0
|
|
%tmp11 = getelementptr inbounds [10 x double], ptr %B, i64 0, i64 0
|
|
%tmp12 = getelementptr inbounds [10 x i8], ptr %C, i64 0, i64 0
|
|
%tmp13 = getelementptr inbounds [10 x i32], ptr %D, i64 0, i64 0
|
|
%tmp14 = getelementptr inbounds [10 x i64], ptr %E, i64 0, i64 0
|
|
call void @foo(ptr %tmp10, ptr %tmp11, ptr %tmp12, ptr %tmp13, ptr %tmp14)
|
|
%tmp15 = getelementptr inbounds [10 x float], ptr %A, i64 0, i64 8
|
|
%tmp16 = load float, ptr %tmp15, align 16
|
|
%tmp17 = fptosi float %tmp16 to i32
|
|
ret i32 %tmp17
|
|
}
|