Test source: git
Comments: Bug in IR semantics or optimization: escapes noescape pointer
Source: <stdin>
-- 1. ModuleToFunctionPassAdaptor
ERROR: Unsupported attribute: noalias
-- 1. PassManager<Function> : Skipping NOP
-- 2. LoopVectorizePass
----------------------------------------
define i32 @foo(ptr nocapture %a, ptr nocapture %b, i32 %n) {
%entry:
%cmp6 = icmp sgt i32 %n, 0
br i1 %cmp6, label %for.body, label %for.end
%for.body:
%indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
%arrayidx = gep inbounds ptr nocapture %b, 4 x i64 %indvars.iv
%0 = load float, ptr %arrayidx, align 4
%mul = fmul float %0, 3.000000, exceptions=ignore
%arrayidx2 = gep inbounds ptr nocapture %a, 4 x i64 %indvars.iv
store float %mul, ptr %arrayidx2, align 4
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, %n
br i1 %exitcond, label %for.end, label %for.body
%for.end:
ret i32 undef
}
Transformation seems to be correct! (syntactically equal)
-- 3. LoopVectorizePass
----------------------------------------
define i32 @foo(ptr nocapture %a, ptr nocapture %b, i32 %n) {
%entry:
%cmp6 = icmp sgt i32 %n, 0
br i1 %cmp6, label %for.body, label %for.end
%for.body:
%indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
%arrayidx = gep inbounds ptr nocapture %b, 4 x i64 %indvars.iv
%0 = load float, ptr %arrayidx, align 4
%mul = fmul float %0, 3.000000, exceptions=ignore
%arrayidx2 = gep inbounds ptr nocapture %a, 4 x i64 %indvars.iv
store float %mul, ptr %arrayidx2, align 4
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, %n
br i1 %exitcond, label %for.end, label %for.body
%for.end:
ret i32 undef
}
=>
define i32 @foo(ptr nocapture %a, ptr nocapture %b, i32 %n) {
%entry:
%b2 = ptrtoint ptr nocapture %b to i64
%a1 = ptrtoint ptr nocapture %a to i64
%cmp6 = icmp sgt i32 %n, 0
br i1 %cmp6, label %for.body.preheader, label %for.end
%for.body.preheader:
%0 = add i32 %n, 4294967295
%1 = zext i32 %0 to i64
%2 = add nsw nuw i64 %1, 1
%min.iters.check = icmp ult i64 %2, 4
br i1 %min.iters.check, label %scalar.ph, label %vector.memcheck
%vector.memcheck:
%3 = sub i64 %a1, %b2
%diff.check = icmp ult i64 %3, 16
br i1 %diff.check, label %scalar.ph, label %vector.ph
%vector.ph:
%n.mod.vf = urem i64 %2, 4
%n.vec = sub i64 %2, %n.mod.vf
br label %vector.body
%vector.body:
%index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
%4 = add i64 %index, 0
%5 = gep inbounds ptr nocapture %b, 4 x i64 %4
%6 = gep inbounds ptr %5, 4 x i32 0
%7 = bitcast ptr %6 to ptr
%wide.load = load <4 x float>, ptr %7, align 4
%8 = fmul <4 x float> %wide.load, { 3.000000, 3.000000, 3.000000, 3.000000 }, exceptions=ignore
%9 = gep inbounds ptr nocapture %a, 4 x i64 %4
%10 = gep inbounds ptr %9, 4 x i32 0
%11 = bitcast ptr %10 to ptr
store <4 x float> %8, ptr %11, align 4
%index.next = add nuw i64 %index, 4
%12 = icmp eq i64 %index.next, %n.vec
br i1 %12, label %middle.block, label %vector.body
%middle.block:
%cmp.n = icmp eq i64 %2, %n.vec
br i1 %cmp.n, label %for.end.loopexit, label %scalar.ph
%scalar.ph:
%bc.resume.val = phi i64 [ %n.vec, %middle.block ], [ 0, %for.body.preheader ], [ 0, %vector.memcheck ]
br label %for.body
%for.body:
%indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ %bc.resume.val, %scalar.ph ]
%arrayidx = gep inbounds ptr nocapture %b, 4 x i64 %indvars.iv
%13 = load float, ptr %arrayidx, align 4
%mul = fmul float %13, 3.000000, exceptions=ignore
%arrayidx2 = gep inbounds ptr nocapture %a, 4 x i64 %indvars.iv
store float %mul, ptr %arrayidx2, align 4
%indvars.iv.next = add i64 %indvars.iv, 1
%lftr.wideiv = trunc i64 %indvars.iv.next to i32
%exitcond = icmp eq i32 %lftr.wideiv, %n
br i1 %exitcond, label %for.end.loopexit, label %for.body
%for.end.loopexit:
br label %for.end
%for.end:
ret i32 undef
}
Transformation doesn't verify! (unsound)
ERROR: Source is more defined than target
Example:
ptr nocapture %a = poison
ptr nocapture %b = poison
i32 %n = #x80000004 (2147483652, -2147483644)
Source:
i1 %cmp6 = #x0 (0)
>> Jump to %for.end
SOURCE MEMORY STATE
===================
NON-LOCAL BLOCKS:
Block 0 > size: 0 align: 4 alloc type: 0 address: 0
Block 1 > size: 1325516685390394441 align: 8589934592 alloc type: 0 address: 9857077814180729874
Block 2 > size: 1283600712156450305 align: 16 alloc type: 0 address: 13418686264881451588
Target:
i64 %b2 = UB triggered!
Pass: LoopVectorizePass
Command line: '/home/nlopes/llvm/build/bin/opt' '-load=/home/nlopes/alive2/build/tv/tv.so' '-load-pass-plugin=/home/nlopes/alive2/build/tv/tv.so' '-tv-exit-on-error' '-aa-pipeline=basic-aa' '-passes=loop-vectorize,dce,instcombine' '-force-vector-interleave=1' '-force-vector-width=4' '-S' '-tv-smt-to=20000' '-tv-report-dir=/home/nlopes/alive2/build/logs' '-tv-smt-stats'
------------------- SMT STATS -------------------
Num queries: 6
Num invalid: 0
Num skips: 0
Num trivial: 2 (25.0%)
Num timeout: 0 (0.0%)
Num errors: 0 (0.0%)
Num SAT: 6 (100.0%)
Num UNSAT: 0 (0.0%)
Alive2: Transform doesn't verify; aborting!
+ : 'RUN: at line 2' + /bitbucket/nlopes/llvm/build/bin/FileCheck /bitbucket/nlopes/llvm/llvm/test/Transforms/LoopVectorize/runtime-check.ll + /home/nlopes/alive2/build/opt-alive.sh -aa-pipeline=basic-aa -passes=loop-vectorize,dce,instcombine -force-vector-interleave=1 -force-vector-width=4 -S FileCheck error: '<stdin>' is empty. FileCheck command line: /bitbucket/nlopes/llvm/build/bin/FileCheck /bitbucket/nlopes/llvm/llvm/test/Transforms/LoopVectorize/runtime-check.ll