From faa1159a6915992d7f035ce06caf952fd4a4e96a Mon Sep 17 00:00:00 2001 From: Michael Liao Date: Thu, 30 Aug 2012 00:30:16 +0000 Subject: [PATCH] Fix PR13727 - The root cause is that target constant materialization in X86 fast-isel creates a PC-rel addressing which may overflow 32-bit range in non-Small code model if .rodata section is allocated too far away from code segment in MCJIT, which uses Large code model so far. - Follow the similar logic to fix non-Small code model in fast-isel by skipping non-Small code model. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@162881 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Target/X86/X86FastISel.cpp | 10 ++-- test/CodeGen/X86/pr13727.ll | 88 ++++++++++++++++++++++++++++++++++ 2 files changed, 95 insertions(+), 3 deletions(-) create mode 100644 test/CodeGen/X86/pr13727.ll diff --git a/lib/Target/X86/X86FastISel.cpp b/lib/Target/X86/X86FastISel.cpp index e5952aae16d..54704d84a9c 100644 --- a/lib/Target/X86/X86FastISel.cpp +++ b/lib/Target/X86/X86FastISel.cpp @@ -2014,13 +2014,17 @@ X86FastISel::TargetSelectInstruction(const Instruction *I) { unsigned X86FastISel::TargetMaterializeConstant(const Constant *C) { MVT VT; if (!isTypeLegal(C->getType(), VT)) - return false; + return 0; + + // Can't handle alternate code models yet. + if (TM.getCodeModel() != CodeModel::Small) + return 0; // Get opcode and regclass of the output for the given load instruction. unsigned Opc = 0; const TargetRegisterClass *RC = NULL; switch (VT.SimpleTy) { - default: return false; + default: return 0; case MVT::i8: Opc = X86::MOV8rm; RC = &X86::GR8RegClass; @@ -2058,7 +2062,7 @@ unsigned X86FastISel::TargetMaterializeConstant(const Constant *C) { break; case MVT::f80: // No f80 support yet. - return false; + return 0; } // Materialize addresses with LEA instructions. diff --git a/test/CodeGen/X86/pr13727.ll b/test/CodeGen/X86/pr13727.ll new file mode 100644 index 00000000000..5fa68f9b872 --- /dev/null +++ b/test/CodeGen/X86/pr13727.ll @@ -0,0 +1,88 @@ +; RUN: %lli -use-mcjit -O0 -disable-lazy-compilation=false %s + +; The intention of this test is to verify that symbols mapped to COMMON in ELF +; work as expected. +; +; Compiled from this C code: +; +; int zero_int; +; double zero_double; +; int zero_arr[10]; +; +; int main() +; { +; zero_arr[zero_int + 5] = 40; +; +; if (zero_double < 1.1) +; zero_arr[zero_int + 2] = 70; +; +; for (int i = 1; i < 10; ++i) { +; zero_arr[i] = zero_arr[i - 1] + zero_arr[i]; +; } +; return zero_arr[9] == 110 ? 0 : -1; +; } + +@zero_int = common global i32 0, align 4 +@zero_arr = common global [10 x i32] zeroinitializer, align 16 +@zero_double = common global double 0.000000e+00, align 8 + +define i32 @main() nounwind { +entry: + %retval = alloca i32, align 4 + %i = alloca i32, align 4 + store i32 0, i32* %retval + %0 = load i32* @zero_int, align 4 + %add = add nsw i32 %0, 5 + %idxprom = sext i32 %add to i64 + %arrayidx = getelementptr inbounds [10 x i32]* @zero_arr, i32 0, i64 %idxprom + store i32 40, i32* %arrayidx, align 4 + %1 = load double* @zero_double, align 8 + %cmp = fcmp olt double %1, 1.100000e+00 + br i1 %cmp, label %if.then, label %if.end + +if.then: ; preds = %entry + %2 = load i32* @zero_int, align 4 + %add1 = add nsw i32 %2, 2 + %idxprom2 = sext i32 %add1 to i64 + %arrayidx3 = getelementptr inbounds [10 x i32]* @zero_arr, i32 0, i64 %idxprom2 + store i32 70, i32* %arrayidx3, align 4 + br label %if.end + +if.end: ; preds = %if.then, %entry + store i32 1, i32* %i, align 4 + br label %for.cond + +for.cond: ; preds = %for.inc, %if.end + %3 = load i32* %i, align 4 + %cmp4 = icmp slt i32 %3, 10 + br i1 %cmp4, label %for.body, label %for.end + +for.body: ; preds = %for.cond + %4 = load i32* %i, align 4 + %sub = sub nsw i32 %4, 1 + %idxprom5 = sext i32 %sub to i64 + %arrayidx6 = getelementptr inbounds [10 x i32]* @zero_arr, i32 0, i64 %idxprom5 + %5 = load i32* %arrayidx6, align 4 + %6 = load i32* %i, align 4 + %idxprom7 = sext i32 %6 to i64 + %arrayidx8 = getelementptr inbounds [10 x i32]* @zero_arr, i32 0, i64 %idxprom7 + %7 = load i32* %arrayidx8, align 4 + %add9 = add nsw i32 %5, %7 + %8 = load i32* %i, align 4 + %idxprom10 = sext i32 %8 to i64 + %arrayidx11 = getelementptr inbounds [10 x i32]* @zero_arr, i32 0, i64 %idxprom10 + store i32 %add9, i32* %arrayidx11, align 4 + br label %for.inc + +for.inc: ; preds = %for.body + %9 = load i32* %i, align 4 + %inc = add nsw i32 %9, 1 + store i32 %inc, i32* %i, align 4 + br label %for.cond + +for.end: ; preds = %for.cond + %10 = load i32* getelementptr inbounds ([10 x i32]* @zero_arr, i32 0, i64 9), align 4 + %cmp12 = icmp eq i32 %10, 110 + %cond = select i1 %cmp12, i32 0, i32 -1 + ret i32 %cond +}