summaryrefslogtreecommitdiffhomepage
path: root/patches/llvm_host
diff options
context:
space:
mode:
Diffstat (limited to 'patches/llvm_host')
-rw-r--r--patches/llvm_host/llvm-0001-Add-Musl-MuslEABI-and-Musl-EABIHF-triples.patch112
-rw-r--r--patches/llvm_host/llvm-0002-Fix-build-with-musl-libc.patch35
-rw-r--r--patches/llvm_host/llvm-0003-Fix-DynamicLibrary-to-build-with-musl-libc.patch33
-rw-r--r--patches/llvm_host/llvm-0004-Fix-ScalarEvolutionExpander-step-scaling-bug.patch99
-rw-r--r--patches/llvm_host/llvm-0005-optimize-store-of-bitcast-from-vector-to-aggregate.patch181
-rw-r--r--patches/llvm_host/llvm-0006-clone-every-functions-debug-info.patch129
-rw-r--r--patches/llvm_host/llvm-0007-reduce-complexity-of-debug-info-clonning-and-fix-correctness.patch102
-rw-r--r--patches/llvm_host/llvm-0008-dont-widen-metadata-on-store-to-load-forwarding.patch107
-rw-r--r--patches/llvm_host/llvm-0009-nm-workaround.patch27
9 files changed, 825 insertions, 0 deletions
diff --git a/patches/llvm_host/llvm-0001-Add-Musl-MuslEABI-and-Musl-EABIHF-triples.patch b/patches/llvm_host/llvm-0001-Add-Musl-MuslEABI-and-Musl-EABIHF-triples.patch
new file mode 100644
index 00000000..c1dfc5bd
--- /dev/null
+++ b/patches/llvm_host/llvm-0001-Add-Musl-MuslEABI-and-Musl-EABIHF-triples.patch
@@ -0,0 +1,112 @@
+From 1cec99f6d53d442d8a8c9d2ed9f4a58f8fdc6412 Mon Sep 17 00:00:00 2001
+From: Natanael Copa <ncopa@alpinelinux.org>
+Date: Thu, 18 Feb 2016 08:30:07 +0100
+Subject: [PATCH 1/3] Add Musl, MuslEABI and Musl EABIHF triples
+
+---
+ include/llvm/ADT/Triple.h | 11 +++++++++++
+ lib/Support/Triple.cpp | 7 +++++++
+ lib/Target/ARM/ARMSubtarget.h | 3 +++
+ lib/Target/ARM/ARMTargetMachine.cpp | 2 ++
+ 4 files changed, 23 insertions(+)
+
+diff --git a/include/llvm/ADT/Triple.h b/include/llvm/ADT/Triple.h
+index e01db0a..2fc4fc1 100644
+--- a/include/llvm/ADT/Triple.h
++++ b/include/llvm/ADT/Triple.h
+@@ -174,6 +174,10 @@ public:
+ EABIHF,
+ Android,
+
++ Musl,
++ MuslEABI,
++ MuslEABIHF,
++
+ MSVC,
+ Itanium,
+ Cygnus,
+@@ -544,6 +548,13 @@ public:
+ /// Tests whether the target is Android
+ bool isAndroid() const { return getEnvironment() == Triple::Android; }
+
++ /// Tests whether the target is musl libc
++ bool isMusl() const {
++ return getEnvironment() == Triple::Musl ||
++ getEnvironment() == Triple::MuslEABI ||
++ getEnvironment() == Triple::MuslEABIHF;
++ }
++
+ /// @}
+ /// @name Mutators
+ /// @{
+diff --git a/lib/Support/Triple.cpp b/lib/Support/Triple.cpp
+index 11afcf7..d90a06a 100644
+--- a/lib/Support/Triple.cpp
++++ b/lib/Support/Triple.cpp
+@@ -199,6 +199,9 @@ const char *Triple::getEnvironmentTypeName(EnvironmentType Kind) {
+ case CODE16: return "code16";
+ case EABI: return "eabi";
+ case EABIHF: return "eabihf";
++ case Musl: return "musl";
++ case MuslEABIHF: return "muslgnueabihf";
++ case MuslEABI: return "muslgnueabi";
+ case Android: return "android";
+ case MSVC: return "msvc";
+ case Itanium: return "itanium";
+@@ -454,6 +457,9 @@ static Triple::EnvironmentType parseEnvironment(StringRef EnvironmentName) {
+ .StartsWith("code16", Triple::CODE16)
+ .StartsWith("gnu", Triple::GNU)
+ .StartsWith("android", Triple::Android)
++ .StartsWith("muslgnueabihf", Triple::MuslEABIHF)
++ .StartsWith("muslgnueabi", Triple::MuslEABI)
++ .StartsWith("musl", Triple::Musl)
+ .StartsWith("msvc", Triple::MSVC)
+ .StartsWith("itanium", Triple::Itanium)
+ .StartsWith("cygnus", Triple::Cygnus)
+@@ -1431,6 +1437,7 @@ StringRef Triple::getARMCPUForArch(StringRef MArch) const {
+ switch (getEnvironment()) {
+ case llvm::Triple::EABIHF:
+ case llvm::Triple::GNUEABIHF:
++ case llvm::Triple::MuslEABIHF:
+ return "arm1176jzf-s";
+ default:
+ return "arm7tdmi";
+diff --git a/lib/Target/ARM/ARMSubtarget.h b/lib/Target/ARM/ARMSubtarget.h
+index 4d54e57..7ed7ab7 100644
+--- a/lib/Target/ARM/ARMSubtarget.h
++++ b/lib/Target/ARM/ARMSubtarget.h
+@@ -405,8 +405,10 @@ public:
+ bool isTargetEHABICompatible() const {
+ return (TargetTriple.getEnvironment() == Triple::EABI ||
+ TargetTriple.getEnvironment() == Triple::GNUEABI ||
++ TargetTriple.getEnvironment() == Triple::MuslEABI ||
+ TargetTriple.getEnvironment() == Triple::EABIHF ||
+ TargetTriple.getEnvironment() == Triple::GNUEABIHF ||
++ TargetTriple.getEnvironment() == Triple::MuslEABIHF ||
+ isTargetAndroid()) &&
+ !isTargetDarwin() && !isTargetWindows();
+ }
+@@ -415,6 +417,7 @@ public:
+ // FIXME: this is invalid for WindowsCE
+ return TargetTriple.getEnvironment() == Triple::GNUEABIHF ||
+ TargetTriple.getEnvironment() == Triple::EABIHF ||
++ TargetTriple.getEnvironment() == Triple::MuslEABIHF ||
+ isTargetWindows() || isAAPCS16_ABI();
+ }
+ bool isTargetAndroid() const { return TargetTriple.isAndroid(); }
+diff --git a/lib/Target/ARM/ARMTargetMachine.cpp b/lib/Target/ARM/ARMTargetMachine.cpp
+index fca1901..25f5cf9 100644
+--- a/lib/Target/ARM/ARMTargetMachine.cpp
++++ b/lib/Target/ARM/ARMTargetMachine.cpp
+@@ -101,6 +101,8 @@ computeTargetABI(const Triple &TT, StringRef CPU,
+ case llvm::Triple::GNUEABIHF:
+ case llvm::Triple::EABIHF:
+ case llvm::Triple::EABI:
++ case llvm::Triple::MuslEABI:
++ case llvm::Triple::MuslEABIHF:
+ TargetABI = ARMBaseTargetMachine::ARM_ABI_AAPCS;
+ break;
+ case llvm::Triple::GNU:
+--
+2.7.3
+
diff --git a/patches/llvm_host/llvm-0002-Fix-build-with-musl-libc.patch b/patches/llvm_host/llvm-0002-Fix-build-with-musl-libc.patch
new file mode 100644
index 00000000..1a690808
--- /dev/null
+++ b/patches/llvm_host/llvm-0002-Fix-build-with-musl-libc.patch
@@ -0,0 +1,35 @@
+From 5c571082fdaf61f6df19d9b7137dc26d71334058 Mon Sep 17 00:00:00 2001
+From: Natanael Copa <ncopa@alpinelinux.org>
+Date: Thu, 18 Feb 2016 10:33:04 +0100
+Subject: [PATCH 2/3] Fix build with musl libc
+
+On musl libc the fopen64 and fopen are the same thing, but for
+compatibility they have a `#define fopen64 fopen`. Same applies for
+fseek64, fstat64, fstatvfs64, ftello64, lstat64, stat64 and tmpfile64.
+---
+ include/llvm/Analysis/TargetLibraryInfo.h | 9 +++++++++
+ 1 file changed, 9 insertions(+)
+
+diff --git a/include/llvm/Analysis/TargetLibraryInfo.h b/include/llvm/Analysis/TargetLibraryInfo.h
+index 7becdf0..7f14427 100644
+--- a/include/llvm/Analysis/TargetLibraryInfo.h
++++ b/include/llvm/Analysis/TargetLibraryInfo.h
+@@ -18,6 +18,15 @@
+ #include "llvm/IR/Module.h"
+ #include "llvm/Pass.h"
+
++#undef fopen64
++#undef fseeko64
++#undef fstat64
++#undef fstatvfs64
++#undef ftello64
++#undef lstat64
++#undef stat64
++#undef tmpfile64
++
+ namespace llvm {
+ /// VecDesc - Describes a possible vectorization of a function.
+ /// Function 'VectorFnName' is equivalent to 'ScalarFnName' vectorized
+--
+2.7.3
+
diff --git a/patches/llvm_host/llvm-0003-Fix-DynamicLibrary-to-build-with-musl-libc.patch b/patches/llvm_host/llvm-0003-Fix-DynamicLibrary-to-build-with-musl-libc.patch
new file mode 100644
index 00000000..d5d7f07b
--- /dev/null
+++ b/patches/llvm_host/llvm-0003-Fix-DynamicLibrary-to-build-with-musl-libc.patch
@@ -0,0 +1,33 @@
+From d12ecb83d01dcb580dd94f4d57828f33d3eb4c35 Mon Sep 17 00:00:00 2001
+From: Natanael Copa <ncopa@alpinelinux.org>
+Date: Thu, 18 Feb 2016 15:33:21 +0100
+Subject: [PATCH 3/3] Fix DynamicLibrary to build with musl libc
+
+stdin/out/err is part of the libc and not the kernel so we check for the
+specific libc that does the unexpected instead of linux.
+
+This is needed for making it build with musl libc.
+---
+ lib/Support/DynamicLibrary.cpp | 4 ++--
+ 1 file changed, 2 insertions(+), 2 deletions(-)
+
+diff --git a/lib/Support/DynamicLibrary.cpp b/lib/Support/DynamicLibrary.cpp
+index 9a7aeb5..0c1c8f8 100644
+--- a/lib/Support/DynamicLibrary.cpp
++++ b/lib/Support/DynamicLibrary.cpp
+@@ -140,10 +140,10 @@ void* DynamicLibrary::SearchForAddressOfSymbol(const char *symbolName) {
+ #define EXPLICIT_SYMBOL(SYM) \
+ if (!strcmp(symbolName, #SYM)) return &SYM
+
+-// On linux we have a weird situation. The stderr/out/in symbols are both
++// On GNU libc we have a weird situation. The stderr/out/in symbols are both
+ // macros and global variables because of standards requirements. So, we
+ // boldly use the EXPLICIT_SYMBOL macro without checking for a #define first.
+-#if defined(__linux__) and !defined(__ANDROID__)
++#if defined(__GLIBC__)
+ {
+ EXPLICIT_SYMBOL(stderr);
+ EXPLICIT_SYMBOL(stdout);
+--
+2.7.3
+
diff --git a/patches/llvm_host/llvm-0004-Fix-ScalarEvolutionExpander-step-scaling-bug.patch b/patches/llvm_host/llvm-0004-Fix-ScalarEvolutionExpander-step-scaling-bug.patch
new file mode 100644
index 00000000..9945bf35
--- /dev/null
+++ b/patches/llvm_host/llvm-0004-Fix-ScalarEvolutionExpander-step-scaling-bug.patch
@@ -0,0 +1,99 @@
+From c8ce9e59a883f452bc538b5f16fb83649472dc3e Mon Sep 17 00:00:00 2001
+From: Keno Fischer <kfischer@college.harvard.edu>
+Date: Wed, 13 Jul 2016 01:28:12 +0000
+Subject: [PATCH] Fix ScalarEvolutionExpander step scaling bug
+
+The expandAddRecExprLiterally function incorrectly transforms
+`[Start + Step * X]` into `Step * [Start + X]` instead of the correct
+transform of `[Step * X] + Start`.
+
+This caused https://github.com/JuliaLang/julia/issues/14704#issuecomment-174126219
+due to what appeared to be sufficiently complicated loop interactions.
+
+Patch by Jameson Nash (jameson@juliacomputing.com).
+
+Reviewers: sanjoy
+Differential Revision: http://reviews.llvm.org/D16505
+
+git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@275239 91177308-0d34-0410-b5e6-96231b3b80d8
+
+Alpine maintainer notes:
+This patch replaces deps/patches/llvm-3.7.1_3.patch from Julia.
+---
+ lib/Analysis/ScalarEvolutionExpander.cpp | 7 ++++
+ .../ScalarEvolution/incorrect-offset-scaling.ll | 48 ++++++++++++++++++++++
+ 2 files changed, 55 insertions(+)
+ create mode 100644 test/Analysis/ScalarEvolution/incorrect-offset-scaling.ll
+
+diff --git a/lib/Analysis/ScalarEvolutionExpander.cpp b/lib/Analysis/ScalarEvolutionExpander.cpp
+index 7716435..77e4ec7 100644
+--- a/lib/Analysis/ScalarEvolutionExpander.cpp
++++ b/lib/Analysis/ScalarEvolutionExpander.cpp
+@@ -1288,6 +1288,13 @@ Value *SCEVExpander::expandAddRecExprLiterally(const SCEVAddRecExpr *S) {
+ if (!SE.dominates(Step, L->getHeader())) {
+ PostLoopScale = Step;
+ Step = SE.getConstant(Normalized->getType(), 1);
++ if (!Start->isZero()) {
++ // The normalization below assumes that Start is constant zero, so if
++ // it isn't re-associate Start to PostLoopOffset.
++ assert(!PostLoopOffset && "Start not-null but PostLoopOffset set?");
++ PostLoopOffset = Start;
++ Start = SE.getConstant(Normalized->getType(), 0);
++ }
+ Normalized =
+ cast<SCEVAddRecExpr>(SE.getAddRecExpr(
+ Start, Step, Normalized->getLoop(),
+diff --git a/test/Analysis/ScalarEvolution/incorrect-offset-scaling.ll b/test/Analysis/ScalarEvolution/incorrect-offset-scaling.ll
+new file mode 100644
+index 0000000..7ffb093
+--- /dev/null
++++ b/test/Analysis/ScalarEvolution/incorrect-offset-scaling.ll
+@@ -0,0 +1,48 @@
++; RUN: opt -S -loop-reduce < %s | FileCheck %s
++
++target triple = "x86_64-unknown-unknown"
++target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
++
++define void @incorrect_offset_scaling(i64, i64*) {
++top:
++ br label %L
++
++L: ; preds = %idxend.10, %idxend, %L2, %top
++ br i1 undef, label %L, label %L1
++
++L1: ; preds = %L1.preheader, %L2
++ %r13 = phi i64 [ %r1, %L2 ], [ 1, %L ]
++; CHECK: %lsr.iv = phi i64 [ 0, %L{{[^ ]+}} ], [ %lsr.iv.next, %L2 ]
++; CHECK-NOT: %lsr.iv = phi i64 [ -1, %L{{[^ ]+}} ], [ %lsr.iv.next, %L2 ]
++; CHECK: br
++ %r0 = add i64 %r13, -1
++ br label %idxend.8
++
++L2: ; preds = %idxend.8
++ %r1 = add i64 %r13, 1
++ br i1 undef, label %L, label %L1
++
++if6: ; preds = %idxend.8
++ %r2 = add i64 %0, -1
++ %r3 = load i64, i64* %1, align 8
++; CHECK-NOT: %r2
++; CHECK: %r3 = load i64
++ br label %ib
++
++idxend.8: ; preds = %L1
++ br i1 undef, label %if6, label %L2
++
++ib: ; preds = %if6
++ %r4 = mul i64 %r3, %r0
++ %r5 = add i64 %r2, %r4
++ %r6 = icmp ult i64 %r5, undef
++; CHECK %2 = mul i64 %lsr.iv, %r3
++; CHECK %3 = add i64 %1, -1
++; CHECK %4 = add i64 %0, %r3
++; CHECK %r6
++ %r7 = getelementptr i64, i64* undef, i64 %r5
++ store i64 1, i64* %r7, align 8
++; CHECK %5 = mul i64 %lsr.iv, %r3
++; CHECK %6 = add i64 %5, -1
++ br label %L
++}
diff --git a/patches/llvm_host/llvm-0005-optimize-store-of-bitcast-from-vector-to-aggregate.patch b/patches/llvm_host/llvm-0005-optimize-store-of-bitcast-from-vector-to-aggregate.patch
new file mode 100644
index 00000000..7f891b66
--- /dev/null
+++ b/patches/llvm_host/llvm-0005-optimize-store-of-bitcast-from-vector-to-aggregate.patch
@@ -0,0 +1,181 @@
+From 1ca1fcaa5b4c75a65a202badfd5df8240a36ca0f Mon Sep 17 00:00:00 2001
+From: "Arch D. Robison" <arch.robison@intel.com>
+Date: Mon, 25 Apr 2016 22:22:39 +0000
+Subject: [PATCH] Optimize store of "bitcast" from vector to aggregate.
+
+This patch is what was the "instcombine" portion of D14185, with an additional
+test added (see julia_pseudovec in test/Transforms/InstCombine/insert-val-extract-elem.ll).
+The patch causes instcombine to replace sequences of extractelement-insertvalue-store
+that act essentially like a bitcast followed by a store.
+
+Differential review: http://reviews.llvm.org/D14260
+
+git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@267482 91177308-0d34-0410-b5e6-96231b3b80d8
+
+Alpine maintainer notes:
+ - Updated for llvm 3.8.1.
+ - This patch replaces llvm-D14260.patch from Julia.
+---
+ .../InstCombine/InstCombineLoadStoreAlloca.cpp | 60 ++++++++++++++++++
+ .../InstCombine/insert-val-extract-elem.ll | 74 ++++++++++++++++++++++
+ 2 files changed, 134 insertions(+)
+ create mode 100644 test/Transforms/InstCombine/insert-val-extract-elem.ll
+
+diff --git a/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp b/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp
+index 96f0908..0ee6045 100644
+--- a/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp
++++ b/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp
+@@ -913,6 +913,61 @@ Instruction *InstCombiner::visitLoadInst(LoadInst &LI) {
+ return nullptr;
+ }
+
++/// \brief Look for extractelement/insertvalue sequence that acts like a bitcast.
++///
++/// \returns underlying value that was "cast", or nullptr otherwise.
++///
++/// For example, if we have:
++///
++/// %E0 = extractelement <2 x double> %U, i32 0
++/// %V0 = insertvalue [2 x double] undef, double %E0, 0
++/// %E1 = extractelement <2 x double> %U, i32 1
++/// %V1 = insertvalue [2 x double] %V0, double %E1, 1
++///
++/// and the layout of a <2 x double> is isomorphic to a [2 x double],
++/// then %V1 can be safely approximated by a conceptual "bitcast" of %U.
++/// Note that %U may contain non-undef values where %V1 has undef.
++static Value *likeBitCastFromVector(InstCombiner &IC, Value *V) {
++ Value *U = nullptr;
++ while (auto *IV = dyn_cast<InsertValueInst>(V)) {
++ auto *E = dyn_cast<ExtractElementInst>(IV->getInsertedValueOperand());
++ if (!E)
++ return nullptr;
++ auto *W = E->getVectorOperand();
++ if (!U)
++ U = W;
++ else if (U != W)
++ return nullptr;
++ auto *CI = dyn_cast<ConstantInt>(E->getIndexOperand());
++ if (!CI || IV->getNumIndices() != 1 || CI->getZExtValue() != *IV->idx_begin())
++ return nullptr;
++ V = IV->getAggregateOperand();
++ }
++ if (!isa<UndefValue>(V) ||!U)
++ return nullptr;
++
++ auto *UT = cast<VectorType>(U->getType());
++ auto *VT = V->getType();
++ // Check that types UT and VT are bitwise isomorphic.
++ const auto &DL = IC.getDataLayout();
++ if (DL.getTypeStoreSizeInBits(UT) != DL.getTypeStoreSizeInBits(VT)) {
++ return nullptr;
++ }
++ if (auto *AT = dyn_cast<ArrayType>(VT)) {
++ if (AT->getNumElements() != UT->getNumElements())
++ return nullptr;
++ } else {
++ auto *ST = cast<StructType>(VT);
++ if (ST->getNumElements() != UT->getNumElements())
++ return nullptr;
++ for (const auto *EltT : ST->elements()) {
++ if (EltT != UT->getElementType())
++ return nullptr;
++ }
++ }
++ return U;
++}
++
+ /// \brief Combine stores to match the type of value being stored.
+ ///
+ /// The core idea here is that the memory does not have any intrinsic type and
+@@ -924,6 +979,11 @@
+ return true;
+ }
+
++ if (Value *U = likeBitCastFromVector(IC, V)) {
++ combineStoreToNewValue(IC, SI, U);
++ return true;
++ }
++
+ // FIXME: We should also canonicalize loads of vectors when their elements are
+ // cast to other types.
+ return false;
+diff --git a/test/Transforms/InstCombine/insert-val-extract-elem.ll b/test/Transforms/InstCombine/insert-val-extract-elem.ll
+new file mode 100644
+index 0000000..db7b403
+--- /dev/null
++++ b/test/Transforms/InstCombine/insert-val-extract-elem.ll
+@@ -0,0 +1,74 @@
++; RUN: opt -S -instcombine %s | FileCheck %s
++
++; CHECK-LABEL: julia_2xdouble
++; CHECK-NOT: insertvalue
++; CHECK-NOT: extractelement
++; CHECK: store <2 x double>
++define void @julia_2xdouble([2 x double]* sret, <2 x double>*) {
++top:
++ %x = load <2 x double>, <2 x double>* %1
++ %x0 = extractelement <2 x double> %x, i32 0
++ %i0 = insertvalue [2 x double] undef, double %x0, 0
++ %x1 = extractelement <2 x double> %x, i32 1
++ %i1 = insertvalue [2 x double] %i0, double %x1, 1
++ store [2 x double] %i1, [2 x double]* %0, align 4
++ ret void
++}
++
++; Test with two inserts to the same index
++; CHECK-LABEL: julia_2xi64
++; CHECK-NOT: insertvalue
++; CHECK-NOT: extractelement
++; CHECK: store <2 x i64>
++define void @julia_2xi64([2 x i64]* sret, <2 x i64>*) {
++top:
++ %x = load <2 x i64>, <2 x i64>* %1
++ %x0 = extractelement <2 x i64> %x, i32 1
++ %i0 = insertvalue [2 x i64] undef, i64 %x0, 0
++ %x1 = extractelement <2 x i64> %x, i32 1
++ %i1 = insertvalue [2 x i64] %i0, i64 %x1, 1
++ %x2 = extractelement <2 x i64> %x, i32 0
++ %i2 = insertvalue [2 x i64] %i1, i64 %x2, 0
++ store [2 x i64] %i2, [2 x i64]* %0, align 4
++ ret void
++}
++
++; CHECK-LABEL: julia_4xfloat
++; CHECK-NOT: insertvalue
++; CHECK-NOT: extractelement
++; CHECK: store <4 x float>
++define void @julia_4xfloat([4 x float]* sret, <4 x float>*) {
++top:
++ %x = load <4 x float>, <4 x float>* %1
++ %x0 = extractelement <4 x float> %x, i32 0
++ %i0 = insertvalue [4 x float] undef, float %x0, 0
++ %x1 = extractelement <4 x float> %x, i32 1
++ %i1 = insertvalue [4 x float] %i0, float %x1, 1
++ %x2 = extractelement <4 x float> %x, i32 2
++ %i2 = insertvalue [4 x float] %i1, float %x2, 2
++ %x3 = extractelement <4 x float> %x, i32 3
++ %i3 = insertvalue [4 x float] %i2, float %x3, 3
++ store [4 x float] %i3, [4 x float]* %0, align 4
++ ret void
++}
++
++%pseudovec = type { float, float, float, float }
++
++; CHECK-LABEL: julia_pseudovec
++; CHECK-NOT: insertvalue
++; CHECK-NOT: extractelement
++; CHECK: store <4 x float>
++define void @julia_pseudovec(%pseudovec* sret, <4 x float>*) {
++top:
++ %x = load <4 x float>, <4 x float>* %1
++ %x0 = extractelement <4 x float> %x, i32 0
++ %i0 = insertvalue %pseudovec undef, float %x0, 0
++ %x1 = extractelement <4 x float> %x, i32 1
++ %i1 = insertvalue %pseudovec %i0, float %x1, 1
++ %x2 = extractelement <4 x float> %x, i32 2
++ %i2 = insertvalue %pseudovec %i1, float %x2, 2
++ %x3 = extractelement <4 x float> %x, i32 3
++ %i3 = insertvalue %pseudovec %i2, float %x3, 3
++ store %pseudovec %i3, %pseudovec* %0, align 4
++ ret void
++}
diff --git a/patches/llvm_host/llvm-0006-clone-every-functions-debug-info.patch b/patches/llvm_host/llvm-0006-clone-every-functions-debug-info.patch
new file mode 100644
index 00000000..567ddcad
--- /dev/null
+++ b/patches/llvm_host/llvm-0006-clone-every-functions-debug-info.patch
@@ -0,0 +1,129 @@
+From 11adcc4de0797c83e61ae0240927f0bafcf041a9 Mon Sep 17 00:00:00 2001
+From: Keno Fischer <kfischer@college.harvard.edu>
+Date: Sat, 13 Feb 2016 02:04:29 +0000
+Subject: [PATCH] [Cloning] Clone every Function's Debug Info
+
+Summary:
+Export the CloneDebugInfoMetadata utility, which clones all debug info
+associated with a function into the first module. Also use this function
+in CloneModule on each function we clone (the CloneFunction entrypoint
+already does this).
+
+Without this, cloning a module will lead to DI quality regressions,
+especially since r252219 reversed the Function <-> DISubprogram edge
+(before we could get lucky and have this edge preserved if the
+DISubprogram itself was, e.g. due to location metadata).
+
+This was verified to fix missing debug information in julia and
+a unittest to verify the new behavior is included.
+
+Patch by Yichao Yu! Thanks!
+
+Reviewers: loladiro, pcc
+Differential Revision: http://reviews.llvm.org/D17165
+
+git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@260791 91177308-0d34-0410-b5e6-96231b3b80d8
+
+Alpine maintainer notes:
+ - Corresponds to llvm-D17165-D18583.patch in Julia.
+---
+ include/llvm/Transforms/Utils/Cloning.h | 5 +++++
+ lib/Transforms/Utils/CloneFunction.cpp | 4 ++--
+ lib/Transforms/Utils/CloneModule.cpp | 1 +
+ unittests/Transforms/Utils/Cloning.cpp | 25 +++++++++++++++++++++++++
+ 4 files changed, 33 insertions(+), 2 deletions(-)
+
+diff --git a/include/llvm/Transforms/Utils/Cloning.h b/include/llvm/Transforms/Utils/Cloning.h
+index 4f006f2..0bae2bd 100644
+--- a/include/llvm/Transforms/Utils/Cloning.h
++++ b/include/llvm/Transforms/Utils/Cloning.h
+@@ -130,6 +130,11 @@ Function *CloneFunction(const Function *F, ValueToValueMapTy &VMap,
+ bool ModuleLevelChanges,
+ ClonedCodeInfo *CodeInfo = nullptr);
+
++/// Clone the module-level debug info associated with OldFunc. The cloned data
++/// will point to NewFunc instead.
++void CloneDebugInfoMetadata(Function *NewFunc, const Function *OldFunc,
++ ValueToValueMapTy &VMap);
++
+ /// Clone OldFunc into NewFunc, transforming the old arguments into references
+ /// to VMap values. Note that if NewFunc already has basic blocks, the ones
+ /// cloned into it will be added to the end of the function. This function
+diff --git a/lib/Transforms/Utils/CloneFunction.cpp b/lib/Transforms/Utils/CloneFunction.cpp
+index 6454afb..8b5692a 100644
+--- a/lib/Transforms/Utils/CloneFunction.cpp
++++ b/lib/Transforms/Utils/CloneFunction.cpp
+@@ -187,8 +187,8 @@ static void AddOperand(DICompileUnit *CU, DISubprogramArray SPs,
+
+ // Clone the module-level debug info associated with OldFunc. The cloned data
+ // will point to NewFunc instead.
+-static void CloneDebugInfoMetadata(Function *NewFunc, const Function *OldFunc,
+- ValueToValueMapTy &VMap) {
++void llvm::CloneDebugInfoMetadata(Function *NewFunc, const Function *OldFunc,
++ ValueToValueMapTy &VMap) {
+ DebugInfoFinder Finder;
+ Finder.processModule(*OldFunc->getParent());
+
+diff --git a/lib/Transforms/Utils/CloneModule.cpp b/lib/Transforms/Utils/CloneModule.cpp
+index 53de62a..b16a02a 100644
+--- a/lib/Transforms/Utils/CloneModule.cpp
++++ b/lib/Transforms/Utils/CloneModule.cpp
+@@ -136,6 +136,7 @@ std::unique_ptr<Module> llvm::CloneModule(
+ VMap[&*J] = &*DestI++;
+ }
+
++ CloneDebugInfoMetadata(F, &*I, VMap);
+ SmallVector<ReturnInst*, 8> Returns; // Ignore returns cloned.
+ CloneFunctionInto(F, &*I, VMap, /*ModuleLevelChanges=*/true, Returns);
+ }
+diff --git a/unittests/Transforms/Utils/Cloning.cpp b/unittests/Transforms/Utils/Cloning.cpp
+index 25e322e..b761e4e 100644
+--- a/unittests/Transforms/Utils/Cloning.cpp
++++ b/unittests/Transforms/Utils/Cloning.cpp
+@@ -423,6 +423,7 @@ class CloneModule : public ::testing::Test {
+ void SetupModule() { OldM = new Module("", C); }
+
+ void CreateOldModule() {
++ DIBuilder DBuilder(*OldM);
+ IRBuilder<> IBuilder(C);
+
+ auto *FuncType = FunctionType::get(Type::getVoidTy(C), false);
+@@ -431,9 +432,25 @@ class CloneModule : public ::testing::Test {
+ auto *F =
+ Function::Create(FuncType, GlobalValue::PrivateLinkage, "f", OldM);
+ F->setPersonalityFn(PersFn);
++
++ // Create debug info
++ auto *File = DBuilder.createFile("filename.c", "/file/dir/");
++ DITypeRefArray ParamTypes = DBuilder.getOrCreateTypeArray(None);
++ DISubroutineType *DFuncType = DBuilder.createSubroutineType(ParamTypes);
++ auto *CU =
++ DBuilder.createCompileUnit(dwarf::DW_LANG_C99, "filename.c",
++ "/file/dir", "CloneModule", false, "", 0);
++ // Function DI
++ auto *Subprogram = DBuilder.createFunction(CU, "f", "f", File, 4, DFuncType,
++ true, true, 3, 0, false);
++ F->setSubprogram(Subprogram);
++
+ auto *Entry = BasicBlock::Create(C, "", F);
+ IBuilder.SetInsertPoint(Entry);
+ IBuilder.CreateRetVoid();
++
++ // Finalize the debug info
++ DBuilder.finalize();
+ }
+
+ void CreateNewModule() { NewM = llvm::CloneModule(OldM).release(); }
+@@ -447,4 +464,12 @@ TEST_F(CloneModule, Verify) {
+ EXPECT_FALSE(verifyModule(*NewM));
+ }
+
++TEST_F(CloneModule, Subprogram) {
++ Function *NewF = NewM->getFunction("f");
++ DISubprogram *SP = NewF->getSubprogram();
++ EXPECT_TRUE(SP != nullptr);
++ EXPECT_EQ(SP->getName(), "f");
++ EXPECT_EQ(SP->getFile()->getFilename(), "filename.c");
++ EXPECT_EQ(SP->getLine(), (unsigned)4);
++}
+ }
diff --git a/patches/llvm_host/llvm-0007-reduce-complexity-of-debug-info-clonning-and-fix-correctness.patch b/patches/llvm_host/llvm-0007-reduce-complexity-of-debug-info-clonning-and-fix-correctness.patch
new file mode 100644
index 00000000..0930c28e
--- /dev/null
+++ b/patches/llvm_host/llvm-0007-reduce-complexity-of-debug-info-clonning-and-fix-correctness.patch
@@ -0,0 +1,102 @@
+From af289e04413504c3bdc252e08c3fe17bf7ea6dc8 Mon Sep 17 00:00:00 2001
+From: Peter Collingbourne <peter@pcc.me.uk>
+Date: Wed, 30 Mar 2016 22:05:13 +0000
+Subject: [PATCH] Cloning: Reduce complexity of debug info cloning and fix
+ correctness issue.
+
+Commit r260791 contained an error in that it would introduce a cross-module
+reference in the old module. It also introduced O(N^2) complexity in the
+module cloner by requiring the entire module to be visited for each function.
+Fix both of these problems by avoiding use of the CloneDebugInfoMetadata
+function (which is only designed to do intra-module cloning) and cloning
+function-attached metadata in the same way that we clone all other metadata.
+
+Differential Revision: http://reviews.llvm.org/D18583
+
+git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@264935 91177308-0d34-0410-b5e6-96231b3b80d8
+
+Alpine maintainer notes:
+ - Corresponds to llvm-D17165-D18583.patch in Julia.
+---
+ include/llvm/Transforms/Utils/Cloning.h | 5 -----
+ lib/Transforms/Utils/CloneFunction.cpp | 13 +++++++++++--
+ lib/Transforms/Utils/CloneModule.cpp | 1 -
+ unittests/Transforms/Utils/Cloning.cpp | 6 ++++++
+ 4 files changed, 17 insertions(+), 8 deletions(-)
+
+diff --git a/include/llvm/Transforms/Utils/Cloning.h b/include/llvm/Transforms/Utils/Cloning.h
+index 0bae2bd..4f006f2 100644
+--- a/include/llvm/Transforms/Utils/Cloning.h
++++ b/include/llvm/Transforms/Utils/Cloning.h
+@@ -130,11 +130,6 @@ Function *CloneFunction(const Function *F, ValueToValueMapTy &VMap,
+ bool ModuleLevelChanges,
+ ClonedCodeInfo *CodeInfo = nullptr);
+
+-/// Clone the module-level debug info associated with OldFunc. The cloned data
+-/// will point to NewFunc instead.
+-void CloneDebugInfoMetadata(Function *NewFunc, const Function *OldFunc,
+- ValueToValueMapTy &VMap);
+-
+ /// Clone OldFunc into NewFunc, transforming the old arguments into references
+ /// to VMap values. Note that if NewFunc already has basic blocks, the ones
+ /// cloned into it will be added to the end of the function. This function
+diff --git a/lib/Transforms/Utils/CloneFunction.cpp b/lib/Transforms/Utils/CloneFunction.cpp
+index 05b0a17..8e1715a 100644
+--- a/lib/Transforms/Utils/CloneFunction.cpp
++++ b/lib/Transforms/Utils/CloneFunction.cpp
+@@ -119,6 +119,15 @@ void llvm::CloneFunctionInto(Function *NewFunc, const Function *OldFunc,
+ .addAttributes(NewFunc->getContext(), AttributeSet::FunctionIndex,
+ OldAttrs.getFnAttributes()));
+
++ SmallVector<std::pair<unsigned, MDNode *>, 1> MDs;
++ OldFunc->getAllMetadata(MDs);
++ for (auto MD : MDs)
++ NewFunc->setMetadata(
++ MD.first,
++ MapMetadata(MD.second, VMap,
++ ModuleLevelChanges ? RF_None : RF_NoModuleLevelChanges,
++ TypeMapper, Materializer));
++
+ // Loop over all of the basic blocks in the function, cloning them as
+ // appropriate. Note that we save BE this way in order to handle cloning of
+ // recursive functions into themselves.
+@@ -187,8 +196,8 @@ static void AddOperand(DICompileUnit *CU, DISubprogramArray SPs,
+
+ // Clone the module-level debug info associated with OldFunc. The cloned data
+ // will point to NewFunc instead.
+-void llvm::CloneDebugInfoMetadata(Function *NewFunc, const Function *OldFunc,
+- ValueToValueMapTy &VMap) {
++static void CloneDebugInfoMetadata(Function *NewFunc, const Function *OldFunc,
++ ValueToValueMapTy &VMap) {
+ DebugInfoFinder Finder;
+ Finder.processModule(*OldFunc->getParent());
+
+diff --git a/lib/Transforms/Utils/CloneModule.cpp b/lib/Transforms/Utils/CloneModule.cpp
+index 494e275..929f51b 100644
+--- a/lib/Transforms/Utils/CloneModule.cpp
++++ b/lib/Transforms/Utils/CloneModule.cpp
+@@ -138,7 +138,6 @@ std::unique_ptr<Module> llvm::CloneModule(
+ VMap[&*J] = &*DestI++;
+ }
+
+- CloneDebugInfoMetadata(F, &*I, VMap);
+ SmallVector<ReturnInst*, 8> Returns; // Ignore returns cloned.
+ CloneFunctionInto(F, &*I, VMap, /*ModuleLevelChanges=*/true, Returns);
+ }
+diff --git a/unittests/Transforms/Utils/Cloning.cpp b/unittests/Transforms/Utils/Cloning.cpp
+index b761e4e..f06a20f 100644
+--- a/unittests/Transforms/Utils/Cloning.cpp
++++ b/unittests/Transforms/Utils/Cloning.cpp
+@@ -464,6 +464,12 @@ TEST_F(CloneModule, Verify) {
+ EXPECT_FALSE(verifyModule(*NewM));
+ }
+
++TEST_F(CloneModule, OldModuleUnchanged) {
++ DebugInfoFinder Finder;
++ Finder.processModule(*OldM);
++ EXPECT_EQ(1U, Finder.subprogram_count());
++}
++
+ TEST_F(CloneModule, Subprogram) {
+ Function *NewF = NewM->getFunction("f");
+ DISubprogram *SP = NewF->getSubprogram();
diff --git a/patches/llvm_host/llvm-0008-dont-widen-metadata-on-store-to-load-forwarding.patch b/patches/llvm_host/llvm-0008-dont-widen-metadata-on-store-to-load-forwarding.patch
new file mode 100644
index 00000000..b66584dc
--- /dev/null
+++ b/patches/llvm_host/llvm-0008-dont-widen-metadata-on-store-to-load-forwarding.patch
@@ -0,0 +1,107 @@
+From 3c80c2658022201214241e9229ac35097cc476d2 Mon Sep 17 00:00:00 2001
+From: Eli Friedman <eli.friedman@gmail.com>
+Date: Thu, 16 Jun 2016 02:33:42 +0000
+Subject: [PATCH] [InstCombine] Don't widen metadata on store-to-load
+ forwarding
+
+The original check for load CSE or store-to-load forwarding is wrong
+when the forwarded stored value happened to be a load.
+
+Ref https://github.com/JuliaLang/julia/issues/16894
+
+Differential Revision: http://reviews.llvm.org/D21271
+
+Patch by Yichao Yu!
+
+git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@272868 91177308-0d34-0410-b5e6-96231b3b80d8
+
+Alpine maintainer notes:
+ - Updated for llvm 3.8.1.
+ - Corresponds to llvm-D21271-instcombine-tbaa-3.8.patch in Julia.
+---
+ include/llvm/Analysis/Loads.h | 3 ++-
+ lib/Analysis/Loads.cpp | 5 ++++-
+ .../InstCombine/InstCombineLoadStoreAlloca.cpp | 6 ++++--
+ test/Transforms/InstCombine/tbaa-store-to-load.ll | 17 +++++++++++++++++
+ 4 files changed, 27 insertions(+), 4 deletions(-)
+ create mode 100644 test/Transforms/InstCombine/tbaa-store-to-load.ll
+
+diff --git a/include/llvm/Analysis/Loads.h b/include/llvm/Analysis/Loads.h
+index e5bd0c8..9d24b7b 100644
+--- a/include/llvm/Analysis/Loads.h
++++ b/include/llvm/Analysis/Loads.h
+@@ -82,7 +82,8 @@ Value *FindAvailableLoadedValue(LoadInst *Load, BasicBlock *ScanBB,
+ BasicBlock::iterator &ScanFrom,
+ unsigned MaxInstsToScan = DefMaxInstsToScan,
+ AliasAnalysis *AA = nullptr,
+- AAMDNodes *AATags = nullptr);
++ AAMDNodes *AATags = nullptr,
++ bool *IsLoadCSE = nullptr);
+
+ }
+
+diff --git a/lib/Analysis/Loads.cpp b/lib/Analysis/Loads.cpp
+index dce243c..7d3fd59 100644
+--- a/lib/Analysis/Loads.cpp
++++ b/lib/Analysis/Loads.cpp
+@@ -322,7 +322,8 @@ llvm::DefMaxInstsToScan("available-load-scan-limit", cl::init(6), cl::Hidden,
+ Value *llvm::FindAvailableLoadedValue(LoadInst *Load, BasicBlock *ScanBB,
+ BasicBlock::iterator &ScanFrom,
+ unsigned MaxInstsToScan,
+- AliasAnalysis *AA, AAMDNodes *AATags) {
++ AliasAnalysis *AA, AAMDNodes *AATags,
++ bool *IsLoadCSE) {
+ if (MaxInstsToScan == 0)
+ MaxInstsToScan = ~0U;
+
+@@ -374,6 +375,8 @@ Value *llvm::FindAvailableLoadedValue(LoadInst *Load, BasicBlock *ScanBB,
+
+ if (AATags)
+ LI->getAAMetadata(*AATags);
++ if (IsLoadCSE)
++ *IsLoadCSE = true;
+ return LI;
+ }
+
+diff --git a/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp b/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp
+index 6a5d5a6..d312983 100644
+--- a/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp
++++ b/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp
+@@ -800,10 +800,12 @@ Instruction *InstCombiner::visitLoadInst(LoadInst &LI) {
+ // separated by a few arithmetic operations.
+ BasicBlock::iterator BBI(LI);
+ AAMDNodes AATags;
++ bool IsLoadCSE = false;
+ if (Value *AvailableVal =
+ FindAvailableLoadedValue(Op, LI.getParent(), BBI,
+- DefMaxInstsToScan, AA, &AATags)) {
+- if (LoadInst *NLI = dyn_cast<LoadInst>(AvailableVal)) {
++ DefMaxInstsToScan, AA, &AATags, &IsLoadCSE)) {
++ if (IsLoadCSE) {
++ LoadInst *NLI = cast<LoadInst>(AvailableVal);
+ unsigned KnownIDs[] = {
+ LLVMContext::MD_tbaa, LLVMContext::MD_alias_scope,
+ LLVMContext::MD_noalias, LLVMContext::MD_range,
+diff --git a/test/Transforms/InstCombine/tbaa-store-to-load.ll b/test/Transforms/InstCombine/tbaa-store-to-load.ll
+new file mode 100644
+index 0000000..707be73
+--- /dev/null
++++ b/test/Transforms/InstCombine/tbaa-store-to-load.ll
+@@ -0,0 +1,17 @@
++; RUN: opt -S -instcombine < %s 2>&1 | FileCheck %s
++
++define i64 @f(i64* %p1, i64* %p2) {
++top:
++ ; check that the tbaa is preserved
++ ; CHECK-LABEL: @f(
++ ; CHECK: %v1 = load i64, i64* %p1, align 8, !tbaa !0
++ ; CHECK: store i64 %v1, i64* %p2, align 8
++ ; CHECK: ret i64 %v1
++ %v1 = load i64, i64* %p1, align 8, !tbaa !0
++ store i64 %v1, i64* %p2, align 8
++ %v2 = load i64, i64* %p2, align 8
++ ret i64 %v2
++}
++
++!0 = !{!1, !1, i64 0}
++!1 = !{!"load_tbaa"}
diff --git a/patches/llvm_host/llvm-0009-nm-workaround.patch b/patches/llvm_host/llvm-0009-nm-workaround.patch
new file mode 100644
index 00000000..5755e57f
--- /dev/null
+++ b/patches/llvm_host/llvm-0009-nm-workaround.patch
@@ -0,0 +1,27 @@
+From https://llvm.org/bugs/show_bug.cgi?id=24115
+
+diff --git a/tools/llvm-nm/llvm-nm.cpp b/tools/llvm-nm/llvm-nm.cpp
+index e7ee312..b9e4779 100644
+--- a/tools/llvm-nm/llvm-nm.cpp
++++ b/tools/llvm-nm/llvm-nm.cpp
+@@ -486,7 +486,7 @@ static void sortAndPrintSymbolList(SymbolicFile &Obj, bool printName,
+ std::string ArchiveName,
+ std::string ArchitectureName) {
+ if (!NoSort) {
+- std::function<bool(const NMSymbol &, const NMSymbol &)> Cmp;
++ bool (*Cmp)(const NMSymbol &, const NMSymbol &);
+ if (NumericSort)
+ Cmp = compareSymbolAddress;
+ else if (SizeSort)
+@@ -495,8 +495,9 @@ static void sortAndPrintSymbolList(SymbolicFile &Obj, bool printName,
+ Cmp = compareSymbolName;
+
+ if (ReverseSort)
+- Cmp = [=](const NMSymbol &A, const NMSymbol &B) { return Cmp(B, A); };
+- std::sort(SymbolList.begin(), SymbolList.end(), Cmp);
++ std::sort(SymbolList.begin(), SymbolList.end(), [=](const NMSymbol &A, const NMSymbol &B) { return Cmp(B, A); });
++ else
++ std::sort(SymbolList.begin(), SymbolList.end(), Cmp);
+ }
+
+ if (!PrintFileName) {