diff --git a/deps/llvm.mk b/deps/llvm.mk
index ee025c46f27ac..5e530f4e2cd7e 100644
--- a/deps/llvm.mk
+++ b/deps/llvm.mk
@@ -497,6 +497,8 @@ $(eval $(call LLVM_PATCH,llvm-PR278923)) # Issue #19976, Remove for 4.0
 $(eval $(call LLVM_PATCH,llvm-D28759-loopclearance))
 $(eval $(call LLVM_PATCH,llvm-D28786-callclearance))
 $(eval $(call LLVM_PATCH,llvm-rL293230-icc17-cmake)) # Remove for 4.0
+$(eval $(call LLVM_PATCH,llvm-D32593))
+$(eval $(call LLVM_PATCH,llvm-D33179))
 endif # LLVM_VER
 
 ifeq ($(LLVM_VER),3.7.1)
diff --git a/deps/patches/llvm-D32593.patch b/deps/patches/llvm-D32593.patch
new file mode 100644
index 0000000000000..19c4acb960f97
--- /dev/null
+++ b/deps/patches/llvm-D32593.patch
@@ -0,0 +1,83 @@
+From 5eeab81d22e07b6e12821067fced590f534c251a Mon Sep 17 00:00:00 2001
+From: Keno Fischer <keno@juliacomputing.com>
+Date: Thu, 27 Apr 2017 14:33:33 -0400
+Subject: [PATCH] [SROA] Fix crash due to bad bitcast
+
+Summary:
+As shown in the test case, SROA was crashing when trying to split
+stores (to the alloca) of loads (from anywhere), because it assumed
+the pointer operand to the loads and stores had to have the same
+address space. This isn't the case. Make sure to use the correct
+pointer type for both the load and the store.
+
+Reviewers: chandlerc, majnemer, sanjoy
+
+Subscribers: arsenm, llvm-commits
+
+Differential Revision: https://reviews.llvm.org/D32593
+---
+ lib/Transforms/Scalar/SROA.cpp         |  7 ++++---
+ test/Transforms/SROA/address-spaces.ll | 18 ++++++++++++++++++
+ 2 files changed, 22 insertions(+), 3 deletions(-)
+
+diff --git a/lib/Transforms/Scalar/SROA.cpp b/lib/Transforms/Scalar/SROA.cpp
+index d01e91a..610d5a8 100644
+--- a/lib/Transforms/Scalar/SROA.cpp
++++ b/lib/Transforms/Scalar/SROA.cpp
+@@ -3697,7 +3697,8 @@ bool SROA::presplitLoadsAndStores(AllocaInst &AI, AllocaSlices &AS) {
+     int Idx = 0, Size = Offsets.Splits.size();
+     for (;;) {
+       auto *PartTy = Type::getIntNTy(Ty->getContext(), PartSize * 8);
+-      auto *PartPtrTy = PartTy->getPointerTo(SI->getPointerAddressSpace());
++      auto *LoadPartPtrTy = PartTy->getPointerTo(LI->getPointerAddressSpace());
++      auto *StorePartPtrTy = PartTy->getPointerTo(SI->getPointerAddressSpace());
+ 
+       // Either lookup a split load or create one.
+       LoadInst *PLoad;
+@@ -3708,7 +3709,7 @@ bool SROA::presplitLoadsAndStores(AllocaInst &AI, AllocaSlices &AS) {
+         PLoad = IRB.CreateAlignedLoad(
+             getAdjustedPtr(IRB, DL, LoadBasePtr,
+                            APInt(DL.getPointerSizeInBits(), PartOffset),
+-                           PartPtrTy, LoadBasePtr->getName() + "."),
++                           LoadPartPtrTy, LoadBasePtr->getName() + "."),
+             getAdjustedAlignment(LI, PartOffset, DL), /*IsVolatile*/ false,
+             LI->getName());
+       }
+@@ -3718,7 +3719,7 @@ bool SROA::presplitLoadsAndStores(AllocaInst &AI, AllocaSlices &AS) {
+       StoreInst *PStore = IRB.CreateAlignedStore(
+           PLoad, getAdjustedPtr(IRB, DL, StoreBasePtr,
+                                 APInt(DL.getPointerSizeInBits(), PartOffset),
+-                                PartPtrTy, StoreBasePtr->getName() + "."),
++                                StorePartPtrTy, StoreBasePtr->getName() + "."),
+           getAdjustedAlignment(SI, PartOffset, DL), /*IsVolatile*/ false);
+ 
+       // Now build a new slice for the alloca.
+diff --git a/test/Transforms/SROA/address-spaces.ll b/test/Transforms/SROA/address-spaces.ll
+index 119f225..8fba30c 100644
+--- a/test/Transforms/SROA/address-spaces.ll
++++ b/test/Transforms/SROA/address-spaces.ll
+@@ -83,3 +83,21 @@ define void @pr27557() {
+   store i32 addrspace(3)* @l, i32 addrspace(3)** %3, align 8
+   ret void
+ }
++
++; Make sure pre-splitting doesn't try to introduce an illegal bitcast
++define float @presplit(i64 addrspace(1)* %p) {
++entry:
++; CHECK-LABEL: @presplit(
++; CHECK: %[[CAST:.*]] = bitcast i64 addrspace(1)* {{.*}} to i32 addrspace(1)*
++; CHECK: load i32, i32 addrspace(1)* %[[CAST]]
++   %b = alloca i64
++   %b.cast = bitcast i64* %b to [2 x float]*
++   %b.gep1 = getelementptr [2 x float], [2 x float]* %b.cast, i32 0, i32 0
++   %b.gep2 = getelementptr [2 x float], [2 x float]* %b.cast, i32 0, i32 1
++   %l = load i64, i64 addrspace(1)* %p
++   store i64 %l, i64* %b
++   %f1 = load float, float* %b.gep1
++   %f2 = load float, float* %b.gep2
++   %ret = fadd float %f1, %f2
++   ret float %ret
++}
+-- 
+2.9.3
+
diff --git a/deps/patches/llvm-D33179.patch b/deps/patches/llvm-D33179.patch
new file mode 100644
index 0000000000000..2be915018de24
--- /dev/null
+++ b/deps/patches/llvm-D33179.patch
@@ -0,0 +1,64 @@
+From b1a005ba688397ca360e89cd6c6f51f232d6c25e Mon Sep 17 00:00:00 2001
+From: Keno Fischer <keno@juliacomputing.com>
+Date: Fri, 19 May 2017 18:42:20 -0400
+Subject: [PATCH] [Sink] Fix predicate in legality check
+
+Summary:
+isSafeToSpeculativelyExecute is the wrong predicate to use here.
+All that checks for is whether it is safe to hoist a value due to
+unaligned/un-dereferencable accesses. However, not only are we doing
+sinking rather than hoisting, our concern is that the location
+we're loading from may have been modified. Instead forbid sinking
+any load across a critical edge.
+
+Reviewers: majnemer
+
+Subscribers: llvm-commits
+
+Differential Revision: https://reviews.llvm.org/D33179
+---
+ lib/Transforms/Scalar/Sink.cpp      |  2 +-
+ test/Transforms/Sink/badloadsink.ll | 18 ++++++++++++++++++
+ 2 files changed, 19 insertions(+), 1 deletion(-)
+ create mode 100644 test/Transforms/Sink/badloadsink.ll
+
+diff --git a/lib/Transforms/Scalar/Sink.cpp b/lib/Transforms/Scalar/Sink.cpp
+index 102e9ea..5210f16 100644
+--- a/lib/Transforms/Scalar/Sink.cpp
++++ b/lib/Transforms/Scalar/Sink.cpp
+@@ -114,7 +114,7 @@ static bool IsAcceptableTarget(Instruction *Inst, BasicBlock *SuccToSinkTo,
+   if (SuccToSinkTo->getUniquePredecessor() != Inst->getParent()) {
+     // We cannot sink a load across a critical edge - there may be stores in
+     // other code paths.
+-    if (!isSafeToSpeculativelyExecute(Inst))
++    if (isa<LoadInst>(Inst))
+       return false;
+ 
+     // We don't want to sink across a critical edge if we don't dominate the
+diff --git a/test/Transforms/Sink/badloadsink.ll b/test/Transforms/Sink/badloadsink.ll
+new file mode 100644
+index 0000000..e3f4884
+--- /dev/null
++++ b/test/Transforms/Sink/badloadsink.ll
+@@ -0,0 +1,18 @@
++; RUN: opt < %s -basicaa -sink -S | FileCheck %s
++declare void @foo(i64 *)
++define i64 @sinkload(i1 %cmp) {
++; CHECK-LABEL: @sinkload
++top:
++    %a = alloca i64
++; CHECK: call void @foo(i64* %a)
++; CHECK-NEXT: %x = load i64, i64* %a
++    call void @foo(i64* %a)
++    %x = load i64, i64* %a
++    br i1 %cmp, label %A, label %B
++A:
++    store i64 0, i64 *%a
++    br label %B
++B:
++; CHECK-NOT: load i64, i64 *%a
++    ret i64 %x
++}
+-- 
+2.9.3
+