diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInstructions.td b/llvm/lib/Target/AMDGPU/AMDGPUInstructions.td
index 4f28d6fa430a91a8b833796bb598b9d531282c45..e719933b052d7f528fef8241aba8362ab9badfd5 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUInstructions.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPUInstructions.td
@@ -248,6 +248,10 @@ class Aligned8Bytes <dag ops, dag frag> : PatFrag <ops, frag, [{
   return cast<MemSDNode>(N)->getAlignment() % 8 == 0;
 }]>;
 
+class Aligned16Bytes <dag ops, dag frag> : PatFrag <ops, frag, [{
+  return cast<MemSDNode>(N)->getAlignment() >= 16;
+}]>;
+
 class LoadFrag <SDPatternOperator op> : PatFrag<(ops node:$ptr), (op node:$ptr)>;
 
 class StoreFrag<SDPatternOperator op> : PatFrag <
@@ -371,6 +375,10 @@ def load_align8_local : Aligned8Bytes <
   (ops node:$ptr), (load_local node:$ptr)
 >;
 
+def load_align16_local : Aligned16Bytes <
+  (ops node:$ptr), (load_local node:$ptr)
+>;
+
 def store_align8_local : Aligned8Bytes <
   (ops node:$val, node:$ptr), (store_local node:$val, node:$ptr)
 >;
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h
index 830c0770706cfe1d50c3b4442cfdf055db88d70c..0006636432fbc6e97ddb219ab5fe56c0e96e564b 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h
@@ -414,6 +414,12 @@ public:
     return FlatForGlobal;
   }
 
+  /// \returns If target supports ds_read/write_b128 and user enables generation
+  /// of ds_read/write_b128.
+  bool useDS128(bool UserEnable) const {
+    return CIInsts && UserEnable;
+  }
+
   /// \returns If MUBUF instructions always perform range checking, even for
   /// buffer resources used for private memory access.
   bool privateMemoryResourceIsRangeChecked() const {
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
index 4292575c6019ad021b5997e37a3225f24c66a3a5..a22c1803c23e01692117e0fbe9c96063262a568e 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
@@ -265,11 +265,11 @@ unsigned AMDGPUTTIImpl::getLoadStoreVecRegBitWidth(unsigned AddrSpace) const {
     return 512;
   }
 
-  if (AddrSpace == AS.FLAT_ADDRESS)
-    return 128;
-  if (AddrSpace == AS.LOCAL_ADDRESS ||
+  if (AddrSpace == AS.FLAT_ADDRESS ||
+      AddrSpace == AS.LOCAL_ADDRESS ||
       AddrSpace == AS.REGION_ADDRESS)
-    return 64;
+    return 128;
+
   if (AddrSpace == AS.PRIVATE_ADDRESS)
     return 8 * ST->getMaxPrivateElementSize();
 
diff --git a/llvm/lib/Target/AMDGPU/DSInstructions.td b/llvm/lib/Target/AMDGPU/DSInstructions.td
index ec85da72bfa5106e8d9955130a0bf265c560a79b..88484c06218a6b924f50166034332e9e46c274aa 100644
--- a/llvm/lib/Target/AMDGPU/DSInstructions.td
+++ b/llvm/lib/Target/AMDGPU/DSInstructions.td
@@ -649,6 +649,7 @@ defm : DSReadPat_mc <DS_READ_B32, i32, "load_local">;
 let AddedComplexity = 100 in {
 
 defm : DSReadPat_mc <DS_READ_B64, v2i32, "load_align8_local">;
+defm : DSReadPat_mc <DS_READ_B128, v4i32, "load_align16_local">;
 
 } // End AddedComplexity = 100
 
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 68a45cb881706b659bc04b5ef0d22cff25b2e4a7..8463b22d60a7e24d2dbe0eece2108a75327a08c6 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -94,6 +94,11 @@ static cl::opt<bool> EnableVGPRIndexMode(
   cl::desc("Use GPR indexing mode instead of movrel for vector indexing"),
   cl::init(false));
 
+static cl::opt<bool> EnableDS128(
+  "amdgpu-ds128",
+  cl::desc("Use DS_read/write_b128"),
+  cl::init(false));
+
 static cl::opt<unsigned> AssumeFrameIndexHighZeroBits(
   "amdgpu-frame-index-zero-bits",
   cl::desc("High bits of frame index assumed to be zero"),
@@ -5425,14 +5430,13 @@ SDValue SITargetLowering::LowerLOAD(SDValue Op, SelectionDAG &DAG) const {
       llvm_unreachable("unsupported private_element_size");
     }
   } else if (AS == AMDGPUASI.LOCAL_ADDRESS) {
-    if (NumElements > 2)
-      return SplitVectorLoad(Op, DAG);
-
-    if (NumElements == 2)
+    // Use ds_read_b128 if possible.
+    if (Subtarget->useDS128(EnableDS128) && Load->getAlignment() >= 16 &&
+        MemVT.getStoreSize() == 16)
       return SDValue();
 
-    // If properly aligned, if we split we might be able to use ds_read_b64.
-    return SplitVectorLoad(Op, DAG);
+    if (NumElements > 2)
+      return SplitVectorLoad(Op, DAG);
   }
   return SDValue();
 }
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.td b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
index 10f5c3bae3ff1184f8d80ae8639cf8a84e20c0e9..fb46d174bffb42c83dadae0eb5c0cd3a14868ef4 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.td
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
@@ -410,6 +410,9 @@ def sextloadi16_glue : PatFrag<(ops node:$ptr), (sextload_glue node:$ptr), [{
 def load_glue_align8 : Aligned8Bytes <
   (ops node:$ptr), (load_glue node:$ptr)
 >;
+def load_glue_align16 : Aligned16Bytes <
+  (ops node:$ptr), (load_glue node:$ptr)
+>;
 
 
 def load_local_m0 : LoadFrag<load_glue>, LocalAddress;
@@ -418,6 +421,7 @@ def sextloadi16_local_m0 : LoadFrag<sextloadi16_glue>, LocalAddress;
 def az_extloadi8_local_m0 : LoadFrag<az_extloadi8_glue>, LocalAddress;
 def az_extloadi16_local_m0 : LoadFrag<az_extloadi16_glue>, LocalAddress;
 def load_align8_local_m0 : LoadFrag <load_glue_align8>, LocalAddress;
+def load_align16_local_m0 : LoadFrag <load_glue_align16>, LocalAddress;
 
 
 def AMDGPUst_glue : SDNode <"ISD::STORE", SDTStore,
diff --git a/llvm/test/CodeGen/AMDGPU/load-local-f32.ll b/llvm/test/CodeGen/AMDGPU/load-local-f32.ll
index f035d22018a71569af8a273a5414c316be4ba0d5..3d8720888c79a25d79abba4b76339e51efa0b755 100644
--- a/llvm/test/CodeGen/AMDGPU/load-local-f32.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-local-f32.ll
@@ -2,6 +2,11 @@
 ; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,SICIVI,FUNC %s
 ; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefixes=EG,FUNC %s
 
+; Testing for ds_read_128
+; RUN: llc -march=amdgcn -mcpu=tahiti -amdgpu-ds128 < %s | FileCheck -check-prefixes=SI,FUNC %s
+; RUN: llc -march=amdgcn -mcpu=tonga -amdgpu-ds128 < %s | FileCheck -check-prefixes=CIVI,FUNC %s
+; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-ds128 < %s | FileCheck -check-prefixes=CIVI,FUNC %s
+
 ; FUNC-LABEL: {{^}}load_f32_local:
 ; SICIVI: s_mov_b32 m0
 ; GFX9-NOT: m0
@@ -122,4 +127,18 @@ entry:
   ret void
 }
 
+; Tests if ds_read_b128 gets generated for the 16 byte aligned load.
+; FUNC-LABEL: {{^}}local_v4f32_to_128:
+; SI-NOT: ds_read_b128
+; CIVI: ds_read_b128
+; EG: LDS_READ_RET
+; EG: LDS_READ_RET
+; EG: LDS_READ_RET
+; EG: LDS_READ_RET
+define amdgpu_kernel void @local_v4f32_to_128(<4 x float> addrspace(3)* %out, <4 x float> addrspace(3)* %in) {
+  %ld = load <4 x float>, <4 x float> addrspace(3)* %in, align 16
+  store <4 x float> %ld, <4 x float> addrspace(3)* %out
+  ret void
+}
+
 attributes #0 = { nounwind }
diff --git a/llvm/test/CodeGen/AMDGPU/load-local-f64.ll b/llvm/test/CodeGen/AMDGPU/load-local-f64.ll
index ffb67101fd786524244a223c38fcf0cbc1f2e814..14c31e67d7f26e2562ee510af202134fcaa81d47 100644
--- a/llvm/test/CodeGen/AMDGPU/load-local-f64.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-local-f64.ll
@@ -4,6 +4,10 @@
 ; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX9,FUNC %s
 ; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefixes=EG,FUNC %s
 
+; Testing for ds_read_b128
+; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs -amdgpu-ds128 < %s | FileCheck -check-prefixes=CIVI,FUNC %s
+; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs -amdgpu-ds128 < %s | FileCheck -check-prefixes=CIVI,FUNC %s
+
 ; FUNC-LABEL: {{^}}local_load_f64:
 ; SICIV: s_mov_b32 m0
 ; GFX9-NOT: m0
@@ -170,4 +174,18 @@ entry:
   ret void
 }
 
+; Tests if ds_read_b128 gets generated for the 16 byte aligned load.
+; FUNC-LABEL: {{^}}local_load_v2f64_to_128:
+; CIVI: ds_read_b128
+; EG: LDS_READ_RET
+; EG: LDS_READ_RET
+; EG: LDS_READ_RET
+; EG: LDS_READ_RET
+define amdgpu_kernel void @local_load_v2f64_to_128(<2 x double> addrspace(3)* %out, <2 x double> addrspace(3)* %in) {
+entry:
+  %ld = load <2 x double>, <2 x double> addrspace(3)* %in, align 16
+  store <2 x double> %ld, <2 x double> addrspace(3)* %out
+  ret void
+}
+
 attributes #0 = { nounwind }
diff --git a/llvm/test/CodeGen/AMDGPU/load-local-i16.ll b/llvm/test/CodeGen/AMDGPU/load-local-i16.ll
index d3557c14540ce64142b335c869255f4f701e9afe..7438fd2681dc8f530fcf96246d68432e9c1ee20c 100644
--- a/llvm/test/CodeGen/AMDGPU/load-local-i16.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-local-i16.ll
@@ -3,6 +3,10 @@
 ; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX9,GFX89,FUNC %s
 ; RUN: llc -march=r600 -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
 
+; Testing for ds_read_b128
+; RUN: llc -march=amdgcn -mcpu=tonga -amdgpu-ds128 < %s | FileCheck -check-prefixes=CIVI,FUNC %s
+; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-ds128 < %s | FileCheck -check-prefixes=CIVI,FUNC %s
+
 ; FUNC-LABEL: {{^}}local_load_i16:
 ; GFX9-NOT: m0
 ; SICIVI: s_mov_b32 m0
@@ -935,4 +939,18 @@ define amdgpu_kernel void @local_sextload_v32i16_to_v32i64(<32 x i64> addrspace(
 ;   ret void
 ; }
 
+; Tests if ds_read_b128 gets generated for the 16 byte aligned load.
+; FUNC-LABEL: {{^}}local_v8i16_to_128:
+; SI-NOT: ds_read_b128
+; CIVI: ds_read_b128
+; EG: LDS_READ_RET
+; EG: LDS_READ_RET
+; EG: LDS_READ_RET
+; EG: LDS_READ_RET
+define amdgpu_kernel void @local_v8i16_to_128(<8 x i16> addrspace(3)* %out, <8 x i16> addrspace(3)* %in) {
+  %ld = load <8 x i16>, <8 x i16> addrspace(3)* %in, align 16
+  store <8 x i16> %ld, <8 x i16> addrspace(3)* %out
+  ret void
+}
+
 attributes #0 = { nounwind }
diff --git a/llvm/test/CodeGen/AMDGPU/load-local-i32.ll b/llvm/test/CodeGen/AMDGPU/load-local-i32.ll
index c736586fa21724bc2f001949c395b89d8a46a556..1dd7daf95ab05427502fbcf0f4119062e031bbba 100644
--- a/llvm/test/CodeGen/AMDGPU/load-local-i32.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-local-i32.ll
@@ -3,6 +3,11 @@
 ; RUN: llc -march=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,VI,FUNC %s
 ; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
 
+; Testing for ds_read_128
+; RUN: llc -march=amdgcn -mcpu=tahiti -amdgpu-ds128 < %s | FileCheck -check-prefixes=SI,FUNC %s
+; RUN: llc -march=amdgcn -mcpu=tonga -amdgpu-ds128 < %s | FileCheck -check-prefixes=CIVI,FUNC %s
+; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-ds128 < %s | FileCheck -check-prefixes=CIVI,FUNC %s
+
 ; FUNC-LABEL: {{^}}local_load_i32:
 ; GCN-NOT: s_wqm_b64
 ; SICIVI: s_mov_b32 m0, -1
@@ -175,6 +180,20 @@ define amdgpu_kernel void @local_sextload_v4i32_to_v4i64(<4 x i64> addrspace(3)*
   ret void
 }
 
+; Tests if ds_read_b128 gets generated for the 16 byte aligned load.
+; FUNC-LABEL: {{^}}local_v4i32_to_128:
+; SI-NOT: ds_read_b128
+; CIVI: ds_read_b128
+; EG: LDS_READ_RET
+; EG: LDS_READ_RET
+; EG: LDS_READ_RET
+; EG: LDS_READ_RET
+define amdgpu_kernel void @local_v4i32_to_128(<4 x i32> addrspace(3)* %out, <4 x i32> addrspace(3)* %in) {
+  %ld = load <4 x i32>, <4 x i32> addrspace(3)* %in, align 16
+  store <4 x i32> %ld, <4 x i32> addrspace(3)* %out
+  ret void
+}
+
 ; FUNC-LABEL: {{^}}local_zextload_v8i32_to_v8i64:
 ; SICIVI: s_mov_b32 m0, -1
 ; GFX9-NOT: m0
diff --git a/llvm/test/CodeGen/AMDGPU/load-local-i64.ll b/llvm/test/CodeGen/AMDGPU/load-local-i64.ll
index 376f6f513c3a2b7cd278137454ef7fe0b126f429..359fbb42f40a9ba5d07b7f892d2324f850a26131 100644
--- a/llvm/test/CodeGen/AMDGPU/load-local-i64.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-local-i64.ll
@@ -4,6 +4,10 @@
 ; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX9,FUNC %s
 ; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefixes=EG,FUNC %s
 
+; Testing for ds_read_b128
+; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs -amdgpu-ds128 < %s | FileCheck -check-prefixes=CIVI,FUNC %s
+; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs -amdgpu-ds128 < %s | FileCheck -check-prefixes=CIVI,FUNC %s
+
 ; FUNC-LABEL: {{^}}local_load_i64:
 ; SICIVI: s_mov_b32 m0
 ; GFX9-NOT: m0
@@ -36,6 +40,16 @@ entry:
   ret void
 }
 
+; Tests if ds_read_b128 gets generated for the 16 byte aligned load.
+; FUNC-LABEL: {{^}}local_load_v2i64_to_128:
+; CIVI: ds_read_b128
+define amdgpu_kernel void @local_load_v2i64_to_128(<2 x i64> addrspace(3)* %out, <2 x i64> addrspace(3)* %in) {
+entry:
+  %ld = load <2 x i64>, <2 x i64> addrspace(3)* %in
+  store <2 x i64> %ld, <2 x i64> addrspace(3)* %out
+  ret void
+}
+
 ; FUNC-LABEL: {{^}}local_load_v3i64:
 ; SICIVI: s_mov_b32 m0
 ; GFX9-NOT: m0
diff --git a/llvm/test/CodeGen/AMDGPU/load-local-i8.ll b/llvm/test/CodeGen/AMDGPU/load-local-i8.ll
index 72f5408675fccb8ec92f5a46b3aa3050d7583a3a..e1931af042bcb2b5f1a20bbe6a33f325993acb7f 100644
--- a/llvm/test/CodeGen/AMDGPU/load-local-i8.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-local-i8.ll
@@ -3,6 +3,9 @@
 ; RUN: llc -march=amdgcn -mtriple=amdgcn---amdgiz -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX9,FUNC %s
 ; RUN: llc -march=r600 -mtriple=r600---amdgiz -mcpu=redwood -verify-machineinstrs < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
 
+; Testing for ds_read_b128
+; RUN: llc -march=amdgcn -mcpu=tonga -amdgpu-ds128 < %s | FileCheck -check-prefixes=CIVI,FUNC %s
+; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-ds128 < %s | FileCheck -check-prefixes=CIVI,FUNC %s
 
 ; FUNC-LABEL: {{^}}local_load_i8:
 ; GCN-NOT: s_wqm_b64
@@ -1021,4 +1024,18 @@ define amdgpu_kernel void @local_sextload_v32i8_to_v32i16(<32 x i16> addrspace(3
 ;   ret void
 ; }
 
+; Tests if ds_read_b128 gets generated for the 16 byte aligned load.
+; FUNC-LABEL: {{^}}local_v16i8_to_128:
+; SI-NOT: ds_read_b128
+; CIVI: ds_read_b128
+; EG: LDS_READ_RET
+; EG: LDS_READ_RET
+; EG: LDS_READ_RET
+; EG: LDS_READ_RET
+define amdgpu_kernel void @local_v16i8_to_128(<16 x i8> addrspace(3)* %out, <16 x i8> addrspace(3)* %in) {
+  %ld = load <16 x i8>, <16 x i8> addrspace(3)* %in, align 16
+  store <16 x i8> %ld, <16 x i8> addrspace(3)* %out
+  ret void
+}
+
 attributes #0 = { nounwind }
diff --git a/llvm/test/Transforms/LoadStoreVectorizer/AMDGPU/merge-stores.ll b/llvm/test/Transforms/LoadStoreVectorizer/AMDGPU/merge-stores.ll
index 19fc44bb6c85f8bc28d743e288b2e2b8bb6e0254..5eb3b25c1dc4491270fa97ed371e5bb64444e5ed 100644
--- a/llvm/test/Transforms/LoadStoreVectorizer/AMDGPU/merge-stores.ll
+++ b/llvm/test/Transforms/LoadStoreVectorizer/AMDGPU/merge-stores.ll
@@ -504,8 +504,7 @@ define amdgpu_kernel void @merge_local_store_2_constants_i32_align_2(i32 addrspa
 }
 
 ; CHECK-LABEL: @merge_local_store_4_constants_i32
-; CHECK: store <2 x i32> <i32 456, i32 333>, <2 x i32> addrspace(3)*
-; CHECK: store <2 x i32> <i32 1234, i32 123>, <2 x i32> addrspace(3)*
+; CHECK: store <4 x i32> <i32 1234, i32 123, i32 456, i32 333>, <4 x i32> addrspace(3)*
 define amdgpu_kernel void @merge_local_store_4_constants_i32(i32 addrspace(3)* %out) #0 {
   %out.gep.1 = getelementptr i32, i32 addrspace(3)* %out, i32 1
   %out.gep.2 = getelementptr i32, i32 addrspace(3)* %out, i32 2
diff --git a/llvm/test/Transforms/LoadStoreVectorizer/AMDGPU/multiple_tails.ll b/llvm/test/Transforms/LoadStoreVectorizer/AMDGPU/multiple_tails.ll
index 8a78f3d7e9bc106df878de42bbf06c1a6e4ec4c1..b684ca8c12ce5d2c88ff8a9f7e3facd638e27510 100644
--- a/llvm/test/Transforms/LoadStoreVectorizer/AMDGPU/multiple_tails.ll
+++ b/llvm/test/Transforms/LoadStoreVectorizer/AMDGPU/multiple_tails.ll
@@ -29,11 +29,10 @@ define amdgpu_kernel void @no_crash(i32 %arg) {
 ; longest chain vectorized
 
 ; CHECK-LABEL: @interleave_get_longest
-; CHECK: load <2 x i32>
+; CHECK: load <4 x i32>
 ; CHECK: load i32
 ; CHECK: store <2 x i32> zeroinitializer
 ; CHECK: load i32
-; CHECK: load <2 x i32>
 ; CHECK: load i32
 ; CHECK: load i32