Index: include/llvm/IR/IntrinsicsAMDGPU.td
===================================================================
--- include/llvm/IR/IntrinsicsAMDGPU.td
+++ include/llvm/IR/IntrinsicsAMDGPU.td
@@ -294,6 +294,19 @@
 def int_amdgcn_atomic_inc : AMDGPUAtomicIncIntrin;
 def int_amdgcn_atomic_dec : AMDGPUAtomicIncIntrin;
 
+class AMDGPUAtomicF32Intrin : Intrinsic<[llvm_float_ty],
+  [LLVMAnyPointerType<llvm_float_ty>,
+  llvm_float_ty,
+  llvm_i32_ty, // ordering
+  llvm_i32_ty, // scope
+  llvm_i1_ty], // isVolatile
+  [IntrArgMemOnly, NoCapture<0>]
+>;
+
+def int_amdgcn_atomic_add : AMDGPUAtomicF32Intrin;
+def int_amdgcn_atomic_min : AMDGPUAtomicF32Intrin;
+def int_amdgcn_atomic_max : AMDGPUAtomicF32Intrin;
+
 class AMDGPUImageLoad<bit NoMem = 0> : Intrinsic <
   [llvm_anyfloat_ty], // vdata(VGPR)
   [llvm_anyint_ty,    // vaddr(VGPR)
Index: lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
===================================================================
--- lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
+++ lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
@@ -450,7 +450,10 @@
   }
 
   if (isa<AtomicSDNode>(N) ||
-      (Opc == AMDGPUISD::ATOMIC_INC || Opc == AMDGPUISD::ATOMIC_DEC))
+      (Opc == AMDGPUISD::ATOMIC_INC || Opc == AMDGPUISD::ATOMIC_DEC ||
+       Opc == AMDGPUISD::ATOMIC_LOAD_FADD ||
+       Opc == AMDGPUISD::ATOMIC_LOAD_FMIN ||
+       Opc == AMDGPUISD::ATOMIC_LOAD_FMAX))
     N = glueCopyToM0(N);
 
   switch (Opc) {
Index: lib/Target/AMDGPU/AMDGPUISelLowering.h
===================================================================
--- lib/Target/AMDGPU/AMDGPUISelLowering.h
+++ lib/Target/AMDGPU/AMDGPUISelLowering.h
@@ -445,6 +445,9 @@
   ATOMIC_CMP_SWAP,
   ATOMIC_INC,
   ATOMIC_DEC,
+  ATOMIC_LOAD_FADD,
+  ATOMIC_LOAD_FMIN,
+  ATOMIC_LOAD_FMAX,
   BUFFER_LOAD,
   BUFFER_LOAD_FORMAT,
   BUFFER_STORE,
Index: lib/Target/AMDGPU/AMDGPUISelLowering.cpp
===================================================================
--- lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -3980,6 +3980,9 @@
   NODE_NAME_CASE(ATOMIC_CMP_SWAP)
   NODE_NAME_CASE(ATOMIC_INC)
   NODE_NAME_CASE(ATOMIC_DEC)
+  NODE_NAME_CASE(ATOMIC_LOAD_FADD)
+  NODE_NAME_CASE(ATOMIC_LOAD_FMIN)
+  NODE_NAME_CASE(ATOMIC_LOAD_FMAX)
   NODE_NAME_CASE(BUFFER_LOAD)
   NODE_NAME_CASE(BUFFER_LOAD_FORMAT)
   NODE_NAME_CASE(BUFFER_STORE)
Index: lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
===================================================================
--- lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
+++ lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
@@ -449,6 +449,9 @@
   case Intrinsic::r600_read_tidig_z:
   case Intrinsic::amdgcn_atomic_inc:
   case Intrinsic::amdgcn_atomic_dec:
+  case Intrinsic::amdgcn_atomic_add:
+  case Intrinsic::amdgcn_atomic_min:
+  case Intrinsic::amdgcn_atomic_max:
   case Intrinsic::amdgcn_image_atomic_swap:
   case Intrinsic::amdgcn_image_atomic_add:
   case Intrinsic::amdgcn_image_atomic_sub:
Index: lib/Target/AMDGPU/DSInstructions.td
===================================================================
--- lib/Target/AMDGPU/DSInstructions.td
+++ lib/Target/AMDGPU/DSInstructions.td
@@ -440,7 +440,7 @@
 defm DS_MSKOR_RTN_B32 : DS_1A2D_RET_mc<"ds_mskor_rtn_b32", VGPR_32, "ds_mskor_b32">;
 defm DS_CMPST_RTN_B32 : DS_1A2D_RET_mc<"ds_cmpst_rtn_b32", VGPR_32, "ds_cmpst_b32">;
 defm DS_CMPST_RTN_F32 : DS_1A2D_RET_mc<"ds_cmpst_rtn_f32", VGPR_32, "ds_cmpst_f32">;
-defm DS_MIN_RTN_F32   : DS_1A1D_RET_mc <"ds_min_rtn_f32", VGPR_32, "ds_min_f32">;
+defm DS_MIN_RTN_F32   : DS_1A1D_RET_mc<"ds_min_rtn_f32", VGPR_32, "ds_min_f32">;
 defm DS_MAX_RTN_F32   : DS_1A1D_RET_mc<"ds_max_rtn_f32", VGPR_32, "ds_max_f32">;
 
 defm DS_WRXCHG_RTN_B32 : DS_1A1D_RET_mc<"ds_wrxchg_rtn_b32">;
@@ -769,6 +769,9 @@
 defm : DSAtomicRetPat_mc<DS_MIN_RTN_U32, i32, "atomic_load_umin_local">;
 defm : DSAtomicRetPat_mc<DS_MAX_RTN_U32, i32, "atomic_load_umax_local">;
 defm : DSAtomicCmpXChg_mc<DS_CMPST_RTN_B32, i32, "atomic_cmp_swap_local">;
+defm : DSAtomicRetPat_mc<DS_MIN_RTN_F32, f32, "atomic_load_fmin_local">;
+defm : DSAtomicRetPat_mc<DS_MAX_RTN_F32, f32, "atomic_load_fmax_local">;
+defm : DSAtomicRetPat_mc<DS_ADD_RTN_F32, f32, "atomic_load_fadd_local">;
 
 // 64-bit atomics.
 defm : DSAtomicRetPat_mc<DS_WRXCHG_RTN_B64, i64, "atomic_swap_local">;
Index: lib/Target/AMDGPU/SIISelLowering.cpp
===================================================================
--- lib/Target/AMDGPU/SIISelLowering.cpp
+++ lib/Target/AMDGPU/SIISelLowering.cpp
@@ -561,7 +561,10 @@
                                           unsigned IntrID) const {
   switch (IntrID) {
   case Intrinsic::amdgcn_atomic_inc:
-  case Intrinsic::amdgcn_atomic_dec: {
+  case Intrinsic::amdgcn_atomic_dec:
+  case Intrinsic::amdgcn_atomic_add:
+  case Intrinsic::amdgcn_atomic_min:
+  case Intrinsic::amdgcn_atomic_max: {
     Info.opc = ISD::INTRINSIC_W_CHAIN;
     Info.memVT = MVT::getVT(CI.getType());
     Info.ptrVal = CI.getOperand(0);
@@ -583,7 +586,10 @@
                                             Type *&AccessTy) const {
   switch (II->getIntrinsicID()) {
   case Intrinsic::amdgcn_atomic_inc:
-  case Intrinsic::amdgcn_atomic_dec: {
+  case Intrinsic::amdgcn_atomic_dec:
+  case Intrinsic::amdgcn_atomic_add:
+  case Intrinsic::amdgcn_atomic_min:
+  case Intrinsic::amdgcn_atomic_max: {
     Value *Ptr = II->getArgOperand(0);
     AccessTy = II->getType();
     Ops.push_back(Ptr);
@@ -4259,10 +4265,31 @@
 
   switch (IntrID) {
   case Intrinsic::amdgcn_atomic_inc:
-  case Intrinsic::amdgcn_atomic_dec: {
+  case Intrinsic::amdgcn_atomic_dec:
+  case Intrinsic::amdgcn_atomic_add:
+  case Intrinsic::amdgcn_atomic_min:
+  case Intrinsic::amdgcn_atomic_max: {
     MemSDNode *M = cast<MemSDNode>(Op);
-    unsigned Opc = (IntrID == Intrinsic::amdgcn_atomic_inc) ?
-      AMDGPUISD::ATOMIC_INC : AMDGPUISD::ATOMIC_DEC;
+    unsigned Opc;
+    switch (IntrID) {
+    case Intrinsic::amdgcn_atomic_inc:
+      Opc = AMDGPUISD::ATOMIC_INC;
+      break;
+    case Intrinsic::amdgcn_atomic_dec:
+      Opc = AMDGPUISD::ATOMIC_DEC;
+      break;
+    case Intrinsic::amdgcn_atomic_add:
+      Opc = AMDGPUISD::ATOMIC_LOAD_FADD;
+      break;
+    case Intrinsic::amdgcn_atomic_min:
+      Opc = AMDGPUISD::ATOMIC_LOAD_FMIN;
+      break;
+    case Intrinsic::amdgcn_atomic_max:
+      Opc = AMDGPUISD::ATOMIC_LOAD_FMAX;
+      break;
+    default:
+      llvm_unreachable("Unknown intrinsic!");
+    }
     SDValue Ops[] = {
       M->getOperand(0), // Chain
       M->getOperand(2), // Ptr
@@ -6519,7 +6546,10 @@
   case ISD::ATOMIC_LOAD_UMIN:
   case ISD::ATOMIC_LOAD_UMAX:
   case AMDGPUISD::ATOMIC_INC:
-  case AMDGPUISD::ATOMIC_DEC: // TODO: Target mem intrinsics.
+  case AMDGPUISD::ATOMIC_DEC:
+  case AMDGPUISD::ATOMIC_LOAD_FADD:
+  case AMDGPUISD::ATOMIC_LOAD_FMIN:
+  case AMDGPUISD::ATOMIC_LOAD_FMAX:  // TODO: Target mem intrinsics.
     if (DCI.isBeforeLegalize())
       break;
     return performMemSDNodeCombine(cast<MemSDNode>(N), DCI);
Index: lib/Target/AMDGPU/SIInstrInfo.td
===================================================================
--- lib/Target/AMDGPU/SIInstrInfo.td
+++ lib/Target/AMDGPU/SIInstrInfo.td
@@ -45,6 +45,22 @@
   [SDNPMayLoad, SDNPMayStore, SDNPMemOperand, SDNPHasChain]
 >;
 
+def SDTAtomic2_f32 : SDTypeProfile<1, 2, [
+  SDTCisSameAs<0,2>, SDTCisFP<0>, SDTCisPtrTy<1>
+]>;
+
+def SIatomic_fadd : SDNode<"AMDGPUISD::ATOMIC_LOAD_FADD", SDTAtomic2_f32,
+  [SDNPMayLoad, SDNPMayStore, SDNPMemOperand, SDNPHasChain]
+>;
+
+def SIatomic_fmin : SDNode<"AMDGPUISD::ATOMIC_LOAD_FMIN", SDTAtomic2_f32,
+  [SDNPMayLoad, SDNPMayStore, SDNPMemOperand, SDNPHasChain]
+>;
+
+def SIatomic_fmax : SDNode<"AMDGPUISD::ATOMIC_LOAD_FMAX", SDTAtomic2_f32,
+  [SDNPMayLoad, SDNPMayStore, SDNPMemOperand, SDNPHasChain]
+>;
+
 def SItbuffer_load : SDNode<"AMDGPUISD::TBUFFER_LOAD_FORMAT",
   SDTypeProfile<1, 9,
     [                     // vdata
@@ -163,6 +179,9 @@
 
 def atomic_inc_local : local_binary_atomic_op<SIatomic_inc>;
 def atomic_dec_local : local_binary_atomic_op<SIatomic_dec>;
+def atomic_load_fadd_local : local_binary_atomic_op<SIatomic_fadd>;
+def atomic_load_fmin_local : local_binary_atomic_op<SIatomic_fmin>;
+def atomic_load_fmax_local : local_binary_atomic_op<SIatomic_fmax>;
 
 //===----------------------------------------------------------------------===//
 // SDNodes PatFrags for loads/stores with a glue input.
@@ -297,10 +316,11 @@
   (shl $src0, $src1)
 >;
 
-multiclass SIAtomicM0Glue2 <string op_name, bit is_amdgpu = 0> {
+multiclass SIAtomicM0Glue2 <string op_name, bit is_amdgpu = 0,
+                            SDTypeProfile tc = SDTAtomic2> {
 
   def _glue : SDNode <
-    !if(is_amdgpu, "AMDGPUISD", "ISD")#"::ATOMIC_"#op_name, SDTAtomic2,
+    !if(is_amdgpu, "AMDGPUISD", "ISD")#"::ATOMIC_"#op_name, tc,
     [SDNPHasChain, SDNPMayStore, SDNPMayLoad, SDNPMemOperand, SDNPInGlue]
   >;
 
@@ -319,6 +339,9 @@
 defm atomic_load_umin : SIAtomicM0Glue2 <"LOAD_UMIN">;
 defm atomic_load_umax : SIAtomicM0Glue2 <"LOAD_UMAX">;
 defm atomic_swap : SIAtomicM0Glue2 <"SWAP">;
+defm atomic_load_fadd : SIAtomicM0Glue2 <"LOAD_FADD", 1, SDTAtomic2_f32>;
+defm atomic_load_fmin : SIAtomicM0Glue2 <"LOAD_FMIN", 1, SDTAtomic2_f32>;
+defm atomic_load_fmax : SIAtomicM0Glue2 <"LOAD_FMAX", 1, SDTAtomic2_f32>;
 
 def atomic_cmp_swap_glue : SDNode <"ISD::ATOMIC_CMP_SWAP", SDTAtomic3,
   [SDNPHasChain, SDNPMayStore, SDNPMayLoad, SDNPMemOperand, SDNPInGlue]
Index: lib/Transforms/Scalar/InferAddressSpaces.cpp
===================================================================
--- lib/Transforms/Scalar/InferAddressSpaces.cpp
+++ lib/Transforms/Scalar/InferAddressSpaces.cpp
@@ -260,7 +260,10 @@
 
   switch (II->getIntrinsicID()) {
   case Intrinsic::amdgcn_atomic_inc:
-  case Intrinsic::amdgcn_atomic_dec:{
+  case Intrinsic::amdgcn_atomic_dec:
+  case Intrinsic::amdgcn_atomic_add:
+  case Intrinsic::amdgcn_atomic_min:
+  case Intrinsic::amdgcn_atomic_max: {
     const ConstantInt *IsVolatile = dyn_cast<ConstantInt>(II->getArgOperand(4));
     if (!IsVolatile || !IsVolatile->isZero())
       return false;
@@ -289,6 +292,9 @@
   case Intrinsic::objectsize:
   case Intrinsic::amdgcn_atomic_inc:
   case Intrinsic::amdgcn_atomic_dec:
+  case Intrinsic::amdgcn_atomic_add:
+  case Intrinsic::amdgcn_atomic_min:
+  case Intrinsic::amdgcn_atomic_max:
     appendsFlatAddressExpressionToPostorderStack(II->getArgOperand(0),
                                                  PostorderStack, Visited);
     break;
Index: test/CodeGen/AMDGPU/lds_atomic_f32.ll
===================================================================
--- /dev/null
+++ test/CodeGen/AMDGPU/lds_atomic_f32.ll
@@ -0,0 +1,47 @@
+; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -enable-var-scope %s
+declare float @llvm.amdgcn.atomic.add.f32(float addrspace(3)* nocapture, float, i32, i32, i1)
+declare float @llvm.amdgcn.atomic.min.f32(float addrspace(3)* nocapture, float, i32, i32, i1)
+declare float @llvm.amdgcn.atomic.max.f32(float addrspace(3)* nocapture, float, i32, i32, i1)
+
+; CHECK-LABEL: {{^}}lds_atomic_add_f32:
+; CHECK: v_mov_b32_e32 [[V0:v[0-9]+]], 0x42280000
+; CHECK: ds_add_rtn_f32 [[V2:v[0-9]+]], [[V1:v[0-9]+]], [[V0]]
+; CHECK: ds_add_f32 [[V1]], [[V0]]
+; CHECK: s_waitcnt lgkmcnt(1)
+; CHECK: ds_add_rtn_f32 v2, [[V1]], [[V2]]
+define amdgpu_kernel void @lds_atomic_add_f32(float addrspace(1)* %out, float addrspace(3)* %ptrf) {
+  %a1 = call float @llvm.amdgcn.atomic.add.f32(float addrspace(3)* %ptrf, float 4.2e+1, i32 0, i32 0, i1 false)
+  %a2 = call float @llvm.amdgcn.atomic.add.f32(float addrspace(3)* %ptrf, float 4.2e+1, i32 0, i32 0, i1 false)
+  %a3 = call float @llvm.amdgcn.atomic.add.f32(float addrspace(3)* %ptrf, float %a1, i32 0, i32 0, i1 false)
+  store float %a3, float addrspace(1)* %out
+  ret void
+}
+
+; CHECK-LABEL: {{^}}lds_atomic_min_f32:
+; CHECK: v_mov_b32_e32 [[V0:v[0-9]+]], 0x42280000
+; CHECK: ds_min_rtn_f32 [[V2:v[0-9]+]], [[V1:v[0-9]+]], [[V0]]
+; CHECK: ds_min_f32 [[V1]], [[V0]]
+; CHECK: s_waitcnt lgkmcnt(1)
+; CHECK: ds_min_rtn_f32 v2, [[V1]], [[V2]]
+define amdgpu_kernel void @lds_atomic_min_f32(float addrspace(1)* %out, float addrspace(3)* %ptrf) {
+  %a1 = call float @llvm.amdgcn.atomic.min.f32(float addrspace(3)* %ptrf, float 4.2e+1, i32 0, i32 0, i1 false)
+  %a2 = call float @llvm.amdgcn.atomic.min.f32(float addrspace(3)* %ptrf, float 4.2e+1, i32 0, i32 0, i1 false)
+  %a3 = call float @llvm.amdgcn.atomic.min.f32(float addrspace(3)* %ptrf, float %a1, i32 0, i32 0, i1 false)
+  store float %a3, float addrspace(1)* %out
+  ret void
+}
+
+; CHECK-LABEL: {{^}}lds_atomic_max_f32:
+; CHECK: v_mov_b32_e32 [[V0:v[0-9]+]], 0x42280000
+; CHECK: ds_max_rtn_f32 [[V2:v[0-9]+]], [[V1:v[0-9]+]], [[V0]]
+; CHECK: ds_max_f32 [[V1]], [[V0]]
+; CHECK: s_waitcnt lgkmcnt(1)
+; CHECK: ds_max_rtn_f32 v2, [[V1]], [[V2]]
+define amdgpu_kernel void @lds_atomic_max_f32(float addrspace(1)* %out, float addrspace(3)* %ptrf) {
+  %a1 = call float @llvm.amdgcn.atomic.max.f32(float addrspace(3)* %ptrf, float 4.2e+1, i32 0, i32 0, i1 false)
+  %a2 = call float @llvm.amdgcn.atomic.max.f32(float addrspace(3)* %ptrf, float 4.2e+1, i32 0, i32 0, i1 false)
+  %a3 = call float @llvm.amdgcn.atomic.max.f32(float addrspace(3)* %ptrf, float %a1, i32 0, i32 0, i1 false)
+  store float %a3, float addrspace(1)* %out
+  ret void
+}
+