diff --git a/llvm/lib/Target/AMDGPU/AMDGPU.td b/llvm/lib/Target/AMDGPU/AMDGPU.td
--- a/llvm/lib/Target/AMDGPU/AMDGPU.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPU.td
@@ -1343,7 +1343,8 @@
 
 def D16PreservesUnusedBits :
   Predicate<"Subtarget->d16PreservesUnusedBits()">,
-  AssemblerPredicate<(all_of FeatureGFX9Insts, (not FeatureSRAMECC))>;
+  AssemblerPredicate<(all_of FeatureGFX9Insts, (not FeatureSRAMECC),
+    (not FeatureGFX90AInsts))>;
 
 def LDSRequiresM0Init : Predicate<"Subtarget->ldsRequiresM0Init()">;
 def NotLDSRequiresM0Init : Predicate<"!Subtarget->ldsRequiresM0Init()">;
diff --git a/llvm/lib/Target/AMDGPU/GCNSubtarget.h b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
--- a/llvm/lib/Target/AMDGPU/GCNSubtarget.h
+++ b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
@@ -591,7 +591,9 @@
   }
 
   bool d16PreservesUnusedBits() const {
-    return hasD16LoadStore() && !TargetID.isSramEccOnOrAny();
+    // gfx90a's d16 loads don't preserve unused bits
+    return hasD16LoadStore() && !TargetID.isSramEccOnOrAny() &&
+           !hasGFX90AInsts();
   }
 
   bool hasD16Images() const {
diff --git a/llvm/test/CodeGen/AMDGPU/load-hi16.ll b/llvm/test/CodeGen/AMDGPU/load-hi16.ll
--- a/llvm/test/CodeGen/AMDGPU/load-hi16.ll
+++ b/llvm/test/CodeGen/AMDGPU/load-hi16.ll
@@ -1,5 +1,6 @@
 ; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-sroa=0 -mattr=-promote-alloca -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX900,GFX900-MUBUF %s
 ; RUN: llc -march=amdgcn -mcpu=gfx906 -amdgpu-sroa=0 -mattr=-promote-alloca -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX906,NO-D16-HI %s
+; RUN: llc -march=amdgcn -mcpu=gfx90a -amdgpu-sroa=0 -mattr=-promote-alloca,-sramecc -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,NO-D16-HI %s
 ; RUN: llc -march=amdgcn -mcpu=fiji -amdgpu-sroa=0 -mattr=-promote-alloca -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX803,NO-D16-HI %s
 ; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-sroa=0 -mattr=-promote-alloca -amdgpu-enable-flat-scratch -verify-machineinstrs < %s | FileCheck -check-prefixes=GCN,GFX900,GFX900-FLATSCR %s
 
diff --git a/llvm/test/CodeGen/AMDGPU/store-hi16.ll b/llvm/test/CodeGen/AMDGPU/store-hi16.ll
--- a/llvm/test/CodeGen/AMDGPU/store-hi16.ll
+++ b/llvm/test/CodeGen/AMDGPU/store-hi16.ll
@@ -1,5 +1,6 @@
 ; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-sroa=0 -mattr=-promote-alloca -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefixes=GCN,GFX900,GFX9,GFX900-MUBUF %s
-; RUN: llc -march=amdgcn -mcpu=gfx906 -amdgpu-sroa=0 -mattr=-promote-alloca,+sram-ecc -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefixes=GCN,GFX906,GFX9,NO-D16-HI %s
+; RUN: llc -march=amdgcn -mcpu=gfx906 -amdgpu-sroa=0 -mattr=-promote-alloca,+sramecc -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefixes=GCN,GFX906PLUS,GFX9,NO-D16-HI %s
+; RUN: llc -march=amdgcn -mcpu=gfx90a -amdgpu-sroa=0 -mattr=-promote-alloca,-sramecc -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefixes=GCN,GFX906PLUS,GFX9,NO-D16-HI %s
 ; RUN: llc -march=amdgcn -mcpu=fiji -amdgpu-sroa=0 -mattr=-promote-alloca -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefixes=GCN,GFX803,NO-D16-HI %s
 ; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-sroa=0 -mattr=-promote-alloca -amdgpu-enable-flat-scratch -verify-machineinstrs < %s | FileCheck -allow-deprecated-dag-overlap -check-prefixes=GCN,GFX900,GFX9,GFX900-FLATSCR %s
 
@@ -10,7 +11,7 @@
 
 ; NO-D16-HI-NEXT: v_lshrrev_b32_e32 v2, 16, v2
 ; GFX803-NEXT: flat_store_short v[0:1], v2
-; GFX906-NEXT: global_store_short v[0:1], v2, off
+; GFX906PLUS-NEXT: global_store_short v[0:1], v2, off
 
 ; GCN-NEXT: s_waitcnt
 ; GCN-NEXT: s_setpc_b64
@@ -30,7 +31,7 @@
 
 ; NO-D16-HI-NEXT: v_lshrrev_b32_e32 v2, 16, v2
 ; GFX803-NEXT: flat_store_short v[0:1], v2
-; GFX906-NEXT: global_store_short v[0:1], v2, off
+; GFX906PLUS-NEXT: global_store_short v[0:1], v2, off
 
 ; GCN-NEXT: s_waitcnt
 ; GCN-NEXT: s_setpc_b64
@@ -50,7 +51,7 @@
 
 ; NO-D16-HI-NEXT: v_lshrrev_b32_e32 v2, 16, v2
 ; GFX803-NEXT: flat_store_short v[0:1], v2
-; GFX906-NEXT: global_store_short v[0:1], v2, off
+; GFX906PLUS-NEXT: global_store_short v[0:1], v2, off
 
 ; GCN-NEXT: s_waitcnt
 ; GCN-NEXT: s_setpc_b64
@@ -69,7 +70,7 @@
 
 ; NO-D16-HI-NEXT: v_lshrrev_b32_e32 v2, 16, v2
 ; GFX803-NEXT: flat_store_byte v[0:1], v2
-; GFX906-NEXT: global_store_byte v[0:1], v2, off
+; GFX906PLUS-NEXT: global_store_byte v[0:1], v2, off
 
 ; GCN-NEXT: s_waitcnt
 ; GCN-NEXT: s_setpc_b64
@@ -89,7 +90,7 @@
 
 ; NO-D16-HI-NEXT: v_lshrrev_b32_e32 v2, 16, v2
 ; GFX803-NEXT: flat_store_byte v[0:1], v2
-; GFX906-NEXT: global_store_byte v[0:1], v2, off
+; GFX906PLUS-NEXT: global_store_byte v[0:1], v2, off
 
 ; GCN-NEXT: s_waitcnt
 ; GCN-NEXT: s_setpc_b64
@@ -110,8 +111,8 @@
 ; GFX803-DAG: v_lshrrev_b32_e32 v2, 16, v2
 ; GFX803: flat_store_short v[0:1], v2{{$}}
 
-; GFX906-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX906-NEXT: global_store_short v[0:1], v2, off
+; GFX906PLUS-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX906PLUS-NEXT: global_store_short v[0:1], v2, off
 
 ; GCN-NEXT: s_waitcnt
 ; GCN-NEXT: s_setpc_b64
@@ -134,8 +135,8 @@
 ; GFX803-DAG: v_lshrrev_b32_e32 v2, 16, v2
 ; GFX803: flat_store_short v[0:1], v{{[0-9]$}}
 
-; GFX906-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX906-NEXT: global_store_short v[0:1], v2, off
+; GFX906PLUS-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX906PLUS-NEXT: global_store_short v[0:1], v2, off
 
 ; GCN-NEXT: s_waitcnt
 ; GCN-NEXT: s_setpc_b64
@@ -157,8 +158,8 @@
 ; GFX803-DAG: v_lshrrev_b32_e32 v2, 16, v2
 ; GFX803: flat_store_byte v[0:1], v{{[0-9]$}}
 
-; GFX906-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX906-NEXT: global_store_byte v[0:1], v2, off
+; GFX906PLUS-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX906PLUS-NEXT: global_store_byte v[0:1], v2, off
 
 ; GCN-NEXT: s_waitcnt
 ; GCN-NEXT: s_setpc_b64
@@ -181,8 +182,8 @@
 ; GFX803-DAG: v_lshrrev_b32_e32 v2, 16, v2
 ; GFX803: flat_store_byte v[0:1], v{{[0-9]$}}
 
-; GFX906-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX906-NEXT: global_store_byte v[0:1], v2, off
+; GFX906PLUS-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX906PLUS-NEXT: global_store_byte v[0:1], v2, off
 
 ; GCN-NEXT: s_waitcnt
 ; GCN-NEXT: s_setpc_b64
@@ -291,8 +292,8 @@
 ; GCN: s_waitcnt
 ; GFX900-NEXT: flat_store_short_d16_hi v[0:1], v2 offset:4094{{$}}
 
-; GFX906-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX906-NEXT: flat_store_short v[0:1], v2 offset:4094
+; GFX906PLUS-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX906PLUS-NEXT: flat_store_short v[0:1], v2 offset:4094
 
 ; GFX803-DAG: v_add_u32_e32
 ; GFX803-DAG: v_addc_u32_e32
@@ -318,8 +319,8 @@
 ; GFX9-DAG: v_add_co_u32_e32 v{{[0-9]+}}, vcc, 0xfffff802, v
 ; GFX9-DAG: v_addc_co_u32_e32 v{{[0-9]+}}, vcc, -1, v
 
-; GFX906-DAG: v_lshrrev_b32_e32
-; GFX906: flat_store_short v[0:1], v2{{$}}
+; GFX906PLUS-DAG: v_lshrrev_b32_e32
+; GFX906PLUS: flat_store_short v[0:1], v2{{$}}
 
 ; GFX900-NEXT: flat_store_short_d16_hi v[0:1], v2{{$}}
 ; GFX803: flat_store_short v[0:1], v2{{$}}
@@ -343,8 +344,8 @@
 ; GFX803-DAG: v_addc_u32_e32
 ; GFX803: flat_store_byte v[0:1], v2{{$}}
 
-; GFX906-NEXT: v_lshrrev_b32_e32 v2, 16, v2
-; GFX906-NEXT: flat_store_byte v[0:1], v2 offset:4095{{$}}
+; GFX906PLUS-NEXT: v_lshrrev_b32_e32 v2, 16, v2
+; GFX906PLUS-NEXT: flat_store_byte v[0:1], v2 offset:4095{{$}}
 
 ; GCN-NEXT: s_waitcnt
 ; GCN-NEXT: s_setpc_b64
@@ -369,8 +370,8 @@
 
 ; GFX900-NEXT: flat_store_byte_d16_hi v[0:1], v2{{$}}
 
-; GFX906-DAG: v_lshrrev_b32_e32 v2, 16, v2
-; GFX906: flat_store_byte v[0:1], v2{{$}}
+; GFX906PLUS-DAG: v_lshrrev_b32_e32 v2, 16, v2
+; GFX906PLUS: flat_store_byte v[0:1], v2{{$}}
 
 ; GFX803-DAG: v_lshrrev_b32_e32 v2, 16, v2
 ; GFX803: flat_store_byte v[0:1], v2{{$}}