Merging r355227 and r355228:

------------------------------------------------------------------------
r355227 | ctopper | 2019-03-01 22:02:34 +0100 (Fri, 01 Mar 2019) | 3 lines

[X86] Add test case for D58805. NFC

This demonstrates dead store elimination removing a store that may alias a gather that uses null as its base.
------------------------------------------------------------------------

------------------------------------------------------------------------
r355228 | ctopper | 2019-03-01 22:02:40 +0100 (Fri, 01 Mar 2019) | 7 lines

[X86] Remove IntrArgMemOnly from target specific gather/scatter intrinsics

IntrArgMemOnly implies that only memory pointed to by pointer typed arguments will be accessed. But these intrinsics allow you to pass null to the pointer argument and put the full address into the index argument. Other passes won't be able to understand this.

A colleague found that ISPC was creating gathers like this and then dead store elimination removed some stores because it didn't understand what the gather was doing since the pointer argument was null.

Differential Revision: https://reviews.llvm.org/D58805
------------------------------------------------------------------------

llvm-svn: 355383

2 files changed, 155 insertions, 120 deletions

diff --git a/llvm/include/llvm/IR/IntrinsicsX86.td b/llvm/include/llvm/IR/IntrinsicsX86.td
index 8d8cc8e97678..4f4bcbc67bfb 100644
--- a/llvm/include/llvm/IR/IntrinsicsX86.td
+++ b/llvm/include/llvm/IR/IntrinsicsX86.td
@@ -1677,71 +1677,73 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
 
 // Gather ops
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
+  // NOTE: These can't be ArgMemOnly because you can put the address completely
+  // in the index register.
   def int_x86_avx2_gather_d_pd : GCCBuiltin<"__builtin_ia32_gatherd_pd">,
       Intrinsic<[llvm_v2f64_ty],
         [llvm_v2f64_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_v2f64_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
   def int_x86_avx2_gather_d_pd_256 : GCCBuiltin<"__builtin_ia32_gatherd_pd256">,
       Intrinsic<[llvm_v4f64_ty],
         [llvm_v4f64_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_v4f64_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
   def int_x86_avx2_gather_q_pd : GCCBuiltin<"__builtin_ia32_gatherq_pd">,
       Intrinsic<[llvm_v2f64_ty],
         [llvm_v2f64_ty, llvm_ptr_ty, llvm_v2i64_ty, llvm_v2f64_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
   def int_x86_avx2_gather_q_pd_256 : GCCBuiltin<"__builtin_ia32_gatherq_pd256">,
       Intrinsic<[llvm_v4f64_ty],
         [llvm_v4f64_ty, llvm_ptr_ty, llvm_v4i64_ty, llvm_v4f64_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
   def int_x86_avx2_gather_d_ps : GCCBuiltin<"__builtin_ia32_gatherd_ps">,
       Intrinsic<[llvm_v4f32_ty],
         [llvm_v4f32_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_v4f32_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
   def int_x86_avx2_gather_d_ps_256 : GCCBuiltin<"__builtin_ia32_gatherd_ps256">,
       Intrinsic<[llvm_v8f32_ty],
         [llvm_v8f32_ty, llvm_ptr_ty, llvm_v8i32_ty, llvm_v8f32_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
   def int_x86_avx2_gather_q_ps : GCCBuiltin<"__builtin_ia32_gatherq_ps">,
       Intrinsic<[llvm_v4f32_ty],
         [llvm_v4f32_ty, llvm_ptr_ty, llvm_v2i64_ty, llvm_v4f32_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
   def int_x86_avx2_gather_q_ps_256 : GCCBuiltin<"__builtin_ia32_gatherq_ps256">,
       Intrinsic<[llvm_v4f32_ty],
         [llvm_v4f32_ty, llvm_ptr_ty, llvm_v4i64_ty, llvm_v4f32_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
 
   def int_x86_avx2_gather_d_q : GCCBuiltin<"__builtin_ia32_gatherd_q">,
       Intrinsic<[llvm_v2i64_ty],
         [llvm_v2i64_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_v2i64_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
   def int_x86_avx2_gather_d_q_256 : GCCBuiltin<"__builtin_ia32_gatherd_q256">,
       Intrinsic<[llvm_v4i64_ty],
         [llvm_v4i64_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_v4i64_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
   def int_x86_avx2_gather_q_q : GCCBuiltin<"__builtin_ia32_gatherq_q">,
       Intrinsic<[llvm_v2i64_ty],
         [llvm_v2i64_ty, llvm_ptr_ty, llvm_v2i64_ty, llvm_v2i64_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
   def int_x86_avx2_gather_q_q_256 : GCCBuiltin<"__builtin_ia32_gatherq_q256">,
       Intrinsic<[llvm_v4i64_ty],
         [llvm_v4i64_ty, llvm_ptr_ty, llvm_v4i64_ty, llvm_v4i64_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
   def int_x86_avx2_gather_d_d : GCCBuiltin<"__builtin_ia32_gatherd_d">,
       Intrinsic<[llvm_v4i32_ty],
         [llvm_v4i32_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_v4i32_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
   def int_x86_avx2_gather_d_d_256 : GCCBuiltin<"__builtin_ia32_gatherd_d256">,
       Intrinsic<[llvm_v8i32_ty],
         [llvm_v8i32_ty, llvm_ptr_ty, llvm_v8i32_ty, llvm_v8i32_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
   def int_x86_avx2_gather_q_d : GCCBuiltin<"__builtin_ia32_gatherq_d">,
       Intrinsic<[llvm_v4i32_ty],
         [llvm_v4i32_ty, llvm_ptr_ty, llvm_v2i64_ty, llvm_v4i32_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
   def int_x86_avx2_gather_q_d_256 : GCCBuiltin<"__builtin_ia32_gatherq_d256">,
       Intrinsic<[llvm_v4i32_ty],
         [llvm_v4i32_ty, llvm_ptr_ty, llvm_v4i64_ty, llvm_v4i32_ty, llvm_i8_ty],
-        [IntrReadMem, IntrArgMemOnly]>;
+        [IntrReadMem]>;
 }
 
 // Misc.
@@ -3564,530 +3566,542 @@ let TargetPrefix = "x86" in {
 // Gather and Scatter ops
 let TargetPrefix = "x86" in {
   // NOTE: These are deprecated in favor of the versions that take a vXi1 mask.
+  // NOTE: These can't be ArgMemOnly because you can put the address completely
+  // in the index register.
   def int_x86_avx512_gather_dpd_512  : GCCBuiltin<"__builtin_ia32_gathersiv8df">,
           Intrinsic<[llvm_v8f64_ty], [llvm_v8f64_ty, llvm_ptr_ty,
                      llvm_v8i32_ty, llvm_i8_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
   def int_x86_avx512_gather_dps_512  : GCCBuiltin<"__builtin_ia32_gathersiv16sf">,
           Intrinsic<[llvm_v16f32_ty], [llvm_v16f32_ty, llvm_ptr_ty,
                      llvm_v16i32_ty, llvm_i16_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
   def int_x86_avx512_gather_qpd_512  : GCCBuiltin<"__builtin_ia32_gatherdiv8df">,
           Intrinsic<[llvm_v8f64_ty], [llvm_v8f64_ty, llvm_ptr_ty,
                      llvm_v8i64_ty, llvm_i8_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
   def int_x86_avx512_gather_qps_512  : GCCBuiltin<"__builtin_ia32_gatherdiv16sf">,
           Intrinsic<[llvm_v8f32_ty], [llvm_v8f32_ty, llvm_ptr_ty,
                      llvm_v8i64_ty, llvm_i8_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
 
 
   def int_x86_avx512_gather_dpq_512  : GCCBuiltin<"__builtin_ia32_gathersiv8di">,
           Intrinsic<[llvm_v8i64_ty], [llvm_v8i64_ty, llvm_ptr_ty,
                      llvm_v8i32_ty, llvm_i8_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
   def int_x86_avx512_gather_dpi_512  : GCCBuiltin<"__builtin_ia32_gathersiv16si">,
           Intrinsic<[llvm_v16i32_ty], [llvm_v16i32_ty, llvm_ptr_ty,
                      llvm_v16i32_ty, llvm_i16_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
   def int_x86_avx512_gather_qpq_512  : GCCBuiltin<"__builtin_ia32_gatherdiv8di">,
           Intrinsic<[llvm_v8i64_ty], [llvm_v8i64_ty, llvm_ptr_ty,
                      llvm_v8i64_ty, llvm_i8_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
   def int_x86_avx512_gather_qpi_512  : GCCBuiltin<"__builtin_ia32_gatherdiv16si">,
           Intrinsic<[llvm_v8i32_ty], [llvm_v8i32_ty, llvm_ptr_ty,
                      llvm_v8i64_ty, llvm_i8_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
 
   def int_x86_avx512_gather3div2_df :
         GCCBuiltin<"__builtin_ia32_gather3div2df">,
           Intrinsic<[llvm_v2f64_ty],
           [llvm_v2f64_ty, llvm_ptr_ty, llvm_v2i64_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_gather3div2_di :
         GCCBuiltin<"__builtin_ia32_gather3div2di">,
           Intrinsic<[llvm_v2i64_ty],
           [llvm_v2i64_ty, llvm_ptr_ty, llvm_v2i64_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_gather3div4_df :
         GCCBuiltin<"__builtin_ia32_gather3div4df">,
           Intrinsic<[llvm_v4f64_ty],
           [llvm_v4f64_ty, llvm_ptr_ty, llvm_v4i64_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_gather3div4_di :
         GCCBuiltin<"__builtin_ia32_gather3div4di">,
           Intrinsic<[llvm_v4i64_ty],
           [llvm_v4i64_ty, llvm_ptr_ty, llvm_v4i64_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_gather3div4_sf :
         GCCBuiltin<"__builtin_ia32_gather3div4sf">,
           Intrinsic<[llvm_v4f32_ty],
           [llvm_v4f32_ty, llvm_ptr_ty, llvm_v2i64_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_gather3div4_si :
         GCCBuiltin<"__builtin_ia32_gather3div4si">,
           Intrinsic<[llvm_v4i32_ty],
           [llvm_v4i32_ty, llvm_ptr_ty, llvm_v2i64_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_gather3div8_sf :
         GCCBuiltin<"__builtin_ia32_gather3div8sf">,
           Intrinsic<[llvm_v4f32_ty],
           [llvm_v4f32_ty, llvm_ptr_ty, llvm_v4i64_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_gather3div8_si :
         GCCBuiltin<"__builtin_ia32_gather3div8si">,
           Intrinsic<[llvm_v4i32_ty],
           [llvm_v4i32_ty, llvm_ptr_ty, llvm_v4i64_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_gather3siv2_df :
         GCCBuiltin<"__builtin_ia32_gather3siv2df">,
           Intrinsic<[llvm_v2f64_ty],
           [llvm_v2f64_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_gather3siv2_di :
         GCCBuiltin<"__builtin_ia32_gather3siv2di">,
           Intrinsic<[llvm_v2i64_ty],
           [llvm_v2i64_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_gather3siv4_df :
         GCCBuiltin<"__builtin_ia32_gather3siv4df">,
           Intrinsic<[llvm_v4f64_ty],
           [llvm_v4f64_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_gather3siv4_di :
         GCCBuiltin<"__builtin_ia32_gather3siv4di">,
           Intrinsic<[llvm_v4i64_ty],
           [llvm_v4i64_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_gather3siv4_sf :
         GCCBuiltin<"__builtin_ia32_gather3siv4sf">,
           Intrinsic<[llvm_v4f32_ty],
           [llvm_v4f32_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_gather3siv4_si :
         GCCBuiltin<"__builtin_ia32_gather3siv4si">,
           Intrinsic<[llvm_v4i32_ty],
           [llvm_v4i32_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_gather3siv8_sf :
         GCCBuiltin<"__builtin_ia32_gather3siv8sf">,
           Intrinsic<[llvm_v8f32_ty],
           [llvm_v8f32_ty, llvm_ptr_ty, llvm_v8i32_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_gather3siv8_si :
         GCCBuiltin<"__builtin_ia32_gather3siv8si">,
           Intrinsic<[llvm_v8i32_ty],
           [llvm_v8i32_ty, llvm_ptr_ty, llvm_v8i32_ty, llvm_i8_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
 // scatter
   // NOTE: These are deprecated in favor of the versions that take a vXi1 mask.
+  // NOTE: These can't be ArgMemOnly because you can put the address completely
+  // in the index register.
   def int_x86_avx512_scatter_dpd_512  : GCCBuiltin<"__builtin_ia32_scattersiv8df">,
           Intrinsic<[], [llvm_ptr_ty, llvm_i8_ty,
                         llvm_v8i32_ty, llvm_v8f64_ty, llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
   def int_x86_avx512_scatter_dps_512  : GCCBuiltin<"__builtin_ia32_scattersiv16sf">,
           Intrinsic<[], [llvm_ptr_ty, llvm_i16_ty,
                        llvm_v16i32_ty, llvm_v16f32_ty, llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
   def int_x86_avx512_scatter_qpd_512  : GCCBuiltin<"__builtin_ia32_scatterdiv8df">,
           Intrinsic<[], [llvm_ptr_ty, llvm_i8_ty,
                      llvm_v8i64_ty, llvm_v8f64_ty, llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
   def int_x86_avx512_scatter_qps_512  : GCCBuiltin<"__builtin_ia32_scatterdiv16sf">,
           Intrinsic<[], [llvm_ptr_ty, llvm_i8_ty,
                      llvm_v8i64_ty, llvm_v8f32_ty, llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
 
 
   def int_x86_avx512_scatter_dpq_512  : GCCBuiltin<"__builtin_ia32_scattersiv8di">,
           Intrinsic<[], [llvm_ptr_ty, llvm_i8_ty,
                          llvm_v8i32_ty, llvm_v8i64_ty, llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
   def int_x86_avx512_scatter_dpi_512  : GCCBuiltin<"__builtin_ia32_scattersiv16si">,
           Intrinsic<[], [llvm_ptr_ty, llvm_i16_ty,
                      llvm_v16i32_ty, llvm_v16i32_ty, llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
   def int_x86_avx512_scatter_qpq_512  : GCCBuiltin<"__builtin_ia32_scatterdiv8di">,
           Intrinsic<[], [llvm_ptr_ty, llvm_i8_ty,llvm_v8i64_ty, llvm_v8i64_ty,
                          llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
   def int_x86_avx512_scatter_qpi_512  : GCCBuiltin<"__builtin_ia32_scatterdiv16si">,
           Intrinsic<[], [llvm_ptr_ty, llvm_i8_ty, llvm_v8i64_ty, llvm_v8i32_ty,
                          llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
 
   def int_x86_avx512_scatterdiv2_df :
        GCCBuiltin<"__builtin_ia32_scatterdiv2df">,
         Intrinsic<[],
         [llvm_ptr_ty, llvm_i8_ty, llvm_v2i64_ty, llvm_v2f64_ty, llvm_i32_ty],
-        [IntrArgMemOnly]>;
+        []>;
 
   def int_x86_avx512_scatterdiv2_di :
         GCCBuiltin<"__builtin_ia32_scatterdiv2di">,
           Intrinsic<[],
           [llvm_ptr_ty, llvm_i8_ty, llvm_v2i64_ty, llvm_v2i64_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_scatterdiv4_df :
         GCCBuiltin<"__builtin_ia32_scatterdiv4df">,
           Intrinsic<[],
           [llvm_ptr_ty, llvm_i8_ty, llvm_v4i64_ty, llvm_v4f64_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_scatterdiv4_di :
         GCCBuiltin<"__builtin_ia32_scatterdiv4di">,
           Intrinsic<[],
           [llvm_ptr_ty, llvm_i8_ty, llvm_v4i64_ty, llvm_v4i64_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_scatterdiv4_sf :
         GCCBuiltin<"__builtin_ia32_scatterdiv4sf">,
           Intrinsic<[],
           [llvm_ptr_ty, llvm_i8_ty, llvm_v2i64_ty, llvm_v4f32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_scatterdiv4_si :
         GCCBuiltin<"__builtin_ia32_scatterdiv4si">,
           Intrinsic<[],
           [llvm_ptr_ty, llvm_i8_ty, llvm_v2i64_ty, llvm_v4i32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_scatterdiv8_sf :
         GCCBuiltin<"__builtin_ia32_scatterdiv8sf">,
           Intrinsic<[],
           [llvm_ptr_ty, llvm_i8_ty, llvm_v4i64_ty, llvm_v4f32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_scatterdiv8_si :
         GCCBuiltin<"__builtin_ia32_scatterdiv8si">,
           Intrinsic<[],
           [llvm_ptr_ty, llvm_i8_ty, llvm_v4i64_ty, llvm_v4i32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_scattersiv2_df :
         GCCBuiltin<"__builtin_ia32_scattersiv2df">,
           Intrinsic<[],
           [llvm_ptr_ty, llvm_i8_ty, llvm_v4i32_ty, llvm_v2f64_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_scattersiv2_di :
         GCCBuiltin<"__builtin_ia32_scattersiv2di">,
           Intrinsic<[],
           [llvm_ptr_ty, llvm_i8_ty, llvm_v4i32_ty, llvm_v2i64_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_scattersiv4_df :
         GCCBuiltin<"__builtin_ia32_scattersiv4df">,
           Intrinsic<[],
           [llvm_ptr_ty, llvm_i8_ty, llvm_v4i32_ty, llvm_v4f64_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_scattersiv4_di :
         GCCBuiltin<"__builtin_ia32_scattersiv4di">,
           Intrinsic<[],
           [llvm_ptr_ty, llvm_i8_ty, llvm_v4i32_ty, llvm_v4i64_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_scattersiv4_sf :
         GCCBuiltin<"__builtin_ia32_scattersiv4sf">,
           Intrinsic<[],
           [llvm_ptr_ty, llvm_i8_ty, llvm_v4i32_ty, llvm_v4f32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_scattersiv4_si :
         GCCBuiltin<"__builtin_ia32_scattersiv4si">,
           Intrinsic<[],
           [llvm_ptr_ty, llvm_i8_ty, llvm_v4i32_ty, llvm_v4i32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_scattersiv8_sf :
         GCCBuiltin<"__builtin_ia32_scattersiv8sf">,
           Intrinsic<[],
           [llvm_ptr_ty, llvm_i8_ty, llvm_v8i32_ty, llvm_v8f32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_scattersiv8_si :
         GCCBuiltin<"__builtin_ia32_scattersiv8si">,
           Intrinsic<[],
           [llvm_ptr_ty, llvm_i8_ty, llvm_v8i32_ty, llvm_v8i32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   // gather prefetch
+  // NOTE: These can't be ArgMemOnly because you can put the address completely
+  // in the index register.
   def int_x86_avx512_gatherpf_dpd_512  : GCCBuiltin<"__builtin_ia32_gatherpfdpd">,
           Intrinsic<[], [llvm_i8_ty, llvm_v8i32_ty, llvm_ptr_ty,
-                     llvm_i32_ty, llvm_i32_ty], [IntrArgMemOnly]>;
+                     llvm_i32_ty, llvm_i32_ty], []>;
   def int_x86_avx512_gatherpf_dps_512  : GCCBuiltin<"__builtin_ia32_gatherpfdps">,
           Intrinsic<[], [llvm_i16_ty, llvm_v16i32_ty, llvm_ptr_ty,
-                     llvm_i32_ty, llvm_i32_ty], [IntrArgMemOnly]>;
+                     llvm_i32_ty, llvm_i32_ty], []>;
   def int_x86_avx512_gatherpf_qpd_512  : GCCBuiltin<"__builtin_ia32_gatherpfqpd">,
           Intrinsic<[], [llvm_i8_ty, llvm_v8i64_ty, llvm_ptr_ty,
-                     llvm_i32_ty, llvm_i32_ty], [IntrArgMemOnly]>;
+                     llvm_i32_ty, llvm_i32_ty], []>;
   def int_x86_avx512_gatherpf_qps_512  : GCCBuiltin<"__builtin_ia32_gatherpfqps">,
           Intrinsic<[], [llvm_i8_ty, llvm_v8i64_ty, llvm_ptr_ty,
-                     llvm_i32_ty, llvm_i32_ty], [IntrArgMemOnly]>;
+                     llvm_i32_ty, llvm_i32_ty], []>;
 
   // scatter prefetch
+  // NOTE: These can't be ArgMemOnly because you can put the address completely
+  // in the index register.
   def int_x86_avx512_scatterpf_dpd_512  : GCCBuiltin<"__builtin_ia32_scatterpfdpd">,
           Intrinsic<[], [llvm_i8_ty, llvm_v8i32_ty, llvm_ptr_ty,
-                     llvm_i32_ty, llvm_i32_ty], [IntrArgMemOnly]>;
+                     llvm_i32_ty, llvm_i32_ty], []>;
   def int_x86_avx512_scatterpf_dps_512  : GCCBuiltin<"__builtin_ia32_scatterpfdps">,
           Intrinsic<[], [llvm_i16_ty, llvm_v16i32_ty, llvm_ptr_ty,
-                     llvm_i32_ty, llvm_i32_ty], [IntrArgMemOnly]>;
+                     llvm_i32_ty, llvm_i32_ty], []>;
   def int_x86_avx512_scatterpf_qpd_512  : GCCBuiltin<"__builtin_ia32_scatterpfqpd">,
           Intrinsic<[], [llvm_i8_ty, llvm_v8i64_ty, llvm_ptr_ty,
-                     llvm_i32_ty, llvm_i32_ty], [IntrArgMemOnly]>;
+                     llvm_i32_ty, llvm_i32_ty], []>;
   def int_x86_avx512_scatterpf_qps_512  : GCCBuiltin<"__builtin_ia32_scatterpfqps">,
           Intrinsic<[], [llvm_i8_ty, llvm_v8i64_ty, llvm_ptr_ty,
-                     llvm_i32_ty, llvm_i32_ty], [IntrArgMemOnly]>;
+                     llvm_i32_ty, llvm_i32_ty], []>;
 }
 
 // AVX512 gather/scatter intrinsics that use vXi1 masks.
 let TargetPrefix = "x86" in {
+  // NOTE: These can't be ArgMemOnly because you can put the address completely
+  // in the index register.
   def int_x86_avx512_mask_gather_dpd_512  :
           Intrinsic<[llvm_v8f64_ty], [llvm_v8f64_ty, llvm_ptr_ty,
                      llvm_v8i32_ty, llvm_v8i1_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
   def int_x86_avx512_mask_gather_dps_512  :
           Intrinsic<[llvm_v16f32_ty], [llvm_v16f32_ty, llvm_ptr_ty,
                      llvm_v16i32_ty, llvm_v16i1_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
   def int_x86_avx512_mask_gather_qpd_512  :
           Intrinsic<[llvm_v8f64_ty], [llvm_v8f64_ty, llvm_ptr_ty,
                      llvm_v8i64_ty, llvm_v8i1_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
   def int_x86_avx512_mask_gather_qps_512  :
           Intrinsic<[llvm_v8f32_ty], [llvm_v8f32_ty, llvm_ptr_ty,
                      llvm_v8i64_ty, llvm_v8i1_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
 
 
   def int_x86_avx512_mask_gather_dpq_512  :
           Intrinsic<[llvm_v8i64_ty], [llvm_v8i64_ty, llvm_ptr_ty,
                      llvm_v8i32_ty, llvm_v8i1_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
   def int_x86_avx512_mask_gather_dpi_512  :
           Intrinsic<[llvm_v16i32_ty], [llvm_v16i32_ty, llvm_ptr_ty,
                      llvm_v16i32_ty, llvm_v16i1_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
   def int_x86_avx512_mask_gather_qpq_512  :
           Intrinsic<[llvm_v8i64_ty], [llvm_v8i64_ty, llvm_ptr_ty,
                      llvm_v8i64_ty, llvm_v8i1_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
   def int_x86_avx512_mask_gather_qpi_512  :
           Intrinsic<[llvm_v8i32_ty], [llvm_v8i32_ty, llvm_ptr_ty,
                      llvm_v8i64_ty, llvm_v8i1_ty, llvm_i32_ty],
-                    [IntrReadMem, IntrArgMemOnly]>;
+                    [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3div2_df :
           Intrinsic<[llvm_v2f64_ty],
           [llvm_v2f64_ty, llvm_ptr_ty, llvm_v2i64_ty, llvm_v2i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3div2_di :
           Intrinsic<[llvm_v2i64_ty],
           [llvm_v2i64_ty, llvm_ptr_ty, llvm_v2i64_ty, llvm_v2i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3div4_df :
           Intrinsic<[llvm_v4f64_ty],
           [llvm_v4f64_ty, llvm_ptr_ty, llvm_v4i64_ty, llvm_v4i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3div4_di :
           Intrinsic<[llvm_v4i64_ty],
           [llvm_v4i64_ty, llvm_ptr_ty, llvm_v4i64_ty, llvm_v4i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3div4_sf :
           Intrinsic<[llvm_v4f32_ty],
           [llvm_v4f32_ty, llvm_ptr_ty, llvm_v2i64_ty, llvm_v2i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3div4_si :
           Intrinsic<[llvm_v4i32_ty],
           [llvm_v4i32_ty, llvm_ptr_ty, llvm_v2i64_ty, llvm_v2i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3div8_sf :
           Intrinsic<[llvm_v4f32_ty],
           [llvm_v4f32_ty, llvm_ptr_ty, llvm_v4i64_ty, llvm_v4i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3div8_si :
           Intrinsic<[llvm_v4i32_ty],
           [llvm_v4i32_ty, llvm_ptr_ty, llvm_v4i64_ty, llvm_v4i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3siv2_df :
           Intrinsic<[llvm_v2f64_ty],
           [llvm_v2f64_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_v2i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3siv2_di :
           Intrinsic<[llvm_v2i64_ty],
           [llvm_v2i64_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_v2i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3siv4_df :
           Intrinsic<[llvm_v4f64_ty],
           [llvm_v4f64_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_v4i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3siv4_di :
           Intrinsic<[llvm_v4i64_ty],
           [llvm_v4i64_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_v4i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3siv4_sf :
           Intrinsic<[llvm_v4f32_ty],
           [llvm_v4f32_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_v4i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3siv4_si :
           Intrinsic<[llvm_v4i32_ty],
           [llvm_v4i32_ty, llvm_ptr_ty, llvm_v4i32_ty, llvm_v4i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3siv8_sf :
           Intrinsic<[llvm_v8f32_ty],
           [llvm_v8f32_ty, llvm_ptr_ty, llvm_v8i32_ty, llvm_v8i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_gather3siv8_si :
           Intrinsic<[llvm_v8i32_ty],
           [llvm_v8i32_ty, llvm_ptr_ty, llvm_v8i32_ty, llvm_v8i1_ty, llvm_i32_ty],
-          [IntrReadMem, IntrArgMemOnly]>;
+          [IntrReadMem]>;
 
   def int_x86_avx512_mask_scatter_dpd_512  :
           Intrinsic<[], [llvm_ptr_ty, llvm_v8i1_ty,
                         llvm_v8i32_ty, llvm_v8f64_ty, llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
   def int_x86_avx512_mask_scatter_dps_512  :
           Intrinsic<[], [llvm_ptr_ty, llvm_v16i1_ty,
                        llvm_v16i32_ty, llvm_v16f32_ty, llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
   def int_x86_avx512_mask_scatter_qpd_512  :
           Intrinsic<[], [llvm_ptr_ty, llvm_v8i1_ty,
                      llvm_v8i64_ty, llvm_v8f64_ty, llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
   def int_x86_avx512_mask_scatter_qps_512  :
           Intrinsic<[], [llvm_ptr_ty, llvm_v8i1_ty,
                      llvm_v8i64_ty, llvm_v8f32_ty, llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
 
 
+  // NOTE: These can't be ArgMemOnly because you can put the address completely
+  // in the index register.
   def int_x86_avx512_mask_scatter_dpq_512  :
           Intrinsic<[], [llvm_ptr_ty, llvm_v8i1_ty,
                          llvm_v8i32_ty, llvm_v8i64_ty, llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
   def int_x86_avx512_mask_scatter_dpi_512  :
           Intrinsic<[], [llvm_ptr_ty, llvm_v16i1_ty,
                      llvm_v16i32_ty, llvm_v16i32_ty, llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
   def int_x86_avx512_mask_scatter_qpq_512  :
           Intrinsic<[], [llvm_ptr_ty, llvm_v8i1_ty,llvm_v8i64_ty, llvm_v8i64_ty,
                          llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
   def int_x86_avx512_mask_scatter_qpi_512  :
           Intrinsic<[], [llvm_ptr_ty, llvm_v8i1_ty, llvm_v8i64_ty, llvm_v8i32_ty,
                          llvm_i32_ty],
-                    [IntrArgMemOnly]>;
+                    []>;
 
   def int_x86_avx512_mask_scatterdiv2_df :
         Intrinsic<[],
         [llvm_ptr_ty, llvm_v2i1_ty, llvm_v2i64_ty, llvm_v2f64_ty, llvm_i32_ty],
-        [IntrArgMemOnly]>;
+        []>;
 
   def int_x86_avx512_mask_scatterdiv2_di :
           Intrinsic<[],
           [llvm_ptr_ty, llvm_v2i1_ty, llvm_v2i64_ty, llvm_v2i64_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_mask_scatterdiv4_df :
           Intrinsic<[],
           [llvm_ptr_ty, llvm_v4i1_ty, llvm_v4i64_ty, llvm_v4f64_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_mask_scatterdiv4_di :
           Intrinsic<[],
           [llvm_ptr_ty, llvm_v4i1_ty, llvm_v4i64_ty, llvm_v4i64_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_mask_scatterdiv4_sf :
           Intrinsic<[],
           [llvm_ptr_ty, llvm_v2i1_ty, llvm_v2i64_ty, llvm_v4f32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_mask_scatterdiv4_si :
           Intrinsic<[],
           [llvm_ptr_ty, llvm_v2i1_ty, llvm_v2i64_ty, llvm_v4i32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_mask_scatterdiv8_sf :
           Intrinsic<[],
           [llvm_ptr_ty, llvm_v4i1_ty, llvm_v4i64_ty, llvm_v4f32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_mask_scatterdiv8_si :
           Intrinsic<[],
           [llvm_ptr_ty, llvm_v4i1_ty, llvm_v4i64_ty, llvm_v4i32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_mask_scattersiv2_df :
           Intrinsic<[],
           [llvm_ptr_ty, llvm_v2i1_ty, llvm_v4i32_ty, llvm_v2f64_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_mask_scattersiv2_di :
           Intrinsic<[],
           [llvm_ptr_ty, llvm_v2i1_ty, llvm_v4i32_ty, llvm_v2i64_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_mask_scattersiv4_df :
           Intrinsic<[],
           [llvm_ptr_ty, llvm_v4i1_ty, llvm_v4i32_ty, llvm_v4f64_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_mask_scattersiv4_di :
           Intrinsic<[],
           [llvm_ptr_ty, llvm_v4i1_ty, llvm_v4i32_ty, llvm_v4i64_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_mask_scattersiv4_sf :
           Intrinsic<[],
           [llvm_ptr_ty, llvm_v4i1_ty, llvm_v4i32_ty, llvm_v4f32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_mask_scattersiv4_si :
           Intrinsic<[],
           [llvm_ptr_ty, llvm_v4i1_ty, llvm_v4i32_ty, llvm_v4i32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_mask_scattersiv8_sf :
           Intrinsic<[],
           [llvm_ptr_ty, llvm_v8i1_ty, llvm_v8i32_ty, llvm_v8f32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 
   def int_x86_avx512_mask_scattersiv8_si :
           Intrinsic<[],
           [llvm_ptr_ty, llvm_v8i1_ty, llvm_v8i32_ty, llvm_v8i32_ty, llvm_i32_ty],
-          [IntrArgMemOnly]>;
+          []>;
 }
 
 // AVX-512 conflict detection instruction
diff --git a/llvm/test/Transforms/DeadStoreElimination/X86/gather-null-pointer.ll b/llvm/test/Transforms/DeadStoreElimination/X86/gather-null-pointer.ll
new file mode 100644
index 000000000000..6a5f4bb9eb25
--- /dev/null
+++ b/llvm/test/Transforms/DeadStoreElimination/X86/gather-null-pointer.ll
@@ -0,0 +1,21 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -dse -S | FileCheck %s
+
+; Both stores should be emitted because we can't tell if the gather aliases.
+
+define <4 x i32> @bar(<4 x i32> %arg, i32* %arg1) {
+; CHECK-LABEL: @bar(
+; CHECK-NEXT:  bb:
+; CHECK-NEXT:    store i32 5, i32* [[ARG1:%.*]]
+; CHECK-NEXT:    [[TMP:%.*]] = tail call <4 x i32> @llvm.x86.avx2.gather.d.d(<4 x i32> zeroinitializer, i8* null, <4 x i32> [[ARG:%.*]], <4 x i32> <i32 -1, i32 -1, i32 -1, i32 -1>, i8 1)
+; CHECK-NEXT:    store i32 10, i32* [[ARG1]]
+; CHECK-NEXT:    ret <4 x i32> [[TMP]]
+;
+bb:
+  store i32 5, i32* %arg1
+  %tmp = tail call <4 x i32> @llvm.x86.avx2.gather.d.d(<4 x i32> zeroinitializer, i8* null, <4 x i32> %arg, <4 x i32> <i32 -1, i32 -1, i32 -1, i32 -1>, i8 1)
+  store i32 10, i32* %arg1
+  ret <4 x i32> %tmp
+}
+
+declare <4 x i32> @llvm.x86.avx2.gather.d.d(<4 x i32>, i8*, <4 x i32>, <4 x i32>, i8)