[GlobalISel] Fold ptrtoint(undef) and inttoptr(undef) to undef. (#119073)

This helps with shuffles a little, and one of the amd gpu tests is now
equivalent to the SDAG version.

Author: davemgreen

llvm/include/llvm/Target/GlobalISel/Combine.td

@@ -428,7 +428,7 @@ def unary_undef_to_zero: GICombineRule<
 // replaced with undef.
 def propagate_undef_any_op: GICombineRule<
   (defs root:$root),
-  (match (wip_match_opcode G_ADD, G_FPTOSI, G_FPTOUI, G_SUB, G_XOR, G_TRUNC, G_BITCAST, G_ANYEXT):$root,
+  (match (wip_match_opcode G_ADD, G_FPTOSI, G_FPTOUI, G_SUB, G_XOR, G_TRUNC, G_BITCAST, G_ANYEXT, G_PTRTOINT, G_INTTOPTR):$root,
          [{ return Helper.matchAnyExplicitUseIsUndef(*${root}); }]),
   (apply [{ Helper.replaceInstWithUndef(*${root}); }])>;
 

llvm/test/CodeGen/AArch64/GlobalISel/combine-cast.mir

@@ -217,3 +217,27 @@ body:             |
     %large:_(<2 x s64>) = G_ANYEXT %bv(<2 x s32>)
     $q0 = COPY %large(<2 x s64>)
     $d0 = COPY %bv(<2 x s32>)
+...
+---
+name:            test_undef_ptrtoint
+legalized: true
+body:             |
+  bb.1:
+    ; CHECK-LABEL: name: test_undef_ptrtoint
+    ; CHECK: [[DEF:%[0-9]+]]:_(<2 x s64>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: $q0 = COPY [[DEF]](<2 x s64>)
+    %0:_(<2 x p0>) = G_IMPLICIT_DEF
+    %1:_(<2 x s64>) = G_PTRTOINT %0(<2 x p0>)
+    $q0 = COPY %1(<2 x s64>)
+...
+---
+name:            test_undef_inttoptr
+legalized: true
+body:             |
+  bb.1:
+    ; CHECK-LABEL: name: test_undef_inttoptr
+    ; CHECK: [[DEF:%[0-9]+]]:_(<2 x p0>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: $q0 = COPY [[DEF]](<2 x p0>)
+    %0:_(<2 x s64>) = G_IMPLICIT_DEF
+    %1:_(<2 x p0>) = G_INTTOPTR %0(<2 x s64>)
+    $q0 = COPY %1(<2 x p0>)

llvm/test/CodeGen/AArch64/neon-reverseshuffle.ll

@@ -13,18 +13,10 @@ entry:
 }
 
 define <2 x ptr> @v2p0(<2 x ptr> %a) {
-; CHECK-SD-LABEL: v2p0:
-; CHECK-SD:       // %bb.0: // %entry
-; CHECK-SD-NEXT:    ext v0.16b, v0.16b, v0.16b, #8
-; CHECK-SD-NEXT:    ret
-;
-; CHECK-GI-LABEL: v2p0:
-; CHECK-GI:       // %bb.0: // %entry
-; CHECK-GI-NEXT:    adrp x8, .LCPI1_0
-; CHECK-GI-NEXT:    // kill: def $q0 killed $q0 def $q0_q1
-; CHECK-GI-NEXT:    ldr q2, [x8, :lo12:.LCPI1_0]
-; CHECK-GI-NEXT:    tbl v0.16b, { v0.16b, v1.16b }, v2.16b
-; CHECK-GI-NEXT:    ret
+; CHECK-LABEL: v2p0:
+; CHECK:       // %bb.0: // %entry
+; CHECK-NEXT:    ext v0.16b, v0.16b, v0.16b, #8
+; CHECK-NEXT:    ret
 entry:
   %V128 = shufflevector <2 x ptr> %a, <2 x ptr> undef, <2 x i32> <i32 1, i32 0>
   ret <2 x ptr> %V128

llvm/test/CodeGen/AMDGPU/lds-global-non-entry-func.ll

@@ -100,59 +100,29 @@ define void @func_use_lds_global() {
 
 ; ERR: warning: <unknown>:0:0: in function func_use_lds_global_constexpr_cast void (ptr addrspace(1)): local memory global used by non-kernel function
 define void @func_use_lds_global_constexpr_cast(ptr addrspace(1) %out) {
-; GFX8-SDAG-LABEL: func_use_lds_global_constexpr_cast:
-; GFX8-SDAG:       ; %bb.0:
-; GFX8-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX8-SDAG-NEXT:    s_mov_b64 s[4:5], 0xc8
-; GFX8-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX8-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-SDAG-NEXT:    s_trap 2
-; GFX8-SDAG-NEXT:    s_setpc_b64 s[30:31]
-;
-; GFX8-GISEL-LABEL: func_use_lds_global_constexpr_cast:
-; GFX8-GISEL:       ; %bb.0:
-; GFX8-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX8-GISEL-NEXT:    s_mov_b64 s[4:5], 0xc8
-; GFX8-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX8-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-GISEL-NEXT:    s_trap 2
-; GFX8-GISEL-NEXT:    flat_store_dword v[0:1], v0
-; GFX8-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-GISEL-NEXT:    s_setpc_b64 s[30:31]
-;
-; GFX9-SDAG-LABEL: func_use_lds_global_constexpr_cast:
-; GFX9-SDAG:       ; %bb.0:
-; GFX9-SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-SDAG-NEXT:    s_trap 2
-; GFX9-SDAG-NEXT:    s_setpc_b64 s[30:31]
+; GFX8-LABEL: func_use_lds_global_constexpr_cast:
+; GFX8:       ; %bb.0:
+; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX8-NEXT:    s_mov_b64 s[4:5], 0xc8
+; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
+; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX8-NEXT:    s_trap 2
+; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
-; GFX9-GISEL-LABEL: func_use_lds_global_constexpr_cast:
-; GFX9-GISEL:       ; %bb.0:
-; GFX9-GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-GISEL-NEXT:    s_trap 2
-; GFX9-GISEL-NEXT:    global_store_dword v[0:1], v0, off
-; GFX9-GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-GISEL-NEXT:    s_setpc_b64 s[30:31]
+; GFX9-LABEL: func_use_lds_global_constexpr_cast:
+; GFX9:       ; %bb.0:
+; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX9-NEXT:    s_trap 2
+; GFX9-NEXT:    s_setpc_b64 s[30:31]
 ;
-; SDAG-LABEL: func_use_lds_global_constexpr_cast:
-; SDAG:       ; %bb.0:
-; SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; SDAG-NEXT:    s_cbranch_execnz .LBB1_2
-; SDAG-NEXT:  ; %bb.1:
-; SDAG-NEXT:    s_setpc_b64 s[30:31]
-; SDAG-NEXT:  .LBB1_2:
-; SDAG-NEXT:    s_endpgm
-;
-; GISEL-LABEL: func_use_lds_global_constexpr_cast:
-; GISEL:       ; %bb.0:
-; GISEL-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GISEL-NEXT:    s_cbranch_execnz .LBB1_2
-; GISEL-NEXT:  ; %bb.1:
-; GISEL-NEXT:    global_store_dword v[0:1], v0, off
-; GISEL-NEXT:    s_waitcnt vmcnt(0)
-; GISEL-NEXT:    s_setpc_b64 s[30:31]
-; GISEL-NEXT:  .LBB1_2:
-; GISEL-NEXT:    s_endpgm
+; CHECK-LABEL: func_use_lds_global_constexpr_cast:
+; CHECK:       ; %bb.0:
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    s_cbranch_execnz .LBB1_2
+; CHECK-NEXT:  ; %bb.1:
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+; CHECK-NEXT:  .LBB1_2:
+; CHECK-NEXT:    s_endpgm
   store i32 ptrtoint (ptr addrspace(3) @lds to i32), ptr addrspace(1) %out, align 4
   ret void
 }
@@ -611,7 +581,3 @@ ret:
   ret i32 %phi
 }
 
-;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
-; CHECK: {{.*}}
-; GFX8: {{.*}}
-; GFX9: {{.*}}