[llvm] [AArch64][CodeGen] Fold tbx(splat(0), table, idxs) to tbl(table, idxs) (PR #214146)

[Benjamin Maxwell via llvm-commits]

================
@@ -0,0 +1,60 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc -mtriple=aarch64 -mattr=+neon < %s | FileCheck %s
+
+define <16 x i8> @test_tbx1_zero_splat(<16 x i8> %tbl, <16 x i8> %idx) {
+; CHECK-LABEL: test_tbx1_zero_splat:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    movi v2.2d, #0000000000000000
+; CHECK-NEXT:    tbx v2.16b, { v0.16b }, v1.16b
+; CHECK-NEXT:    mov v0.16b, v2.16b
+; CHECK-NEXT:    ret
+  %res = call <16 x i8> @llvm.aarch64.neon.tbx1.v16i8(<16 x i8> zeroinitializer, <16 x i8> %tbl, <16 x i8> %idx)
+  ret <16 x i8> %res
+}
+
+define <16 x i8> @test_tbx2_zero_splat(<16 x i8> %tbl1, <16 x i8> %tbl2, <16 x i8> %idx) {
+; CHECK-LABEL: test_tbx2_zero_splat:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    movi v3.2d, #0000000000000000
+; CHECK-NEXT:    // kill: def $q1 killed $q1 killed $q0_q1 def $q0_q1
+; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $q0_q1 def $q0_q1
+; CHECK-NEXT:    tbx v3.16b, { v0.16b, v1.16b }, v2.16b
+; CHECK-NEXT:    mov v0.16b, v3.16b
+; CHECK-NEXT:    ret
+  %res = call <16 x i8> @llvm.aarch64.neon.tbx2.v16i8(<16 x i8> zeroinitializer, <16 x i8> %tbl1, <16 x i8> %tbl2, <16 x i8> %idx)
+  ret <16 x i8> %res
+}
+
+define <16 x i8> @test_tbx3_zero_splat(<16 x i8> %tbl1, <16 x i8> %tbl2, <16 x i8> %tbl3, <16 x i8> %idx) {
+; CHECK-LABEL: test_tbx3_zero_splat:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    movi v4.2d, #0000000000000000
+; CHECK-NEXT:    // kill: def $q2 killed $q2 killed $q0_q1_q2 def $q0_q1_q2
+; CHECK-NEXT:    // kill: def $q1 killed $q1 killed $q0_q1_q2 def $q0_q1_q2
+; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $q0_q1_q2 def $q0_q1_q2
+; CHECK-NEXT:    tbx v4.16b, { v0.16b, v1.16b, v2.16b }, v3.16b
+; CHECK-NEXT:    mov v0.16b, v4.16b
+; CHECK-NEXT:    ret
+  %res = call <16 x i8> @llvm.aarch64.neon.tbx3.v16i8(<16 x i8> zeroinitializer, <16 x i8> %tbl1, <16 x i8> %tbl2, <16 x i8> %tbl3, <16 x i8> %idx)
+  ret <16 x i8> %res
+}
+
+define <16 x i8> @test_tbx4_zero_splat(<16 x i8> %tbl1, <16 x i8> %tbl2, <16 x i8> %tbl3, <16 x i8> %tbl4, <16 x i8> %idx) {
+; CHECK-LABEL: test_tbx4_zero_splat:
+; CHECK:       // %bb.0:
+; CHECK-NEXT:    movi v5.2d, #0000000000000000
+; CHECK-NEXT:    // kill: def $q3 killed $q3 killed $q0_q1_q2_q3 def $q0_q1_q2_q3
+; CHECK-NEXT:    // kill: def $q2 killed $q2 killed $q0_q1_q2_q3 def $q0_q1_q2_q3
+; CHECK-NEXT:    // kill: def $q1 killed $q1 killed $q0_q1_q2_q3 def $q0_q1_q2_q3
+; CHECK-NEXT:    // kill: def $q0 killed $q0 killed $q0_q1_q2_q3 def $q0_q1_q2_q3
+; CHECK-NEXT:    tbx v5.16b, { v0.16b, v1.16b, v2.16b, v3.16b }, v4.16b
+; CHECK-NEXT:    mov v0.16b, v5.16b
+; CHECK-NEXT:    ret
+  %res = call <16 x i8> @llvm.aarch64.neon.tbx4.v16i8(<16 x i8> zeroinitializer, <16 x i8> %tbl1, <16 x i8> %tbl2, <16 x i8> %tbl3, <16 x i8> %tbl4, <16 x i8> %idx)
+  ret <16 x i8> %res
+}
+
+declare <16 x i8> @llvm.aarch64.neon.tbx1.v16i8(<16 x i8>, <16 x i8>, <16 x i8>)
+declare <16 x i8> @llvm.aarch64.neon.tbx2.v16i8(<16 x i8>, <16 x i8>, <16 x i8>, <16 x i8>)
+declare <16 x i8> @llvm.aarch64.neon.tbx3.v16i8(<16 x i8>, <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8>)
+declare <16 x i8> @llvm.aarch64.neon.tbx4.v16i8(<16 x i8>, <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8>, <16 x i8>)
----------------
MacDue wrote:

You don't need to declare instrinics 
```suggestion
```

https://github.com/llvm/llvm-project/pull/214146