Remove patterns for cases where the extract index must be explicitly scaled.

Author: paulwalker-arm

llvm/lib/Target/AArch64/SVEInstrFormats.td

@@ -1491,26 +1491,20 @@ multiclass sve_int_perm_dup_i<string asm> {
 
   // When extracting from an unpacked vector the index must be scaled to account
   // for the "holes" in the underlying packed vector type. We get the scaling
-  // for free by "promoting" the element type to one whose underlying vector type
-  // is packed.
+  // for free by "promoting" the element type to one whose underlying vector
+  // type is packed. This is only valid when extracting from a vector whose
+  // length is the same or bigger than the result of the splat.
 
-  foreach VT = [nxv2f16, nxv4f16, nxv8f16] in {
-    def : Pat<(VT (splat_vector (f16 (vector_extract (nxv4f16 ZPR:$vec), sve_elm_idx_extdup_s:$index)))),
+  foreach VT = [nxv4f16, nxv4bf16] in {
+    def : Pat<(SVEType<VT>.HalfLength (splat_vector (SVEType<VT>.EltAsScalar (vector_extract (VT ZPR:$vec), sve_elm_idx_extdup_s:$index)))),
               (!cast<Instruction>(NAME # _S) ZPR:$vec, sve_elm_idx_extdup_s:$index)>;
-    def : Pat<(VT (splat_vector (f16 (vector_extract (nxv2f16 ZPR:$vec), sve_elm_idx_extdup_d:$index)))),
-              (!cast<Instruction>(NAME # _D) ZPR:$vec, sve_elm_idx_extdup_d:$index)>;
-  }
-
-  foreach VT = [nxv2bf16, nxv4bf16, nxv8bf16] in {
-    def : Pat<(VT (splat_vector (bf16 (vector_extract (nxv4bf16 ZPR:$vec), sve_elm_idx_extdup_s:$index)))),
+    def : Pat<(VT (splat_vector (SVEType<VT>.EltAsScalar (vector_extract (VT ZPR:$vec), sve_elm_idx_extdup_s:$index)))),
               (!cast<Instruction>(NAME # _S) ZPR:$vec, sve_elm_idx_extdup_s:$index)>;
-    def : Pat<(VT (splat_vector (bf16 (vector_extract (nxv2bf16 ZPR:$vec), sve_elm_idx_extdup_d:$index)))),
-              (!cast<Instruction>(NAME # _D) ZPR:$vec, sve_elm_idx_extdup_d:$index)>;
   }
 
-  foreach VT = [nxv2f32, nxv4f32] in {
-    def : Pat<(VT (splat_vector (f32 (vector_extract (nxv2f32 ZPR:$vec), sve_elm_idx_extdup_d:$index)))),
-            (!cast<Instruction>(NAME # _D) ZPR:$vec, sve_elm_idx_extdup_d:$index)>;
+  foreach VT = [nxv2f16, nxv2f32, nxv2bf16] in {
+    def : Pat<(VT (splat_vector (SVEType<VT>.EltAsScalar (vector_extract (VT ZPR:$vec), sve_elm_idx_extdup_d:$index)))),
+              (!cast<Instruction>(NAME # _D) ZPR:$vec, sve_elm_idx_extdup_d:$index)>;
   }
 
   // Duplicate an indexed 128-bit segment across a vector.

llvm/test/CodeGen/AArch64/aarch64-dup-extract-scalable.ll

@@ -155,6 +155,7 @@ define <vscale x 8 x half> @dup_extract_nxv8f16_nxv4f16(<vscale x 4 x half> %dat
 ; CHECK-LABEL: dup_extract_nxv8f16_nxv4f16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z0.s, z0.s[1]
+; CHECK-NEXT:    mov z0.h, h0
 ; CHECK-NEXT:    ret
   %1 = extractelement <vscale x 4 x half> %data, i16 1
   %.splatinsert = insertelement <vscale x 8 x half> poison, half %1, i32 0
@@ -166,6 +167,7 @@ define <vscale x 8 x half> @dup_extract_nxv8f16_nxv2f16(<vscale x 2 x half> %dat
 ; CHECK-LABEL: dup_extract_nxv8f16_nxv2f16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z0.d, z0.d[1]
+; CHECK-NEXT:    mov z0.h, h0
 ; CHECK-NEXT:    ret
   %1 = extractelement <vscale x 2 x half> %data, i16 1
   %.splatinsert = insertelement <vscale x 8 x half> poison, half %1, i32 0
@@ -223,6 +225,7 @@ define <vscale x 4 x half> @dup_extract_nxv4f16_nxv2f16(<vscale x 2 x half> %dat
 ; CHECK-LABEL: dup_extract_nxv4f16_nxv2f16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z0.d, z0.d[1]
+; CHECK-NEXT:    mov z0.h, h0
 ; CHECK-NEXT:    ret
   %1 = extractelement <vscale x 2 x half> %data, i16 1
   %.splatinsert = insertelement <vscale x 4 x half> poison, half %1, i32 0
@@ -326,6 +329,7 @@ define <vscale x 4 x float> @dup_extract_nxv4f32_nxv2f32(<vscale x 2 x float> %d
 ; CHECK-LABEL: dup_extract_nxv4f32_nxv2f32:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z0.d, z0.d[1]
+; CHECK-NEXT:    mov z0.s, s0
 ; CHECK-NEXT:    ret
   %1 = extractelement <vscale x 2 x float> %data, i32 1
   %.splatinsert = insertelement <vscale x 4 x float> poison, float %1, i32 0
@@ -452,6 +456,7 @@ define <vscale x 8 x bfloat> @dup_extract_nxv8bf16_nxv4bf16(<vscale x 4 x bfloat
 ; CHECK-LABEL: dup_extract_nxv8bf16_nxv4bf16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z0.s, z0.s[1]
+; CHECK-NEXT:    mov z0.h, h0
 ; CHECK-NEXT:    ret
   %1 = extractelement <vscale x 4 x bfloat> %data, i16 1
   %.splatinsert = insertelement <vscale x 8 x bfloat> poison, bfloat %1, i32 0
@@ -463,6 +468,7 @@ define <vscale x 8 x bfloat> @dup_extract_nxv8bf16_nxv2bf16(<vscale x 2 x bfloat
 ; CHECK-LABEL: dup_extract_nxv8bf16_nxv2bf16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z0.d, z0.d[1]
+; CHECK-NEXT:    mov z0.h, h0
 ; CHECK-NEXT:    ret
   %1 = extractelement <vscale x 2 x bfloat> %data, i16 1
   %.splatinsert = insertelement <vscale x 8 x bfloat> poison, bfloat %1, i32 0
@@ -520,6 +526,7 @@ define <vscale x 4 x bfloat> @dup_extract_nxv4bf16_nxv2bf16(<vscale x 2 x bfloat
 ; CHECK-LABEL: dup_extract_nxv4bf16_nxv2bf16:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    mov z0.d, z0.d[1]
+; CHECK-NEXT:    mov z0.h, h0
 ; CHECK-NEXT:    ret
   %1 = extractelement <vscale x 2 x bfloat> %data, i16 1
   %.splatinsert = insertelement <vscale x 4 x bfloat> poison, bfloat %1, i32 0