From: Chad Rosier <mcrosier@codeaurora.org>
Date: Mon, 2 Dec 2013 21:07:27 +0000 (+0000)
Subject: [AArch64] Implemented vcopy_lane patterns using scalar DUP instruction.
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=5e2f8fdda97247644715af72b2473711cf2b0df2;p=clang

[AArch64] Implemented vcopy_lane patterns using scalar DUP instruction.
Patch by Ana Pazos!

git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@196153 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/include/clang/Basic/arm_neon.td b/include/clang/Basic/arm_neon.td
index 4ae0859665..3ad1fc84cc 100644
--- a/include/clang/Basic/arm_neon.td
+++ b/include/clang/Basic/arm_neon.td
@@ -836,13 +836,13 @@ def GET_LANE : IInst<"vget_lane", "sdi",
 def SET_LANE : IInst<"vset_lane", "dsdi",
                      "csilPcPsUcUsUiUlQcQsQiQlQUcQUsQUiQUlPcPsQPcQPsfdQfQdPlQPl">;
 def COPY_LANE : IOpInst<"vcopy_lane", "ddidi",
-                        "csiPcPsUcUsUiPcPsfPl", OP_COPY_LN>;
+                        "csilPcPsUcUsUiUlPcPsPlfd", OP_COPY_LN>;
 def COPYQ_LANE : IOpInst<"vcopy_lane", "ddigi",
                         "QcQsQiQlQUcQUsQUiQUlQPcQPsQfQdQPl", OP_COPYQ_LN>;
 def COPY_LANEQ : IOpInst<"vcopy_laneq", "ddiki",
-                     "csiPcPsUcUsUif", OP_COPY_LNQ>;
+                     "csilPcPsPlUcUsUiUlf", OP_COPY_LNQ>;
 def COPYQ_LANEQ : IOpInst<"vcopy_laneq", "ddidi",
-                     "QcQsQiQlQUcQUsQUiQUlQPcQPsQfdQPl", OP_COPY_LN>;
+                     "QcQsQiQlQUcQUsQUiQUlQPcQPsQfdQdQPl", OP_COPY_LN>;
 
 ////////////////////////////////////////////////////////////////////////////////
 // Set all lanes to same value
diff --git a/test/CodeGen/aarch64-neon-copy.c b/test/CodeGen/aarch64-neon-copy.c
index 7c77b177af..6bc6f005f3 100644
--- a/test/CodeGen/aarch64-neon-copy.c
+++ b/test/CodeGen/aarch64-neon-copy.c
@@ -1192,3 +1192,56 @@ float64x2_t test_vmovq_n_f64(float64_t v1) {
   return vmovq_n_f64(v1);
   // CHECK: dup {{v[0-9]+}}.2d, {{v[0-9]+}}.d[0]
 }
+
+// CHECK: test_vcopy_lane_s64
+int64x1_t test_vcopy_lane_s64(int64x1_t a, int64x1_t c) {
+  return vcopy_lane_s64(a, 0, c, 0);
+// CHECK: fmov {{d[0-9]+}}, {{d[0-9]+}}
+// CHECK-NOT: dup {{d[0-9]+}}, {{v[0-9]+}}.d[0]
+}
+
+// CHECK: test_vcopy_lane_u64
+uint64x1_t test_vcopy_lane_u64(uint64x1_t a, uint64x1_t c) {
+  return vcopy_lane_u64(a, 0, c, 0);
+// CHECK: fmov {{d[0-9]+}}, {{d[0-9]+}}
+// CHECK-NOT: dup {{d[0-9]+}}, {{v[0-9]+}}.d[0]
+}
+
+// CHECK: test_vcopy_lane_f64
+float64x1_t test_vcopy_lane_f64(float64x1_t a, float64x1_t c) {
+  return vcopy_lane_f64(a, 0, c, 0);
+// CHECK: fmov {{d[0-9]+}}, {{d[0-9]+}}
+// CHECK-NOT: dup {{d[0-9]+}}, {{v[0-9]+}}.d[0]
+}
+
+// CHECK: test_vcopy_laneq_s64
+int64x1_t test_vcopy_laneq_s64(int64x1_t a, int64x2_t c) {
+  return vcopy_laneq_s64(a, 0, c, 1);
+// CHECK: dup {{d[0-9]+}}, {{v[0-9]+}}.d[1]
+}
+
+// CHECK: test_vcopy_laneq_u64
+uint64x1_t test_vcopy_laneq_u64(uint64x1_t a, uint64x2_t c) {
+  return vcopy_laneq_u64(a, 0, c, 1);
+// CHECK: dup {{d[0-9]+}}, {{v[0-9]+}}.d[1]
+}
+
+// CHECK: test_vcopy_laneq_f64
+float64x1_t test_vcopy_laneq_f64(float64x1_t a, float64x1_t c) {
+  return vcopy_laneq_f64(a, 0, c, 0);
+// CHECK: fmov {{d[0-9]+}}, {{d[0-9]+}}
+// CHECK-NOT: dup {{d[0-9]+}}, {{v[0-9]+}}.d[0]
+}
+
+// CHECK: test_vcopy_laneq_p64
+poly64x1_t test_vcopy_laneq_p64(poly64x1_t a, poly64x2_t c) {
+  return vcopy_laneq_p64(a, 0, c, 1);
+// CHECK: dup {{d[0-9]+}}, {{v[0-9]+}}.d[1]
+}
+
+// CHECK: test_vcopyq_laneq_f64
+float64x2_t test_vcopyq_laneq_f64(float64x2_t a, float64x2_t c) {
+// CHECK: ins  {{v[0-9]+}}.d[1], {{v[0-9]+}}.d[1]
+  return vcopyq_laneq_f64(a, 1, c, 1);
+}
+