tcg/aarch64: Implement INDEX_op_rotl{i,v}_vec

For immediate rotate , we can implement this in two instructions,
using SLI.  For variable rotate, the oddness of aarch64 right-shift-
as-negative-left-shift means a backend-specific expansion works best.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

2 files changed, 52 insertions, 2 deletions

diff --git a/tcg/aarch64/tcg-target.inc.c b/tcg/aarch64/tcg-target.inc.c
index 843fd0ca69..760b0e742d 100644
--- a/tcg/aarch64/tcg-target.inc.c
+++ b/tcg/aarch64/tcg-target.inc.c
@@ -557,6 +557,7 @@ typedef enum {
     I3614_SSHR      = 0x0f000400,
     I3614_SSRA      = 0x0f001400,
     I3614_SHL       = 0x0f005400,
+    I3614_SLI       = 0x2f005400,
     I3614_USHR      = 0x2f000400,
     I3614_USRA      = 0x2f001400,
 
@@ -2411,6 +2412,9 @@ static void tcg_out_vec_op(TCGContext *s, TCGOpcode opc,
     case INDEX_op_sari_vec:
         tcg_out_insn(s, 3614, SSHR, is_q, a0, a1, (16 << vece) - a2);
         break;
+    case INDEX_op_aa64_sli_vec:
+        tcg_out_insn(s, 3614, SLI, is_q, a0, a2, args[3] + (8 << vece));
+        break;
     case INDEX_op_shlv_vec:
         tcg_out_insn(s, 3616, USHL, is_q, vece, a0, a1, a2);
         break;
@@ -2498,8 +2502,11 @@ int tcg_can_emit_vec_op(TCGOpcode opc, TCGType type, unsigned vece)
     case INDEX_op_shlv_vec:
     case INDEX_op_bitsel_vec:
         return 1;
+    case INDEX_op_rotli_vec:
     case INDEX_op_shrv_vec:
     case INDEX_op_sarv_vec:
+    case INDEX_op_rotlv_vec:
+    case INDEX_op_rotrv_vec:
         return -1;
     case INDEX_op_mul_vec:
     case INDEX_op_smax_vec:
@@ -2517,14 +2524,24 @@ void tcg_expand_vec_op(TCGOpcode opc, TCGType type, unsigned vece,
                        TCGArg a0, ...)
 {
     va_list va;
-    TCGv_vec v0, v1, v2, t1;
+    TCGv_vec v0, v1, v2, t1, t2;
+    TCGArg a2;
 
     va_start(va, a0);
     v0 = temp_tcgv_vec(arg_temp(a0));
     v1 = temp_tcgv_vec(arg_temp(va_arg(va, TCGArg)));
-    v2 = temp_tcgv_vec(arg_temp(va_arg(va, TCGArg)));
+    a2 = va_arg(va, TCGArg);
+    v2 = temp_tcgv_vec(arg_temp(a2));
 
     switch (opc) {
+    case INDEX_op_rotli_vec:
+        t1 = tcg_temp_new_vec(type);
+        tcg_gen_shri_vec(vece, t1, v1, -a2 & ((8 << vece) - 1));
+        vec_gen_4(INDEX_op_aa64_sli_vec, type, vece,
+                  tcgv_vec_arg(v0), tcgv_vec_arg(t1), tcgv_vec_arg(v1), a2);
+        tcg_temp_free_vec(t1);
+        break;
+
     case INDEX_op_shrv_vec:
     case INDEX_op_sarv_vec:
         /* Right shifts are negative left shifts for AArch64.  */
@@ -2537,6 +2554,35 @@ void tcg_expand_vec_op(TCGOpcode opc, TCGType type, unsigned vece,
         tcg_temp_free_vec(t1);
         break;
 
+    case INDEX_op_rotlv_vec:
+        t1 = tcg_temp_new_vec(type);
+        tcg_gen_dupi_vec(vece, t1, 8 << vece);
+        tcg_gen_sub_vec(vece, t1, v2, t1);
+        /* Right shifts are negative left shifts for AArch64.  */
+        vec_gen_3(INDEX_op_shlv_vec, type, vece, tcgv_vec_arg(t1),
+                  tcgv_vec_arg(v1), tcgv_vec_arg(t1));
+        vec_gen_3(INDEX_op_shlv_vec, type, vece, tcgv_vec_arg(v0),
+                  tcgv_vec_arg(v1), tcgv_vec_arg(v2));
+        tcg_gen_or_vec(vece, v0, v0, t1);
+        tcg_temp_free_vec(t1);
+        break;
+
+    case INDEX_op_rotrv_vec:
+        t1 = tcg_temp_new_vec(type);
+        t2 = tcg_temp_new_vec(type);
+        tcg_gen_neg_vec(vece, t1, v2);
+        tcg_gen_dupi_vec(vece, t2, 8 << vece);
+        tcg_gen_add_vec(vece, t2, t1, t2);
+        /* Right shifts are negative left shifts for AArch64.  */
+        vec_gen_3(INDEX_op_shlv_vec, type, vece, tcgv_vec_arg(t1),
+                  tcgv_vec_arg(v1), tcgv_vec_arg(t1));
+        vec_gen_3(INDEX_op_shlv_vec, type, vece, tcgv_vec_arg(t2),
+                  tcgv_vec_arg(v1), tcgv_vec_arg(t2));
+        tcg_gen_or_vec(vece, v0, t1, t2);
+        tcg_temp_free_vec(t1);
+        tcg_temp_free_vec(t2);
+        break;
+
     default:
         g_assert_not_reached();
     }
@@ -2557,6 +2603,7 @@ static const TCGTargetOpDef *tcg_target_op_def(TCGOpcode op)
     static const TCGTargetOpDef lZ_l = { .args_ct_str = { "lZ", "l" } };
     static const TCGTargetOpDef r_r_r = { .args_ct_str = { "r", "r", "r" } };
     static const TCGTargetOpDef w_w_w = { .args_ct_str = { "w", "w", "w" } };
+    static const TCGTargetOpDef w_0_w = { .args_ct_str = { "w", "0", "w" } };
     static const TCGTargetOpDef w_w_wO = { .args_ct_str = { "w", "w", "wO" } };
     static const TCGTargetOpDef w_w_wN = { .args_ct_str = { "w", "w", "wN" } };
     static const TCGTargetOpDef w_w_wZ = { .args_ct_str = { "w", "w", "wZ" } };
@@ -2751,6 +2798,8 @@ static const TCGTargetOpDef *tcg_target_op_def(TCGOpcode op)
         return &w_w_wZ;
     case INDEX_op_bitsel_vec:
         return &w_w_w_w;
+    case INDEX_op_aa64_sli_vec:
+        return &w_0_w;
 
     default:
         return NULL;
diff --git a/tcg/aarch64/tcg-target.opc.h b/tcg/aarch64/tcg-target.opc.h
index 26bfd9c460..bce30accd9 100644
--- a/tcg/aarch64/tcg-target.opc.h
+++ b/tcg/aarch64/tcg-target.opc.h
@@ -12,3 +12,4 @@
  */
 
 DEF(aa64_sshl_vec, 1, 2, 0, IMPLVEC)
+DEF(aa64_sli_vec, 1, 2, 1, IMPLVEC)