[PATCH v2 16/37] target/sparc: Implement FMEAN16

[Richard Henderson <richard.henderson@linaro.org>]

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/sparc/helper.h     |  1 +
 target/sparc/insns.decode |  1 +
 target/sparc/translate.c  | 30 ++++++++++++++++++++++++++++++
 target/sparc/vis_helper.c | 21 +++++++++++++++++++++
 4 files changed, 53 insertions(+)

diff --git a/target/sparc/helper.h b/target/sparc/helper.h
index ab79954bb5..f1b84dc9b3 100644
--- a/target/sparc/helper.h
+++ b/target/sparc/helper.h
@@ -118,6 +118,7 @@ DEF_HELPER_FLAGS_2(cmask8, TCG_CALL_NO_RWG_SE, i64, i64, i64)
 DEF_HELPER_FLAGS_2(cmask16, TCG_CALL_NO_RWG_SE, i64, i64, i64)
 DEF_HELPER_FLAGS_2(cmask32, TCG_CALL_NO_RWG_SE, i64, i64, i64)
 DEF_HELPER_FLAGS_2(fchksm16, TCG_CALL_NO_RWG_SE, i64, i64, i64)
+DEF_HELPER_FLAGS_2(fmean16, TCG_CALL_NO_RWG_SE, i64, i64, i64)
 #define VIS_CMPHELPER(name)                                              \
     DEF_HELPER_FLAGS_2(f##name##16, TCG_CALL_NO_RWG_SE,      \
                        i64, i64, i64)                                    \
diff --git a/target/sparc/insns.decode b/target/sparc/insns.decode
index de29996304..febd1a4a13 100644
--- a/target/sparc/insns.decode
+++ b/target/sparc/insns.decode
@@ -421,6 +421,7 @@ FCMPEq      10 000 cc:2 110101 .....  0 0101 0111 .....    \
     PDIST       10 ..... 110110 ..... 0 0011 1110 .....    \
                 &r_r_r_r rd=%dfp_rd rs1=%dfp_rd rs2=%dfp_rs1 rs3=%dfp_rs2
 
+    FMEAN16     10 ..... 110110 ..... 0 0100 0000 .....    @d_d_d
     FCHKSM16    10 ..... 110110 ..... 0 0100 0100 .....    @d_d_d
     FALIGNDATAg 10 ..... 110110 ..... 0 0100 1000 .....    @d_d_d
     FPMERGE     10 ..... 110110 ..... 0 0100 1011 .....    @d_r_r
diff --git a/target/sparc/translate.c b/target/sparc/translate.c
index f53ecf7c41..cf5f1ea4d0 100644
--- a/target/sparc/translate.c
+++ b/target/sparc/translate.c
@@ -815,8 +815,37 @@ static void gen_op_fchksm16(unsigned vece, uint32_t dofs, uint32_t aofs,
     };
     tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &op);
 }
+
+static void gen_vec_fmean16(unsigned vece, TCGv_vec dst,
+                            TCGv_vec src1, TCGv_vec src2)
+{
+    TCGv_vec t = tcg_temp_new_vec_matching(dst);
+
+    tcg_gen_or_vec(vece, t, src1, src2);
+    tcg_gen_and_vec(vece, t, t, tcg_constant_vec_matching(dst, vece, 1));
+    tcg_gen_sari_vec(vece, src1, src1, 1);
+    tcg_gen_sari_vec(vece, src2, src2, 1);
+    tcg_gen_add_vec(vece, dst, src1, src2);
+    tcg_gen_add_vec(vece, dst, dst, t);
+}
+
+static void gen_op_fmean16(unsigned vece, uint32_t dofs, uint32_t aofs,
+                           uint32_t bofs, uint32_t oprsz, uint32_t maxsz)
+{
+    static const TCGOpcode vecop_list[] = {
+        INDEX_op_add_vec, INDEX_op_sari_vec,
+    };
+    static const GVecGen3 op = {
+        .fni8 = gen_helper_fmean16,
+        .fniv = gen_vec_fmean16,
+        .opt_opc = vecop_list,
+        .vece = MO_16,
+    };
+    tcg_gen_gvec_3(dofs, aofs, bofs, oprsz, maxsz, &op);
+}
 #else
 #define gen_op_fchksm16   ({ qemu_build_not_reached(); NULL; })
+#define gen_op_fmean16    ({ qemu_build_not_reached(); NULL; })
 #endif
 
 static void finishing_insn(DisasContext *dc)
@@ -4840,6 +4869,7 @@ TRANS(FPADD32, VIS1, do_gvec_ddd, a, MO_32, tcg_gen_gvec_add)
 TRANS(FPSUB16, VIS1, do_gvec_ddd, a, MO_16, tcg_gen_gvec_sub)
 TRANS(FPSUB32, VIS1, do_gvec_ddd, a, MO_32, tcg_gen_gvec_sub)
 TRANS(FCHKSM16, VIS3, do_gvec_ddd, a, MO_16, gen_op_fchksm16)
+TRANS(FMEAN16, VIS3, do_gvec_ddd, a, MO_16, gen_op_fmean16)
 
 static bool do_ddd(DisasContext *dc, arg_r_r_r *a,
                    void (*func)(TCGv_i64, TCGv_i64, TCGv_i64))
diff --git a/target/sparc/vis_helper.c b/target/sparc/vis_helper.c
index fa607375d2..6ef36755c3 100644
--- a/target/sparc/vis_helper.c
+++ b/target/sparc/vis_helper.c
@@ -412,3 +412,24 @@ uint64_t helper_fchksm16(uint64_t src1, uint64_t src2)
 
     return r.ll;
 }
+
+static inline int16_t do_fmean16(int16_t src1, int16_t src2)
+{
+    return (src1 + src2 + 1) / 2;
+}
+
+uint64_t helper_fmean16(uint64_t src1, uint64_t src2)
+{
+    VIS64 r, s1, s2;
+
+    s1.ll = src1;
+    s2.ll = src2;
+    r.ll = 0;
+
+    r.VIS_SW64(0) = do_fmean16(s1.VIS_SW64(0), s2.VIS_SW64(0));
+    r.VIS_SW64(1) = do_fmean16(s1.VIS_SW64(1), s2.VIS_SW64(1));
+    r.VIS_SW64(2) = do_fmean16(s1.VIS_SW64(2), s2.VIS_SW64(2));
+    r.VIS_SW64(3) = do_fmean16(s1.VIS_SW64(3), s2.VIS_SW64(3));
+
+    return r.ll;
+}
-- 
2.34.1