target/loongarch: Implement vavg/vavgr

This patch includes:
- VAVG.{B/H/W/D}[U];
- VAVGR.{B/H/W/D}[U].

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Message-Id: <20230504122810.4094787-11-gaosong@loongson.cn>

diff --git a/target/loongarch/disas.c b/target/loongarch/disas.c
index 8ee14916f315ba3b3a7d0e7104ee157251804f3f..e7592e7a346c46c6135c107f6743b55e6bb725b5 100644 (file)
--- a/target/loongarch/disas.c
+++ b/target/loongarch/disas.c
@@ -908,3 +908,20 @@ INSN_LSX(vaddwod_h_bu_b,   vvv)
 INSN_LSX(vaddwod_w_hu_h,   vvv)
 INSN_LSX(vaddwod_d_wu_w,   vvv)
 INSN_LSX(vaddwod_q_du_d,   vvv)
+
+INSN_LSX(vavg_b,           vvv)
+INSN_LSX(vavg_h,           vvv)
+INSN_LSX(vavg_w,           vvv)
+INSN_LSX(vavg_d,           vvv)
+INSN_LSX(vavg_bu,          vvv)
+INSN_LSX(vavg_hu,          vvv)
+INSN_LSX(vavg_wu,          vvv)
+INSN_LSX(vavg_du,          vvv)
+INSN_LSX(vavgr_b,          vvv)
+INSN_LSX(vavgr_h,          vvv)
+INSN_LSX(vavgr_w,          vvv)
+INSN_LSX(vavgr_d,          vvv)
+INSN_LSX(vavgr_bu,         vvv)
+INSN_LSX(vavgr_hu,         vvv)
+INSN_LSX(vavgr_wu,         vvv)
+INSN_LSX(vavgr_du,         vvv)

diff --git a/target/loongarch/helper.h b/target/loongarch/helper.h
index 505c336ff384c66a70b012967a488ecefbb98eae..af0f9f9b0d9c1988ff4255acb4cf3683f3bba61b 100644 (file)
--- a/target/loongarch/helper.h
+++ b/target/loongarch/helper.h
@@ -193,3 +193,21 @@ DEF_HELPER_FLAGS_4(vaddwod_h_bu_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
 DEF_HELPER_FLAGS_4(vaddwod_w_hu_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
 DEF_HELPER_FLAGS_4(vaddwod_d_wu_w, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
 DEF_HELPER_FLAGS_4(vaddwod_q_du_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+
+DEF_HELPER_FLAGS_4(vavg_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(vavg_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(vavg_w, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(vavg_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(vavg_bu, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(vavg_hu, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(vavg_wu, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(vavg_du, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+
+DEF_HELPER_FLAGS_4(vavgr_b, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(vavgr_h, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(vavgr_w, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(vavgr_d, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(vavgr_bu, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(vavgr_hu, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(vavgr_wu, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(vavgr_du, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)

diff --git a/target/loongarch/insn_trans/trans_lsx.c.inc b/target/loongarch/insn_trans/trans_lsx.c.inc
index 8edff83157e44dbb94d1f8dc6222c843587de8d2..5fa4792305c5a5ab0fc666dfe67918eded0e2514 100644 (file)
--- a/target/loongarch/insn_trans/trans_lsx.c.inc
+++ b/target/loongarch/insn_trans/trans_lsx.c.inc
@@ -969,3 +969,200 @@ TRANS(vaddwod_h_bu_b, gvec_vvv, MO_8, do_vaddwod_u_s)
 TRANS(vaddwod_w_hu_h, gvec_vvv, MO_16, do_vaddwod_u_s)
 TRANS(vaddwod_d_wu_w, gvec_vvv, MO_32, do_vaddwod_u_s)
 TRANS(vaddwod_q_du_d, gvec_vvv, MO_64, do_vaddwod_u_s)
+
+static void do_vavg(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b,
+                    void (*gen_shr_vec)(unsigned, TCGv_vec,
+                                        TCGv_vec, int64_t),
+                    void (*gen_round_vec)(unsigned, TCGv_vec,
+                                          TCGv_vec, TCGv_vec))
+{
+    TCGv_vec tmp = tcg_temp_new_vec_matching(t);
+    gen_round_vec(vece, tmp, a, b);
+    tcg_gen_and_vec(vece, tmp, tmp, tcg_constant_vec_matching(t, vece, 1));
+    gen_shr_vec(vece, a, a, 1);
+    gen_shr_vec(vece, b, b, 1);
+    tcg_gen_add_vec(vece, t, a, b);
+    tcg_gen_add_vec(vece, t, t, tmp);
+}
+
+static void gen_vavg_s(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b)
+{
+    do_vavg(vece, t, a, b, tcg_gen_sari_vec, tcg_gen_and_vec);
+}
+
+static void gen_vavg_u(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b)
+{
+    do_vavg(vece, t, a, b, tcg_gen_shri_vec, tcg_gen_and_vec);
+}
+
+static void gen_vavgr_s(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b)
+{
+    do_vavg(vece, t, a, b, tcg_gen_sari_vec, tcg_gen_or_vec);
+}
+
+static void gen_vavgr_u(unsigned vece, TCGv_vec t, TCGv_vec a, TCGv_vec b)
+{
+    do_vavg(vece, t, a, b, tcg_gen_shri_vec, tcg_gen_or_vec);
+}
+
+static void do_vavg_s(unsigned vece, uint32_t vd_ofs, uint32_t vj_ofs,
+                      uint32_t vk_ofs, uint32_t oprsz, uint32_t maxsz)
+{
+    static const TCGOpcode vecop_list[] = {
+        INDEX_op_sari_vec, INDEX_op_add_vec, 0
+        };
+    static const GVecGen3 op[4] = {
+        {
+            .fniv = gen_vavg_s,
+            .fno = gen_helper_vavg_b,
+            .opt_opc = vecop_list,
+            .vece = MO_8
+        },
+        {
+            .fniv = gen_vavg_s,
+            .fno = gen_helper_vavg_h,
+            .opt_opc = vecop_list,
+            .vece = MO_16
+        },
+        {
+            .fniv = gen_vavg_s,
+            .fno = gen_helper_vavg_w,
+            .opt_opc = vecop_list,
+            .vece = MO_32
+        },
+        {
+            .fniv = gen_vavg_s,
+            .fno = gen_helper_vavg_d,
+            .opt_opc = vecop_list,
+            .vece = MO_64
+        },
+    };
+
+    tcg_gen_gvec_3(vd_ofs, vj_ofs, vk_ofs, oprsz, maxsz, &op[vece]);
+}
+
+static void do_vavg_u(unsigned vece, uint32_t vd_ofs, uint32_t vj_ofs,
+                      uint32_t vk_ofs, uint32_t oprsz, uint32_t maxsz)
+{
+    static const TCGOpcode vecop_list[] = {
+        INDEX_op_shri_vec, INDEX_op_add_vec, 0
+        };
+    static const GVecGen3 op[4] = {
+        {
+            .fniv = gen_vavg_u,
+            .fno = gen_helper_vavg_bu,
+            .opt_opc = vecop_list,
+            .vece = MO_8
+        },
+        {
+            .fniv = gen_vavg_u,
+            .fno = gen_helper_vavg_hu,
+            .opt_opc = vecop_list,
+            .vece = MO_16
+        },
+        {
+            .fniv = gen_vavg_u,
+            .fno = gen_helper_vavg_wu,
+            .opt_opc = vecop_list,
+            .vece = MO_32
+        },
+        {
+            .fniv = gen_vavg_u,
+            .fno = gen_helper_vavg_du,
+            .opt_opc = vecop_list,
+            .vece = MO_64
+        },
+    };
+
+    tcg_gen_gvec_3(vd_ofs, vj_ofs, vk_ofs, oprsz, maxsz, &op[vece]);
+}
+
+TRANS(vavg_b, gvec_vvv, MO_8, do_vavg_s)
+TRANS(vavg_h, gvec_vvv, MO_16, do_vavg_s)
+TRANS(vavg_w, gvec_vvv, MO_32, do_vavg_s)
+TRANS(vavg_d, gvec_vvv, MO_64, do_vavg_s)
+TRANS(vavg_bu, gvec_vvv, MO_8, do_vavg_u)
+TRANS(vavg_hu, gvec_vvv, MO_16, do_vavg_u)
+TRANS(vavg_wu, gvec_vvv, MO_32, do_vavg_u)
+TRANS(vavg_du, gvec_vvv, MO_64, do_vavg_u)
+
+static void do_vavgr_s(unsigned vece, uint32_t vd_ofs, uint32_t vj_ofs,
+                       uint32_t vk_ofs, uint32_t oprsz, uint32_t maxsz)
+{
+    static const TCGOpcode vecop_list[] = {
+        INDEX_op_sari_vec, INDEX_op_add_vec, 0
+        };
+    static const GVecGen3 op[4] = {
+        {
+            .fniv = gen_vavgr_s,
+            .fno = gen_helper_vavgr_b,
+            .opt_opc = vecop_list,
+            .vece = MO_8
+        },
+        {
+            .fniv = gen_vavgr_s,
+            .fno = gen_helper_vavgr_h,
+            .opt_opc = vecop_list,
+            .vece = MO_16
+        },
+        {
+            .fniv = gen_vavgr_s,
+            .fno = gen_helper_vavgr_w,
+            .opt_opc = vecop_list,
+            .vece = MO_32
+        },
+        {
+            .fniv = gen_vavgr_s,
+            .fno = gen_helper_vavgr_d,
+            .opt_opc = vecop_list,
+            .vece = MO_64
+        },
+    };
+
+    tcg_gen_gvec_3(vd_ofs, vj_ofs, vk_ofs, oprsz, maxsz, &op[vece]);
+}
+
+static void do_vavgr_u(unsigned vece, uint32_t vd_ofs, uint32_t vj_ofs,
+                       uint32_t vk_ofs, uint32_t oprsz, uint32_t maxsz)
+{
+    static const TCGOpcode vecop_list[] = {
+        INDEX_op_shri_vec, INDEX_op_add_vec, 0
+        };
+    static const GVecGen3 op[4] = {
+        {
+            .fniv = gen_vavgr_u,
+            .fno = gen_helper_vavgr_bu,
+            .opt_opc = vecop_list,
+            .vece = MO_8
+        },
+        {
+            .fniv = gen_vavgr_u,
+            .fno = gen_helper_vavgr_hu,
+            .opt_opc = vecop_list,
+            .vece = MO_16
+        },
+        {
+            .fniv = gen_vavgr_u,
+            .fno = gen_helper_vavgr_wu,
+            .opt_opc = vecop_list,
+            .vece = MO_32
+        },
+        {
+            .fniv = gen_vavgr_u,
+            .fno = gen_helper_vavgr_du,
+            .opt_opc = vecop_list,
+            .vece = MO_64
+        },
+    };
+
+    tcg_gen_gvec_3(vd_ofs, vj_ofs, vk_ofs, oprsz, maxsz, &op[vece]);
+}
+
+TRANS(vavgr_b, gvec_vvv, MO_8, do_vavgr_s)
+TRANS(vavgr_h, gvec_vvv, MO_16, do_vavgr_s)
+TRANS(vavgr_w, gvec_vvv, MO_32, do_vavgr_s)
+TRANS(vavgr_d, gvec_vvv, MO_64, do_vavgr_s)
+TRANS(vavgr_bu, gvec_vvv, MO_8, do_vavgr_u)
+TRANS(vavgr_hu, gvec_vvv, MO_16, do_vavgr_u)
+TRANS(vavgr_wu, gvec_vvv, MO_32, do_vavgr_u)
+TRANS(vavgr_du, gvec_vvv, MO_64, do_vavgr_u)

diff --git a/target/loongarch/insns.decode b/target/loongarch/insns.decode
index ee16155b31898404ea3a624f815994d3fc4fbed0..4a44380259838eb5d20d0c57731803efdf66a2c1 100644 (file)
--- a/target/loongarch/insns.decode
+++ b/target/loongarch/insns.decode
@@ -602,3 +602,20 @@ vaddwod_h_bu_b   0111 00000100 00000 ..... ..... .....    @vvv
 vaddwod_w_hu_h   0111 00000100 00001 ..... ..... .....    @vvv
 vaddwod_d_wu_w   0111 00000100 00010 ..... ..... .....    @vvv
 vaddwod_q_du_d   0111 00000100 00011 ..... ..... .....    @vvv
+
+vavg_b           0111 00000110 01000 ..... ..... .....    @vvv
+vavg_h           0111 00000110 01001 ..... ..... .....    @vvv
+vavg_w           0111 00000110 01010 ..... ..... .....    @vvv
+vavg_d           0111 00000110 01011 ..... ..... .....    @vvv
+vavg_bu          0111 00000110 01100 ..... ..... .....    @vvv
+vavg_hu          0111 00000110 01101 ..... ..... .....    @vvv
+vavg_wu          0111 00000110 01110 ..... ..... .....    @vvv
+vavg_du          0111 00000110 01111 ..... ..... .....    @vvv
+vavgr_b          0111 00000110 10000 ..... ..... .....    @vvv
+vavgr_h          0111 00000110 10001 ..... ..... .....    @vvv
+vavgr_w          0111 00000110 10010 ..... ..... .....    @vvv
+vavgr_d          0111 00000110 10011 ..... ..... .....    @vvv
+vavgr_bu         0111 00000110 10100 ..... ..... .....    @vvv
+vavgr_hu         0111 00000110 10101 ..... ..... .....    @vvv
+vavgr_wu         0111 00000110 10110 ..... ..... .....    @vvv
+vavgr_du         0111 00000110 10111 ..... ..... .....    @vvv

diff --git a/target/loongarch/lsx_helper.c b/target/loongarch/lsx_helper.c
index 6d391f711bad89b7c1b9b32813488d28d87a53b7..b683a1cd0a1117f29ce5adcd62320874f9566a49 100644 (file)
--- a/target/loongarch/lsx_helper.c
+++ b/target/loongarch/lsx_helper.c
@@ -275,3 +275,35 @@ void HELPER(vaddwod_q_du_d)(void *vd, void *vj, void *vk, uint32_t v)
 DO_ODD_U_S(vaddwod_h_bu_b, 16, H, UH, B, UB, DO_ADD)
 DO_ODD_U_S(vaddwod_w_hu_h, 32, W, UW, H, UH, DO_ADD)
 DO_ODD_U_S(vaddwod_d_wu_w, 64, D, UD, W, UW, DO_ADD)
+
+#define DO_VAVG(a, b)  ((a >> 1) + (b >> 1) + (a & b & 1))
+#define DO_VAVGR(a, b) ((a >> 1) + (b >> 1) + ((a | b) & 1))
+
+#define DO_3OP(NAME, BIT, E, DO_OP)                         \
+void HELPER(NAME)(void *vd, void *vj, void *vk, uint32_t v) \
+{                                                           \
+    int i;                                                  \
+    VReg *Vd = (VReg *)vd;                                  \
+    VReg *Vj = (VReg *)vj;                                  \
+    VReg *Vk = (VReg *)vk;                                  \
+    for (i = 0; i < LSX_LEN/BIT; i++) {                     \
+        Vd->E(i) = DO_OP(Vj->E(i), Vk->E(i));               \
+    }                                                       \
+}
+
+DO_3OP(vavg_b, 8, B, DO_VAVG)
+DO_3OP(vavg_h, 16, H, DO_VAVG)
+DO_3OP(vavg_w, 32, W, DO_VAVG)
+DO_3OP(vavg_d, 64, D, DO_VAVG)
+DO_3OP(vavgr_b, 8, B, DO_VAVGR)
+DO_3OP(vavgr_h, 16, H, DO_VAVGR)
+DO_3OP(vavgr_w, 32, W, DO_VAVGR)
+DO_3OP(vavgr_d, 64, D, DO_VAVGR)
+DO_3OP(vavg_bu, 8, UB, DO_VAVG)
+DO_3OP(vavg_hu, 16, UH, DO_VAVG)
+DO_3OP(vavg_wu, 32, UW, DO_VAVG)
+DO_3OP(vavg_du, 64, UD, DO_VAVG)
+DO_3OP(vavgr_bu, 8, UB, DO_VAVGR)
+DO_3OP(vavgr_hu, 16, UH, DO_VAVGR)
+DO_3OP(vavgr_wu, 32, UW, DO_VAVGR)
+DO_3OP(vavgr_du, 64, UD, DO_VAVGR)