[Qemu-devel] [PATCH v2 32/32] s390x/tcg: Implement VECTOR UNPACK *

[David Hildenbrand <david@redhat.com>]

Combine all variant in a single handler. As source and destination
have different element sizes, we can't use gvec expansion. Expand
manually. Also watch out for overlapping source and destination
registers. Use a safe evaluation order depending on the operation.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: David Hildenbrand <david@redhat.com>
---
 target/s390x/insn-data.def      |  8 +++++++
 target/s390x/translate_vx.inc.c | 38 +++++++++++++++++++++++++++++++++
 2 files changed, 46 insertions(+)

diff --git a/target/s390x/insn-data.def b/target/s390x/insn-data.def
index 00ba9444a3..71fa9b8d6c 100644
--- a/target/s390x/insn-data.def
+++ b/target/s390x/insn-data.def
@@ -1045,6 +1045,14 @@
     F(0xe73e, VSTM,    VRS_a, V,   la2, 0, 0, 0, vstm, 0, IF_VEC)
 /* VECTOR STORE WITH LENGTH */
     F(0xe73f, VSTL,    VRS_b, V,   la2, r3_32u, 0, 0, vstl, 0, IF_VEC)
+/* VECTOR UNPACK HIGH */
+    F(0xe7d7, VUPH,    VRR_a, V,   0, 0, 0, 0, vup, 0, IF_VEC)
+/* VECTOR UNPACK LOGICAL HIGH */
+    F(0xe7d5, VUPLH,   VRR_a, V,   0, 0, 0, 0, vup, 0, IF_VEC)
+/* VECTOR UNPACK LOW */
+    F(0xe7d6, VUPL,    VRR_a, V,   0, 0, 0, 0, vup, 0, IF_VEC)
+/* VECTOR UNPACK LOGICAL LOW */
+    F(0xe7d4, VUPLL,   VRR_a, V,   0, 0, 0, 0, vup, 0, IF_VEC)
 
 #ifndef CONFIG_USER_ONLY
 /* COMPARE AND SWAP AND PURGE */
diff --git a/target/s390x/translate_vx.inc.c b/target/s390x/translate_vx.inc.c
index 2e7e5391c5..23e4afd643 100644
--- a/target/s390x/translate_vx.inc.c
+++ b/target/s390x/translate_vx.inc.c
@@ -891,3 +891,41 @@ static DisasJumpType op_vstl(DisasContext *s, DisasOps *o)
     tcg_temp_free_ptr(a0);
     return DISAS_NEXT;
 }
+
+static DisasJumpType op_vup(DisasContext *s, DisasOps *o)
+{
+    const bool logical = s->fields->op2 == 0xd4 || s->fields->op2 == 0xd5;
+    const uint8_t v1 = get_field(s->fields, v1);
+    const uint8_t v2 = get_field(s->fields, v2);
+    const uint8_t src_es = get_field(s->fields, m3);
+    const uint8_t dst_es = src_es + 1;
+    int dst_idx, src_idx;
+    TCGv_i64 tmp;
+
+    if (src_es > ES_32) {
+        gen_program_exception(s, PGM_SPECIFICATION);
+        return DISAS_NORETURN;
+    }
+
+    tmp = tcg_temp_new_i64();
+    if (s->fields->op2 == 0xd7 || s->fields->op2 == 0xd5) {
+        /* iterate backwards to avoid overwriting data we might need later */
+        for (dst_idx = NUM_VEC_ELEMENTS(dst_es) - 1; dst_idx >= 0; dst_idx--) {
+            src_idx = dst_idx;
+            read_vec_element_i64(tmp, v2, src_idx,
+                                 src_es | (logical ? 0 : MO_SIGN));
+            write_vec_element_i64(tmp, v1, dst_idx, dst_es);
+        }
+
+    } else {
+        /* iterate forward to avoid overwriting data we might need later */
+        for (dst_idx = 0; dst_idx < NUM_VEC_ELEMENTS(dst_es); dst_idx++) {
+            src_idx = dst_idx + NUM_VEC_ELEMENTS(src_es) / 2;
+            read_vec_element_i64(tmp, v2, src_idx,
+                                 src_es | (logical ? 0 : MO_SIGN));
+            write_vec_element_i64(tmp, v1, dst_idx, dst_es);
+        }
+    }
+    tcg_temp_free_i64(tmp);
+    return DISAS_NEXT;
+}
-- 
2.17.2