From: Richard Henderson <[email protected]>
Reviewed-by: Peter Maydell <[email protected]>
Signed-off-by: Richard Henderson <[email protected]>
Message-id: [email protected]
Signed-off-by: Peter Maydell <[email protected]>
---
target/arm/tcg/helper-sme-defs.h | 3 +++
target/arm/tcg/sme.decode | 2 ++
target/arm/tcg/sme_helper.c | 16 ++++++++++++++++
target/arm/tcg/translate-sme.c | 18 ++++++++++++++++++
4 files changed, 39 insertions(+)
diff --git a/target/arm/tcg/helper-sme-defs.h b/target/arm/tcg/helper-sme-defs.h
index 780b8f162a..ddf771e446 100644
--- a/target/arm/tcg/helper-sme-defs.h
+++ b/target/arm/tcg/helper-sme-defs.h
@@ -379,3 +379,6 @@ DEF_HELPER_FLAGS_5(sme_ah_bfmop4s_sh, TCG_CALL_NO_RWG,
void, ptr, ptr, ptr, env,
DEF_HELPER_FLAGS_5(sme_fmop4a_sh, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, env,
i32)
DEF_HELPER_FLAGS_5(sme_fmop4s_sh, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, env,
i32)
DEF_HELPER_FLAGS_5(sme_ah_fmop4s_sh, TCG_CALL_NO_RWG, void, ptr, ptr, ptr,
env, i32)
+
+DEF_HELPER_FLAGS_4(sme_smop4a_sh, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
+DEF_HELPER_FLAGS_4(sme_smop4s_sh, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, i32)
diff --git a/target/arm/tcg/sme.decode b/target/arm/tcg/sme.decode
index 4b7a99e203..20a2febd63 100644
--- a/target/arm/tcg/sme.decode
+++ b/target/arm/tcg/sme.decode
@@ -1116,3 +1116,5 @@ FMOP4_sh 1000 0001 001. ...0 0000 00.. ..0. 00..
@mop4_o2
FMOP4A_sb 1000 0000 001. ...0 0000 00.. ..00 00.. @mop4_o2_s0
FMOP4A_hb 1000 0000 001. ...0 0000 00.. ..00 100. @mop4_o1_s0
+
+SMOP4_sh 1000 0000 000. ...0 1000 00.. ..0. 10.. @mop4_o2
diff --git a/target/arm/tcg/sme_helper.c b/target/arm/tcg/sme_helper.c
index 123f459342..32fe867a3b 100644
--- a/target/arm/tcg/sme_helper.c
+++ b/target/arm/tcg/sme_helper.c
@@ -2938,3 +2938,19 @@ void HELPER(sme_ah_fmop4s_sh)(void *vza, void *vzn, void
*vzm,
{
sme_mop4(vza, vzn, vzm, env, desc, sizeof(float32), inner_ah_fmop4s_sh);
}
+
+#define IMOP4_2WAY(NAME, OP, TYPED, TYPEN, TYPEM) \
+static void inner_##NAME(void *vd, void *vn, void *vm, void *vinfo) \
+{ \
+ TYPEN *n = vn; TYPEM *m = vm; TYPED *d = vd; \
+ *d OP##= (TYPED)n[0] * m[0] + (TYPED)n[1] * m[1]; \
+} \
+void HELPER(sme_##NAME)(void *vza, void *vzn, void *vzm, uint32_t desc) \
+{ \
+ sme_mop4(vza, vzn, vzm, NULL, desc, sizeof(TYPED), inner_##NAME); \
+}
+
+IMOP4_2WAY(smop4a_sh, +, int32_t, int16_t, int16_t)
+IMOP4_2WAY(smop4s_sh, -, int32_t, int16_t, int16_t)
+
+#undef IMOP4_2WAY
diff --git a/target/arm/tcg/translate-sme.c b/target/arm/tcg/translate-sme.c
index b198c7c855..b948591dd3 100644
--- a/target/arm/tcg/translate-sme.c
+++ b/target/arm/tcg/translate-sme.c
@@ -2092,3 +2092,21 @@ TRANS_FEAT(FMOP4A_sb, aa64_sme_mop4_f8f32,
do_mop4_fp8, a, MO_32, gen_helper_sme_fmop4a_sb)
TRANS_FEAT(FMOP4A_hb, aa64_sme_mop4_f8f16,
do_mop4_fp8, a, MO_16, gen_helper_sme_fmop4a_hb)
+
+static bool do_mop4_int(DisasContext *s, arg_mop4 *a, MemOp esz,
+ gen_helper_gvec_3 *fn)
+{
+ if (sme_smza_enabled_check(s)) {
+ int svl = streaming_vec_reg_size(s);
+ uint32_t desc = simd_desc(svl, svl, (a->m << 1) | a->n);
+ TCGv_ptr za = get_tile(s, esz, a->zad);
+ TCGv_ptr zn = vec_full_reg_ptr(s, a->zn);
+ TCGv_ptr zm = vec_full_reg_ptr(s, a->zm);
+
+ fn(za, zn, zm, tcg_constant_i32(desc));
+ }
+ return true;
+}
+
+TRANS_FEAT(SMOP4_sh, aa64_sme_mop4, do_mop4_int, a, MO_32,
+ a->s ? gen_helper_sme_smop4s_sh : gen_helper_sme_smop4a_sh)
--
2.43.0
