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/cpu-features.h | 5 +++++
target/arm/tcg/fp8_helper.c | 16 ++++++++++++++++
target/arm/tcg/helper-fp8-defs.h | 2 ++
target/arm/tcg/sme.decode | 5 +++++
target/arm/tcg/translate-sme.c | 18 ++++++++++++++++++
5 files changed, 46 insertions(+)
diff --git a/target/arm/cpu-features.h b/target/arm/cpu-features.h
index b4138e976b..e577880ab0 100644
--- a/target/arm/cpu-features.h
+++ b/target/arm/cpu-features.h
@@ -1822,6 +1822,11 @@ static inline bool
isar_feature_aa64_sme_mop4_f64f64(const ARMISARegisters *id)
return isar_feature_aa64_sme_mop4(id) && isar_feature_aa64_sme_f64f64(id);
}
+static inline bool isar_feature_aa64_sme_mop4_f8f32(const ARMISARegisters *id)
+{
+ return isar_feature_aa64_sme_mop4(id) && isar_feature_aa64_sme_f8f32(id);
+}
+
/*
* Feature tests for "does this exist in either 32-bit or 64-bit?"
*/
diff --git a/target/arm/tcg/fp8_helper.c b/target/arm/tcg/fp8_helper.c
index 902e4b0ae9..f899a6cbf7 100644
--- a/target/arm/tcg/fp8_helper.c
+++ b/target/arm/tcg/fp8_helper.c
@@ -973,3 +973,19 @@ void HELPER(sme_fvdot_idx_hb)(void *vd, void *vn, void *vm,
} while (++i & 7);
} while (i < elements);
}
+
+static void inner_fmop4a_sb(void *vd, void *vn, void *vm, void *vinfo)
+{
+ float32 *d = vd;
+ uint32_t *n = vn, *m = vm;
+ FP8MulContext *ctx = vinfo;
+
+ *d = f8dotadd_s(*n, *m, 4, *d, ctx);
+}
+
+void HELPER(sme_fmop4a_sb)(void *vza, void *vzn, void *vzm,
+ CPUArchState *env, uint32_t desc)
+{
+ FP8MulContext ctx = fp8_mul_start(env, -1);
+ sme_mop4(vza, vzn, vzm, &ctx, desc, sizeof(float32), inner_fmop4a_sb);
+}
diff --git a/target/arm/tcg/helper-fp8-defs.h b/target/arm/tcg/helper-fp8-defs.h
index 126dcadf77..47f35a82dc 100644
--- a/target/arm/tcg/helper-fp8-defs.h
+++ b/target/arm/tcg/helper-fp8-defs.h
@@ -44,3 +44,5 @@ DEF_HELPER_FLAGS_7(sme_fmopa_hb, TCG_CALL_NO_RWG, void, ptr,
ptr, ptr, ptr, ptr,
DEF_HELPER_FLAGS_5(sme_fvdot_idx_sb, TCG_CALL_NO_RWG, void, ptr, ptr, ptr,
env, i32)
DEF_HELPER_FLAGS_5(sme_fvdot_idx_hb, TCG_CALL_NO_RWG, void, ptr, ptr, ptr,
env, i32)
+
+DEF_HELPER_FLAGS_5(sme_fmop4a_sb, TCG_CALL_NO_RWG, void, ptr, ptr, ptr, env,
i32)
diff --git a/target/arm/tcg/sme.decode b/target/arm/tcg/sme.decode
index ea9f459e38..b8b507d8fb 100644
--- a/target/arm/tcg/sme.decode
+++ b/target/arm/tcg/sme.decode
@@ -1101,6 +1101,9 @@ LUTI4_s_4b 1100 0000 1001 101 1 00 00 00 ....0
zd:5 \
@mop4_o3 .... .... ... m:1 .... .... .. n:1 ... . s:1 . zad:3 \
&mop4 zm=%mop4_zm zn=%mop4_zn
+@mop4_o2_s0 .... .... ... m:1 .... .... .. n:1 ... . ... zad:2 \
+ &mop4 zm=%mop4_zm zn=%mop4_zn s=0
+
BFMOP4_hh 1000 0001 001. ...0 0000 00.. ..0. 100. @mop4_o1
FMOP4_hh 1000 0001 000. ...0 0000 00.. ..0. 100. @mop4_o1
FMOP4_ss 1000 0000 000. ...0 0000 00.. ..0. 00.. @mop4_o2
@@ -1108,3 +1111,5 @@ FMOP4_dd 1000 0000 110. ...0 0000 00.. ..0. 1...
@mop4_o3
BFMOP4_sh 1000 0001 000. ...0 0000 00.. ..0. 00.. @mop4_o2
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
diff --git a/target/arm/tcg/translate-sme.c b/target/arm/tcg/translate-sme.c
index 544a825595..386ded1b90 100644
--- a/target/arm/tcg/translate-sme.c
+++ b/target/arm/tcg/translate-sme.c
@@ -2072,3 +2072,21 @@ static gen_helper_gvec_3_ptr * const fmop4_sh[3] = {
gen_helper_sme_ah_fmop4s_sh
};
TRANS_FEAT(FMOP4_sh, aa64_sme_mop4, do_mop4_fp, a, MO_32, FPST_ENV, fmop4_sh)
+
+static bool do_mop4_fp8(DisasContext *s, arg_mop4 *a, MemOp esz,
+ gen_helper_gvec_3_ptr *fn)
+{
+ if (fpmr_access_check(s) && 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_env, tcg_constant_i32(desc));
+ }
+ return true;
+}
+
+TRANS_FEAT(FMOP4A_sb, aa64_sme_mop4_f8f32,
+ do_mop4_fp8, a, MO_32, gen_helper_sme_fmop4a_sb)
--
2.43.0
