@@ -2189,3 +2189,11 @@ DEF_HELPER_6(th_vmford_vv_d, void, ptr, ptr, ptr, ptr, env, i32)
DEF_HELPER_6(th_vmford_vf_h, void, ptr, ptr, i64, ptr, env, i32)
DEF_HELPER_6(th_vmford_vf_w, void, ptr, ptr, i64, ptr, env, i32)
DEF_HELPER_6(th_vmford_vf_d, void, ptr, ptr, i64, ptr, env, i32)
+
+DEF_HELPER_5(th_vfclass_v_h, void, ptr, ptr, ptr, env, i32)
+DEF_HELPER_5(th_vfclass_v_w, void, ptr, ptr, ptr, env, i32)
+DEF_HELPER_5(th_vfclass_v_d, void, ptr, ptr, ptr, env, i32)
+
+DEF_HELPER_6(th_vfmerge_vfm_h, void, ptr, ptr, i64, ptr, env, i32)
+DEF_HELPER_6(th_vfmerge_vfm_w, void, ptr, ptr, i64, ptr, env, i32)
+DEF_HELPER_6(th_vfmerge_vfm_d, void, ptr, ptr, i64, ptr, env, i32)
@@ -2143,15 +2143,60 @@ GEN_OPFVF_TRANS_TH(th_vmfgt_vf, opfvf_cmp_check_th)
GEN_OPFVF_TRANS_TH(th_vmfge_vf, opfvf_cmp_check_th)
GEN_OPFVF_TRANS_TH(th_vmford_vf, opfvf_cmp_check_th)
+/* Vector Floating-Point Classify Instruction */
+GEN_OPFV_TRANS_TH(th_vfclass_v, opfv_check_th)
+
+/* Vector Floating-Point Merge Instruction */
+GEN_OPFVF_TRANS_TH(th_vfmerge_vfm, opfvf_check_th)
+
+/* Besides of check function, th_vfmv_v_f just reuse the helper_th_vmv_v_x */
+static bool trans_th_vfmv_v_f(DisasContext *s, arg_th_vfmv_v_f *a)
+{
+ if (require_xtheadvector(s) &&
+ vext_check_isa_ill(s) &&
+ th_check_reg(s, a->rd, false) &&
+ (s->sew != 0)) {
+
+ TCGv_i64 t1;
+
+ if (s->vl_eq_vlmax) {
+ t1 = tcg_temp_new_i64();
+ /* NaN-box f[rs1] */
+ do_nanbox(s, t1, cpu_fpr[a->rs1]);
+ tcg_gen_gvec_dup_i64(s->sew, vreg_ofs(s, a->rd),
+ MAXSZ(s), MAXSZ(s), t1);
+ } else {
+ TCGv_ptr dest;
+ TCGv_i32 desc;
+ uint32_t data = FIELD_DP32(0, VDATA_TH, LMUL, s->lmul);
+ static gen_helper_vmv_vx_th * const fns[3] = {
+ gen_helper_th_vmv_v_x_h,
+ gen_helper_th_vmv_v_x_w,
+ gen_helper_th_vmv_v_x_d,
+ };
+
+ t1 = tcg_temp_new_i64();
+ /* NaN-box f[rs1] */
+ do_nanbox(s, t1, cpu_fpr[a->rs1]);
+
+ dest = tcg_temp_new_ptr();
+ desc = tcg_constant_i32(simd_desc(s->cfg_ptr->vlenb,
+ s->cfg_ptr->vlenb, data));
+ tcg_gen_addi_ptr(dest, tcg_env, vreg_ofs(s, a->rd));
+ fns[s->sew - 1](dest, t1, tcg_env, desc);
+ }
+ finalize_rvv_inst(s);
+ return true;
+ }
+ return false;
+}
+
#define TH_TRANS_STUB(NAME) \
static bool trans_##NAME(DisasContext *s, arg_##NAME *a) \
{ \
return require_xtheadvector(s); \
}
-TH_TRANS_STUB(th_vfclass_v)
-TH_TRANS_STUB(th_vfmerge_vfm)
-TH_TRANS_STUB(th_vfmv_v_f)
TH_TRANS_STUB(th_vfcvt_xu_f_v)
TH_TRANS_STUB(th_vfcvt_x_f_v)
TH_TRANS_STUB(th_vfcvt_f_xu_v)
@@ -3147,3 +3147,61 @@ GEN_TH_CMP_VV_ENV(th_vmford_vv_d, uint64_t, H8, !float64_unordered_quiet)
GEN_TH_CMP_VF(th_vmford_vf_h, uint16_t, H2, !float16_unordered_quiet)
GEN_TH_CMP_VF(th_vmford_vf_w, uint32_t, H4, !float32_unordered_quiet)
GEN_TH_CMP_VF(th_vmford_vf_d, uint64_t, H8, !float64_unordered_quiet)
+
+/* Vector Floating-Point Classify Instruction */
+#define TH_OPIVV1(NAME, TD, T2, TX2, HD, HS2, OP) \
+ OPIVV1(NAME, TD, T2, TX2, HD, HS2, OP)
+
+#define GEN_TH_V(NAME, ESZ, DSZ, CLEAR_FN) \
+void HELPER(NAME)(void *vd, void *v0, void *vs2, \
+ CPURISCVState *env, uint32_t desc) \
+{ \
+ uint32_t vlmax = th_maxsz(desc) / ESZ; \
+ uint32_t mlen = th_mlen(desc); \
+ uint32_t vm = th_vm(desc); \
+ uint32_t vl = env->vl; \
+ uint32_t i; \
+ \
+ VSTART_CHECK_EARLY_EXIT(env); \
+ for (i = env->vstart; i < vl; i++) { \
+ if (!vm && !th_elem_mask(v0, mlen, i)) { \
+ continue; \
+ } \
+ do_##NAME(vd, vs2, i); \
+ } \
+ env->vstart = 0; \
+ CLEAR_FN(vd, vl, vl * DSZ, vlmax * DSZ); \
+}
+
+THCALL(TH_OPIVV1, th_vfclass_v_h, OP_UU_H, H2, H2, fclass_h)
+THCALL(TH_OPIVV1, th_vfclass_v_w, OP_UU_W, H4, H4, fclass_s)
+THCALL(TH_OPIVV1, th_vfclass_v_d, OP_UU_D, H8, H8, fclass_d)
+GEN_TH_V(th_vfclass_v_h, 2, 2, clearh_th)
+GEN_TH_V(th_vfclass_v_w, 4, 4, clearl_th)
+GEN_TH_V(th_vfclass_v_d, 8, 8, clearq_th)
+
+/* Vector Floating-Point Merge Instruction */
+#define GEN_VFMERGE_VF_TH(NAME, ETYPE, H, CLEAR_FN) \
+void HELPER(NAME)(void *vd, void *v0, uint64_t s1, void *vs2, \
+ CPURISCVState *env, uint32_t desc) \
+{ \
+ uint32_t mlen = th_mlen(desc); \
+ uint32_t vm = th_vm(desc); \
+ uint32_t vl = env->vl; \
+ uint32_t esz = sizeof(ETYPE); \
+ uint32_t vlmax = th_maxsz(desc) / esz; \
+ uint32_t i; \
+ \
+ VSTART_CHECK_EARLY_EXIT(env); \
+ for (i = env->vstart; i < vl; i++) { \
+ ETYPE s2 = *((ETYPE *)vs2 + H(i)); \
+ *((ETYPE *)vd + H(i)) \
+ = (!vm && !th_elem_mask(v0, mlen, i) ? s2 : s1); \
+ } \
+ env->vstart = 0; \
+ CLEAR_FN(vd, vl, vl * esz, vlmax * esz); \
+}
+
+GEN_VFMERGE_VF_TH(th_vfmerge_vfm_h, int16_t, H2, clearh_th)
+GEN_VFMERGE_VF_TH(th_vfmerge_vfm_w, int32_t, H4, clearl_th)
+GEN_VFMERGE_VF_TH(th_vfmerge_vfm_d, int64_t, H8, clearq_th)
The instructions have the same function as RVV1.0. Overall there are only general differences between XTheadVector and RVV1.0. Signed-off-by: Huang Tao <eric.huang@linux.alibaba.com> --- target/riscv/helper.h | 8 +++ .../riscv/insn_trans/trans_xtheadvector.c.inc | 51 +++++++++++++++- target/riscv/xtheadvector_helper.c | 58 +++++++++++++++++++ 3 files changed, 114 insertions(+), 3 deletions(-)