Re: [PATCH v6 57/61] target/riscv: floating-point scalar move instructions

[LIU Zhiwei]

On 2020/3/28 11:44, Richard Henderson wrote:

On 3/17/20 8:06 AM, LIU Zhiwei wrote:

+/* Floating-Point Scalar Move Instructions */
+static bool trans_vfmv_f_s(DisasContext *s, arg_vfmv_f_s *a)
+{
+    if (!s->vill && has_ext(s, RVF) &&
+        (s->mstatus_fs != 0) && (s->sew != 0)) {
+#ifdef HOST_WORDS_BIGENDIAN
+        int ofs = vreg_ofs(s, a->rs2) + ((7 >> s->sew) << s->sew);
+#else
+        int ofs = vreg_ofs(s, a->rs2);
+#endif

Use endian_ofs from patch 55.

Yes, I forgot it.

+        switch (s->sew) {
+        case MO_8:
+            tcg_gen_ld8u_i64(cpu_fpr[a->rd], cpu_env, ofs);
+            tcg_gen_ori_i64(cpu_fpr[a->rd], cpu_fpr[a->rd],
+                            0xffffffffffffff00ULL);
+            break;

MO_8 should be illegal.

Yes, it has been checked in s->sew != 0.

+        case MO_16:
+            tcg_gen_ld16u_i64(cpu_fpr[a->rd], cpu_env, ofs);
+            tcg_gen_ori_i64(cpu_fpr[a->rd], cpu_fpr[a->rd],
+                            0xffffffffffff0000ULL);
+            break;
+        case MO_32:
+            tcg_gen_ld32u_i64(cpu_fpr[a->rd], cpu_env, ofs);
+            tcg_gen_ori_i64(cpu_fpr[a->rd], cpu_fpr[a->rd],
+                            0xffffffff00000000ULL);
+            break;
+        default:
+            if (has_ext(s, RVD)) {
+                tcg_gen_ld_i64(cpu_fpr[a->rd], cpu_env, ofs);
+            } else {
+                tcg_gen_ld32u_i64(cpu_fpr[a->rd], cpu_env, ofs);
+                tcg_gen_ori_i64(cpu_fpr[a->rd], cpu_fpr[a->rd],
+                                0xffffffff00000000ULL);
+            }
+            break;

Maybe better with MO_64 and default: g_assert_not_reached().

+static bool trans_vfmv_s_f(DisasContext *s, arg_vfmv_s_f *a)
+{
+    if (!s->vill && has_ext(s, RVF) && (s->sew != 0)) {
+        TCGv_ptr dest;
+        TCGv_i64 src1;
+        static gen_helper_vfmv_s_f * const fns[3] = {
+            gen_helper_vfmv_s_f_h,
+            gen_helper_vfmv_s_f_w,
+            gen_helper_vfmv_s_f_d

You shouldn't need to duplicate the vmv_s_x_* helpers.

All these helpers will be away in v7.

Zhiwei

r~