target-mips: add new Floating Point Comparison instructions

Signed-off-by: Yongbok Kim <yongbok.kim@imgtec.com>
Signed-off-by: Leon Alrae <leon.alrae@imgtec.com>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>

4 files changed, 386 insertions, 2 deletions

diff --git a/disas/mips.c b/disas/mips.c
index fed2e5e07d..5ebb5fd6f8 100644
--- a/disas/mips.c
+++ b/disas/mips.c
@@ -1317,6 +1317,50 @@ const struct mips_opcode mips_builtin_opcodes[] =
 {"bc1nez",  "T,p",      0x45a00000, 0xffe00000, CBD|RD_T|FP_S|FP_D,   0, I32R6},
 {"bc2eqz",  "E,p",      0x49200000, 0xffe00000, CBD|RD_C2,            0, I32R6},
 {"bc2nez",  "E,p",      0x49a00000, 0xffe00000, CBD|RD_C2,            0, I32R6},
+{"cmp.af.s",   "D,S,T", 0x46800000, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.un.s",   "D,S,T", 0x46800001, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.eq.s",   "D,S,T", 0x46800002, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.ueq.s",  "D,S,T", 0x46800003, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.lt.s",   "D,S,T", 0x46800004, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.ult.s",  "D,S,T", 0x46800005, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.le.s",   "D,S,T", 0x46800006, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.ule.s",  "D,S,T", 0x46800007, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.saf.s",  "D,S,T", 0x46800008, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.sun.s",  "D,S,T", 0x46800009, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.seq.s",  "D,S,T", 0x4680000a, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.sueq.s", "D,S,T", 0x4680000b, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.slt.s",  "D,S,T", 0x4680000c, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.sult.s", "D,S,T", 0x4680000d, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.sle.s",  "D,S,T", 0x4680000e, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.sule.s", "D,S,T", 0x4680000f, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.or.s",   "D,S,T", 0x46800011, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.une.s",  "D,S,T", 0x46800012, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.ne.s",   "D,S,T", 0x46800013, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.sor.s",  "D,S,T", 0x46800019, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.sune.s", "D,S,T", 0x4680001a, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.sne.s",  "D,S,T", 0x4680001b, 0xffe0003f, RD_S|RD_T|WR_D|FP_S,  0, I32R6},
+{"cmp.af.d",   "D,S,T", 0x46a00000, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.un.d",   "D,S,T", 0x46a00001, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.eq.d",   "D,S,T", 0x46a00002, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.ueq.d",  "D,S,T", 0x46a00003, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.lt.d",   "D,S,T", 0x46a00004, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.ult.d",  "D,S,T", 0x46a00005, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.le.d",   "D,S,T", 0x46a00006, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.ule.d",  "D,S,T", 0x46a00007, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.saf.d",  "D,S,T", 0x46a00008, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.sun.d",  "D,S,T", 0x46a00009, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.seq.d",  "D,S,T", 0x46a0000a, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.sueq.d", "D,S,T", 0x46a0000b, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.slt.d",  "D,S,T", 0x46a0000c, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.sult.d", "D,S,T", 0x46a0000d, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.sle.d",  "D,S,T", 0x46a0000e, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.sule.d", "D,S,T", 0x46a0000f, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.or.d",   "D,S,T", 0x46a00011, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.une.d",  "D,S,T", 0x46a00012, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.ne.d",   "D,S,T", 0x46a00013, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.sor.d",  "D,S,T", 0x46a00019, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.sune.d", "D,S,T", 0x46a0001a, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
+{"cmp.sne.d",  "D,S,T", 0x46a0001b, 0xffe0003f, RD_S|RD_T|WR_D|FP_D,  0, I32R6},
 {"pref",    "k,o(b)",   0xcc000000, 0xfc000000, RD_b,           	0,		I4|I32|G3	},
 {"prefx",   "h,t(b)",	0x4c00000f, 0xfc0007ff, RD_b|RD_t,		0,		I4|I33	},
 {"nop",     "",         0x00000000, 0xffffffff, 0,              	INSN2_ALIAS,	I1      }, /* sll */
diff --git a/target-mips/helper.h b/target-mips/helper.h
index 9020c7bd6a..a127db55c2 100644
--- a/target-mips/helper.h
+++ b/target-mips/helper.h
@@ -304,6 +304,33 @@ FOP_PROTO(le)
 FOP_PROTO(ngt)
 #undef FOP_PROTO
 
+#define FOP_PROTO(op) \
+DEF_HELPER_3(r6_cmp_d_ ## op, i64, env, i64, i64) \
+DEF_HELPER_3(r6_cmp_s_ ## op, i32, env, i32, i32)
+FOP_PROTO(af)
+FOP_PROTO(un)
+FOP_PROTO(eq)
+FOP_PROTO(ueq)
+FOP_PROTO(lt)
+FOP_PROTO(ult)
+FOP_PROTO(le)
+FOP_PROTO(ule)
+FOP_PROTO(saf)
+FOP_PROTO(sun)
+FOP_PROTO(seq)
+FOP_PROTO(sueq)
+FOP_PROTO(slt)
+FOP_PROTO(sult)
+FOP_PROTO(sle)
+FOP_PROTO(sule)
+FOP_PROTO(or)
+FOP_PROTO(une)
+FOP_PROTO(ne)
+FOP_PROTO(sor)
+FOP_PROTO(sune)
+FOP_PROTO(sne)
+#undef FOP_PROTO
+
 /* Special functions */
 #ifndef CONFIG_USER_ONLY
 DEF_HELPER_1(tlbwi, void, env)
diff --git a/target-mips/op_helper.c b/target-mips/op_helper.c
index 3f449ae428..c48ee7f198 100644
--- a/target-mips/op_helper.c
+++ b/target-mips/op_helper.c
@@ -3388,3 +3388,114 @@ FOP_COND_PS(le,  float32_le(fst0, fst1, &env->active_fpu.fp_status),
                  float32_le(fsth0, fsth1, &env->active_fpu.fp_status))
 FOP_COND_PS(ngt, float32_unordered(fst1, fst0, &env->active_fpu.fp_status)    || float32_le(fst0, fst1, &env->active_fpu.fp_status),
                  float32_unordered(fsth1, fsth0, &env->active_fpu.fp_status)  || float32_le(fsth0, fsth1, &env->active_fpu.fp_status))
+
+/* R6 compare operations */
+#define FOP_CONDN_D(op, cond)                                       \
+uint64_t helper_r6_cmp_d_ ## op(CPUMIPSState * env, uint64_t fdt0,  \
+                         uint64_t fdt1)                             \
+{                                                                   \
+    uint64_t c;                                                     \
+    c = cond;                                                       \
+    update_fcr31(env, GETPC());                                     \
+    if (c) {                                                        \
+        return -1;                                                  \
+    } else {                                                        \
+        return 0;                                                   \
+    }                                                               \
+}
+
+/* NOTE: the comma operator will make "cond" to eval to false,
+ * but float64_unordered_quiet() is still called. */
+FOP_CONDN_D(af,  (float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status), 0))
+FOP_CONDN_D(un,  (float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status)))
+FOP_CONDN_D(eq,  (float64_eq_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(ueq, (float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status)
+                  || float64_eq_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(lt,  (float64_lt_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(ult, (float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status)
+                  || float64_lt_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(le,  (float64_le_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(ule, (float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status)
+                  || float64_le_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+/* NOTE: the comma operator will make "cond" to eval to false,
+ * but float64_unordered() is still called. */
+FOP_CONDN_D(saf,  (float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status), 0))
+FOP_CONDN_D(sun,  (float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status)))
+FOP_CONDN_D(seq,  (float64_eq(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(sueq, (float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_eq(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(slt,  (float64_lt(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(sult, (float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_lt(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(sle,  (float64_le(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(sule, (float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_le(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(or,   (float64_le_quiet(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_le_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(une,  (float64_unordered_quiet(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_lt_quiet(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_lt_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(ne,   (float64_lt_quiet(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_lt_quiet(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(sor,  (float64_le(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_le(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(sune, (float64_unordered(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_lt(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_lt(fdt0, fdt1, &env->active_fpu.fp_status)))
+FOP_CONDN_D(sne,  (float64_lt(fdt1, fdt0, &env->active_fpu.fp_status)
+                   || float64_lt(fdt0, fdt1, &env->active_fpu.fp_status)))
+
+#define FOP_CONDN_S(op, cond)                                       \
+uint32_t helper_r6_cmp_s_ ## op(CPUMIPSState * env, uint32_t fst0,  \
+                         uint32_t fst1)                             \
+{                                                                   \
+    uint64_t c;                                                     \
+    c = cond;                                                       \
+    update_fcr31(env, GETPC());                                     \
+    if (c) {                                                        \
+        return -1;                                                  \
+    } else {                                                        \
+        return 0;                                                   \
+    }                                                               \
+}
+
+/* NOTE: the comma operator will make "cond" to eval to false,
+ * but float32_unordered_quiet() is still called. */
+FOP_CONDN_S(af,   (float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status), 0))
+FOP_CONDN_S(un,   (float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)))
+FOP_CONDN_S(eq,   (float32_eq_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(ueq,  (float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_eq_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(lt,   (float32_lt_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(ult,  (float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(le,   (float32_le_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(ule,  (float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_le_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+/* NOTE: the comma operator will make "cond" to eval to false,
+ * but float32_unordered() is still called. */
+FOP_CONDN_S(saf,  (float32_unordered(fst1, fst0, &env->active_fpu.fp_status), 0))
+FOP_CONDN_S(sun,  (float32_unordered(fst1, fst0, &env->active_fpu.fp_status)))
+FOP_CONDN_S(seq,  (float32_eq(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(sueq, (float32_unordered(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_eq(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(slt,  (float32_lt(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(sult, (float32_unordered(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(sle,  (float32_le(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(sule, (float32_unordered(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_le(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(or,   (float32_le_quiet(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_le_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(une,  (float32_unordered_quiet(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt_quiet(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(ne,   (float32_lt_quiet(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt_quiet(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(sor,  (float32_le(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_le(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(sune, (float32_unordered(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt(fst0, fst1, &env->active_fpu.fp_status)))
+FOP_CONDN_S(sne,  (float32_lt(fst1, fst0, &env->active_fpu.fp_status)
+                   || float32_lt(fst0, fst1, &env->active_fpu.fp_status)))
diff --git a/target-mips/translate.c b/target-mips/translate.c
index b6713b2568..6f57171912 100644
--- a/target-mips/translate.c
+++ b/target-mips/translate.c
@@ -1622,6 +1622,98 @@ FOP_CONDS(abs, 1, s, FMT_S, 32)
 FOP_CONDS(, 0, ps, FMT_PS, 64)
 FOP_CONDS(abs, 1, ps, FMT_PS, 64)
 #undef FOP_CONDS
+
+#define FOP_CONDNS(fmt, ifmt, bits, STORE)                              \
+static inline void gen_r6_cmp_ ## fmt(DisasContext * ctx, int n,        \
+                                      int ft, int fs, int fd)           \
+{                                                                       \
+    TCGv_i ## bits fp0 = tcg_temp_new_i ## bits();                      \
+    TCGv_i ## bits fp1 = tcg_temp_new_i ## bits();                      \
+    switch (ifmt) {                                                     \
+    case FMT_D:                                                         \
+        check_cp1_registers(ctx, fs | ft | fd);                         \
+        break;                                                          \
+    }                                                                   \
+    gen_ldcmp_fpr ## bits(ctx, fp0, fs);                                \
+    gen_ldcmp_fpr ## bits(ctx, fp1, ft);                                \
+    switch (n) {                                                        \
+    case  0:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _af(fp0, cpu_env, fp0, fp1);       \
+        break;                                                          \
+    case  1:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _un(fp0, cpu_env, fp0, fp1);       \
+        break;                                                          \
+    case  2:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _eq(fp0, cpu_env, fp0, fp1);       \
+        break;                                                          \
+    case  3:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _ueq(fp0, cpu_env, fp0, fp1);      \
+        break;                                                          \
+    case  4:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _lt(fp0, cpu_env, fp0, fp1);       \
+        break;                                                          \
+    case  5:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _ult(fp0, cpu_env, fp0, fp1);      \
+        break;                                                          \
+    case  6:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _le(fp0, cpu_env, fp0, fp1);       \
+        break;                                                          \
+    case  7:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _ule(fp0, cpu_env, fp0, fp1);      \
+        break;                                                          \
+    case  8:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _saf(fp0, cpu_env, fp0, fp1);      \
+        break;                                                          \
+    case  9:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _sun(fp0, cpu_env, fp0, fp1);      \
+        break;                                                          \
+    case 10:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _seq(fp0, cpu_env, fp0, fp1);      \
+        break;                                                          \
+    case 11:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _sueq(fp0, cpu_env, fp0, fp1);     \
+        break;                                                          \
+    case 12:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _slt(fp0, cpu_env, fp0, fp1);      \
+        break;                                                          \
+    case 13:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _sult(fp0, cpu_env, fp0, fp1);     \
+        break;                                                          \
+    case 14:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _sle(fp0, cpu_env, fp0, fp1);      \
+        break;                                                          \
+    case 15:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _sule(fp0, cpu_env, fp0, fp1);     \
+        break;                                                          \
+    case 17:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _or(fp0, cpu_env, fp0, fp1);       \
+        break;                                                          \
+    case 18:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _une(fp0, cpu_env, fp0, fp1);      \
+        break;                                                          \
+    case 19:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _ne(fp0, cpu_env, fp0, fp1);       \
+        break;                                                          \
+    case 25:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _sor(fp0, cpu_env, fp0, fp1);      \
+        break;                                                          \
+    case 26:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _sune(fp0, cpu_env, fp0, fp1);     \
+        break;                                                          \
+    case 27:                                                            \
+        gen_helper_r6_cmp_ ## fmt ## _sne(fp0, cpu_env, fp0, fp1);      \
+        break;                                                          \
+    default:                                                            \
+        abort();                                                        \
+    }                                                                   \
+    STORE;                                                              \
+    tcg_temp_free_i ## bits (fp0);                                      \
+    tcg_temp_free_i ## bits (fp1);                                      \
+}
+
+FOP_CONDNS(d, FMT_D, 64, gen_store_fpr64(ctx, fp0, fd))
+FOP_CONDNS(s, FMT_S, 32, gen_store_fpr32(fp0, fd))
+#undef FOP_CONDNS
 #undef gen_ldcmp_fpr32
 #undef gen_ldcmp_fpr64
 
@@ -7792,6 +7884,53 @@ enum fopcode {
     OPC_CMP_NGT_PS = FOP (63, FMT_PS),
 };
 
+enum r6_f_cmp_op {
+    R6_OPC_CMP_AF_S   = FOP(0, FMT_W),
+    R6_OPC_CMP_UN_S   = FOP(1, FMT_W),
+    R6_OPC_CMP_EQ_S   = FOP(2, FMT_W),
+    R6_OPC_CMP_UEQ_S  = FOP(3, FMT_W),
+    R6_OPC_CMP_LT_S   = FOP(4, FMT_W),
+    R6_OPC_CMP_ULT_S  = FOP(5, FMT_W),
+    R6_OPC_CMP_LE_S   = FOP(6, FMT_W),
+    R6_OPC_CMP_ULE_S  = FOP(7, FMT_W),
+    R6_OPC_CMP_SAF_S  = FOP(8, FMT_W),
+    R6_OPC_CMP_SUN_S  = FOP(9, FMT_W),
+    R6_OPC_CMP_SEQ_S  = FOP(10, FMT_W),
+    R6_OPC_CMP_SEUQ_S = FOP(11, FMT_W),
+    R6_OPC_CMP_SLT_S  = FOP(12, FMT_W),
+    R6_OPC_CMP_SULT_S = FOP(13, FMT_W),
+    R6_OPC_CMP_SLE_S  = FOP(14, FMT_W),
+    R6_OPC_CMP_SULE_S = FOP(15, FMT_W),
+    R6_OPC_CMP_OR_S   = FOP(17, FMT_W),
+    R6_OPC_CMP_UNE_S  = FOP(18, FMT_W),
+    R6_OPC_CMP_NE_S   = FOP(19, FMT_W),
+    R6_OPC_CMP_SOR_S  = FOP(25, FMT_W),
+    R6_OPC_CMP_SUNE_S = FOP(26, FMT_W),
+    R6_OPC_CMP_SNE_S  = FOP(27, FMT_W),
+
+    R6_OPC_CMP_AF_D   = FOP(0, FMT_L),
+    R6_OPC_CMP_UN_D   = FOP(1, FMT_L),
+    R6_OPC_CMP_EQ_D   = FOP(2, FMT_L),
+    R6_OPC_CMP_UEQ_D  = FOP(3, FMT_L),
+    R6_OPC_CMP_LT_D   = FOP(4, FMT_L),
+    R6_OPC_CMP_ULT_D  = FOP(5, FMT_L),
+    R6_OPC_CMP_LE_D   = FOP(6, FMT_L),
+    R6_OPC_CMP_ULE_D  = FOP(7, FMT_L),
+    R6_OPC_CMP_SAF_D  = FOP(8, FMT_L),
+    R6_OPC_CMP_SUN_D  = FOP(9, FMT_L),
+    R6_OPC_CMP_SEQ_D  = FOP(10, FMT_L),
+    R6_OPC_CMP_SEUQ_D = FOP(11, FMT_L),
+    R6_OPC_CMP_SLT_D  = FOP(12, FMT_L),
+    R6_OPC_CMP_SULT_D = FOP(13, FMT_L),
+    R6_OPC_CMP_SLE_D  = FOP(14, FMT_L),
+    R6_OPC_CMP_SULE_D = FOP(15, FMT_L),
+    R6_OPC_CMP_OR_D   = FOP(17, FMT_L),
+    R6_OPC_CMP_UNE_D  = FOP(18, FMT_L),
+    R6_OPC_CMP_NE_D   = FOP(19, FMT_L),
+    R6_OPC_CMP_SOR_D  = FOP(25, FMT_L),
+    R6_OPC_CMP_SUNE_D = FOP(26, FMT_L),
+    R6_OPC_CMP_SNE_D  = FOP(27, FMT_L),
+};
 static void gen_cp1 (DisasContext *ctx, uint32_t opc, int rt, int fs)
 {
     const char *opn = "cp1 move";
@@ -17069,11 +17208,74 @@ static void decode_opc (CPUMIPSState *env, DisasContext *ctx)
                 check_insn_opc_removed(ctx, ISA_MIPS32R6);
             case OPC_S_FMT:
             case OPC_D_FMT:
-            case OPC_W_FMT:
-            case OPC_L_FMT:
                 gen_farith(ctx, ctx->opcode & FOP(0x3f, 0x1f), rt, rd, sa,
                            (imm >> 8) & 0x7);
                 break;
+            case OPC_W_FMT:
+            case OPC_L_FMT:
+            {
+                int r6_op = ctx->opcode & FOP(0x3f, 0x1f);
+                if (ctx->insn_flags & ISA_MIPS32R6) {
+                    switch (r6_op) {
+                    case R6_OPC_CMP_AF_S:
+                    case R6_OPC_CMP_UN_S:
+                    case R6_OPC_CMP_EQ_S:
+                    case R6_OPC_CMP_UEQ_S:
+                    case R6_OPC_CMP_LT_S:
+                    case R6_OPC_CMP_ULT_S:
+                    case R6_OPC_CMP_LE_S:
+                    case R6_OPC_CMP_ULE_S:
+                    case R6_OPC_CMP_SAF_S:
+                    case R6_OPC_CMP_SUN_S:
+                    case R6_OPC_CMP_SEQ_S:
+                    case R6_OPC_CMP_SEUQ_S:
+                    case R6_OPC_CMP_SLT_S:
+                    case R6_OPC_CMP_SULT_S:
+                    case R6_OPC_CMP_SLE_S:
+                    case R6_OPC_CMP_SULE_S:
+                    case R6_OPC_CMP_OR_S:
+                    case R6_OPC_CMP_UNE_S:
+                    case R6_OPC_CMP_NE_S:
+                    case R6_OPC_CMP_SOR_S:
+                    case R6_OPC_CMP_SUNE_S:
+                    case R6_OPC_CMP_SNE_S:
+                        gen_r6_cmp_s(ctx, ctx->opcode & 0x1f, rt, rd, sa);
+                        break;
+                    case R6_OPC_CMP_AF_D:
+                    case R6_OPC_CMP_UN_D:
+                    case R6_OPC_CMP_EQ_D:
+                    case R6_OPC_CMP_UEQ_D:
+                    case R6_OPC_CMP_LT_D:
+                    case R6_OPC_CMP_ULT_D:
+                    case R6_OPC_CMP_LE_D:
+                    case R6_OPC_CMP_ULE_D:
+                    case R6_OPC_CMP_SAF_D:
+                    case R6_OPC_CMP_SUN_D:
+                    case R6_OPC_CMP_SEQ_D:
+                    case R6_OPC_CMP_SEUQ_D:
+                    case R6_OPC_CMP_SLT_D:
+                    case R6_OPC_CMP_SULT_D:
+                    case R6_OPC_CMP_SLE_D:
+                    case R6_OPC_CMP_SULE_D:
+                    case R6_OPC_CMP_OR_D:
+                    case R6_OPC_CMP_UNE_D:
+                    case R6_OPC_CMP_NE_D:
+                    case R6_OPC_CMP_SOR_D:
+                    case R6_OPC_CMP_SUNE_D:
+                    case R6_OPC_CMP_SNE_D:
+                        gen_r6_cmp_d(ctx, ctx->opcode & 0x1f, rt, rd, sa);
+                        break;
+                    default:
+                        gen_farith(ctx, ctx->opcode & FOP(0x3f, 0x1f), rt, rd, sa,
+                                                       (imm >> 8) & 0x7);
+                        break;
+                    }
+                } else {
+                    gen_farith(ctx, ctx->opcode & FOP(0x3f, 0x1f), rt, rd, sa,
+                               (imm >> 8) & 0x7);
+                }
+                break;
+            }
             default:
                 MIPS_INVAL("cp1");
                 generate_exception (ctx, EXCP_RI);