target/mips: msa: Split helpers for AVER_<S|U>.<B|H|W|D>

Achieves clearer code and slightly better performance.

Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Aleksandar Rikalo <arikalo@wavecomp.com>
Message-Id: <1569415572-19635-14-git-send-email-aleksandar.markovic@rt-rk.com>

3 files changed, 188 insertions, 24 deletions

diff --git a/target/mips/helper.h b/target/mips/helper.h
index 455dd25..9d4c9f1 100644
--- a/target/mips/helper.h
+++ b/target/mips/helper.h
@@ -832,6 +832,16 @@ DEF_HELPER_4(msa_ave_u_h, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_ave_u_w, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_ave_u_d, void, env, i32, i32, i32)
 
+DEF_HELPER_4(msa_aver_s_b, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_aver_s_h, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_aver_s_w, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_aver_s_d, void, env, i32, i32, i32)
+
+DEF_HELPER_4(msa_aver_u_b, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_aver_u_h, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_aver_u_w, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_aver_u_d, void, env, i32, i32, i32)
+
 
 DEF_HELPER_4(msa_andi_b, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_ori_b, void, env, i32, i32, i32)
@@ -890,8 +900,6 @@ DEF_HELPER_5(msa_add_a_df, void, env, i32, i32, i32, i32)
 DEF_HELPER_5(msa_adds_a_df, void, env, i32, i32, i32, i32)
 DEF_HELPER_5(msa_adds_s_df, void, env, i32, i32, i32, i32)
 DEF_HELPER_5(msa_adds_u_df, void, env, i32, i32, i32, i32)
-DEF_HELPER_5(msa_aver_s_df, void, env, i32, i32, i32, i32)
-DEF_HELPER_5(msa_aver_u_df, void, env, i32, i32, i32, i32)
 DEF_HELPER_5(msa_subs_s_df, void, env, i32, i32, i32, i32)
 DEF_HELPER_5(msa_subs_u_df, void, env, i32, i32, i32, i32)
 DEF_HELPER_5(msa_subsus_u_df, void, env, i32, i32, i32, i32)
diff --git a/target/mips/msa_helper.c b/target/mips/msa_helper.c
index 1a7f90c..17443b9 100644
--- a/target/mips/msa_helper.c
+++ b/target/mips/msa_helper.c
@@ -978,6 +978,152 @@ void helper_msa_ave_u_d(CPUMIPSState *env,
     pwd->d[1]  = msa_ave_u_df(DF_DOUBLE, pws->d[1],  pwt->d[1]);
 }
 
+static inline int64_t msa_aver_s_df(uint32_t df, int64_t arg1, int64_t arg2)
+{
+    /* signed shift */
+    return (arg1 >> 1) + (arg2 >> 1) + ((arg1 | arg2) & 1);
+}
+
+void helper_msa_aver_s_b(CPUMIPSState *env,
+                         uint32_t wd, uint32_t ws, uint32_t wt)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    wr_t *pws = &(env->active_fpu.fpr[ws].wr);
+    wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
+
+    pwd->b[0]  = msa_aver_s_df(DF_BYTE, pws->b[0],  pwt->b[0]);
+    pwd->b[1]  = msa_aver_s_df(DF_BYTE, pws->b[1],  pwt->b[1]);
+    pwd->b[2]  = msa_aver_s_df(DF_BYTE, pws->b[2],  pwt->b[2]);
+    pwd->b[3]  = msa_aver_s_df(DF_BYTE, pws->b[3],  pwt->b[3]);
+    pwd->b[4]  = msa_aver_s_df(DF_BYTE, pws->b[4],  pwt->b[4]);
+    pwd->b[5]  = msa_aver_s_df(DF_BYTE, pws->b[5],  pwt->b[5]);
+    pwd->b[6]  = msa_aver_s_df(DF_BYTE, pws->b[6],  pwt->b[6]);
+    pwd->b[7]  = msa_aver_s_df(DF_BYTE, pws->b[7],  pwt->b[7]);
+    pwd->b[8]  = msa_aver_s_df(DF_BYTE, pws->b[8],  pwt->b[8]);
+    pwd->b[9]  = msa_aver_s_df(DF_BYTE, pws->b[9],  pwt->b[9]);
+    pwd->b[10] = msa_aver_s_df(DF_BYTE, pws->b[10], pwt->b[10]);
+    pwd->b[11] = msa_aver_s_df(DF_BYTE, pws->b[11], pwt->b[11]);
+    pwd->b[12] = msa_aver_s_df(DF_BYTE, pws->b[12], pwt->b[12]);
+    pwd->b[13] = msa_aver_s_df(DF_BYTE, pws->b[13], pwt->b[13]);
+    pwd->b[14] = msa_aver_s_df(DF_BYTE, pws->b[14], pwt->b[14]);
+    pwd->b[15] = msa_aver_s_df(DF_BYTE, pws->b[15], pwt->b[15]);
+}
+
+void helper_msa_aver_s_h(CPUMIPSState *env,
+                         uint32_t wd, uint32_t ws, uint32_t wt)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    wr_t *pws = &(env->active_fpu.fpr[ws].wr);
+    wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
+
+    pwd->h[0]  = msa_aver_s_df(DF_HALF, pws->h[0],  pwt->h[0]);
+    pwd->h[1]  = msa_aver_s_df(DF_HALF, pws->h[1],  pwt->h[1]);
+    pwd->h[2]  = msa_aver_s_df(DF_HALF, pws->h[2],  pwt->h[2]);
+    pwd->h[3]  = msa_aver_s_df(DF_HALF, pws->h[3],  pwt->h[3]);
+    pwd->h[4]  = msa_aver_s_df(DF_HALF, pws->h[4],  pwt->h[4]);
+    pwd->h[5]  = msa_aver_s_df(DF_HALF, pws->h[5],  pwt->h[5]);
+    pwd->h[6]  = msa_aver_s_df(DF_HALF, pws->h[6],  pwt->h[6]);
+    pwd->h[7]  = msa_aver_s_df(DF_HALF, pws->h[7],  pwt->h[7]);
+}
+
+void helper_msa_aver_s_w(CPUMIPSState *env,
+                         uint32_t wd, uint32_t ws, uint32_t wt)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    wr_t *pws = &(env->active_fpu.fpr[ws].wr);
+    wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
+
+    pwd->w[0]  = msa_aver_s_df(DF_WORD, pws->w[0],  pwt->w[0]);
+    pwd->w[1]  = msa_aver_s_df(DF_WORD, pws->w[1],  pwt->w[1]);
+    pwd->w[2]  = msa_aver_s_df(DF_WORD, pws->w[2],  pwt->w[2]);
+    pwd->w[3]  = msa_aver_s_df(DF_WORD, pws->w[3],  pwt->w[3]);
+}
+
+void helper_msa_aver_s_d(CPUMIPSState *env,
+                         uint32_t wd, uint32_t ws, uint32_t wt)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    wr_t *pws = &(env->active_fpu.fpr[ws].wr);
+    wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
+
+    pwd->d[0]  = msa_aver_s_df(DF_DOUBLE, pws->d[0],  pwt->d[0]);
+    pwd->d[1]  = msa_aver_s_df(DF_DOUBLE, pws->d[1],  pwt->d[1]);
+}
+
+static inline uint64_t msa_aver_u_df(uint32_t df, uint64_t arg1, uint64_t arg2)
+{
+    uint64_t u_arg1 = UNSIGNED(arg1, df);
+    uint64_t u_arg2 = UNSIGNED(arg2, df);
+    /* unsigned shift */
+    return (u_arg1 >> 1) + (u_arg2 >> 1) + ((u_arg1 | u_arg2) & 1);
+}
+
+void helper_msa_aver_u_b(CPUMIPSState *env,
+                         uint32_t wd, uint32_t ws, uint32_t wt)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    wr_t *pws = &(env->active_fpu.fpr[ws].wr);
+    wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
+
+    pwd->b[0]  = msa_aver_u_df(DF_BYTE, pws->b[0],  pwt->b[0]);
+    pwd->b[1]  = msa_aver_u_df(DF_BYTE, pws->b[1],  pwt->b[1]);
+    pwd->b[2]  = msa_aver_u_df(DF_BYTE, pws->b[2],  pwt->b[2]);
+    pwd->b[3]  = msa_aver_u_df(DF_BYTE, pws->b[3],  pwt->b[3]);
+    pwd->b[4]  = msa_aver_u_df(DF_BYTE, pws->b[4],  pwt->b[4]);
+    pwd->b[5]  = msa_aver_u_df(DF_BYTE, pws->b[5],  pwt->b[5]);
+    pwd->b[6]  = msa_aver_u_df(DF_BYTE, pws->b[6],  pwt->b[6]);
+    pwd->b[7]  = msa_aver_u_df(DF_BYTE, pws->b[7],  pwt->b[7]);
+    pwd->b[8]  = msa_aver_u_df(DF_BYTE, pws->b[8],  pwt->b[8]);
+    pwd->b[9]  = msa_aver_u_df(DF_BYTE, pws->b[9],  pwt->b[9]);
+    pwd->b[10] = msa_aver_u_df(DF_BYTE, pws->b[10], pwt->b[10]);
+    pwd->b[11] = msa_aver_u_df(DF_BYTE, pws->b[11], pwt->b[11]);
+    pwd->b[12] = msa_aver_u_df(DF_BYTE, pws->b[12], pwt->b[12]);
+    pwd->b[13] = msa_aver_u_df(DF_BYTE, pws->b[13], pwt->b[13]);
+    pwd->b[14] = msa_aver_u_df(DF_BYTE, pws->b[14], pwt->b[14]);
+    pwd->b[15] = msa_aver_u_df(DF_BYTE, pws->b[15], pwt->b[15]);
+}
+
+void helper_msa_aver_u_h(CPUMIPSState *env,
+                         uint32_t wd, uint32_t ws, uint32_t wt)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    wr_t *pws = &(env->active_fpu.fpr[ws].wr);
+    wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
+
+    pwd->h[0]  = msa_aver_u_df(DF_HALF, pws->h[0],  pwt->h[0]);
+    pwd->h[1]  = msa_aver_u_df(DF_HALF, pws->h[1],  pwt->h[1]);
+    pwd->h[2]  = msa_aver_u_df(DF_HALF, pws->h[2],  pwt->h[2]);
+    pwd->h[3]  = msa_aver_u_df(DF_HALF, pws->h[3],  pwt->h[3]);
+    pwd->h[4]  = msa_aver_u_df(DF_HALF, pws->h[4],  pwt->h[4]);
+    pwd->h[5]  = msa_aver_u_df(DF_HALF, pws->h[5],  pwt->h[5]);
+    pwd->h[6]  = msa_aver_u_df(DF_HALF, pws->h[6],  pwt->h[6]);
+    pwd->h[7]  = msa_aver_u_df(DF_HALF, pws->h[7],  pwt->h[7]);
+}
+
+void helper_msa_aver_u_w(CPUMIPSState *env,
+                         uint32_t wd, uint32_t ws, uint32_t wt)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    wr_t *pws = &(env->active_fpu.fpr[ws].wr);
+    wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
+
+    pwd->w[0]  = msa_aver_u_df(DF_WORD, pws->w[0],  pwt->w[0]);
+    pwd->w[1]  = msa_aver_u_df(DF_WORD, pws->w[1],  pwt->w[1]);
+    pwd->w[2]  = msa_aver_u_df(DF_WORD, pws->w[2],  pwt->w[2]);
+    pwd->w[3]  = msa_aver_u_df(DF_WORD, pws->w[3],  pwt->w[3]);
+}
+
+void helper_msa_aver_u_d(CPUMIPSState *env,
+                         uint32_t wd, uint32_t ws, uint32_t wt)
+{
+    wr_t *pwd = &(env->active_fpu.fpr[wd].wr);
+    wr_t *pws = &(env->active_fpu.fpr[ws].wr);
+    wr_t *pwt = &(env->active_fpu.fpr[wt].wr);
+
+    pwd->d[0]  = msa_aver_u_df(DF_DOUBLE, pws->d[0],  pwt->d[0]);
+    pwd->d[1]  = msa_aver_u_df(DF_DOUBLE, pws->d[1],  pwt->d[1]);
+}
+
 
 /*
  * Int Compare
@@ -1744,20 +1890,6 @@ static inline uint64_t msa_adds_u_df(uint32_t df, uint64_t arg1, uint64_t arg2)
     return (u_arg1 < max_uint - u_arg2) ? u_arg1 + u_arg2 : max_uint;
 }
 
-static inline int64_t msa_aver_s_df(uint32_t df, int64_t arg1, int64_t arg2)
-{
-    /* signed shift */
-    return (arg1 >> 1) + (arg2 >> 1) + ((arg1 | arg2) & 1);
-}
-
-static inline uint64_t msa_aver_u_df(uint32_t df, uint64_t arg1, uint64_t arg2)
-{
-    uint64_t u_arg1 = UNSIGNED(arg1, df);
-    uint64_t u_arg2 = UNSIGNED(arg2, df);
-    /* unsigned shift */
-    return (u_arg1 >> 1) + (u_arg2 >> 1) + ((u_arg1 | u_arg2) & 1);
-}
-
 static inline int64_t msa_subs_s_df(uint32_t df, int64_t arg1, int64_t arg2)
 {
     int64_t max_int = DF_MAX_INT(df);
@@ -2065,8 +2197,6 @@ MSA_BINOP_DF(add_a)
 MSA_BINOP_DF(adds_a)
 MSA_BINOP_DF(adds_s)
 MSA_BINOP_DF(adds_u)
-MSA_BINOP_DF(aver_s)
-MSA_BINOP_DF(aver_u)
 MSA_BINOP_DF(subs_s)
 MSA_BINOP_DF(subs_u)
 MSA_BINOP_DF(subsus_u)
diff --git a/target/mips/translate.c b/target/mips/translate.c
index 2b0abbb..9b186d3 100644
--- a/target/mips/translate.c
+++ b/target/mips/translate.c
@@ -28498,6 +28498,38 @@ static void gen_msa_3r(CPUMIPSState *env, DisasContext *ctx)
             break;
         }
         break;
+    case OPC_AVER_S_df:
+        switch (df) {
+        case DF_BYTE:
+            gen_helper_msa_aver_s_b(cpu_env, twd, tws, twt);
+            break;
+        case DF_HALF:
+            gen_helper_msa_aver_s_h(cpu_env, twd, tws, twt);
+            break;
+        case DF_WORD:
+            gen_helper_msa_aver_s_w(cpu_env, twd, tws, twt);
+            break;
+        case DF_DOUBLE:
+            gen_helper_msa_aver_s_d(cpu_env, twd, tws, twt);
+            break;
+        }
+        break;
+    case OPC_AVER_U_df:
+        switch (df) {
+        case DF_BYTE:
+            gen_helper_msa_aver_u_b(cpu_env, twd, tws, twt);
+            break;
+        case DF_HALF:
+            gen_helper_msa_aver_u_h(cpu_env, twd, tws, twt);
+            break;
+        case DF_WORD:
+            gen_helper_msa_aver_u_w(cpu_env, twd, tws, twt);
+            break;
+        case DF_DOUBLE:
+            gen_helper_msa_aver_u_d(cpu_env, twd, tws, twt);
+            break;
+        }
+        break;
     case OPC_SLL_df:
         gen_helper_msa_sll_df(cpu_env, tdf, twd, tws, twt);
         break;
@@ -28615,9 +28647,6 @@ static void gen_msa_3r(CPUMIPSState *env, DisasContext *ctx)
     case OPC_MAX_A_df:
         gen_helper_msa_max_a_df(cpu_env, tdf, twd, tws, twt);
         break;
-    case OPC_AVER_S_df:
-        gen_helper_msa_aver_s_df(cpu_env, tdf, twd, tws, twt);
-        break;
     case OPC_MOD_S_df:
         gen_helper_msa_mod_s_df(cpu_env, tdf, twd, tws, twt);
         break;
@@ -28627,9 +28656,6 @@ static void gen_msa_3r(CPUMIPSState *env, DisasContext *ctx)
     case OPC_MIN_A_df:
         gen_helper_msa_min_a_df(cpu_env, tdf, twd, tws, twt);
         break;
-    case OPC_AVER_U_df:
-        gen_helper_msa_aver_u_df(cpu_env, tdf, twd, tws, twt);
-        break;
     case OPC_MOD_U_df:
         gen_helper_msa_mod_u_df(cpu_env, tdf, twd, tws, twt);
         break;