[PULL 13/18] target/mips: msa: Split helpers for SUBSUS_U.<B|H|W|D>

[Aleksandar Markovic]

Achieves clearer code and slightly better performance.

Reviewed-by: Aleksandar Rikalo <aleksandar.rikalo@syrmia.com>
Signed-off-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Message-Id: <20200613152133.8964-12-aleksandar.qemu.devel@gmail.com>
---
 target/mips/helper.h     |   6 ++-
 target/mips/msa_helper.c | 102 ++++++++++++++++++++++++++++++++++++++---------
 target/mips/translate.c  |  15 ++++++-
 3 files changed, 103 insertions(+), 20 deletions(-)

diff --git a/target/mips/helper.h b/target/mips/helper.h
index 61dc1ed..227ff76 100644
--- a/target/mips/helper.h
+++ b/target/mips/helper.h
@@ -988,6 +988,11 @@ DEF_HELPER_4(msa_subs_u_h, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_subs_u_w, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_subs_u_d, void, env, i32, i32, i32)
 
+DEF_HELPER_4(msa_subsus_u_b, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_subsus_u_h, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_subsus_u_w, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_subsus_u_d, void, env, i32, i32, i32)
+
 DEF_HELPER_4(msa_ilvev_b, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_ilvev_h, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_ilvev_w, void, env, i32, i32, i32)
@@ -1084,7 +1089,6 @@ DEF_HELPER_5(msa_srlri_df, void, env, i32, i32, i32, i32)
 DEF_HELPER_5(msa_binsl_df, void, env, i32, i32, i32, i32)
 DEF_HELPER_5(msa_binsr_df, void, env, i32, i32, i32, i32)
 DEF_HELPER_5(msa_subv_df, void, env, i32, i32, i32, i32)
-DEF_HELPER_5(msa_subsus_u_df, void, env, i32, i32, i32, i32)
 DEF_HELPER_5(msa_subsuu_s_df, void, env, i32, i32, i32, i32)
 DEF_HELPER_5(msa_mulv_df, void, env, i32, i32, i32, i32)
 
diff --git a/target/mips/msa_helper.c b/target/mips/msa_helper.c
index bce32ab..f7e5c01 100644
--- a/target/mips/msa_helper.c
+++ b/target/mips/msa_helper.c
@@ -3802,6 +3802,90 @@ void helper_msa_subs_u_d(CPUMIPSState *env,
 }
 
 
+static inline int64_t msa_subsus_u_df(uint32_t df, int64_t arg1, int64_t arg2)
+{
+    uint64_t u_arg1 = UNSIGNED(arg1, df);
+    uint64_t max_uint = DF_MAX_UINT(df);
+    if (arg2 >= 0) {
+        uint64_t u_arg2 = (uint64_t)arg2;
+        return (u_arg1 > u_arg2) ?
+            (int64_t)(u_arg1 - u_arg2) :
+            0;
+    } else {
+        uint64_t u_arg2 = (uint64_t)(-arg2);
+        return (u_arg1 < max_uint - u_arg2) ?
+            (int64_t)(u_arg1 + u_arg2) :
+            (int64_t)max_uint;
+    }
+}
+
+void helper_msa_subsus_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_subsus_u_df(DF_BYTE, pws->b[0],  pwt->b[0]);
+    pwd->b[1]  = msa_subsus_u_df(DF_BYTE, pws->b[1],  pwt->b[1]);
+    pwd->b[2]  = msa_subsus_u_df(DF_BYTE, pws->b[2],  pwt->b[2]);
+    pwd->b[3]  = msa_subsus_u_df(DF_BYTE, pws->b[3],  pwt->b[3]);
+    pwd->b[4]  = msa_subsus_u_df(DF_BYTE, pws->b[4],  pwt->b[4]);
+    pwd->b[5]  = msa_subsus_u_df(DF_BYTE, pws->b[5],  pwt->b[5]);
+    pwd->b[6]  = msa_subsus_u_df(DF_BYTE, pws->b[6],  pwt->b[6]);
+    pwd->b[7]  = msa_subsus_u_df(DF_BYTE, pws->b[7],  pwt->b[7]);
+    pwd->b[8]  = msa_subsus_u_df(DF_BYTE, pws->b[8],  pwt->b[8]);
+    pwd->b[9]  = msa_subsus_u_df(DF_BYTE, pws->b[9],  pwt->b[9]);
+    pwd->b[10] = msa_subsus_u_df(DF_BYTE, pws->b[10], pwt->b[10]);
+    pwd->b[11] = msa_subsus_u_df(DF_BYTE, pws->b[11], pwt->b[11]);
+    pwd->b[12] = msa_subsus_u_df(DF_BYTE, pws->b[12], pwt->b[12]);
+    pwd->b[13] = msa_subsus_u_df(DF_BYTE, pws->b[13], pwt->b[13]);
+    pwd->b[14] = msa_subsus_u_df(DF_BYTE, pws->b[14], pwt->b[14]);
+    pwd->b[15] = msa_subsus_u_df(DF_BYTE, pws->b[15], pwt->b[15]);
+}
+
+void helper_msa_subsus_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_subsus_u_df(DF_HALF, pws->h[0],  pwt->h[0]);
+    pwd->h[1]  = msa_subsus_u_df(DF_HALF, pws->h[1],  pwt->h[1]);
+    pwd->h[2]  = msa_subsus_u_df(DF_HALF, pws->h[2],  pwt->h[2]);
+    pwd->h[3]  = msa_subsus_u_df(DF_HALF, pws->h[3],  pwt->h[3]);
+    pwd->h[4]  = msa_subsus_u_df(DF_HALF, pws->h[4],  pwt->h[4]);
+    pwd->h[5]  = msa_subsus_u_df(DF_HALF, pws->h[5],  pwt->h[5]);
+    pwd->h[6]  = msa_subsus_u_df(DF_HALF, pws->h[6],  pwt->h[6]);
+    pwd->h[7]  = msa_subsus_u_df(DF_HALF, pws->h[7],  pwt->h[7]);
+}
+
+void helper_msa_subsus_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_subsus_u_df(DF_WORD, pws->w[0],  pwt->w[0]);
+    pwd->w[1]  = msa_subsus_u_df(DF_WORD, pws->w[1],  pwt->w[1]);
+    pwd->w[2]  = msa_subsus_u_df(DF_WORD, pws->w[2],  pwt->w[2]);
+    pwd->w[3]  = msa_subsus_u_df(DF_WORD, pws->w[3],  pwt->w[3]);
+}
+
+void helper_msa_subsus_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_subsus_u_df(DF_DOUBLE, pws->d[0],  pwt->d[0]);
+    pwd->d[1]  = msa_subsus_u_df(DF_DOUBLE, pws->d[1],  pwt->d[1]);
+}
+
+
 /*
  * Interleave
  * ----------
@@ -5212,23 +5296,6 @@ MSA_TEROP_IMMU_DF(binsli, binsl)
 MSA_TEROP_IMMU_DF(binsri, binsr)
 #undef MSA_TEROP_IMMU_DF
 
-static inline int64_t msa_subsus_u_df(uint32_t df, int64_t arg1, int64_t arg2)
-{
-    uint64_t u_arg1 = UNSIGNED(arg1, df);
-    uint64_t max_uint = DF_MAX_UINT(df);
-    if (arg2 >= 0) {
-        uint64_t u_arg2 = (uint64_t)arg2;
-        return (u_arg1 > u_arg2) ?
-            (int64_t)(u_arg1 - u_arg2) :
-            0;
-    } else {
-        uint64_t u_arg2 = (uint64_t)(-arg2);
-        return (u_arg1 < max_uint - u_arg2) ?
-            (int64_t)(u_arg1 + u_arg2) :
-            (int64_t)max_uint;
-    }
-}
-
 static inline int64_t msa_subsuu_s_df(uint32_t df, int64_t arg1, int64_t arg2)
 {
     uint64_t u_arg1 = UNSIGNED(arg1, df);
@@ -5369,7 +5436,6 @@ void helper_msa_ ## func ## _df(CPUMIPSState *env, 
uint32_t df,         \
 }
 
 MSA_BINOP_DF(subv)
-MSA_BINOP_DF(subsus_u)
 MSA_BINOP_DF(subsuu_s)
 MSA_BINOP_DF(mulv)
 
diff --git a/target/mips/translate.c b/target/mips/translate.c
index 7671a49..b273f4d 100644
--- a/target/mips/translate.c
+++ b/target/mips/translate.c
@@ -29346,7 +29346,20 @@ static void gen_msa_3r(CPUMIPSState *env, DisasContext 
*ctx)
         gen_helper_msa_splat_df(cpu_env, tdf, twd, tws, twt);
         break;
     case OPC_SUBSUS_U_df:
-        gen_helper_msa_subsus_u_df(cpu_env, tdf, twd, tws, twt);
+        switch (df) {
+        case DF_BYTE:
+            gen_helper_msa_subsus_u_b(cpu_env, twd, tws, twt);
+            break;
+        case DF_HALF:
+            gen_helper_msa_subsus_u_h(cpu_env, twd, tws, twt);
+            break;
+        case DF_WORD:
+            gen_helper_msa_subsus_u_w(cpu_env, twd, tws, twt);
+            break;
+        case DF_DOUBLE:
+            gen_helper_msa_subsus_u_d(cpu_env, twd, tws, twt);
+            break;
+        }
         break;
     case OPC_SUBSUU_S_df:
         gen_helper_msa_subsuu_s_df(cpu_env, tdf, twd, tws, twt);
-- 
2.7.4