[PULL 10/18] target/mips: msa: Split helpers for DOTP_U.<H|W|D>

[Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>]

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: <20200613152133.8964-9-aleksandar.qemu.devel@gmail.com>
---
 target/mips/helper.h     |  4 ++-
 target/mips/msa_helper.c | 65 +++++++++++++++++++++++++++++++++++++++---------
 target/mips/translate.c  | 12 ++++++++-
 3 files changed, 67 insertions(+), 14 deletions(-)

diff --git a/target/mips/helper.h b/target/mips/helper.h
index 06df3de..05d5533 100644
--- a/target/mips/helper.h
+++ b/target/mips/helper.h
@@ -1083,7 +1083,9 @@ DEF_HELPER_5(msa_mulv_df, void, env, i32, i32, i32, i32)
 DEF_HELPER_4(msa_dotp_s_h, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_dotp_s_w, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_dotp_s_d, void, env, i32, i32, i32)
-DEF_HELPER_5(msa_dotp_u_df, void, env, i32, i32, i32, i32)
+DEF_HELPER_4(msa_dotp_u_h, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_dotp_u_w, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_dotp_u_d, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_dpadd_s_h, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_dpadd_s_w, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_dpadd_s_d, void, env, i32, i32, i32)
diff --git a/target/mips/msa_helper.c b/target/mips/msa_helper.c
index 201283f..84d0073 100644
--- a/target/mips/msa_helper.c
+++ b/target/mips/msa_helper.c
@@ -2290,6 +2290,59 @@ void helper_msa_dotp_s_d(CPUMIPSState *env,
 }
 
 
+static inline int64_t msa_dotp_u_df(uint32_t df, int64_t arg1, int64_t arg2)
+{
+    int64_t even_arg1;
+    int64_t even_arg2;
+    int64_t odd_arg1;
+    int64_t odd_arg2;
+    UNSIGNED_EXTRACT(even_arg1, odd_arg1, arg1, df);
+    UNSIGNED_EXTRACT(even_arg2, odd_arg2, arg2, df);
+    return (even_arg1 * even_arg2) + (odd_arg1 * odd_arg2);
+}
+
+void helper_msa_dotp_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_dotp_u_df(DF_HALF, pws->h[0],  pwt->h[0]);
+    pwd->h[1]  = msa_dotp_u_df(DF_HALF, pws->h[1],  pwt->h[1]);
+    pwd->h[2]  = msa_dotp_u_df(DF_HALF, pws->h[2],  pwt->h[2]);
+    pwd->h[3]  = msa_dotp_u_df(DF_HALF, pws->h[3],  pwt->h[3]);
+    pwd->h[4]  = msa_dotp_u_df(DF_HALF, pws->h[4],  pwt->h[4]);
+    pwd->h[5]  = msa_dotp_u_df(DF_HALF, pws->h[5],  pwt->h[5]);
+    pwd->h[6]  = msa_dotp_u_df(DF_HALF, pws->h[6],  pwt->h[6]);
+    pwd->h[7]  = msa_dotp_u_df(DF_HALF, pws->h[7],  pwt->h[7]);
+}
+
+void helper_msa_dotp_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_dotp_u_df(DF_WORD, pws->w[0],  pwt->w[0]);
+    pwd->w[1]  = msa_dotp_u_df(DF_WORD, pws->w[1],  pwt->w[1]);
+    pwd->w[2]  = msa_dotp_u_df(DF_WORD, pws->w[2],  pwt->w[2]);
+    pwd->w[3]  = msa_dotp_u_df(DF_WORD, pws->w[3],  pwt->w[3]);
+}
+
+void helper_msa_dotp_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_dotp_u_df(DF_DOUBLE, pws->d[0],  pwt->d[0]);
+    pwd->d[1]  = msa_dotp_u_df(DF_DOUBLE, pws->d[1],  pwt->d[1]);
+}
+
+
 static inline int64_t msa_dpadd_s_df(uint32_t df, int64_t dest, int64_t arg1,
                                      int64_t arg2)
 {
@@ -5064,17 +5117,6 @@ static inline int64_t msa_mulv_df(uint32_t df, int64_t arg1, int64_t arg2)
     return arg1 * arg2;
 }
 
-static inline int64_t msa_dotp_u_df(uint32_t df, int64_t arg1, int64_t arg2)
-{
-    int64_t even_arg1;
-    int64_t even_arg2;
-    int64_t odd_arg1;
-    int64_t odd_arg2;
-    UNSIGNED_EXTRACT(even_arg1, odd_arg1, arg1, df);
-    UNSIGNED_EXTRACT(even_arg2, odd_arg2, arg2, df);
-    return (even_arg1 * even_arg2) + (odd_arg1 * odd_arg2);
-}
-
 #define CONCATENATE_AND_SLIDE(s, k)             \
     do {                                        \
         for (i = 0; i < s; i++) {               \
@@ -5198,7 +5240,6 @@ MSA_BINOP_DF(subs_u)
 MSA_BINOP_DF(subsus_u)
 MSA_BINOP_DF(subsuu_s)
 MSA_BINOP_DF(mulv)
-MSA_BINOP_DF(dotp_u)
 
 MSA_BINOP_DF(mul_q)
 MSA_BINOP_DF(mulr_q)
diff --git a/target/mips/translate.c b/target/mips/translate.c
index e150454..77e2d95 100644
--- a/target/mips/translate.c
+++ b/target/mips/translate.c
@@ -29407,7 +29407,17 @@ static void gen_msa_3r(CPUMIPSState *env, DisasContext *ctx)
             }
             break;
         case OPC_DOTP_U_df:
-            gen_helper_msa_dotp_u_df(cpu_env, tdf, twd, tws, twt);
+            switch (df) {
+            case DF_HALF:
+                gen_helper_msa_dotp_u_h(cpu_env, twd, tws, twt);
+                break;
+            case DF_WORD:
+                gen_helper_msa_dotp_u_w(cpu_env, twd, tws, twt);
+                break;
+            case DF_DOUBLE:
+                gen_helper_msa_dotp_u_d(cpu_env, twd, tws, twt);
+                break;
+            }
             break;
         case OPC_DPADD_S_df:
             switch (df) {
-- 
2.7.4