target/mips: msa: Unroll loops and demacro <BMNZ|BMZ|BSEL>.V

Achieves clearer code and slightly better performance.

Signed-off-by: Aleksandar Markovic <amarkovic@wavecomp.com>
Reviewed-by: Aleksandar Rikalo <arikalo@wavecomp.com>
Message-Id: <1569415572-19635-11-git-send-email-aleksandar.markovic@rt-rk.com>

diff --git a/target/mips/helper.h b/target/mips/helper.h
index 93494824cd23fb61780e18e1956f3560e836ed38..27544a13135ed43204053badbb06d9900ab68a47 100644 (file)
--- a/target/mips/helper.h
+++ b/target/mips/helper.h
@@ -803,6 +803,10 @@ DEF_HELPER_4(msa_binsr_h, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_binsr_w, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_binsr_d, void, env, i32, i32, i32)
 
+DEF_HELPER_4(msa_bmnz_v, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_bmz_v, void, env, i32, i32, i32)
+DEF_HELPER_4(msa_bsel_v, 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)
@@ -957,9 +961,6 @@ DEF_HELPER_4(msa_and_v, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_or_v, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_nor_v, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_xor_v, void, env, i32, i32, i32)
-DEF_HELPER_4(msa_bmnz_v, void, env, i32, i32, i32)
-DEF_HELPER_4(msa_bmz_v, void, env, i32, i32, i32)
-DEF_HELPER_4(msa_bsel_v, void, env, i32, i32, i32)
 DEF_HELPER_4(msa_fill_df, void, env, i32, i32, i32)
 
 DEF_HELPER_4(msa_copy_s_b, void, env, i32, i32, i32)

diff --git a/target/mips/msa_helper.c b/target/mips/msa_helper.c
index c9b0583664ebe3ef37efb15b6f1e4f90bf6fa2da..82c8871b14e5e095a744867bcbd5595bce797fbc 100644 (file)
--- a/target/mips/msa_helper.c
+++ b/target/mips/msa_helper.c
@@ -466,6 +466,42 @@ void helper_msa_binsr_d(CPUMIPSState *env,
     pwd->d[1]  = msa_binsr_df(DF_DOUBLE, pwd->d[1],  pws->d[1],  pwt->d[1]);
 }
 
+void helper_msa_bmnz_v(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] = UNSIGNED(                                                     \
+        ((pwd->d[0] & (~pwt->d[0])) | (pws->d[0] & pwt->d[0])), DF_DOUBLE);
+    pwd->d[1] = UNSIGNED(                                                     \
+        ((pwd->d[1] & (~pwt->d[1])) | (pws->d[1] & pwt->d[1])), DF_DOUBLE);
+}
+
+void helper_msa_bmz_v(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] = UNSIGNED(                                                     \
+        ((pwd->d[0] & pwt->d[0]) | (pws->d[0] & (~pwt->d[0]))), DF_DOUBLE);
+    pwd->d[1] = UNSIGNED(                                                     \
+        ((pwd->d[1] & pwt->d[1]) | (pws->d[1] & (~pwt->d[1]))), DF_DOUBLE);
+}
+
+void helper_msa_bsel_v(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] = UNSIGNED(                                                     \
+        (pws->d[0] & (~pwd->d[0])) | (pwt->d[0] & pwd->d[0]), DF_DOUBLE);
+    pwd->d[1] = UNSIGNED(                                                     \
+        (pws->d[1] & (~pwd->d[1])) | (pwt->d[1] & pwd->d[1]), DF_DOUBLE);
+}
+
 
 /*
  * Bit Set
@@ -946,6 +982,9 @@ MSA_FN_IMM8(bmzi_b, pwd->b[i],
 MSA_FN_IMM8(bseli_b, pwd->b[i],
         BIT_SELECT(pwd->b[i], pws->b[i], i8, DF_BYTE))
 
+#undef BIT_SELECT
+#undef BIT_MOVE_IF_ZERO
+#undef BIT_MOVE_IF_NOT_ZERO
 #undef MSA_FN_IMM8
 
 #define SHF_POS(i, imm) (((i) & 0xfc) + (((imm) >> (2 * ((i) & 0x03))) & 0x03))
@@ -980,30 +1019,6 @@ void helper_msa_shf_df(CPUMIPSState *env, uint32_t df, uint32_t wd,
     msa_move_v(pwd, pwx);
 }
 
-#define MSA_FN_VECTOR(FUNC, DEST, OPERATION)                            \
-void helper_msa_ ## FUNC(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);                          \
-    uint32_t i;                                                         \
-    for (i = 0; i < DF_ELEMENTS(DF_DOUBLE); i++) {                      \
-        DEST = OPERATION;                                               \
-    }                                                                   \
-}
-
-MSA_FN_VECTOR(bmnz_v, pwd->d[i],
-        BIT_MOVE_IF_NOT_ZERO(pwd->d[i], pws->d[i], pwt->d[i], DF_DOUBLE))
-MSA_FN_VECTOR(bmz_v, pwd->d[i],
-        BIT_MOVE_IF_ZERO(pwd->d[i], pws->d[i], pwt->d[i], DF_DOUBLE))
-MSA_FN_VECTOR(bsel_v, pwd->d[i],
-        BIT_SELECT(pwd->d[i], pws->d[i], pwt->d[i], DF_DOUBLE))
-#undef BIT_MOVE_IF_NOT_ZERO
-#undef BIT_MOVE_IF_ZERO
-#undef BIT_SELECT
-#undef MSA_FN_VECTOR
-
 void helper_msa_and_v(CPUMIPSState *env, uint32_t wd, uint32_t ws, uint32_t wt)
 {
     wr_t *pwd = &(env->active_fpu.fpr[wd].wr);