Have nclip_{wv/wx/wi} use different macros

This allows them to share PARAM macro with narrowing right-shift instructions.
Rename VV_NSHIFT_PARAMS -> VV_NARROW_PARAMS so nclip, nsra, nsrl can share it.
(Same goes to VX_NSHIFT_PARAMS and VI_NSHIFT_PARAMS)

riscv/decode.h

@@ -769,19 +769,19 @@ static inline bool is_aligned(const unsigned val, const unsigned pos)
   auto &vd = P.VU.elt<type_sew_t<x>::type>(rd_num, i, true); \
   auto vs2 = P.VU.elt<type_sew_t<x>::type>(rs2_num, i - offset);
 
-#define VI_NSHIFT_PARAMS(sew1, sew2) \
+#define VI_NARROW_PARAMS(sew1, sew2) \
   auto &vd = P.VU.elt<type_usew_t<sew1>::type>(rd_num, i, true); \
   auto vs2_u = P.VU.elt<type_usew_t<sew2>::type>(rs2_num, i); \
   auto vs2 = P.VU.elt<type_sew_t<sew2>::type>(rs2_num, i); \
   auto zimm5 = (type_usew_t<sew1>::type)insn.v_zimm5();
 
-#define VX_NSHIFT_PARAMS(sew1, sew2) \
+#define VX_NARROW_PARAMS(sew1, sew2) \
   auto &vd = P.VU.elt<type_usew_t<sew1>::type>(rd_num, i, true); \
   auto vs2_u = P.VU.elt<type_usew_t<sew2>::type>(rs2_num, i); \
   auto vs2 = P.VU.elt<type_sew_t<sew2>::type>(rs2_num, i); \
   auto rs1 = (type_sew_t<sew1>::type)RS1;
 
-#define VV_NSHIFT_PARAMS(sew1, sew2) \
+#define VV_NARROW_PARAMS(sew1, sew2) \
   auto &vd = P.VU.elt<type_usew_t<sew1>::type>(rd_num, i, true); \
   auto vs2_u = P.VU.elt<type_usew_t<sew2>::type>(rs2_num, i); \
   auto vs2 = P.VU.elt<type_sew_t<sew2>::type>(rs2_num, i); \
@@ -1114,39 +1114,46 @@ static inline bool is_aligned(const unsigned val, const unsigned pos)
 
 // narrow operation loop
 #define VI_VV_LOOP_NARROW(BODY) \
-VI_NARROW_CHECK_COMMON; \
+  VI_CHECK_SDS(true); \
-VI_LOOP_BASE \
-if (sew == e8){ \
-  VI_NARROW_SHIFT(e8, e16) \
-  BODY; \
-}else if(sew == e16){ \
-  VI_NARROW_SHIFT(e16, e32) \
-  BODY; \
-}else if(sew == e32){ \
-  VI_NARROW_SHIFT(e32, e64) \
-  BODY; \
-} \
-VI_LOOP_END 
-
-#define VI_NARROW_SHIFT(sew1, sew2) \
-  type_usew_t<sew1>::type &vd = P.VU.elt<type_usew_t<sew1>::type>(rd_num, i, true); \
-  type_usew_t<sew2>::type vs2_u = P.VU.elt<type_usew_t<sew2>::type>(rs2_num, i); \
-  type_usew_t<sew1>::type zimm5 = (type_usew_t<sew1>::type)insn.v_zimm5(); \
-  type_sew_t<sew2>::type vs2 = P.VU.elt<type_sew_t<sew2>::type>(rs2_num, i); \
-  type_sew_t<sew1>::type vs1 = P.VU.elt<type_sew_t<sew1>::type>(rs1_num, i); \
-  type_sew_t<sew1>::type rs1 = (type_sew_t<sew1>::type)RS1; 
-
-#define VI_VVXI_LOOP_NARROW(BODY, is_vs1) \
-  VI_CHECK_SDS(is_vs1); \
   VI_LOOP_BASE \
   if (sew == e8){ \
-    VI_NARROW_SHIFT(e8, e16) \
+    VV_NARROW_PARAMS(e8, e16) \
     BODY; \
-  } else if (sew == e16) { \
+  }else if(sew == e16){ \
-    VI_NARROW_SHIFT(e16, e32) \
+    VV_NARROW_PARAMS(e16, e32) \
     BODY; \
-  } else if (sew == e32) { \
+  }else if(sew == e32){ \
-    VI_NARROW_SHIFT(e32, e64) \
+    VV_NARROW_PARAMS(e32, e64) \
+    BODY; \
+  } \
+  VI_LOOP_END
+
+#define VI_VX_LOOP_NARROW(BODY) \
+  VI_CHECK_SDS(false); \
+  VI_LOOP_BASE \
+  if (sew == e8){ \
+    VX_NARROW_PARAMS(e8, e16) \
+    BODY; \
+  }else if(sew == e16){ \
+    VX_NARROW_PARAMS(e16, e32) \
+    BODY; \
+  }else if(sew == e32){ \
+    VX_NARROW_PARAMS(e32, e64) \
+    BODY; \
+  } \
+  VI_LOOP_END
+
+#define VI_VI_LOOP_NARROW(BODY) \
+  VI_CHECK_SDS(false); \
+  VI_LOOP_BASE \
+  if (sew == e8){ \
+    VI_NARROW_PARAMS(e8, e16) \
+    BODY; \
+  }else if(sew == e16){ \
+    VI_NARROW_PARAMS(e16, e32) \
+    BODY; \
+  }else if(sew == e32){ \
+    VI_NARROW_PARAMS(e32, e64) \
     BODY; \
   } \
   VI_LOOP_END
@@ -1155,13 +1162,13 @@ VI_LOOP_END
   VI_CHECK_SDS(false); \
   VI_LOOP_NSHIFT_BASE \
   if (sew == e8){ \
-    VI_NSHIFT_PARAMS(e8, e16) \
+    VI_NARROW_PARAMS(e8, e16) \
     BODY; \
   } else if (sew == e16) { \
-    VI_NSHIFT_PARAMS(e16, e32) \
+    VI_NARROW_PARAMS(e16, e32) \
     BODY; \
   } else if (sew == e32) { \
-    VI_NSHIFT_PARAMS(e32, e64) \
+    VI_NARROW_PARAMS(e32, e64) \
     BODY; \
   } \
   VI_LOOP_END
@@ -1170,13 +1177,13 @@ VI_LOOP_END
   VI_CHECK_SDS(false); \
   VI_LOOP_NSHIFT_BASE \
   if (sew == e8){ \
-    VX_NSHIFT_PARAMS(e8, e16) \
+    VX_NARROW_PARAMS(e8, e16) \
     BODY; \
   } else if (sew == e16) { \
-    VX_NSHIFT_PARAMS(e16, e32) \
+    VX_NARROW_PARAMS(e16, e32) \
     BODY; \
   } else if (sew == e32) { \
-    VX_NSHIFT_PARAMS(e32, e64) \
+    VX_NARROW_PARAMS(e32, e64) \
     BODY; \
   } \
   VI_LOOP_END
@@ -1185,13 +1192,13 @@ VI_LOOP_END
   VI_CHECK_SDS(true); \
   VI_LOOP_NSHIFT_BASE \
   if (sew == e8){ \
-    VV_NSHIFT_PARAMS(e8, e16) \
+    VV_NARROW_PARAMS(e8, e16) \
     BODY; \
   } else if (sew == e16) { \
-    VV_NSHIFT_PARAMS(e16, e32) \
+    VV_NARROW_PARAMS(e16, e32) \
     BODY; \
   } else if (sew == e32) { \
-    VV_NSHIFT_PARAMS(e32, e64) \
+    VV_NARROW_PARAMS(e32, e64) \
     BODY; \
   } \
   VI_LOOP_END

riscv/insns/vnclip_wi.h

@@ -2,7 +2,7 @@
 VRM xrm = P.VU.get_vround_mode();
 int64_t int_max = INT64_MAX >> (64 - P.VU.vsew);
 int64_t int_min = INT64_MIN >> (64 - P.VU.vsew);
-VI_VVXI_LOOP_NARROW
+VI_VI_LOOP_NARROW
 ({
   int128_t result = vs2;
   unsigned shift = zimm5 & ((sew * 2) - 1);
@@ -22,4 +22,4 @@ VI_VVXI_LOOP_NARROW
   }
 
   vd = result;
-}, false)
+})

riscv/insns/vnclip_wv.h

@@ -2,7 +2,7 @@
 VRM xrm = P.VU.get_vround_mode();
 int64_t int_max = INT64_MAX >> (64 - P.VU.vsew);
 int64_t int_min = INT64_MIN >> (64 - P.VU.vsew);
-VI_VVXI_LOOP_NARROW
+VI_VV_LOOP_NARROW
 ({
   int128_t result = vs2;
   unsigned shift = vs1 & ((sew * 2) - 1);
@@ -22,4 +22,4 @@ VI_VVXI_LOOP_NARROW
   }
 
   vd = result;
-}, true)
+})

riscv/insns/vnclip_wx.h

@@ -2,7 +2,7 @@
 VRM xrm = P.VU.get_vround_mode();
 int64_t int_max = INT64_MAX >> (64 - P.VU.vsew);
 int64_t int_min = INT64_MIN >> (64 - P.VU.vsew);
-VI_VVXI_LOOP_NARROW
+VI_VX_LOOP_NARROW
 ({
   int128_t result = vs2;
   unsigned shift = rs1 & ((sew * 2) - 1);
@@ -22,4 +22,4 @@ VI_VVXI_LOOP_NARROW
   }
 
   vd = result;
-}, false)
+})

riscv/insns/vnclipu_wi.h

@@ -2,7 +2,7 @@
 VRM xrm = P.VU.get_vround_mode();
 uint64_t uint_max = UINT64_MAX >> (64 - P.VU.vsew);
 uint64_t sign_mask = UINT64_MAX << P.VU.vsew;
-VI_VVXI_LOOP_NARROW
+VI_VI_LOOP_NARROW
 ({
   uint128_t result = vs2_u;
   unsigned shift = zimm5 & ((sew * 2) - 1);
@@ -20,4 +20,4 @@ VI_VVXI_LOOP_NARROW
   }
 
   vd = result;
-}, false)
+})

riscv/insns/vnclipu_wv.h

@@ -2,7 +2,7 @@
 VRM xrm = P.VU.get_vround_mode();
 uint64_t uint_max = UINT64_MAX >> (64 - P.VU.vsew);
 uint64_t sign_mask = UINT64_MAX << P.VU.vsew;
-VI_VVXI_LOOP_NARROW
+VI_VV_LOOP_NARROW
 ({
   uint128_t result = vs2_u;
   unsigned shift = vs1 & ((sew * 2) - 1);
@@ -19,4 +19,4 @@ VI_VVXI_LOOP_NARROW
   }
 
   vd = result;
-}, true)
+})

riscv/insns/vnclipu_wx.h

@@ -2,7 +2,7 @@
 VRM xrm = P.VU.get_vround_mode();
 uint64_t uint_max = UINT64_MAX >> (64 - P.VU.vsew);
 uint64_t sign_mask = UINT64_MAX << P.VU.vsew;
-VI_VVXI_LOOP_NARROW
+VI_VX_LOOP_NARROW
 ({
   uint128_t result = vs2_u;
   unsigned shift = rs1 & ((sew * 2) - 1);
@@ -19,4 +19,4 @@ VI_VVXI_LOOP_NARROW
   }
 
   vd = result;
-}, false)
+})