optimize sort by inlining comparison functions

This makes special-cased forms such as sort { $b <=> $a }
even faster.

Also, since this commit removes PL_sort_RealCmp, it fixes the
issue with nested sort calls mentioned in gh Perl#16129

embed.fnc

@@ -2879,17 +2879,26 @@ i	|HV*	|opmethod_stash	|NN SV* meth
 #endif
 
 #if defined(PERL_IN_PP_SORT_C)
-S	|I32	|sv_ncmp	|NN SV *const a|NN SV *const b
-S	|I32	|sv_i_ncmp	|NN SV *const a|NN SV *const b
-S	|I32	|amagic_ncmp	|NN SV *const a|NN SV *const b
-S	|I32	|amagic_i_ncmp	|NN SV *const a|NN SV *const b
-S	|I32	|amagic_cmp	|NN SV *const str1|NN SV *const str2
+I	|I32	|sv_ncmp	|NN SV *const a|NN SV *const b
+I	|I32	|sv_ncmp_desc	|NN SV *const a|NN SV *const b
+I	|I32	|sv_i_ncmp	|NN SV *const a|NN SV *const b
+I	|I32	|sv_i_ncmp_desc	|NN SV *const a|NN SV *const b
+I	|I32	|amagic_ncmp	|NN SV *const a|NN SV *const b
+I	|I32	|amagic_ncmp_desc	|NN SV *const a|NN SV *const b
+I	|I32	|amagic_i_ncmp	|NN SV *const a|NN SV *const b
+I	|I32	|amagic_i_ncmp_desc	|NN SV *const a|NN SV *const b
+I	|I32	|amagic_cmp	|NN SV *const str1|NN SV *const str2
+I	|I32	|amagic_cmp_desc	|NN SV *const str1|NN SV *const str2
+I	|I32	|cmp_desc	|NN SV *const str1|NN SV *const str2
 #  ifdef USE_LOCALE_COLLATE
-S	|I32	|amagic_cmp_locale|NN SV *const str1|NN SV *const str2
+I	|I32	|amagic_cmp_locale     |NN SV *const str1|NN SV *const str2
+I	|I32	|amagic_cmp_locale_desc|NN SV *const str1|NN SV *const str2
+I	|I32	|cmp_locale_desc|NN SV *const str1|NN SV *const str2
 #  endif
 S	|I32	|sortcv		|NN SV *const a|NN SV *const b
 S	|I32	|sortcv_xsub	|NN SV *const a|NN SV *const b
 S	|I32	|sortcv_stacked	|NN SV *const a|NN SV *const b
+I	|void	|sortsv_flags_impl	|NULLOK SV** array|size_t num_elts|NN SVCOMPARE_t cmp|U32 flags
 #endif
 
 #if defined(PERL_IN_PP_SYS_C)

embed.h

@@ -1842,15 +1842,24 @@
 #  endif
 #  if defined(PERL_IN_PP_SORT_C)
 #define amagic_cmp(a,b)		S_amagic_cmp(aTHX_ a,b)
+#define amagic_cmp_desc(a,b)	S_amagic_cmp_desc(aTHX_ a,b)
 #define amagic_i_ncmp(a,b)	S_amagic_i_ncmp(aTHX_ a,b)
+#define amagic_i_ncmp_desc(a,b)	S_amagic_i_ncmp_desc(aTHX_ a,b)
 #define amagic_ncmp(a,b)	S_amagic_ncmp(aTHX_ a,b)
+#define amagic_ncmp_desc(a,b)	S_amagic_ncmp_desc(aTHX_ a,b)
+#define cmp_desc(a,b)		S_cmp_desc(aTHX_ a,b)
 #define sortcv(a,b)		S_sortcv(aTHX_ a,b)
 #define sortcv_stacked(a,b)	S_sortcv_stacked(aTHX_ a,b)
 #define sortcv_xsub(a,b)	S_sortcv_xsub(aTHX_ a,b)
+#define sortsv_flags_impl(a,b,c,d)	S_sortsv_flags_impl(aTHX_ a,b,c,d)
 #define sv_i_ncmp(a,b)		S_sv_i_ncmp(aTHX_ a,b)
+#define sv_i_ncmp_desc(a,b)	S_sv_i_ncmp_desc(aTHX_ a,b)
 #define sv_ncmp(a,b)		S_sv_ncmp(aTHX_ a,b)
+#define sv_ncmp_desc(a,b)	S_sv_ncmp_desc(aTHX_ a,b)
 #    if defined(USE_LOCALE_COLLATE)
 #define amagic_cmp_locale(a,b)	S_amagic_cmp_locale(aTHX_ a,b)
+#define amagic_cmp_locale_desc(a,b)	S_amagic_cmp_locale_desc(aTHX_ a,b)
+#define cmp_locale_desc(a,b)	S_cmp_locale_desc(aTHX_ a,b)
 #    endif
 #  endif
 #  if defined(PERL_IN_PP_SYS_C)

embedvar.h

@@ -296,7 +296,6 @@
 #define PL_sighandlerp		(vTHX->Isighandlerp)
 #define PL_signalhook		(vTHX->Isignalhook)
 #define PL_signals		(vTHX->Isignals)
-#define PL_sort_RealCmp		(vTHX->Isort_RealCmp)
 #define PL_sortcop		(vTHX->Isortcop)
 #define PL_sortstash		(vTHX->Isortstash)
 #define PL_splitstr		(vTHX->Isplitstr)

intrpvar.h

@@ -793,8 +793,6 @@ PERLVARI(I, known_layers, PerlIO_list_t *, NULL)
 PERLVARI(I, def_layerlist, PerlIO_list_t *, NULL)
 #endif
 
-PERLVAR(I, sort_RealCmp, SVCOMPARE_t)
-
 PERLVARI(I, checkav_save, AV *, NULL)	/* save CHECK{}s when compiling */
 PERLVARI(I, unitcheckav_save, AV *, NULL)
 					/* save UNITCHECK{}s when compiling */