add vrev8.v byteswap

Author: camel-cdr

bench/byteswap.S

@@ -1,4 +1,20 @@
 #if HAS_RVV_1_0
+
+#if defined(MX) && __riscv_zvbb
+.global MX(byteswap32_rvv_vrev8_)
+MX(byteswap32_rvv_vrev8_):
+1:
+	vsetvli t0, a1, e32, MX(), ta, ma
+	vle32.v v0, (a0)
+	vrev8.v v8, v0
+	vse32.v v8, (a0)
+	sub a1, a1, t0
+	slli t1, t0, 2
+	add a0, a0, t1
+	bnez a1, 1b
+	ret
+#endif
+
 #if MX_N == 4 || MX_N == 2 || MX_N == 1
 
 # a0 = ptr, a1 = len

bench/byteswap.c

@@ -53,10 +53,18 @@ byteswap32_SWAR_rev8(uint32_t *ptr, size_t n)
 	f(rvv_m1_gatherei16s_m8)
 #endif
 
+#if __riscv_zvbb
+#define IMPLS_ZVBB(f) MX(f,rvv_vrev8)
+#else
+#define IMPLS_ZVBB(f)
+#endif
+
+
 #define IMPLS(f) \
 	f(scalar) \
 	f(scalar_autovec) \
 	REV8(f) \
+	IMPLS_ZVBB(f) \
 	IMPLS_RVV(f)
 
 typedef void Func(uint32_t *ptr, size_t n);