internal/bytealg: improve compare on Power10/PPC64

Handle comparisons of 15 or less bytes more efficiently
with Power10 instructions when building with GOPPC64=power10.

name            old time/op  new time/op  delta
BytesCompare/1  2.53ns ± 0%  2.17ns ± 0%  -14.17%
BytesCompare/2  2.70ns ± 0%  2.17ns ± 0%  -19.77%
BytesCompare/4  2.59ns ± 0%  2.17ns ± 0%  -16.20%
BytesCompare/8  2.66ns ± 0%  2.17ns ± 0%  -18.63%

Change-Id: I6d7c6af0a58ea3e03acc3930c54b77f2ac1dfbd5
Reviewed-on: https://go-review.googlesource.com/c/go/+/522315
Reviewed-by: Joedian Reid <[email protected]>
Run-TryBot: Paul Murphy <[email protected]>
Reviewed-by: Lynn Boger <[email protected]>
Reviewed-by: Bryan Mills <[email protected]>
TryBot-Result: Gopher Robot <[email protected]>

Author: pmur

src/internal/bytealg/compare_ppc64x.s

@@ -274,7 +274,16 @@ lower:
 	RET
 
 	PCALIGN $16
-cmp8:	// 8 - 15B
+cmp8:	// 8 - 15B (0 - 15B if GOPPC64_power10)
+#ifdef GOPPC64_power10
+	SLD	$56,R9,R9
+	LXVLL	R5,R9,V3	// Load bytes starting from MSB to LSB, unused are zero filled.
+	LXVLL	R6,R9,V4
+	VCMPUQ	V3,V4,CR0	// Compare as a 128b integer.
+	SETB_CR0(R6)
+	ISEL	CR0EQ,R3,R6,R3	// If equal, length determines the return value.
+	RET
+#else
 	CMP	R9,$8
 	BLT	cmp4
 	ANDCC	$7,R9,R9
@@ -330,3 +339,4 @@ cmp0:
 	SETB_CR0(R6)
 	ISEL	CR0EQ,R3,R6,R3
 	RET
+#endif