kselftest/arm64: Add pidbench for floating point syscall cases

Since it's likely to be useful for performance work with SVE let's have a
pidbench that gives us some numbers for consideration. In order to ensure
that we test exactly the scenario we want this is written in assembly - if
system libraries use SVE this would stop us exercising the case where the
process has never used SVE.

We exercise three cases:

 - Never having used SVE.
 - Having used SVE once.
 - Using SVE after each syscall.

by spinning running getpid() for a fixed number of iterations with the
time measured using CNTVCT_EL0 reported on the console. This is obviously
a totally unrealistic benchmark which will show the extremes of any
performance variation but equally given the potential gotchas with use of
FP instructions by system libraries it's good to have some concrete code
shared to make it easier to compare notes on results.

Testing over multiple SVE vector lengths will need to be done with vlset
currently, the test could be extended to iterate over all of them if
desired.
Signed-off-by: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20211202165107.1075259-1-broonie@kernel.orgSigned-off-by: Catalin Marinas <catalin.marinas@arm.com>

tools/testing/selftests/arm64/fp/.gitignore

+fp-pidbench
 fpsimd-test
 rdvl-sve
 sve-probe-vls

tools/testing/selftests/arm64/fp/Makefile

@@ -2,13 +2,15 @@
 
 CFLAGS += -I../../../../../usr/include/
 TEST_GEN_PROGS := sve-ptrace sve-probe-vls vec-syscfg
-TEST_PROGS_EXTENDED := fpsimd-test fpsimd-stress \
+TEST_PROGS_EXTENDED := fp-pidbench fpsimd-test fpsimd-stress \
 	rdvl-sve \
 	sve-test sve-stress \
 	vlset
 
 all: $(TEST_GEN_PROGS) $(TEST_PROGS_EXTENDED)
 
+fp-pidbench: fp-pidbench.S asm-utils.o
+	$(CC) -nostdlib $^ -o $@
 fpsimd-test: fpsimd-test.o asm-utils.o
 	$(CC) -nostdlib $^ -o $@
 rdvl-sve: rdvl-sve.o rdvl.o

tools/testing/selftests/arm64/fp/fp-pidbench.S

+// SPDX-License-Identifier: GPL-2.0-only
+// Copyright (C) 2021 ARM Limited.
+// Original author: Mark Brown <broonie@kernel.org>
+//
+// Trivial syscall overhead benchmark.
+//
+// This is implemented in asm to ensure that we don't have any issues with
+// system libraries using instructions that disrupt the test.
+
+#include <asm/unistd.h>
+#include "assembler.h"
+
+.arch_extension sve
+
+.macro test_loop per_loop
+	mov	x10, x20
+	mov	x8, #__NR_getpid
+	mrs	x11, CNTVCT_EL0
+1:
+	\per_loop
+	svc	#0
+	sub	x10, x10, #1
+	cbnz	x10, 1b
+
+	mrs	x12, CNTVCT_EL0
+	sub	x0, x12, x11
+	bl	putdec
+	puts	"\n"
+.endm
+
+// Main program entry point
+.globl _start
+function _start
+_start:
+	puts	"Iterations per test: "
+	mov	x20, #10000
+	lsl	x20, x20, #8
+	mov	x0, x20
+	bl	putdec
+	puts	"\n"
+
+	// Test having never used SVE
+	puts	"No SVE: "
+	test_loop
+
+	// Check for SVE support - should use hwcap but that's hard in asm
+	mrs	x0, ID_AA64PFR0_EL1
+	ubfx	x0, x0, #32, #4
+	cbnz	x0, 1f
+	puts	"System does not support SVE\n"
+	b	out
+1:
+
+	// Execute a SVE instruction
+	puts	"SVE VL: "
+	rdvl	x0, #8
+	bl	putdec
+	puts	"\n"
+
+	puts	"SVE used once: "
+	test_loop
+
+	// Use SVE per syscall
+	puts	"SVE used per syscall: "
+	test_loop "rdvl x0, #8"
+
+	//  And we're done
+out:
+	mov	x0, #0
+	mov	x8, #__NR_exit
+	svc	#0