Use PMCjd/fast_mix

2 files changed, 28 insertions, 29 deletions

diff --git a/main.c b/main.c
index cab65b3..4c4d8bb 100644
--- a/main.c
+++ b/main.c
@@ -9,6 +9,7 @@
 #include <linux/slab.h>
 #include <linux/sort.h>
 #include <linux/random.h>
+#include <linux/perf_event.h>
 #include <asm/cpufeature.h>
 #include <asm/processor.h>
 #include <asm/fpu/api.h>
@@ -23,11 +24,12 @@ module_param(stamp, ulong, 0);
 
 #define do_it(name) do { \
 	u32 eax = 0, ebx = 0, ecx = 0, edx = 0; \
-	for (i = 0; i < WARMUP; ++i) \
+	for (i = 0; i < TRIALS; ++i) \
+		asm volatile("cpuid" : "+a" (eax), "=b" (ebx), "=d" (edx), "+c" (ecx)); \
+	for (i = 0; i < TRIALS; ++i) \
 		ret |= mix_ ##name(pool, input); \
-	asm volatile("cpuid" : "+a" (eax), "=b" (ebx), "=d" (edx), "+c" (ecx)); \
 	for (i = 0; i <= TRIALS; ++i) { \
-		trial_times[i] = get_cycles(); \
+		trial_times[i] = native_read_pmc(pmc_index); \
 		ret |= mix_ ##name(pool, input); \
 	} \
 	for (i = 0; i < TRIALS; ++i) \
@@ -61,23 +63,44 @@ static int compare_cycles(const void *a, const void *b)
 
 static int __init mod_init(void)
 {
-	enum { WARMUP = 6000, TRIALS = 100000, IDLE = 1 * 1000 };
+	enum { TRIALS = 300000 };
 	int ret = 0, i;
 	cycles_t *trial_times;
 	forall(local_it)
 	unsigned long flags;
+	struct perf_event *cycles_event;
+	int pmc_index;
+	static struct perf_event_attr perf_cycles_attr = {
+		.type = PERF_TYPE_HARDWARE,
+		.config = PERF_COUNT_HW_CPU_CYCLES,
+		.size = sizeof(struct perf_event_attr),
+		.pinned = 1,
+		.disabled = 0,
+		.exclude_user = 1
+	};
 	DEFINE_SPINLOCK(lock);
 	u32 pool[4] = { 0 }, input[4];
 
+
 	get_random_bytes(input, sizeof(input));
 
 	trial_times = kcalloc(TRIALS + 1, sizeof(cycles_t), GFP_KERNEL);
 	if (!trial_times)
 		return -ENOMEM;
 
-	msleep(IDLE);
+	msleep(1000);
 	spin_lock_irqsave(&lock, flags);
+	cycles_event = perf_event_create_kernel_counter(&perf_cycles_attr, raw_smp_processor_id(), NULL, NULL, NULL);
+	if (IS_ERR(cycles_event)) {
+		pr_err("unable to create perf counter: %ld\n", PTR_ERR(cycles_event));
+		goto skip;
+	}
+	pmc_index = cycles_event->hw.event_base_rdpmc;
+
 	forall(do_it)
+
+	perf_event_release_kernel(cycles_event);
+skip:
 	spin_unlock_irqrestore(&lock, flags);
 	forall(report_it)
 	kfree(trial_times);
diff --git a/run.sh b/run.sh
index 39c4719..7e94e4c 100755
--- a/run.sh
+++ b/run.sh
@@ -1,32 +1,8 @@
 #!/bin/bash
 set -e
 
-nob_cpus() {
-	echo "[+] Setting non-boot CPUs to status $1"
-	for i in /sys/devices/system/cpu/*/online; do
-		echo "$1" > "$i"
-	done
-}
-
-noturbo() {
-	echo "[+] Setting no-turbo to status $1"
-	if [[ -e /sys/devices/system/cpu/intel_pstate/no_turbo ]]; then
-		echo "$1" > /sys/devices/system/cpu/intel_pstate/no_turbo
-	else
-		local val
-		[[ $1 == 0 ]] && val=0x850089
-		[[ $1 == 1 ]] && val=0x4000850089
-		[[ -n $val ]] || return 0
-		wrmsr -a 0x1a0 $val
-	fi
-}
-
 [[ -e kbench9000.ko ]]
 
-trap "nob_cpus 1; noturbo 0;" INT TERM EXIT
-noturbo 1
-nob_cpus 0
-
 echo "[+] Inserting module to run tests"
 stamp="$(date +%s)"
 insmod kbench9000.ko stamp="$stamp"