Optimized memset() implementation for POWER

This uses the dcbz instruction to clear cacheline at a time rather than
byte at a time. This means that even without high levels of optimization,
we *dramatically* improve boot performance with SKIBOOT_GCOV=1 and probably
ever so slightly speed things up for normal builds.

We currently just hard-code 128 as cacheline size as all CPUs that skiboot
currently boots on have 128 byte cachelines.

Signed-off-by: Stewart Smith <stewart@linux.vnet.ibm.com>

3 files changed, 55 insertions, 0 deletions

diff --git a/libc/string/memset.c b/libc/string/memset.c
index f8dfbf5..18b7619 100644
--- a/libc/string/memset.c
+++ b/libc/string/memset.c
@@ -12,11 +12,33 @@
 
 #include "string.h"
 
+#define CACHE_LINE_SIZE 128
+
 void *
 memset(void *dest, int c, size_t size)
 {
 	unsigned char *d = (unsigned char *)dest;
 
+#if defined(__powerpc__) || defined(__powerpc64__)
+	if (size > CACHE_LINE_SIZE && c==0) {
+		while ((unsigned long long)d % CACHE_LINE_SIZE) {
+			*d++ = (unsigned char)c;
+			size--;
+		}
+		while (size >= CACHE_LINE_SIZE) {
+			asm volatile ("dcbz 0,%0\n" : : "r"(d) : "memory");
+			d+= CACHE_LINE_SIZE;
+			size-= CACHE_LINE_SIZE;
+		}
+	}
+#endif
+
+	while (size >= 8 && c == 0) {
+		*((unsigned long long*)d) = 0ULL;
+		d+=8;
+		size-=8;
+	}
+
 	while (size-- > 0) {
 		*d++ = (unsigned char)c;
 	}
diff --git a/libc/test/run-memops-test.c b/libc/test/run-memops-test.c
index 0979994..429cfdc 100644
--- a/libc/test/run-memops-test.c
+++ b/libc/test/run-memops-test.c
@@ -41,3 +41,19 @@
 #include "../string/strncpy.c"
 #include "../string/strstr.c"
 #include "../string/strtok.c"
+#include <stdlib.h>
+
+int test_memset(char* buf, int c, size_t s);
+
+int test_memset(char* buf, int c, size_t s)
+{
+	int i;
+	int r= 0;
+
+	memset(buf, c, s);
+	for(i=0; i<s; i++)
+		if (buf[i] != c)
+			r = -1;
+
+	return r;
+}
diff --git a/libc/test/run-memops.c b/libc/test/run-memops.c
index 15ec9bd..80822df 100644
--- a/libc/test/run-memops.c
+++ b/libc/test/run-memops.c
@@ -21,7 +21,24 @@
 #include <string.h>
 #include <stdio.h>
 
+int test_memset(char* buf, int c, size_t s);
+
 int main(void)
 {
+	char* buf;
+
+	buf = malloc(100);
+	assert(test_memset(buf, 0x42, 100) == 0);
+	free(buf);
+
+	buf = malloc(128);
+	assert(test_memset(buf, 0, 128) == 0);
+	assert(test_memset(buf+1, 0, 127) == 0);
+	free(buf);
+
+	buf = malloc(1024);
+	assert(test_memset(buf, 0, 1024) == 0);
+	free(buf);
+
 	return 0;
 }