hi !
this patch is necessary to get latencies < 1ms for ep93xx armv4t with
2.6.21.5-rt18.
Index: linux-2.6/arch/arm/mach-ep93xx/core.c
===================================================================
--- linux-2.6.orig/arch/arm/mach-ep93xx/core.c
+++ linux-2.6/arch/arm/mach-ep93xx/core.c
@@ -108,7 +108,17 @@ static int ep93xx_timer_interrupt(int ir
static int ep93xx_set_next_event(unsigned long evt,
struct clock_event_device *unused)
{
+ u32 tmode;
+ tmode = __raw_readl(EP93XX_TIMER1_CONTROL);
+ tmode &= ~EP93XX_TC123_ENABLE;
+
+ /* stop timer */
+ __raw_writel(tmode, EP93XX_TIMER1_CONTROL);
+ /* program timer */
__raw_writel(evt, EP93XX_TIMER1_LOAD);
+ /* start timer */
+ tmode |= EP93XX_TC123_ENABLE;
+ __raw_writel(tmode, EP93XX_TIMER1_CONTROL);
return 0;
}
i have now disabled all kernel hacking things, and get this latencies with
2.6.21.5-rt18. i run top with delay 0 over serial and ping -f. cyclictest is
run over ssh.
[root@micro9]\# ./cyclictest_new -i 10000 -n -p 80 -t 10
5.56 4.16 1.94 1/69 1199
T: 0(1122) P:80 I:10000 C: 56041 Min: 24 Act: 106 Avg: 176 Max: 466
T: 1(1123) P:79 I:10500 C: 53372 Min: 24 Act: 309 Avg: 190 Max: 668
T: 2(1124) P:78 I:11000 C: 50946 Min: 23 Act: 241 Avg: 186 Max: 645
T: 3(1125) P:77 I:11500 C: 48731 Min: 24 Act: 228 Avg: 186 Max: 663
T: 4(1126) P:76 I:12000 C: 46701 Min: 24 Act: 36 Avg: 197 Max: 683
T: 5(1127) P:75 I:12500 C: 44833 Min: 25 Act: 34 Avg: 189 Max: 659
T: 6(1128) P:74 I:13000 C: 43109 Min: 24 Act: 293 Avg: 192 Max: 630
T: 7(1129) P:73 I:13500 C: 41512 Min: 23 Act: 31 Avg: 195 Max: 773
T: 8(1130) P:72 I:14000 C: 40029 Min: 24 Act: 223 Avg: 179 Max: 712
T: 9(1131) P:71 I:14500 C: 38649 Min: 24 Act: 101 Avg: 206 Max: 738
does someone have some latencies to compare? what do you think about this
latencies, is this the the end what is possible ?
--
regards
Manfred Gruber
-
To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
