I think all XScale processors have CP14 timestamp (see <http://download.intel.com/design/intelxscale/XScaleDatasheet4.pdf>) but it's in register 1, not register 0. The original report is right in that StrongArm processors do not have a CP14. Therefore the following *should* work:
--- liboil-0.3.10.orig/liboil/liboilcpu.c 2006-11-06 01:37:03.000000000
+0000
+++ liboil-0.3.10/liboil/liboilcpu.c 2007-02-22 13:41:31.000000000 +0000
@@ -620,7 +620,7 @@
{
unsigned int ts;
__asm__ __volatile__ (
- " mrc p14, 0, %0, c0, c0, 0 \n"
+ " mrc p14, 0, %0, c1, c0, 0 \n"
: "=r" (ts));
return ts;
}
@@ -629,7 +629,8 @@
oil_cpu_detect_arm(void)
{
#ifdef __linux__
- int arm_implementer;
+ int arm_implementer = 0;
+ char arm_architecture = 0;
char *cpuinfo;
char *s;
@@ -640,18 +641,16 @@
arm_implementer = strtoul (s, NULL, 0);
free(s);
}
+ s = get_cpuinfo_line(cpuinfo, "CPU architecture");
+ if (s) {
+ char *p = s + strspn(s, " \t");
+ arm_architecture = *p;
+ free(s);
+ }
- switch(arm_implementer) {
- case 0x69: /* Intel */
- /* assume that all Intel chips support CP14 timestamp */
- _oil_profile_stamp = oil_profile_stamp_xscale;
- break;
- case 0x41: /* ARM */
- /* ARM chips are known to not have timestamping available from
- * user space */
- break;
- default:
- break;
+ /* only Intel XScale processors support CP14 timestamp */
+ if (arm_implementer == 0x69 && arm_architecture == '5') {
+ _oil_profile_stamp = oil_profile_stamp_xscale;
}
free (cpuinfo);
-- END --
In practice it still results in SIGILL, so I agree we should never use
oil_profile_stamp_xscale().
Ben.
--
Ben Hutchings
Beware of bugs in the above code;
I have only proved it correct, not tried it. - Donald Knuth
signature.asc
Description: This is a digitally signed message part
