On 30 March 2015 at 14:01, Stefan Hajnoczi <stefa...@redhat.com> wrote:
> On Sat, Mar 28, 2015 at 06:22:11PM +0200, Emil Condrea wrote:
>> diff --git a/target-arm/helper.c b/target-arm/helper.c
>> index 10886c5..504530a 100644
>> --- a/target-arm/helper.c
>> +++ b/target-arm/helper.c
>> @@ -649,7 +649,7 @@ void pmccntr_sync(CPUARMState *env)
>> uint64_t temp_ticks;
>>
>> temp_ticks = muldiv64(qemu_clock_get_us(QEMU_CLOCK_VIRTUAL),
>> - get_ticks_per_sec(), 1000000);
>> + NSEC_PER_SEC, 1000000);
>>
>> if (env->cp15.c9_pmcr & PMCRD) {
>> /* Increment once every 64 processor clock cycles */
>> @@ -688,7 +688,7 @@ static uint64_t pmccntr_read(CPUARMState *env, const
>> ARMCPRegInfo *ri)
>> }
>>
>> total_ticks = muldiv64(qemu_clock_get_us(QEMU_CLOCK_VIRTUAL),
>> - get_ticks_per_sec(), 1000000);
>> + NSEC_PER_SEC, 1000000);
>>
>> if (env->cp15.c9_pmcr & PMCRD) {
>> /* Increment once every 64 processor clock cycles */
>> @@ -709,7 +709,7 @@ static void pmccntr_write(CPUARMState *env, const
>> ARMCPRegInfo *ri,
>> }
>>
>> total_ticks = muldiv64(qemu_clock_get_us(QEMU_CLOCK_VIRTUAL),
>> - get_ticks_per_sec(), 1000000);
>> + NSEC_PER_SEC, 1000000);
>
> Peter: I wonder why the muldiv64() expression is necessary. Is it
> intentionally returning the microsecond clock converted to nanoseconds?
>
> Perhaps the expression should be replaced with just
> qemu_get_clock_ns(QEMU_CLOCK_VIRTUAL)?
>
> Or if microsecond precision is required, use qemu_get_clock_us() * 1000.
Ask Peter C -- IIRC this is his code.
-- PMM
