[v6,RESEND] eal: add tsc_hz to rte_mem_config
Checks
Commit Message
This ensures secondary processes never have to
calculate the TSC rate themselves, which can be
noticeable in VMs that don't have access to
arch-specific detection mechanism (such as
CPUID leaf 0x15 or MSR 0xCE on x86).
Since rte_mem_config is now internal to the rte_eal
library, we can add tsc_hz without ABI breakage
concerns.
Reduces rte_eal_init() execution time in a secondary
process from 165ms to 66ms on my test system.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
---
lib/librte_eal/common/eal_common_timer.c | 15 +++++++++++++++
lib/librte_eal/common/eal_memcfg.h | 3 +++
2 files changed, 18 insertions(+)
Comments
On Mon, Oct 07, 2019 at 08:28:21AM -0700, Jim Harris wrote:
> This ensures secondary processes never have to
> calculate the TSC rate themselves, which can be
> noticeable in VMs that don't have access to
> arch-specific detection mechanism (such as
> CPUID leaf 0x15 or MSR 0xCE on x86).
>
> Since rte_mem_config is now internal to the rte_eal
> library, we can add tsc_hz without ABI breakage
> concerns.
>
> Reduces rte_eal_init() execution time in a secondary
> process from 165ms to 66ms on my test system.
>
> Signed-off-by: Jim Harris <james.r.harris@intel.com>
> ---
This seems a good idea to me.
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
On Tue, Oct 8, 2019 at 10:39 AM Bruce Richardson
<bruce.richardson@intel.com> wrote:
>
> On Mon, Oct 07, 2019 at 08:28:21AM -0700, Jim Harris wrote:
> > This ensures secondary processes never have to
> > calculate the TSC rate themselves, which can be
> > noticeable in VMs that don't have access to
> > arch-specific detection mechanism (such as
> > CPUID leaf 0x15 or MSR 0xCE on x86).
> >
> > Since rte_mem_config is now internal to the rte_eal
> > library, we can add tsc_hz without ABI breakage
> > concerns.
> >
> > Reduces rte_eal_init() execution time in a secondary
> > process from 165ms to 66ms on my test system.
> >
> > Signed-off-by: Jim Harris <james.r.harris@intel.com>
> > ---
> This seems a good idea to me.
>
> Acked-by: Bruce Richardson <bruce.richardson@intel.com>
I wonder if we can get rid of eal_tsc_resolution_hz and just rely on
the shared memory to get/store this info.
Feel free to look at this later and send a followup patch :-).
Applied, thanks.
--
David Marchand
@@ -15,8 +15,10 @@
#include <rte_log.h>
#include <rte_cycles.h>
#include <rte_pause.h>
+#include <rte_eal.h>
#include "eal_private.h"
+#include "eal_memcfg.h"
/* The frequency of the RDTSC timer resolution */
static uint64_t eal_tsc_resolution_hz;
@@ -77,8 +79,20 @@ estimate_tsc_freq(void)
void
set_tsc_freq(void)
{
+ struct rte_mem_config *mcfg = rte_eal_get_configuration()->mem_config;
uint64_t freq;
+ if (rte_eal_process_type() == RTE_PROC_SECONDARY) {
+ /*
+ * Just use the primary process calculated TSC rate in any
+ * secondary process. It avoids any unnecessary overhead on
+ * systems where arch-specific frequency detection is not
+ * available.
+ */
+ eal_tsc_resolution_hz = mcfg->tsc_hz;
+ return;
+ }
+
freq = get_tsc_freq_arch();
if (!freq)
freq = get_tsc_freq();
@@ -87,6 +101,7 @@ set_tsc_freq(void)
RTE_LOG(DEBUG, EAL, "TSC frequency is ~%" PRIu64 " KHz\n", freq / 1000);
eal_tsc_resolution_hz = freq;
+ mcfg->tsc_hz = freq;
}
void rte_delay_us_callback_register(void (*userfunc)(unsigned int))
@@ -70,6 +70,9 @@ struct rte_mem_config {
uint32_t single_file_segments;
/**< stored single file segments parameter. */
+ uint64_t tsc_hz;
+ /**< TSC rate */
+
uint8_t dma_maskbits; /**< Keeps the more restricted dma mask. */
};