From: Sivaprasad Tummala <sivaprasad.tummala@amd.com>
To: <david.hunt@intel.com>
Cc: <dev@dpdk.org>
Subject: [PATCH v2 3/3] power: amd power monitor support
Date: Thu, 13 Apr 2023 04:53:34 -0700 [thread overview]
Message-ID: <20230413115334.43172-3-sivaprasad.tummala@amd.com> (raw)
In-Reply-To: <20230413115334.43172-1-sivaprasad.tummala@amd.com>
mwaitx allows epyc processors to enter a implementation dependent
power/performance optimized state (C1 state) for a specific period
or until a store to the monitored address range.
Signed-off-by: Sivaprasad Tummala <sivaprasad.tummala@amd.com>
---
lib/eal/x86/rte_power_intrinsics.c | 83 ++++++++++++++++++++++++++++--
lib/power/rte_power_pmd_mgmt.c | 3 +-
2 files changed, 82 insertions(+), 4 deletions(-)
diff --git a/lib/eal/x86/rte_power_intrinsics.c b/lib/eal/x86/rte_power_intrinsics.c
index f749da9b85..d688066b3a 100644
--- a/lib/eal/x86/rte_power_intrinsics.c
+++ b/lib/eal/x86/rte_power_intrinsics.c
@@ -30,6 +30,7 @@ __umwait_wakeup(volatile void *addr)
static bool wait_supported;
static bool wait_multi_supported;
+static bool amd_mwaitx_supported;
static inline uint64_t
__get_umwait_val(const volatile void *p, const uint8_t sz)
@@ -65,6 +66,76 @@ __check_val_size(const uint8_t sz)
}
}
+/**
+ * This function uses MONITORX/MWAITX instructions and will enter C1 state.
+ * For more information about usage of these instructions, please refer to
+ * AMD64 Architecture Programmer’s Manual.
+ */
+static inline int
+amd_power_monitorx(const struct rte_power_monitor_cond *pmc,
+ const uint64_t tsc_timestamp)
+{
+ const unsigned int lcore_id = rte_lcore_id();
+ struct power_wait_status *s;
+ uint64_t cur_value;
+
+ RTE_SET_USED(tsc_timestamp);
+
+ /* prevent non-EAL thread from using this API */
+ if (lcore_id >= RTE_MAX_LCORE)
+ return -EINVAL;
+
+ if (pmc == NULL)
+ return -EINVAL;
+
+ if (__check_val_size(pmc->size) < 0)
+ return -EINVAL;
+
+ if (pmc->fn == NULL)
+ return -EINVAL;
+
+ s = &wait_status[lcore_id];
+
+ /* update sleep address */
+ rte_spinlock_lock(&s->lock);
+ s->monitor_addr = pmc->addr;
+
+ /*
+ * we're using raw byte codes for now as only the newest compiler
+ * versions support this instruction natively.
+ */
+ /* set address for MONITORX */
+ asm volatile(".byte 0x0f, 0x01, 0xfa;"
+ :
+ : "a"(pmc->addr),
+ "c"(0), /* no extensions */
+ "d"(0)); /* no hints */
+
+ /* now that we've put this address into monitor, we can unlock */
+ rte_spinlock_unlock(&s->lock);
+
+ cur_value = __get_umwait_val(pmc->addr, pmc->size);
+
+ /* check if callback indicates we should abort */
+ if (pmc->fn(cur_value, pmc->opaque) != 0)
+ goto end;
+
+ /* execute MWAITX */
+ asm volatile(".byte 0x0f, 0x01, 0xfb;"
+ : /* ignore rflags */
+ : "a"(0), /* enter C1 */
+ "c"(0), /* no time-out */
+ "b"(0));
+
+end:
+ /* erase sleep address */
+ rte_spinlock_lock(&s->lock);
+ s->monitor_addr = NULL;
+ rte_spinlock_unlock(&s->lock);
+
+ return 0;
+}
+
/**
* This function uses UMONITOR/UMWAIT instructions and will enter C0.2 state.
* For more information about usage of these instructions, please refer to
@@ -81,8 +152,12 @@ rte_power_monitor(const struct rte_power_monitor_cond *pmc,
uint64_t cur_value;
/* prevent user from running this instruction if it's not supported */
- if (!wait_supported)
- return -ENOTSUP;
+ if (!wait_supported) {
+ if (amd_mwaitx_supported)
+ return amd_power_monitorx(pmc, tsc_timestamp);
+ else
+ return -ENOTSUP;
+ }
/* prevent non-EAL thread from using this API */
if (lcore_id >= RTE_MAX_LCORE)
@@ -170,6 +245,8 @@ RTE_INIT(rte_power_intrinsics_init) {
wait_supported = 1;
if (i.power_monitor_multi)
wait_multi_supported = 1;
+ if (i.amd_power_monitorx)
+ amd_mwaitx_supported = 1;
}
int
@@ -178,7 +255,7 @@ rte_power_monitor_wakeup(const unsigned int lcore_id)
struct power_wait_status *s;
/* prevent user from running this instruction if it's not supported */
- if (!wait_supported)
+ if (!wait_supported && !amd_mwaitx_supported)
return -ENOTSUP;
/* prevent buffer overrun */
diff --git a/lib/power/rte_power_pmd_mgmt.c b/lib/power/rte_power_pmd_mgmt.c
index ca1840387c..048a41dc29 100644
--- a/lib/power/rte_power_pmd_mgmt.c
+++ b/lib/power/rte_power_pmd_mgmt.c
@@ -447,7 +447,8 @@ check_monitor(struct pmd_core_cfg *cfg, const union queue *qdata)
bool multimonitor_supported;
/* check if rte_power_monitor is supported */
- if (!global_data.intrinsics_support.power_monitor) {
+ if ((!global_data.intrinsics_support.power_monitor) &&
+ (!global_data.intrinsics_support.amd_power_monitorx)) {
RTE_LOG(DEBUG, POWER, "Monitoring intrinsics are not supported\n");
return -ENOTSUP;
}
--
2.34.1
next prev parent reply other threads:[~2023-04-13 11:54 UTC|newest]
Thread overview: 51+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-04-13 11:53 [PATCH v2 1/3] eal: add x86 cpuid support for monitorx Sivaprasad Tummala
2023-04-13 11:53 ` [PATCH v2 2/3] doc: announce new cpu flag added to rte_cpu_flag_t Sivaprasad Tummala
2023-04-17 4:31 ` [PATCH v3 1/4] " Sivaprasad Tummala
2023-04-18 8:14 ` [PATCH v4 0/4] power: monitor support for AMD EPYC processors Sivaprasad Tummala
2023-04-18 8:25 ` Sivaprasad Tummala
2023-04-18 8:25 ` [PATCH v4 1/4] doc: announce new cpu flag added to rte_cpu_flag_t Sivaprasad Tummala
2023-04-18 8:52 ` Ferruh Yigit
2023-04-18 9:22 ` Bruce Richardson
2023-06-01 9:23 ` David Marchand
2023-07-05 11:32 ` Konstantin Ananyev
2023-08-16 18:59 ` [PATCH v5 1/3] eal: add x86 cpuid support for monitorx Sivaprasad Tummala
2023-08-16 18:59 ` [PATCH v5 2/3] eal: removed unnecessary checks in x86 power monitor APIs Sivaprasad Tummala
2023-08-16 18:59 ` [PATCH v5 3/3] power: amd power monitor support Sivaprasad Tummala
2023-08-16 19:27 ` Tyler Retzlaff
2023-08-17 11:34 ` Tummala, Sivaprasad
2023-08-17 14:18 ` Konstantin Ananyev
2023-08-18 13:25 ` Ferruh Yigit
2023-08-18 13:48 ` Bruce Richardson
2023-08-21 15:42 ` Tyler Retzlaff
2023-08-22 22:30 ` Konstantin Ananyev
2023-08-23 9:19 ` Ferruh Yigit
2023-08-23 16:03 ` Tyler Retzlaff
2023-08-24 9:04 ` Ferruh Yigit
2023-08-25 16:00 ` Tyler Retzlaff
2023-08-30 22:45 ` Konstantin Ananyev
2023-09-27 10:38 ` Tummala, Sivaprasad
2023-09-28 10:11 ` Konstantin Ananyev
2023-10-06 8:26 ` David Marchand
2023-10-09 8:02 ` Tummala, Sivaprasad
2023-10-09 14:05 ` [PATCH v6 1/3] eal: add x86 cpuid support for monitorx Sivaprasad Tummala
2023-10-09 14:05 ` [PATCH v6 2/3] eal: removed unnecessary checks in x86 power monitor APIs Sivaprasad Tummala
2023-10-09 14:05 ` [PATCH v6 3/3] power: amd power monitor support Sivaprasad Tummala
2023-10-10 8:59 ` David Marchand
2023-10-11 9:33 ` Tummala, Sivaprasad
2023-10-10 10:14 ` Konstantin Ananyev
2023-10-09 16:23 ` [PATCH v6 1/3] eal: add x86 cpuid support for monitorx Patrick Robb
2023-10-10 8:21 ` David Marchand
2023-04-18 8:25 ` [PATCH v4 2/4] " Sivaprasad Tummala
2023-06-14 13:15 ` Burakov, Anatoly
2023-04-18 8:25 ` [PATCH v4 3/4] eal: removed unnecessary checks in x86 power monitor APIs Sivaprasad Tummala
2023-06-14 13:14 ` Burakov, Anatoly
2023-04-18 8:25 ` [PATCH v4 4/4] power: amd power monitor support Sivaprasad Tummala
2023-06-14 13:14 ` Burakov, Anatoly
2023-04-13 11:53 ` Sivaprasad Tummala [this message]
2023-04-17 4:34 ` [PATCH v3 " Sivaprasad Tummala
2023-04-13 11:59 ` [PATCH v2 1/3] eal: add x86 cpuid support for monitorx David Marchand
2023-04-13 17:50 ` Tummala, Sivaprasad
2023-04-14 7:05 ` David Marchand
2023-04-14 8:51 ` Tummala, Sivaprasad
2023-04-14 11:48 ` Ferruh Yigit
2023-04-17 4:32 ` [PATCH v3 2/4] " Sivaprasad Tummala
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20230413115334.43172-3-sivaprasad.tummala@amd.com \
--to=sivaprasad.tummala@amd.com \
--cc=david.hunt@intel.com \
--cc=dev@dpdk.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).