From: Madhavan Srinivasan <maddy@linux.vnet.ibm.com>
To: Anton Blanchard <anton@ozlabs.org>,
benh@kernel.crashing.org, paulus@samba.org, mpe@ellerman.id.au,
mikey@neuling.org
Cc: linuxppc-dev@lists.ozlabs.org
Subject: Re: [PATCH] powerpc/perf: Fix POWER9 branch event
Date: Mon, 19 Jun 2017 19:35:27 +0530 [thread overview]
Message-ID: <adadaf41-5a8b-0632-beff-c942081dbfe6@linux.vnet.ibm.com> (raw)
In-Reply-To: <20170619035526.7070-1-anton@ozlabs.org>
On Monday 19 June 2017 09:25 AM, Anton Blanchard wrote:
> From: Anton Blanchard <anton@samba.org>
>
> The POWER9 branch event is wrong, and we always get a count of zero:
>
> ...
> 0 branches
> 3844 branch-misses # 0.00% of all branches
>
> Replace it with the correct event.
Anton,
How about this patch, this will force DD1 to use 0x10012 for
branch and later to use 0x4d05e ?
diff --git a/arch/powerpc/perf/power9-events-list.h
b/arch/powerpc/perf/power9-events-list.h
index 71a6bfee5c02..13bd159db300 100644
--- a/arch/powerpc/perf/power9-events-list.h
+++ b/arch/powerpc/perf/power9-events-list.h
@@ -16,7 +16,7 @@ EVENT(PM_CYC, 0x0001e)
EVENT(PM_ICT_NOSLOT_CYC, 0x100f8)
EVENT(PM_CMPLU_STALL, 0x1e054)
EVENT(PM_INST_CMPL, 0x00002)
-EVENT(PM_BRU_CMPL, 0x10012)
+EVENT(PM_BR_CMPL, 0x4d05e)
EVENT(PM_BR_MPRED_CMPL, 0x400f6)
/* All L1 D cache load references counted at finish, gated by reject */
@@ -56,3 +56,4 @@ EVENT(PM_RUN_CYC, 0x600f4)
/* Instruction Dispatched */
EVENT(PM_INST_DISP, 0x200f2)
EVENT(PM_INST_DISP_ALT, 0x300f2)
+EVENT(PM_BR_CMPL_ALT, 0x10012)
diff --git a/arch/powerpc/perf/power9-pmu.c b/arch/powerpc/perf/power9-pmu.c
index bb28e1a41257..485dacf30152 100644
--- a/arch/powerpc/perf/power9-pmu.c
+++ b/arch/powerpc/perf/power9-pmu.c
@@ -123,7 +123,7 @@ GENERIC_EVENT_ATTR(cpu-cycles, PM_CYC);
GENERIC_EVENT_ATTR(stalled-cycles-frontend, PM_ICT_NOSLOT_CYC);
GENERIC_EVENT_ATTR(stalled-cycles-backend, PM_CMPLU_STALL);
GENERIC_EVENT_ATTR(instructions, PM_INST_CMPL);
-GENERIC_EVENT_ATTR(branch-instructions, PM_BRU_CMPL);
+GENERIC_EVENT_ATTR(branch-instructions, PM_BR_CMPL);
GENERIC_EVENT_ATTR(branch-misses, PM_BR_MPRED_CMPL);
GENERIC_EVENT_ATTR(cache-references, PM_LD_REF_L1);
GENERIC_EVENT_ATTR(cache-misses, PM_LD_MISS_L1_FIN);
@@ -141,7 +141,7 @@ CACHE_EVENT_ATTR(LLC-prefetches, PM_L3_PREF_ALL);
CACHE_EVENT_ATTR(LLC-store-misses, PM_L2_ST_MISS);
CACHE_EVENT_ATTR(LLC-stores, PM_L2_ST);
CACHE_EVENT_ATTR(branch-load-misses, PM_BR_MPRED_CMPL);
-CACHE_EVENT_ATTR(branch-loads, PM_BRU_CMPL);
+CACHE_EVENT_ATTR(branch-loads, PM_BR_CMPL);
CACHE_EVENT_ATTR(dTLB-load-misses, PM_DTLB_MISS);
CACHE_EVENT_ATTR(iTLB-load-misses, PM_ITLB_MISS);
@@ -150,7 +150,7 @@ static struct attribute *power9_events_attr[] = {
GENERIC_EVENT_PTR(PM_ICT_NOSLOT_CYC),
GENERIC_EVENT_PTR(PM_CMPLU_STALL),
GENERIC_EVENT_PTR(PM_INST_CMPL),
- GENERIC_EVENT_PTR(PM_BRU_CMPL),
+ GENERIC_EVENT_PTR(PM_BR_CMPL),
GENERIC_EVENT_PTR(PM_BR_MPRED_CMPL),
GENERIC_EVENT_PTR(PM_LD_REF_L1),
GENERIC_EVENT_PTR(PM_LD_MISS_L1_FIN),
@@ -167,7 +167,7 @@ static struct attribute *power9_events_attr[] = {
CACHE_EVENT_PTR(PM_L2_ST_MISS),
CACHE_EVENT_PTR(PM_L2_ST),
CACHE_EVENT_PTR(PM_BR_MPRED_CMPL),
- CACHE_EVENT_PTR(PM_BRU_CMPL),
+ CACHE_EVENT_PTR(PM_BR_CMPL),
CACHE_EVENT_PTR(PM_DTLB_MISS),
CACHE_EVENT_PTR(PM_ITLB_MISS),
NULL
@@ -231,7 +231,7 @@ static int power9_generic_events_dd1[] = {
[PERF_COUNT_HW_STALLED_CYCLES_FRONTEND] = PM_ICT_NOSLOT_CYC,
[PERF_COUNT_HW_STALLED_CYCLES_BACKEND] = PM_CMPLU_STALL,
[PERF_COUNT_HW_INSTRUCTIONS] = PM_INST_DISP,
- [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = PM_BRU_CMPL,
+ [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = PM_BR_CMPL_ALT,
[PERF_COUNT_HW_BRANCH_MISSES] = PM_BR_MPRED_CMPL,
[PERF_COUNT_HW_CACHE_REFERENCES] = PM_LD_REF_L1,
[PERF_COUNT_HW_CACHE_MISSES] = PM_LD_MISS_L1_FIN,
@@ -242,7 +242,7 @@ static int power9_generic_events[] = {
[PERF_COUNT_HW_STALLED_CYCLES_FRONTEND] = PM_ICT_NOSLOT_CYC,
[PERF_COUNT_HW_STALLED_CYCLES_BACKEND] = PM_CMPLU_STALL,
[PERF_COUNT_HW_INSTRUCTIONS] = PM_INST_CMPL,
- [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = PM_BRU_CMPL,
+ [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = PM_BR_CMPL,
[PERF_COUNT_HW_BRANCH_MISSES] = PM_BR_MPRED_CMPL,
[PERF_COUNT_HW_CACHE_REFERENCES] = PM_LD_REF_L1,
[PERF_COUNT_HW_CACHE_MISSES] = PM_LD_MISS_L1_FIN,
@@ -368,7 +368,7 @@ static int
power9_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = {
},
[ C(BPU) ] = {
[ C(OP_READ) ] = {
- [ C(RESULT_ACCESS) ] = PM_BRU_CMPL,
+ [ C(RESULT_ACCESS) ] = PM_BR_CMPL,
[ C(RESULT_MISS) ] = PM_BR_MPRED_CMPL,
},
[ C(OP_WRITE) ] = {
@@ -453,6 +453,10 @@ static int __init init_power9_pmu(void)
* sampling scenarios in power9 DD1, instead use PM_INST_DISP.
*/
EVENT_VAR(PM_INST_CMPL, _g).id = PM_INST_DISP;
+ /*
+ * This is a hack, but then this is only for DD1
+ */
+ EVENT_VAR(PM_BR_CMPL, _g).id = PM_BR_CMPL_ALT;
rc = register_power_pmu(&power9_isa207_pmu);
} else {
rc = register_power_pmu(&power9_pmu);
prev parent reply other threads:[~2017-06-19 14:05 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-06-19 3:55 [PATCH] powerpc/perf: Fix POWER9 branch event Anton Blanchard
2017-06-19 14:05 ` Madhavan Srinivasan [this message]
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=adadaf41-5a8b-0632-beff-c942081dbfe6@linux.vnet.ibm.com \
--to=maddy@linux.vnet.ibm.com \
--cc=anton@ozlabs.org \
--cc=benh@kernel.crashing.org \
--cc=linuxppc-dev@lists.ozlabs.org \
--cc=mikey@neuling.org \
--cc=mpe@ellerman.id.au \
--cc=paulus@samba.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).