Skip to contentSkip to conversation

perf/x86/intel: Add support for MISPREDICT bit on Knights Landing cpus

PATCH Superseded Jacek Tomaka 2 participants 3 messages Jul 30, 2018 2 messages not in the archive

Subsystems (2)

v2 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 perf/x86/intel: Add support for MISPREDICT bit on Knights Landing cpus

From: Jacek Tomaka <[email protected]>

Problem: perf did not show branch predicted/mispredicted bit in brstack.

Output of perf -F brstack for profile collected

Before:
0x4fdbcd/0x4fdc03/-/-/-/0
0x45f4c1/0x4fdba0/-/-/-/0
0x45f544/0x45f4bb/-/-/-/0
0x45f555/0x45f53c/-/-/-/0
0x7f66901cc24b/0x45f555/-/-/-/0
0x7f66901cc22e/0x7f66901cc23d/-/-/-/0
0x7f66901cc1ff/0x7f66901cc20f/-/-/-/0
0x7f66901cc1e8/0x7f66901cc1fc/-/-/-/0

After:
0x4fdbcd/0x4fdc03/P/-/-/0
0x45f4c1/0x4fdba0/P/-/-/0
0x45f544/0x45f4bb/P/-/-/0
0x45f555/0x45f53c/P/-/-/0
0x7f66901cc24b/0x45f555/P/-/-/0
0x7f66901cc22e/0x7f66901cc23d/P/-/-/0
0x7f66901cc1ff/0x7f66901cc20f/P/-/-/0
0x7f66901cc1e8/0x7f66901cc1fc/P/-/-/0

Cause:
As mentioned in Software Development Manual vol 3, 17.4.8.1, IA32_PERF_CAPABILITIES[5:0] indicates the format of the address that is stored in the LBR stack. Knights Landing reports 1 (LBR_FORMAT_LIP) as its format. Despite that, registers containing FROM address of the branch, do have MISPREDICT bit but because of the format indicated in IA32_PERF_CAPABILITIES[5:0], LBR did not read MISPREDICT bit.

Solution:
Teach LBR about above Knights Landing quirk and make it read MISPREDICT bit.

---
 arch/x86/events/intel/lbr.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
arch/x86/events/intel/lbr.c +5 -1
index cf372b9055..043aa09f3a 100644
@@ -19,7 +19,7 @@ enum {
 	LBR_FORMAT_MAX_KNOWN    = LBR_FORMAT_TIME,
 };
 
-static const enum {
+static enum {
 	LBR_EIP_FLAGS		= 1,
 	LBR_TSX			= 2,
 } lbr_desc[LBR_FORMAT_MAX_KNOWN + 1] = {
@@ -1230,4 +1230,8 @@ void intel_pmu_lbr_init_knl(void)
 
 	x86_pmu.lbr_sel_mask = LBR_SEL_MASK;
 	x86_pmu.lbr_sel_map  = snb_lbr_sel_map;
+
+	/* Knights Landing does have MISPREDICT bit */
+	if (x86_pmu.intel_cap.lbr_format == LBR_FORMAT_LIP)
+		lbr_desc[LBR_FORMAT_LIP] |= LBR_EIP_FLAGS;
 }
Signature · 1 line
2.17.0
maintainer
#1 lore raw
Quoted 54 lines · Jacek Tomaka

From: Jacek Tomaka <[email protected]>

Problem: perf did not show branch predicted/mispredicted bit in brstack.

Output of perf -F brstack for profile collected

Before:
0x4fdbcd/0x4fdc03/-/-/-/0
0x45f4c1/0x4fdba0/-/-/-/0
0x45f544/0x45f4bb/-/-/-/0
0x45f555/0x45f53c/-/-/-/0
0x7f66901cc24b/0x45f555/-/-/-/0
0x7f66901cc22e/0x7f66901cc23d/-/-/-/0
0x7f66901cc1ff/0x7f66901cc20f/-/-/-/0
0x7f66901cc1e8/0x7f66901cc1fc/-/-/-/0

After:
0x4fdbcd/0x4fdc03/P/-/-/0
0x45f4c1/0x4fdba0/P/-/-/0
0x45f544/0x45f4bb/P/-/-/0
0x45f555/0x45f53c/P/-/-/0
0x7f66901cc24b/0x45f555/P/-/-/0
0x7f66901cc22e/0x7f66901cc23d/P/-/-/0
0x7f66901cc1ff/0x7f66901cc20f/P/-/-/0
0x7f66901cc1e8/0x7f66901cc1fc/P/-/-/0

Cause:
As mentioned in Software Development Manual vol 3, 17.4.8.1, IA32_PERF_CAPABILITIES[5:0] indicates the format of the address that is stored in the LBR stack. Knights Landing reports 1 (LBR_FORMAT_LIP) as its format. Despite that, registers containing FROM address of the branch, do have MISPREDICT bit but because of the format indicated in IA32_PERF_CAPABILITIES[5:0], LBR did not read MISPREDICT bit.

Solution:
Teach LBR about above Knights Landing quirk and make it read MISPREDICT bit.

---
 arch/x86/events/intel/lbr.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
arch/x86/events/intel/lbr.c +1 -1
index cf372b9055..043aa09f3a 100644
@@ -19,7 +19,7 @@ enum {
 	LBR_FORMAT_MAX_KNOWN    = LBR_FORMAT_TIME,
 };
 
-static const enum {
+static enum {
 	LBR_EIP_FLAGS		= 1,
 	LBR_TSX			= 2,
 } lbr_desc[LBR_FORMAT_MAX_KNOWN + 1] = {

arch/x86/events/intel/lbr.c void intel_pmu_lbr_init_knl(void) · line 1237

Quoted 8 lines
arch/x86/events/intel/lbr.c +4 -0
@@ -1230,4 +1230,8 @@ void intel_pmu_lbr_init_knl(void)
 
 	x86_pmu.lbr_sel_mask = LBR_SEL_MASK;
 	x86_pmu.lbr_sel_map  = snb_lbr_sel_map;
+
+	/* Knights Landing does have MISPREDICT bit */
+	if (x86_pmu.intel_cap.lbr_format == LBR_FORMAT_LIP)
+		lbr_desc[LBR_FORMAT_LIP] |= LBR_EIP_FLAGS;
 }

So why not set lbr_format to LBR_FORMAT_EIP_FLAGS ?

A message that is not in the archive (sent off-list or not yet received): CAKVxXCVjBaW1kFA48r-bKwrnzR9+SaFXtgaBTb-tr6QHeu1srg@mail.gmail.com

A message that is not in the archive (sent off-list or not yet received): CAKVxXCVoRb8Hf18AmUF7RcyNy7JACKByj2FsVOSiO31=ZHFHYQ@mail.gmail.com

maintainer
#4 lore raw

A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.

Quoted 8 lines · Jacek Tomaka

Ah, right:
/*
* Due to lack of segmentation in Linux the effective address (offset)
* is the same as the linear address, allowing us to merge the LIP and EIP
* LBR formats.
*/

Yeah, LBR_FORMAT_EIP_FLAGS is ok as well. Would it be preffered?

Yeah, think so, that keeps the lookup table const.