2022-02-22 05:52:34

by Ali Saidi

[permalink] [raw]
Subject: [PATCH v2 1/2] perf arm-spe: Use SPE data source for neoverse cores

When synthesizing data from SPE, augment the type with source information
for Arm Neoverse cores. The field is IMPLDEF but the Neoverse cores all use
the same encoding. I can't find encoding information for any other SPE
implementations to unify their choices with Arm's thus that is left for
future work.

This change populates the mem_lvl_num for Neoverse cores instead of the
deprecated mem_lvl namespace.

Signed-off-by: Ali Saidi <[email protected]>
---
.../util/arm-spe-decoder/arm-spe-decoder.c | 1 +
.../util/arm-spe-decoder/arm-spe-decoder.h | 12 ++
tools/perf/util/arm-spe.c | 106 +++++++++++++++---
3 files changed, 104 insertions(+), 15 deletions(-)

diff --git a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c
index 5e390a1a79ab..091987dd3966 100644
--- a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c
+++ b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c
@@ -220,6 +220,7 @@ static int arm_spe_read_record(struct arm_spe_decoder *decoder)

break;
case ARM_SPE_DATA_SOURCE:
+ decoder->record.source = payload;
break;
case ARM_SPE_BAD:
break;
diff --git a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h
index 69b31084d6be..c81bf90c0996 100644
--- a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h
+++ b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h
@@ -29,6 +29,17 @@ enum arm_spe_op_type {
ARM_SPE_ST = 1 << 1,
};

+enum arm_spe_neoverse_data_source {
+ ARM_SPE_NV_L1D = 0x0,
+ ARM_SPE_NV_L2 = 0x8,
+ ARM_SPE_NV_PEER_CORE = 0x9,
+ ARM_SPE_NV_LCL_CLSTR = 0xa,
+ ARM_SPE_NV_SYS_CACHE = 0xb,
+ ARM_SPE_NV_PEER_CLSTR = 0xc,
+ ARM_SPE_NV_REMOTE = 0xd,
+ ARM_SPE_NV_DRAM = 0xe,
+};
+
struct arm_spe_record {
enum arm_spe_sample_type type;
int err;
@@ -40,6 +51,7 @@ struct arm_spe_record {
u64 virt_addr;
u64 phys_addr;
u64 context_id;
+ u16 source;
};

struct arm_spe_insn;
diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
index d2b64e3f588b..e0243c2fed5f 100644
--- a/tools/perf/util/arm-spe.c
+++ b/tools/perf/util/arm-spe.c
@@ -34,6 +34,7 @@
#include "arm-spe-decoder/arm-spe-decoder.h"
#include "arm-spe-decoder/arm-spe-pkt-decoder.h"

+#include <../../../arch/arm64/include/asm/cputype.h>
#define MAX_TIMESTAMP (~0ULL)

struct arm_spe {
@@ -45,6 +46,7 @@ struct arm_spe {
struct perf_session *session;
struct machine *machine;
u32 pmu_type;
+ u64 midr;

struct perf_tsc_conversion tc;

@@ -399,33 +401,103 @@ static bool arm_spe__is_memory_event(enum arm_spe_sample_type type)
return false;
}

-static u64 arm_spe__synth_data_source(const struct arm_spe_record *record)
-{
- union perf_mem_data_src data_src = { 0 };
+static const struct midr_range neoverse_spe[] = {
+ MIDR_ALL_VERSIONS(MIDR_NEOVERSE_N1),
+ MIDR_ALL_VERSIONS(MIDR_NEOVERSE_N2),
+ {},
+};

- if (record->op == ARM_SPE_LD)
- data_src.mem_op = PERF_MEM_OP_LOAD;
- else
- data_src.mem_op = PERF_MEM_OP_STORE;

+static void arm_spe__synth_data_source_neoverse(const struct arm_spe_record *record,
+ union perf_mem_data_src *data_src)
+{
+ switch (record->source) {
+ case ARM_SPE_NV_L1D:
+ data_src->mem_lvl = PERF_MEM_LVL_HIT;
+ data_src->mem_lvl_num = PERF_MEM_LVLNUM_L1;
+ break;
+ case ARM_SPE_NV_L2:
+ data_src->mem_lvl = PERF_MEM_LVL_HIT;
+ data_src->mem_lvl_num = PERF_MEM_LVLNUM_L2;
+ break;
+ case ARM_SPE_NV_PEER_CORE:
+ data_src->mem_lvl = PERF_MEM_LVL_HIT;
+ data_src->mem_snoop = PERF_MEM_SNOOP_HITM;
+ data_src->mem_lvl_num = PERF_MEM_LVLNUM_ANY_CACHE;
+ break;
+ /*
+ * We don't know if this is L1, L2, or even L3 (for the cases the system
+ * has an L3, but we do know it was a cache-2-cache transfer, so set
+ * SNOOP_HITM
+ */
+ case ARM_SPE_NV_LCL_CLSTR:
+ case ARM_SPE_NV_PEER_CLSTR:
+ data_src->mem_lvl = PERF_MEM_LVL_HIT;
+ data_src->mem_snoop = PERF_MEM_SNOOP_HITM;
+ data_src->mem_lvl_num = PERF_MEM_LVLNUM_ANY_CACHE;
+ break;
+ /*
+ * System cache is assumed to be L4, as cluster cache (if it exists)
+ * would be L3 cache on Neoverse platforms
+ */
+ case ARM_SPE_NV_SYS_CACHE:
+ data_src->mem_lvl = PERF_MEM_LVL_HIT;
+ data_src->mem_lvl_num = PERF_MEM_LVLNUM_L4;
+ break;
+ /*
+ * We don't know what level it hit in, except it came from the other
+ * socket
+ */
+ case ARM_SPE_NV_REMOTE:
+ data_src->mem_snoop = PERF_MEM_SNOOP_HITM;
+ data_src->mem_remote = PERF_MEM_REMOTE_REMOTE;
+ break;
+ case ARM_SPE_NV_DRAM:
+ data_src->mem_lvl = PERF_MEM_LVL_HIT;
+ data_src->mem_lvl_num = PERF_MEM_LVLNUM_RAM;
+ break;
+ default:
+ break;
+ }
+}
+
+static void arm_spe__synth_data_source_generic(const struct arm_spe_record *record,
+ union perf_mem_data_src *data_src)
+{
if (record->type & (ARM_SPE_LLC_ACCESS | ARM_SPE_LLC_MISS)) {
- data_src.mem_lvl = PERF_MEM_LVL_L3;
+ data_src->mem_lvl = PERF_MEM_LVL_L3;

if (record->type & ARM_SPE_LLC_MISS)
- data_src.mem_lvl |= PERF_MEM_LVL_MISS;
+ data_src->mem_lvl |= PERF_MEM_LVL_MISS;
else
- data_src.mem_lvl |= PERF_MEM_LVL_HIT;
+ data_src->mem_lvl |= PERF_MEM_LVL_HIT;
} else if (record->type & (ARM_SPE_L1D_ACCESS | ARM_SPE_L1D_MISS)) {
- data_src.mem_lvl = PERF_MEM_LVL_L1;
+ data_src->mem_lvl = PERF_MEM_LVL_L1;

if (record->type & ARM_SPE_L1D_MISS)
- data_src.mem_lvl |= PERF_MEM_LVL_MISS;
+ data_src->mem_lvl |= PERF_MEM_LVL_MISS;
else
- data_src.mem_lvl |= PERF_MEM_LVL_HIT;
+ data_src->mem_lvl |= PERF_MEM_LVL_HIT;
}

if (record->type & ARM_SPE_REMOTE_ACCESS)
- data_src.mem_lvl |= PERF_MEM_LVL_REM_CCE1;
+ data_src->mem_lvl |= PERF_MEM_LVL_REM_CCE1;
+}
+
+static u64 arm_spe__synth_data_source(const struct arm_spe_record *record, u64 midr)
+{
+ union perf_mem_data_src data_src = { 0 };
+ bool is_neoverse = is_midr_in_range(midr, neoverse_spe);
+
+ if (record->op & ARM_SPE_LD)
+ data_src.mem_op = PERF_MEM_OP_LOAD;
+ else
+ data_src.mem_op = PERF_MEM_OP_STORE;
+
+ if (is_neoverse)
+ arm_spe__synth_data_source_neoverse(record, &data_src);
+ else
+ arm_spe__synth_data_source_generic(record, &data_src);

if (record->type & (ARM_SPE_TLB_ACCESS | ARM_SPE_TLB_MISS)) {
data_src.mem_dtlb = PERF_MEM_TLB_WK;
@@ -446,7 +518,7 @@ static int arm_spe_sample(struct arm_spe_queue *speq)
u64 data_src;
int err;

- data_src = arm_spe__synth_data_source(record);
+ data_src = arm_spe__synth_data_source(record, spe->midr);

if (spe->sample_flc) {
if (record->type & ARM_SPE_L1D_MISS) {
@@ -796,6 +868,10 @@ static int arm_spe_process_event(struct perf_session *session,
u64 timestamp;
struct arm_spe *spe = container_of(session->auxtrace,
struct arm_spe, auxtrace);
+ const char *cpuid = perf_env__cpuid(session->evlist->env);
+ u64 midr = strtol(cpuid, NULL, 16);
+
+ spe->midr = midr;

if (dump_trace)
return 0;
--
2.32.0


2022-03-02 13:36:13

by German Gomez

[permalink] [raw]
Subject: Re: [PATCH v2 1/2] perf arm-spe: Use SPE data source for neoverse cores

Hi Ali,

On 21/02/2022 22:47, Ali Saidi wrote:
> When synthesizing data from SPE, augment the type with source information
> for Arm Neoverse cores. The field is IMPLDEF but the Neoverse cores all use
> the same encoding. I can't find encoding information for any other SPE
> implementations to unify their choices with Arm's thus that is left for
> future work.
>
> This change populates the mem_lvl_num for Neoverse cores instead of the
> deprecated mem_lvl namespace.
>
> Signed-off-by: Ali Saidi <[email protected]>
> ---
> .../util/arm-spe-decoder/arm-spe-decoder.c | 1 +
> .../util/arm-spe-decoder/arm-spe-decoder.h | 12 ++
> tools/perf/util/arm-spe.c | 106 +++++++++++++++---
> 3 files changed, 104 insertions(+), 15 deletions(-)
>
> diff --git a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c
> index 5e390a1a79ab..091987dd3966 100644
> --- a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c
> +++ b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c
> @@ -220,6 +220,7 @@ static int arm_spe_read_record(struct arm_spe_decoder *decoder)
>
> break;
> case ARM_SPE_DATA_SOURCE:
> + decoder->record.source = payload;
> break;
> case ARM_SPE_BAD:
> break;
> diff --git a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h
> index 69b31084d6be..c81bf90c0996 100644
> --- a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h
> +++ b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h
> @@ -29,6 +29,17 @@ enum arm_spe_op_type {
> ARM_SPE_ST = 1 << 1,
> };
>
> +enum arm_spe_neoverse_data_source {
> + ARM_SPE_NV_L1D = 0x0,
> + ARM_SPE_NV_L2 = 0x8,
> + ARM_SPE_NV_PEER_CORE = 0x9,
> + ARM_SPE_NV_LCL_CLSTR = 0xa,
> + ARM_SPE_NV_SYS_CACHE = 0xb,
> + ARM_SPE_NV_PEER_CLSTR = 0xc,
> + ARM_SPE_NV_REMOTE = 0xd,
> + ARM_SPE_NV_DRAM = 0xe,
> +};
> +
> struct arm_spe_record {
> enum arm_spe_sample_type type;
> int err;
> @@ -40,6 +51,7 @@ struct arm_spe_record {
> u64 virt_addr;
> u64 phys_addr;
> u64 context_id;
> + u16 source;
> };
>
> struct arm_spe_insn;
> diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
> index d2b64e3f588b..e0243c2fed5f 100644
> --- a/tools/perf/util/arm-spe.c
> +++ b/tools/perf/util/arm-spe.c
> @@ -34,6 +34,7 @@
> #include "arm-spe-decoder/arm-spe-decoder.h"
> #include "arm-spe-decoder/arm-spe-pkt-decoder.h"
>
> +#include <../../../arch/arm64/include/asm/cputype.h>
> #define MAX_TIMESTAMP (~0ULL)
>
> struct arm_spe {
> @@ -45,6 +46,7 @@ struct arm_spe {
> struct perf_session *session;
> struct machine *machine;
> u32 pmu_type;
> + u64 midr;
>
> struct perf_tsc_conversion tc;
>
> @@ -399,33 +401,103 @@ static bool arm_spe__is_memory_event(enum arm_spe_sample_type type)
> return false;
> }
>
> -static u64 arm_spe__synth_data_source(const struct arm_spe_record *record)
> -{
> - union perf_mem_data_src data_src = { 0 };
> +static const struct midr_range neoverse_spe[] = {
> + MIDR_ALL_VERSIONS(MIDR_NEOVERSE_N1),
> + MIDR_ALL_VERSIONS(MIDR_NEOVERSE_N2),
> + {},
> +};
>
> - if (record->op == ARM_SPE_LD)
> - data_src.mem_op = PERF_MEM_OP_LOAD;
> - else
> - data_src.mem_op = PERF_MEM_OP_STORE;
>
> +static void arm_spe__synth_data_source_neoverse(const struct arm_spe_record *record,
> + union perf_mem_data_src *data_src)
> +{
> + switch (record->source) {
> + case ARM_SPE_NV_L1D:
> + data_src->mem_lvl = PERF_MEM_LVL_HIT;

I understand mem_lvl is deprecated but shouldn't we add the level bits here as well for backwards compat?

> + data_src->mem_lvl_num = PERF_MEM_LVLNUM_L1;
> + break;
> + case ARM_SPE_NV_L2:
> + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> + data_src->mem_lvl_num = PERF_MEM_LVLNUM_L2;
> + break;
> + case ARM_SPE_NV_PEER_CORE:
> + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> + data_src->mem_snoop = PERF_MEM_SNOOP_HITM;
> + data_src->mem_lvl_num = PERF_MEM_LVLNUM_ANY_CACHE;
> + break;
> + /*
> + * We don't know if this is L1, L2, or even L3 (for the cases the system
> + * has an L3, but we do know it was a cache-2-cache transfer, so set
> + * SNOOP_HITM
> + */
> + case ARM_SPE_NV_LCL_CLSTR:
> + case ARM_SPE_NV_PEER_CLSTR:
> + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> + data_src->mem_snoop = PERF_MEM_SNOOP_HITM;
> + data_src->mem_lvl_num = PERF_MEM_LVLNUM_ANY_CACHE;
> + break;
> + /*
> + * System cache is assumed to be L4, as cluster cache (if it exists)
> + * would be L3 cache on Neoverse platforms
> + */
> + case ARM_SPE_NV_SYS_CACHE:
> + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> + data_src->mem_lvl_num = PERF_MEM_LVLNUM_L4;
> + break;
> + /*
> + * We don't know what level it hit in, except it came from the other
> + * socket
> + */
> + case ARM_SPE_NV_REMOTE:
> + data_src->mem_snoop = PERF_MEM_SNOOP_HITM;
> + data_src->mem_remote = PERF_MEM_REMOTE_REMOTE;
> + break;
> + case ARM_SPE_NV_DRAM:
> + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> + data_src->mem_lvl_num = PERF_MEM_LVLNUM_RAM;
> + break;
> + default:
> + break;
> + }
> +}
> +
> +static void arm_spe__synth_data_source_generic(const struct arm_spe_record *record,
> + union perf_mem_data_src *data_src)
> +{
> if (record->type & (ARM_SPE_LLC_ACCESS | ARM_SPE_LLC_MISS)) {
> - data_src.mem_lvl = PERF_MEM_LVL_L3;
> + data_src->mem_lvl = PERF_MEM_LVL_L3;
>
> if (record->type & ARM_SPE_LLC_MISS)
> - data_src.mem_lvl |= PERF_MEM_LVL_MISS;
> + data_src->mem_lvl |= PERF_MEM_LVL_MISS;
> else
> - data_src.mem_lvl |= PERF_MEM_LVL_HIT;
> + data_src->mem_lvl |= PERF_MEM_LVL_HIT;
> } else if (record->type & (ARM_SPE_L1D_ACCESS | ARM_SPE_L1D_MISS)) {
> - data_src.mem_lvl = PERF_MEM_LVL_L1;
> + data_src->mem_lvl = PERF_MEM_LVL_L1;
>
> if (record->type & ARM_SPE_L1D_MISS)
> - data_src.mem_lvl |= PERF_MEM_LVL_MISS;
> + data_src->mem_lvl |= PERF_MEM_LVL_MISS;
> else
> - data_src.mem_lvl |= PERF_MEM_LVL_HIT;
> + data_src->mem_lvl |= PERF_MEM_LVL_HIT;
> }
>
> if (record->type & ARM_SPE_REMOTE_ACCESS)
> - data_src.mem_lvl |= PERF_MEM_LVL_REM_CCE1;
> + data_src->mem_lvl |= PERF_MEM_LVL_REM_CCE1;
> +}
> +
> +static u64 arm_spe__synth_data_source(const struct arm_spe_record *record, u64 midr)
> +{
> + union perf_mem_data_src data_src = { 0 };
> + bool is_neoverse = is_midr_in_range(midr, neoverse_spe);
> +
> + if (record->op & ARM_SPE_LD)
> + data_src.mem_op = PERF_MEM_OP_LOAD;
> + else
> + data_src.mem_op = PERF_MEM_OP_STORE;
> +
> + if (is_neoverse)
> + arm_spe__synth_data_source_neoverse(record, &data_src);
> + else
> + arm_spe__synth_data_source_generic(record, &data_src);
>
> if (record->type & (ARM_SPE_TLB_ACCESS | ARM_SPE_TLB_MISS)) {
> data_src.mem_dtlb = PERF_MEM_TLB_WK;
> @@ -446,7 +518,7 @@ static int arm_spe_sample(struct arm_spe_queue *speq)
> u64 data_src;
> int err;
>
> - data_src = arm_spe__synth_data_source(record);
> + data_src = arm_spe__synth_data_source(record, spe->midr);
>
> if (spe->sample_flc) {
> if (record->type & ARM_SPE_L1D_MISS) {
> @@ -796,6 +868,10 @@ static int arm_spe_process_event(struct perf_session *session,
> u64 timestamp;
> struct arm_spe *spe = container_of(session->auxtrace,
> struct arm_spe, auxtrace);
> + const char *cpuid = perf_env__cpuid(session->evlist->env);
> + u64 midr = strtol(cpuid, NULL, 16);
> +
> + spe->midr = midr;

I think this midr setup belongs in the arm_spe_process_auxtrace_info callback instead.

Thanks,
German

>
> if (dump_trace)
> return 0;

2022-03-14 10:34:58

by Leo Yan

[permalink] [raw]
Subject: Re: [PATCH v2 1/2] perf arm-spe: Use SPE data source for neoverse cores

On Sun, Mar 13, 2022 at 07:06:19PM +0000, Ali Saidi wrote:

[...]

> > > > +static void arm_spe__synth_data_source_neoverse(const struct arm_spe_record *record,
> > > > + union perf_mem_data_src *data_src)
> > > > +{
> > > > + switch (record->source) {
> > > > + case ARM_SPE_NV_L1D:
> > > > + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> > >
> > > I understand mem_lvl is deprecated but shouldn't we add the level bits here as well for backwards compat?
> >
> > Thanks for pointing out this. Yeah, I think German's suggestion is
> > valid, the commit 6ae5fa61d27d ("perf/x86: Fix data source decoding
> > for Skylake") introduces new field 'mem_lvl_num', but it also keeps
> > backwards compatible for the field 'mem_lvl'.
> >
> I thought about that, but then I'm making some assumption about how to fit
> this into the old LVL framework, which is perhaps OK (afaik there are no
> Neoverse systems with more than 3 cache levels). What stopped me was that
> perf_mem__lvl_scnprintf() does the wrong thing when both are set so I
> assumed that setting both was not the right course of action.

Thanks for pointing out this. I looked at perf_mem__lvl_scnprintf()
and it prints both for fields 'mem_lvl' and 'mem_lvl_num'. Thus I can
see the output result shows the duplicate info for memory access like
"L1 or L1 hit", "L3 or L3 hit", etc. This would be a common issue
crossing archs. Do I miss any other issues?

> > > > + data_src->mem_lvl_num = PERF_MEM_LVLNUM_L1;
> > > > + break;
> > > > + case ARM_SPE_NV_L2:
> > > > + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> > > > + data_src->mem_lvl_num = PERF_MEM_LVLNUM_L2;
> > > > + break;
> > > > + case ARM_SPE_NV_PEER_CORE:
> > > > + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> > > > + data_src->mem_snoop = PERF_MEM_SNOOP_HITM;
> > > > + data_src->mem_lvl_num = PERF_MEM_LVLNUM_ANY_CACHE;
> >
> > For PEER_CORE data source, we don't know if it's coming from peer
> > core's L1 cache or L2 cache, right?
>
> We don't.
>
> > If so, do you think if it's possible to retrieve more accurate info
> > from the field "record->type"?
>
> No, we just don't know and it really doesn't matter. The main reason to
> understand the source is to understand the penalty of data coming from
> the source and that it's coming from a core should be sufficient.

Okay, the question is Neoverse has three different data sources
"ARM_SPE_NV_PEER_CORE", "ARM_SPE_NV_LCL_CLSTR" and
"ARM_SPE_NV_PEER_CLSTR", but the patch only uses the same attribution
for all of them.

To be honest, I don't have precise understanding the definition for
these three types, seems to me "ARM_SPE_NV_PEER_CORE" means to fetch
data cache from peer core (like SMT things), "ARM_SPE_NV_LCL_CLSTR"
means cache conherency within the same cluster with SCU,
"ARM_SPE_NV_PEER_CLSTR" means the conherency happens with external bus
(like CCI or CMN). So I'd like to suggest to consider to extend the
level definitions so can allow us to express the data source for Arm
arch precisely.

It's important to understand current cache level definitions which is
derived from x86 arch and think what's the good way to match and
extend for Arm memory hierarchy. I will think a bit more for this,
and if have any idea will share back.

Thanks,
Leo

2022-03-14 11:39:47

by Ali Saidi

[permalink] [raw]
Subject: Re: [PATCH v2 1/2] perf arm-spe: Use SPE data source for neoverse cores

On Sun, 13 Mar 2022 11:47:58 +0000 Leo Yan wrote:
> Hi Ali, German,
>
> On Wed, Mar 02, 2022 at 11:59:05AM +0000, German Gomez wrote:
> > Hi Ali,
> >
> > On 21/02/2022 22:47, Ali Saidi wrote:
> > > When synthesizing data from SPE, augment the type with source information
> > > for Arm Neoverse cores. The field is IMPLDEF but the Neoverse cores all use
> > > the same encoding. I can't find encoding information for any other SPE
> > > implementations to unify their choices with Arm's thus that is left for
> > > future work.
[snip]
> > >
> > > +static void arm_spe__synth_data_source_neoverse(const struct arm_spe_record *record,
> > > + union perf_mem_data_src *data_src)
> > > +{
> > > + switch (record->source) {
> > > + case ARM_SPE_NV_L1D:
> > > + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> >
> > I understand mem_lvl is deprecated but shouldn't we add the level bits here as well for backwards compat?
>
> Thanks for pointing out this. Yeah, I think German's suggestion is
> valid, the commit 6ae5fa61d27d ("perf/x86: Fix data source decoding
> for Skylake") introduces new field 'mem_lvl_num', but it also keeps
> backwards compatible for the field 'mem_lvl'.
I thought about that, but then I'm making some assumption about how to fit
this into the old LVL framework, which is perhaps OK (afaik there are no
Neoverse systems with more than 3 cache levels). What stopped me was that
perf_mem__lvl_scnprintf() does the wrong thing when both are set so I
assumed that setting both was not the right course of action.

>
> > > + data_src->mem_lvl_num = PERF_MEM_LVLNUM_L1;
> > > + break;
> > > + case ARM_SPE_NV_L2:
> > > + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> > > + data_src->mem_lvl_num = PERF_MEM_LVLNUM_L2;
> > > + break;
> > > + case ARM_SPE_NV_PEER_CORE:
> > > + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> > > + data_src->mem_snoop = PERF_MEM_SNOOP_HITM;
> > > + data_src->mem_lvl_num = PERF_MEM_LVLNUM_ANY_CACHE;
>
> For PEER_CORE data source, we don't know if it's coming from peer
> core's L1 cache or L2 cache, right?
We don't.

>
> If so, do you think if it's possible to retrieve more accurate info
> from the field "record->type"?
No, we just don't know and it really doesn't matter. The main reason to
understand the source is to understand the penalty of data coming from
the source and that it's coming from a core should be sufficient.

>
> > > + break;
> > > + /*
> > > + * We don't know if this is L1, L2, or even L3 (for the cases the system
> > > + * has an L3, but we do know it was a cache-2-cache transfer, so set
> > > + * SNOOP_HITM
> > > + */
> > > + case ARM_SPE_NV_LCL_CLSTR:
> > > + case ARM_SPE_NV_PEER_CLSTR:
> > > + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> > > + data_src->mem_snoop = PERF_MEM_SNOOP_HITM;
> > > + data_src->mem_lvl_num = PERF_MEM_LVLNUM_ANY_CACHE;
>
> Seems to me, we need to add attribution to indicate the difference
> between ARM_SPE_NV_PEER_CORE and ARM_SPE_NV_LCL_CLSTR.
I don't think we really do, see my reasoning above.

>
> For ARM_SPE_NV_PEER_CLSTR data source, should we set any "remote"
> attribution as well?

No, we should leave remote for data coming from another chip/socket
which is really impactful.

>
> > > + break;
> > > + /*
> > > + * System cache is assumed to be L4, as cluster cache (if it exists)
> > > + * would be L3 cache on Neoverse platforms
> > > + */
> > > + case ARM_SPE_NV_SYS_CACHE:
> > > + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> > > + data_src->mem_lvl_num = PERF_MEM_LVLNUM_L4;
> > > + break;
> > > + /*
> > > + * We don't know what level it hit in, except it came from the other
> > > + * socket
> > > + */
> > > + case ARM_SPE_NV_REMOTE:
> > > + data_src->mem_snoop = PERF_MEM_SNOOP_HITM;
> > > + data_src->mem_remote = PERF_MEM_REMOTE_REMOTE;
> > > + break;
>
> Just curious, is it possible that 'record->source' combines multiple
> bits? Like we can get a data source value with:
>
> ARM_SPE_NV_REMOTE | ARM_SPE_NV_REMOTE

source encodes a single value (not bits that represent flags) on Neoverse
cores.

[snip]
> > > @@ -796,6 +868,10 @@ static int arm_spe_process_event(struct perf_session *session,
> > > u64 timestamp;
> > > struct arm_spe *spe = container_of(session->auxtrace,
> > > struct arm_spe, auxtrace);
> > > + const char *cpuid = perf_env__cpuid(session->evlist->env);
> > > + u64 midr = strtol(cpuid, NULL, 16);
> > > +
> > > + spe->midr = midr;
> >
> > I think this midr setup belongs in the arm_spe_process_auxtrace_info callback instead.
>
> Yeah, arm_spe_process_event() would be invoked for multiple times for
> processing perf events. arm_spe_process_auxtrace_info() would be a
> good place to initialize midr.

Will do.

Thanks,
Ali

2022-03-14 16:38:31

by Leo Yan

[permalink] [raw]
Subject: Re: [PATCH v2 1/2] perf arm-spe: Use SPE data source for neoverse cores

Hi Ali, German,

On Wed, Mar 02, 2022 at 11:59:05AM +0000, German Gomez wrote:
> Hi Ali,
>
> On 21/02/2022 22:47, Ali Saidi wrote:
> > When synthesizing data from SPE, augment the type with source information
> > for Arm Neoverse cores. The field is IMPLDEF but the Neoverse cores all use
> > the same encoding. I can't find encoding information for any other SPE
> > implementations to unify their choices with Arm's thus that is left for
> > future work.
> >
> > This change populates the mem_lvl_num for Neoverse cores instead of the
> > deprecated mem_lvl namespace.
> >
> > Signed-off-by: Ali Saidi <[email protected]>
> > ---
> > .../util/arm-spe-decoder/arm-spe-decoder.c | 1 +
> > .../util/arm-spe-decoder/arm-spe-decoder.h | 12 ++
> > tools/perf/util/arm-spe.c | 106 +++++++++++++++---
> > 3 files changed, 104 insertions(+), 15 deletions(-)
> >
> > diff --git a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c
> > index 5e390a1a79ab..091987dd3966 100644
> > --- a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c
> > +++ b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.c
> > @@ -220,6 +220,7 @@ static int arm_spe_read_record(struct arm_spe_decoder *decoder)
> >
> > break;
> > case ARM_SPE_DATA_SOURCE:
> > + decoder->record.source = payload;
> > break;
> > case ARM_SPE_BAD:
> > break;
> > diff --git a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h
> > index 69b31084d6be..c81bf90c0996 100644
> > --- a/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h
> > +++ b/tools/perf/util/arm-spe-decoder/arm-spe-decoder.h
> > @@ -29,6 +29,17 @@ enum arm_spe_op_type {
> > ARM_SPE_ST = 1 << 1,
> > };
> >
> > +enum arm_spe_neoverse_data_source {
> > + ARM_SPE_NV_L1D = 0x0,
> > + ARM_SPE_NV_L2 = 0x8,
> > + ARM_SPE_NV_PEER_CORE = 0x9,
> > + ARM_SPE_NV_LCL_CLSTR = 0xa,
> > + ARM_SPE_NV_SYS_CACHE = 0xb,
> > + ARM_SPE_NV_PEER_CLSTR = 0xc,
> > + ARM_SPE_NV_REMOTE = 0xd,
> > + ARM_SPE_NV_DRAM = 0xe,
> > +};
> > +
> > struct arm_spe_record {
> > enum arm_spe_sample_type type;
> > int err;
> > @@ -40,6 +51,7 @@ struct arm_spe_record {
> > u64 virt_addr;
> > u64 phys_addr;
> > u64 context_id;
> > + u16 source;
> > };
> >
> > struct arm_spe_insn;
> > diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
> > index d2b64e3f588b..e0243c2fed5f 100644
> > --- a/tools/perf/util/arm-spe.c
> > +++ b/tools/perf/util/arm-spe.c
> > @@ -34,6 +34,7 @@
> > #include "arm-spe-decoder/arm-spe-decoder.h"
> > #include "arm-spe-decoder/arm-spe-pkt-decoder.h"
> >
> > +#include <../../../arch/arm64/include/asm/cputype.h>
> > #define MAX_TIMESTAMP (~0ULL)
> >
> > struct arm_spe {
> > @@ -45,6 +46,7 @@ struct arm_spe {
> > struct perf_session *session;
> > struct machine *machine;
> > u32 pmu_type;
> > + u64 midr;
> >
> > struct perf_tsc_conversion tc;
> >
> > @@ -399,33 +401,103 @@ static bool arm_spe__is_memory_event(enum arm_spe_sample_type type)
> > return false;
> > }
> >
> > -static u64 arm_spe__synth_data_source(const struct arm_spe_record *record)
> > -{
> > - union perf_mem_data_src data_src = { 0 };
> > +static const struct midr_range neoverse_spe[] = {
> > + MIDR_ALL_VERSIONS(MIDR_NEOVERSE_N1),
> > + MIDR_ALL_VERSIONS(MIDR_NEOVERSE_N2),
> > + {},
> > +};
> >
> > - if (record->op == ARM_SPE_LD)
> > - data_src.mem_op = PERF_MEM_OP_LOAD;
> > - else
> > - data_src.mem_op = PERF_MEM_OP_STORE;
> >
> > +static void arm_spe__synth_data_source_neoverse(const struct arm_spe_record *record,
> > + union perf_mem_data_src *data_src)
> > +{
> > + switch (record->source) {
> > + case ARM_SPE_NV_L1D:
> > + data_src->mem_lvl = PERF_MEM_LVL_HIT;
>
> I understand mem_lvl is deprecated but shouldn't we add the level bits here as well for backwards compat?

Thanks for pointing out this. Yeah, I think German's suggestion is
valid, the commit 6ae5fa61d27d ("perf/x86: Fix data source decoding
for Skylake") introduces new field 'mem_lvl_num', but it also keeps
backwards compatible for the field 'mem_lvl'.

> > + data_src->mem_lvl_num = PERF_MEM_LVLNUM_L1;
> > + break;
> > + case ARM_SPE_NV_L2:
> > + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> > + data_src->mem_lvl_num = PERF_MEM_LVLNUM_L2;
> > + break;
> > + case ARM_SPE_NV_PEER_CORE:
> > + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> > + data_src->mem_snoop = PERF_MEM_SNOOP_HITM;
> > + data_src->mem_lvl_num = PERF_MEM_LVLNUM_ANY_CACHE;

For PEER_CORE data source, we don't know if it's coming from peer
core's L1 cache or L2 cache, right?

If so, do you think if it's possible to retrieve more accurate info
from the field "record->type"?

> > + break;
> > + /*
> > + * We don't know if this is L1, L2, or even L3 (for the cases the system
> > + * has an L3, but we do know it was a cache-2-cache transfer, so set
> > + * SNOOP_HITM
> > + */
> > + case ARM_SPE_NV_LCL_CLSTR:
> > + case ARM_SPE_NV_PEER_CLSTR:
> > + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> > + data_src->mem_snoop = PERF_MEM_SNOOP_HITM;
> > + data_src->mem_lvl_num = PERF_MEM_LVLNUM_ANY_CACHE;

Seems to me, we need to add attribution to indicate the difference
between ARM_SPE_NV_PEER_CORE and ARM_SPE_NV_LCL_CLSTR.

For ARM_SPE_NV_PEER_CLSTR data source, should we set any "remote"
attribution as well?

> > + break;
> > + /*
> > + * System cache is assumed to be L4, as cluster cache (if it exists)
> > + * would be L3 cache on Neoverse platforms
> > + */
> > + case ARM_SPE_NV_SYS_CACHE:
> > + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> > + data_src->mem_lvl_num = PERF_MEM_LVLNUM_L4;
> > + break;
> > + /*
> > + * We don't know what level it hit in, except it came from the other
> > + * socket
> > + */
> > + case ARM_SPE_NV_REMOTE:
> > + data_src->mem_snoop = PERF_MEM_SNOOP_HITM;
> > + data_src->mem_remote = PERF_MEM_REMOTE_REMOTE;
> > + break;

Just curious, is it possible that 'record->source' combines multiple
bits? Like we can get a data source value with:

ARM_SPE_NV_REMOTE | ARM_SPE_NV_REMOTE

> > + case ARM_SPE_NV_DRAM:
> > + data_src->mem_lvl = PERF_MEM_LVL_HIT;
> > + data_src->mem_lvl_num = PERF_MEM_LVLNUM_RAM;
> > + break;
> > + default:
> > + break;
> > + }
> > +}
> > +
> > +static void arm_spe__synth_data_source_generic(const struct arm_spe_record *record,
> > + union perf_mem_data_src *data_src)
> > +{
> > if (record->type & (ARM_SPE_LLC_ACCESS | ARM_SPE_LLC_MISS)) {
> > - data_src.mem_lvl = PERF_MEM_LVL_L3;
> > + data_src->mem_lvl = PERF_MEM_LVL_L3;
> >
> > if (record->type & ARM_SPE_LLC_MISS)
> > - data_src.mem_lvl |= PERF_MEM_LVL_MISS;
> > + data_src->mem_lvl |= PERF_MEM_LVL_MISS;
> > else
> > - data_src.mem_lvl |= PERF_MEM_LVL_HIT;
> > + data_src->mem_lvl |= PERF_MEM_LVL_HIT;
> > } else if (record->type & (ARM_SPE_L1D_ACCESS | ARM_SPE_L1D_MISS)) {
> > - data_src.mem_lvl = PERF_MEM_LVL_L1;
> > + data_src->mem_lvl = PERF_MEM_LVL_L1;
> >
> > if (record->type & ARM_SPE_L1D_MISS)
> > - data_src.mem_lvl |= PERF_MEM_LVL_MISS;
> > + data_src->mem_lvl |= PERF_MEM_LVL_MISS;
> > else
> > - data_src.mem_lvl |= PERF_MEM_LVL_HIT;
> > + data_src->mem_lvl |= PERF_MEM_LVL_HIT;
> > }
> >
> > if (record->type & ARM_SPE_REMOTE_ACCESS)
> > - data_src.mem_lvl |= PERF_MEM_LVL_REM_CCE1;
> > + data_src->mem_lvl |= PERF_MEM_LVL_REM_CCE1;
> > +}
> > +
> > +static u64 arm_spe__synth_data_source(const struct arm_spe_record *record, u64 midr)
> > +{
> > + union perf_mem_data_src data_src = { 0 };
> > + bool is_neoverse = is_midr_in_range(midr, neoverse_spe);
> > +
> > + if (record->op & ARM_SPE_LD)
> > + data_src.mem_op = PERF_MEM_OP_LOAD;
> > + else
> > + data_src.mem_op = PERF_MEM_OP_STORE;
> > +
> > + if (is_neoverse)
> > + arm_spe__synth_data_source_neoverse(record, &data_src);
> > + else
> > + arm_spe__synth_data_source_generic(record, &data_src);
> >
> > if (record->type & (ARM_SPE_TLB_ACCESS | ARM_SPE_TLB_MISS)) {
> > data_src.mem_dtlb = PERF_MEM_TLB_WK;
> > @@ -446,7 +518,7 @@ static int arm_spe_sample(struct arm_spe_queue *speq)
> > u64 data_src;
> > int err;
> >
> > - data_src = arm_spe__synth_data_source(record);
> > + data_src = arm_spe__synth_data_source(record, spe->midr);
> >
> > if (spe->sample_flc) {
> > if (record->type & ARM_SPE_L1D_MISS) {
> > @@ -796,6 +868,10 @@ static int arm_spe_process_event(struct perf_session *session,
> > u64 timestamp;
> > struct arm_spe *spe = container_of(session->auxtrace,
> > struct arm_spe, auxtrace);
> > + const char *cpuid = perf_env__cpuid(session->evlist->env);
> > + u64 midr = strtol(cpuid, NULL, 16);
> > +
> > + spe->midr = midr;
>
> I think this midr setup belongs in the arm_spe_process_auxtrace_info callback instead.

Yeah, arm_spe_process_event() would be invoked for multiple times for
processing perf events. arm_spe_process_auxtrace_info() would be a
good place to initialize midr.

Thanks,
Leo