
Topdown level 1 and 2 for Intel Ice Lake (ICX). Limiting topdown support to THREAD for the moment on Ice Lake, as NODE support is still unreliable. Also removing Topdown Level 1 from Sapphire Rapids onwards, as Topdown LeveL 2 also shows Level 1 on Sapphire, and it reduces the overall number of bundles. Type: improvement Signed-off-by: Ray Kinsella <mdr@ashroe.eu> Change-Id: Iaa68b711dc8b6fb1090880b411debadb3c37f8bc
177 lines
4.6 KiB
C
177 lines
4.6 KiB
C
/*
|
|
* Copyright (c) 2022 Intel and/or its affiliates.
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at:
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
*/
|
|
#include <perfmon/perfmon.h>
|
|
#include <perfmon/intel/core.h>
|
|
|
|
static int
|
|
is_icelake ()
|
|
{
|
|
return clib_cpu_supports_avx512_bitalg () && !clib_cpu_supports_movdir64b ();
|
|
}
|
|
|
|
static perfmon_cpu_supports_t topdown_lvl2_cpu_supports_icx[] = {
|
|
{ is_icelake, PERFMON_BUNDLE_TYPE_THREAD }
|
|
};
|
|
|
|
#define GET_METRIC(m, i) (f64) (((m) >> (i * 8)) & 0xff)
|
|
|
|
enum
|
|
{
|
|
TD_SLOTS = 0,
|
|
STALLS_MEM_ANY,
|
|
STALLS_TOTAL,
|
|
BOUND_ON_STORES,
|
|
RECOVERY_CYCLES,
|
|
UOP_DROPPING,
|
|
UOP_NOT_DELIVERED,
|
|
TD_RETIRING,
|
|
TD_BAD_SPEC,
|
|
TD_FE_BOUND,
|
|
TD_BE_BOUND,
|
|
};
|
|
|
|
static_always_inline f64
|
|
memory_bound_fraction (perfmon_reading_t *ss)
|
|
{
|
|
return (ss->value[STALLS_MEM_ANY] + ss->value[BOUND_ON_STORES]) /
|
|
(f64) (ss->value[STALLS_TOTAL] + ss->value[BOUND_ON_STORES]);
|
|
}
|
|
|
|
static_always_inline f64
|
|
perf_metrics_sum (perfmon_reading_t *ss)
|
|
{
|
|
return ss->value[TD_RETIRING] + ss->value[TD_BAD_SPEC] +
|
|
ss->value[TD_FE_BOUND] + ss->value[TD_BE_BOUND];
|
|
}
|
|
|
|
static_always_inline f64
|
|
retiring (perfmon_reading_t *ss)
|
|
{
|
|
return ss->value[TD_RETIRING] / perf_metrics_sum (ss);
|
|
}
|
|
|
|
static_always_inline f64
|
|
bad_speculation (perfmon_reading_t *ss)
|
|
{
|
|
return ss->value[TD_BAD_SPEC] / perf_metrics_sum (ss);
|
|
}
|
|
|
|
static_always_inline f64
|
|
frontend_bound (perfmon_reading_t *ss)
|
|
{
|
|
return (ss->value[TD_FE_BOUND] / perf_metrics_sum (ss)) -
|
|
(ss->value[UOP_DROPPING] / perf_metrics_sum (ss));
|
|
}
|
|
|
|
static_always_inline f64
|
|
backend_bound (perfmon_reading_t *ss)
|
|
{
|
|
return (ss->value[TD_BE_BOUND] / perf_metrics_sum (ss)) +
|
|
((5 * ss->value[RECOVERY_CYCLES]) / perf_metrics_sum (ss));
|
|
}
|
|
|
|
static_always_inline f64
|
|
fetch_latency (perfmon_reading_t *ss)
|
|
{
|
|
f64 r = ((5 * ss->value[UOP_NOT_DELIVERED] - ss->value[UOP_DROPPING]) /
|
|
(f64) ss->value[TD_SLOTS]);
|
|
return r;
|
|
}
|
|
|
|
static_always_inline f64
|
|
fetch_bandwidth (perfmon_reading_t *ss)
|
|
{
|
|
return clib_max (0, frontend_bound (ss) - fetch_latency (ss));
|
|
}
|
|
|
|
static_always_inline f64
|
|
memory_bound (perfmon_reading_t *ss)
|
|
{
|
|
return backend_bound (ss) * memory_bound_fraction (ss);
|
|
}
|
|
|
|
static_always_inline f64
|
|
core_bound (perfmon_reading_t *ss)
|
|
{
|
|
return backend_bound (ss) - memory_bound (ss);
|
|
}
|
|
|
|
static u8 *
|
|
format_topdown_lvl2_icx (u8 *s, va_list *args)
|
|
{
|
|
perfmon_reading_t *ss = va_arg (*args, perfmon_reading_t *);
|
|
u64 idx = va_arg (*args, int);
|
|
f64 sv = 0;
|
|
|
|
switch (idx)
|
|
{
|
|
case 0:
|
|
sv = retiring (ss);
|
|
break;
|
|
case 1:
|
|
sv = bad_speculation (ss);
|
|
break;
|
|
case 2:
|
|
sv = frontend_bound (ss);
|
|
break;
|
|
case 3:
|
|
sv = backend_bound (ss);
|
|
break;
|
|
case 4:
|
|
sv = fetch_latency (ss);
|
|
break;
|
|
case 5:
|
|
sv = fetch_bandwidth (ss);
|
|
break;
|
|
case 6:
|
|
sv = memory_bound (ss);
|
|
break;
|
|
case 7:
|
|
sv = core_bound (ss);
|
|
break;
|
|
}
|
|
|
|
s = format (s, "%f", sv * 100);
|
|
|
|
return s;
|
|
}
|
|
|
|
PERFMON_REGISTER_BUNDLE (topdown_lvl2_metric_icx) = {
|
|
.name = "topdown",
|
|
.description = "Top-down Microarchitecture Analysis Level 1 & 2",
|
|
.source = "intel-core",
|
|
.events[0] = INTEL_CORE_E_TOPDOWN_SLOTS,
|
|
.events[1] = INTEL_CORE_E_CYCLE_ACTIVITY_STALLS_MEM_ANY,
|
|
.events[2] = INTEL_CORE_E_CYCLE_ACTIVITY_STALLS_TOTAL,
|
|
.events[3] = INTEL_CORE_E_EXE_ACTIVITY_BOUND_ON_STORES,
|
|
.events[4] = INTEL_CORE_E_INT_MISC_RECOVERY_CYCLES,
|
|
.events[5] = INTEL_CORE_E_INT_MISC_UOP_DROPPING,
|
|
.events[6] = INTEL_CORE_E_IDQ_UOPS_NOT_DELIVERED_CORE,
|
|
.events[7] = INTEL_CORE_E_TOPDOWN_L1_RETIRING_METRIC,
|
|
.events[8] = INTEL_CORE_E_TOPDOWN_L1_BAD_SPEC_METRIC,
|
|
.events[9] = INTEL_CORE_E_TOPDOWN_L1_FE_BOUND_METRIC,
|
|
.events[10] = INTEL_CORE_E_TOPDOWN_L1_BE_BOUND_METRIC,
|
|
.n_events = 11,
|
|
.cpu_supports = topdown_lvl2_cpu_supports_icx,
|
|
.n_cpu_supports = ARRAY_LEN (topdown_lvl2_cpu_supports_icx),
|
|
.format_fn = format_topdown_lvl2_icx,
|
|
.column_headers = PERFMON_STRINGS ("% RT", "% BS", "% FE", "% BE", "% FE.FL",
|
|
"% FE.FB", "% BE.MB", "% BE.CB"),
|
|
.footer = "Retiring (RT), Bad Speculation (BS),\n"
|
|
" FrontEnd bound (FE), BackEnd bound (BE),\n"
|
|
" Fetch Latency (FL), Fetch Bandwidth (FB),\n"
|
|
" Memory Bound (MB), Core Bound (CB)",
|
|
};
|