q3k/linux
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity
linux/arch/powerpc/kernel/power5-pmu.c

571 lines
15 KiB
C
Raw Normal View History

perfcounters/powerpc: Add support for POWER5 processors This adds the back-end for the PMU on the POWER5 processor. This knows how to use the fixed-function PMC5 and PMC6 (instructions completed and run cycles). Unlike POWER6, PMC5/6 obey the freeze conditions and can generate interrupts, so their use doesn't impose any extra restrictions. POWER5+ is different and is not supported by this patch. Signed-off-by: Paul Mackerras <paulus@samba.org>
2009-02-24 00:33:56 +00:00
/*
* Performance counter support for POWER5 (not POWER5++) processors.
*
* Copyright 2009 Paul Mackerras, IBM Corporation.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version
* 2 of the License, or (at your option) any later version.
*/
#include <linux/kernel.h>
#include <linux/perf_counter.h>
#include <asm/reg.h>
/*
* Bits in event code for POWER5 (not POWER5++)
*/
#define PM_PMC_SH 20 /* PMC number (1-based) for direct events */
#define PM_PMC_MSK 0xf
#define PM_PMC_MSKS (PM_PMC_MSK << PM_PMC_SH)
#define PM_UNIT_SH 16 /* TTMMUX number and setting - unit select */
#define PM_UNIT_MSK 0xf
#define PM_BYTE_SH 12 /* Byte number of event bus to use */
#define PM_BYTE_MSK 7
#define PM_GRS_SH 8 /* Storage subsystem mux select */
#define PM_GRS_MSK 7
#define PM_BUSEVENT_MSK 0x80 /* Set if event uses event bus */
#define PM_PMCSEL_MSK 0x7f
/* Values in PM_UNIT field */
#define PM_FPU 0
#define PM_ISU0 1
#define PM_IFU 2
#define PM_ISU1 3
#define PM_IDU 4
#define PM_ISU0_ALT 6
#define PM_GRS 7
#define PM_LSU0 8
#define PM_LSU1 0xc
#define PM_LASTUNIT 0xc
/*
* Bits in MMCR1 for POWER5
*/
#define MMCR1_TTM0SEL_SH 62
#define MMCR1_TTM1SEL_SH 60
#define MMCR1_TTM2SEL_SH 58
#define MMCR1_TTM3SEL_SH 56
#define MMCR1_TTMSEL_MSK 3
#define MMCR1_TD_CP_DBG0SEL_SH 54
#define MMCR1_TD_CP_DBG1SEL_SH 52
#define MMCR1_TD_CP_DBG2SEL_SH 50
#define MMCR1_TD_CP_DBG3SEL_SH 48
#define MMCR1_GRS_L2SEL_SH 46
#define MMCR1_GRS_L2SEL_MSK 3
#define MMCR1_GRS_L3SEL_SH 44
#define MMCR1_GRS_L3SEL_MSK 3
#define MMCR1_GRS_MCSEL_SH 41
#define MMCR1_GRS_MCSEL_MSK 7
#define MMCR1_GRS_FABSEL_SH 39
#define MMCR1_GRS_FABSEL_MSK 3
#define MMCR1_PMC1_ADDER_SEL_SH 35
#define MMCR1_PMC2_ADDER_SEL_SH 34
#define MMCR1_PMC3_ADDER_SEL_SH 33
#define MMCR1_PMC4_ADDER_SEL_SH 32
#define MMCR1_PMC1SEL_SH 25
#define MMCR1_PMC2SEL_SH 17
#define MMCR1_PMC3SEL_SH 9
#define MMCR1_PMC4SEL_SH 1
#define MMCR1_PMCSEL_SH(n) (MMCR1_PMC1SEL_SH - (n) * 8)
#define MMCR1_PMCSEL_MSK 0x7f
/*
* Bits in MMCRA
*/
/*
* Layout of constraint bits:
* 6666555555555544444444443333333333222222222211111111110000000000
* 3210987654321098765432109876543210987654321098765432109876543210
* <><>[ ><><>< ><> [ >[ >[ >< >< >< >< ><><><><><><>
* T0T1 NC G0G1G2 G3 UC PS1PS2 B0 B1 B2 B3 P6P5P4P3P2P1
*
* T0 - TTM0 constraint
* 54-55: TTM0SEL value (0=FPU, 2=IFU, 3=ISU1) 0xc0_0000_0000_0000
*
* T1 - TTM1 constraint
* 52-53: TTM1SEL value (0=IDU, 3=GRS) 0x30_0000_0000_0000
*
* NC - number of counters
* 51: NC error 0x0008_0000_0000_0000
* 48-50: number of events needing PMC1-4 0x0007_0000_0000_0000
*
* G0..G3 - GRS mux constraints
* 46-47: GRS_L2SEL value
* 44-45: GRS_L3SEL value
* 41-44: GRS_MCSEL value
* 39-40: GRS_FABSEL value
* Note that these match up with their bit positions in MMCR1
*
* UC - unit constraint: can't have all three of FPU|IFU|ISU1, ISU0, IDU|GRS
* 37: UC3 error 0x20_0000_0000
* 36: FPU|IFU|ISU1 events needed 0x10_0000_0000
* 35: ISU0 events needed 0x08_0000_0000
* 34: IDU|GRS events needed 0x04_0000_0000
*
* PS1
* 33: PS1 error 0x2_0000_0000
* 31-32: count of events needing PMC1/2 0x1_8000_0000
*
* PS2
* 30: PS2 error 0x4000_0000
* 28-29: count of events needing PMC3/4 0x3000_0000
*
* B0
* 24-27: Byte 0 event source 0x0f00_0000
* Encoding as for the event code
*
* B1, B2, B3
* 20-23, 16-19, 12-15: Byte 1, 2, 3 event sources
*
* P1..P6
* 0-11: Count of events needing PMC1..PMC6
*/
static const int grsel_shift[8] = {
MMCR1_GRS_L2SEL_SH, MMCR1_GRS_L2SEL_SH, MMCR1_GRS_L2SEL_SH,
MMCR1_GRS_L3SEL_SH, MMCR1_GRS_L3SEL_SH, MMCR1_GRS_L3SEL_SH,
MMCR1_GRS_MCSEL_SH, MMCR1_GRS_FABSEL_SH
};
/* Masks and values for using events from the various units */
static u64 unit_cons[PM_LASTUNIT+1][2] = {
[PM_FPU] = { 0xc0002000000000ull, 0x00001000000000ull },
[PM_ISU0] = { 0x00002000000000ull, 0x00000800000000ull },
[PM_ISU1] = { 0xc0002000000000ull, 0xc0001000000000ull },
[PM_IFU] = { 0xc0002000000000ull, 0x80001000000000ull },
[PM_IDU] = { 0x30002000000000ull, 0x00000400000000ull },
[PM_GRS] = { 0x30002000000000ull, 0x30000400000000ull },
};
static int power5_get_constraint(unsigned int event, u64 *maskp, u64 *valp)
{
int pmc, byte, unit, sh;
int bit, fmask;
u64 mask = 0, value = 0;
int grp = -1;
pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
if (pmc) {
if (pmc > 6)
return -1;
sh = (pmc - 1) * 2;
mask |= 2 << sh;
value |= 1 << sh;
if (pmc <= 4)
grp = (pmc - 1) >> 1;
else if (event != 0x500009 && event != 0x600005)
return -1;
}
if (event & PM_BUSEVENT_MSK) {
unit = (event >> PM_UNIT_SH) & PM_UNIT_MSK;
if (unit > PM_LASTUNIT)
return -1;
if (unit == PM_ISU0_ALT)
unit = PM_ISU0;
mask |= unit_cons[unit][0];
value |= unit_cons[unit][1];
byte = (event >> PM_BYTE_SH) & PM_BYTE_MSK;
if (byte >= 4) {
if (unit != PM_LSU1)
return -1;
/* Map LSU1 low word (bytes 4-7) to unit LSU1+1 */
++unit;
byte &= 3;
}
if (unit == PM_GRS) {
bit = event & 7;
fmask = (bit == 6)? 7: 3;
sh = grsel_shift[bit];
mask |= (u64)fmask << sh;
value |= (u64)((event >> PM_GRS_SH) & fmask) << sh;
}
/*
* Bus events on bytes 0 and 2 can be counted
* on PMC1/2; bytes 1 and 3 on PMC3/4.
*/
if (!pmc)
grp = byte & 1;
/* Set byte lane select field */
mask |= 0xfULL << (24 - 4 * byte);
value |= (u64)unit << (24 - 4 * byte);
}
if (grp == 0) {
/* increment PMC1/2 field */
mask |= 0x200000000ull;
value |= 0x080000000ull;
} else if (grp == 1) {
/* increment PMC3/4 field */
mask |= 0x40000000ull;
value |= 0x10000000ull;
}
if (pmc < 5) {
/* need a counter from PMC1-4 set */
mask |= 0x8000000000000ull;
value |= 0x1000000000000ull;
}
*maskp = mask;
*valp = value;
return 0;
}
#define MAX_ALT 3 /* at most 3 alternatives for any event */
static const unsigned int event_alternatives[][MAX_ALT] = {
{ 0x120e4, 0x400002 }, /* PM_GRP_DISP_REJECT */
{ 0x410c7, 0x441084 }, /* PM_THRD_L2MISS_BOTH_CYC */
{ 0x100005, 0x600005 }, /* PM_RUN_CYC */
{ 0x100009, 0x200009, 0x500009 }, /* PM_INST_CMPL */
{ 0x300009, 0x400009 }, /* PM_INST_DISP */
};
/*
* Scan the alternatives table for a match and return the
* index into the alternatives table if found, else -1.
*/
static int find_alternative(unsigned int event)
{
int i, j;
for (i = 0; i < ARRAY_SIZE(event_alternatives); ++i) {
if (event < event_alternatives[i][0])
break;
for (j = 0; j < MAX_ALT && event_alternatives[i][j]; ++j)
if (event == event_alternatives[i][j])
return i;
}
return -1;
}
static const unsigned char bytedecode_alternatives[4][4] = {
/* PMC 1 */ { 0x21, 0x23, 0x25, 0x27 },
/* PMC 2 */ { 0x07, 0x17, 0x0e, 0x1e },
/* PMC 3 */ { 0x20, 0x22, 0x24, 0x26 },
/* PMC 4 */ { 0x07, 0x17, 0x0e, 0x1e }
};
/*
* Some direct events for decodes of event bus byte 3 have alternative
* PMCSEL values on other counters. This returns the alternative
* event code for those that do, or -1 otherwise.
*/
static int find_alternative_bdecode(unsigned int event)
{
int pmc, altpmc, pp, j;
pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
if (pmc == 0 || pmc > 4)
return -1;
altpmc = 5 - pmc; /* 1 <-> 4, 2 <-> 3 */
pp = event & PM_PMCSEL_MSK;
for (j = 0; j < 4; ++j) {
if (bytedecode_alternatives[pmc - 1][j] == pp) {
return (event & ~(PM_PMC_MSKS | PM_PMCSEL_MSK)) |
(altpmc << PM_PMC_SH) |
bytedecode_alternatives[altpmc - 1][j];
}
}
return -1;
}
perf_counter: powerpc: allow use of limited-function counters POWER5+ and POWER6 have two hardware counters with limited functionality: PMC5 counts instructions completed in run state and PMC6 counts cycles in run state. (Run state is the state when a hardware RUN bit is 1; the idle task clears RUN while waiting for work to do and sets it when there is work to do.) These counters can't be written to by the kernel, can't generate interrupts, and don't obey the freeze conditions. That means we can only use them for per-task counters (where we know we'll always be in run state; we can't put a per-task counter on an idle task), and only if we don't want interrupts and we do want to count in all processor modes. Obviously some counters can't go on a limited hardware counter, but there are also situations where we can only put a counter on a limited hardware counter - if there are already counters on that exclude some processor modes and we want to put on a per-task cycle or instruction counter that doesn't exclude any processor mode, it could go on if it can use a limited hardware counter. To keep track of these constraints, this adds a flags argument to the processor-specific get_alternatives() functions, with three bits defined: one to say that we can accept alternative event codes that go on limited counters, one to say we only want alternatives on limited counters, and one to say that this is a per-task counter and therefore events that are gated by run state are equivalent to those that aren't (e.g. a "cycles" event is equivalent to a "cycles in run state" event). These flags are computed for each counter and stored in the counter->hw.counter_base field (slightly wonky name for what it does, but it was an existing unused field). Since the limited counters don't freeze when we freeze the other counters, we need some special handling to avoid getting skew between things counted on the limited counters and those counted on normal counters. To minimize this skew, if we are using any limited counters, we read PMC5 and PMC6 immediately after setting and clearing the freeze bit. This is done in a single asm in the new write_mmcr0() function. The code here is specific to PMC5 and PMC6 being the limited hardware counters. Being more general (e.g. having a bitmap of limited hardware counter numbers) would have meant more complex code to read the limited counters when freezing and unfreezing the normal counters, with conditional branches, which would have increased the skew. Since it isn't necessary for the code to be more general at this stage, it isn't. This also extends the back-ends for POWER5+ and POWER6 to be able to handle up to 6 counters rather than the 4 they previously handled. Signed-off-by: Paul Mackerras <paulus@samba.org> Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Robert Richter <robert.richter@amd.com> LKML-Reference: <18936.19035.163066.892208@cargo.ozlabs.ibm.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-29 12:38:51 +00:00
static int power5_get_alternatives(unsigned int event, unsigned int flags,
unsigned int alt[])
perfcounters/powerpc: Add support for POWER5 processors This adds the back-end for the PMU on the POWER5 processor. This knows how to use the fixed-function PMC5 and PMC6 (instructions completed and run cycles). Unlike POWER6, PMC5/6 obey the freeze conditions and can generate interrupts, so their use doesn't impose any extra restrictions. POWER5+ is different and is not supported by this patch. Signed-off-by: Paul Mackerras <paulus@samba.org>
2009-02-24 00:33:56 +00:00
{
int i, j, ae, nalt = 1;
alt[0] = event;
nalt = 1;
i = find_alternative(event);
if (i >= 0) {
for (j = 0; j < MAX_ALT; ++j) {
ae = event_alternatives[i][j];
if (ae && ae != event)
alt[nalt++] = ae;
}
} else {
ae = find_alternative_bdecode(event);
if (ae > 0)
alt[nalt++] = ae;
}
return nalt;
}
perf_counter: powerpc: set sample enable bit for marked instruction events Impact: enable access to hardware feature POWER processors have the ability to "mark" a subset of the instructions and provide more detailed information on what happens to the marked instructions as they flow through the pipeline. This marking is enabled by the "sample enable" bit in MMCRA, and there are synchronization requirements around setting and clearing the bit. This adds logic to the processor-specific back-ends so that they know which events relate to marked instructions and set the sampling enable bit if any event that we want to put on the PMU is a marked instruction event. It also adds logic to the generic powerpc code to do the necessary synchronization if that bit is set. Signed-off-by: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <18908.31930.1024.228867@cargo.ozlabs.ibm.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-08 10:30:18 +00:00
/*
* Map of which direct events on which PMCs are marked instruction events.
* Indexed by PMCSEL value, bit i (LE) set if PMC i is a marked event.
* Bit 0 is set if it is marked for all PMCs.
* The 0x80 bit indicates a byte decode PMCSEL value.
*/
static unsigned char direct_event_is_marked[0x28] = {
0, /* 00 */
0x1f, /* 01 PM_IOPS_CMPL */
0x2, /* 02 PM_MRK_GRP_DISP */
0xe, /* 03 PM_MRK_ST_CMPL, PM_MRK_ST_GPS, PM_MRK_ST_CMPL_INT */
0, /* 04 */
0x1c, /* 05 PM_MRK_BRU_FIN, PM_MRK_INST_FIN, PM_MRK_CRU_FIN */
0x80, /* 06 */
0x80, /* 07 */
0, 0, 0,/* 08 - 0a */
0x18, /* 0b PM_THRESH_TIMEO, PM_MRK_GRP_TIMEO */
0, /* 0c */
0x80, /* 0d */
0x80, /* 0e */
0, /* 0f */
0, /* 10 */
0x14, /* 11 PM_MRK_GRP_BR_REDIR, PM_MRK_GRP_IC_MISS */
0, /* 12 */
0x10, /* 13 PM_MRK_GRP_CMPL */
0x1f, /* 14 PM_GRP_MRK, PM_MRK_{FXU,FPU,LSU}_FIN */
0x2, /* 15 PM_MRK_GRP_ISSUED */
0x80, /* 16 */
0x80, /* 17 */
0, 0, 0, 0, 0,
0x80, /* 1d */
0x80, /* 1e */
0, /* 1f */
0x80, /* 20 */
0x80, /* 21 */
0x80, /* 22 */
0x80, /* 23 */
0x80, /* 24 */
0x80, /* 25 */
0x80, /* 26 */
0x80, /* 27 */
};
/*
* Returns 1 if event counts things relating to marked instructions
* and thus needs the MMCRA_SAMPLE_ENABLE bit set, or 0 if not.
*/
static int power5_marked_instr_event(unsigned int event)
{
int pmc, psel;
int bit, byte, unit;
u32 mask;
pmc = (event >> PM_PMC_SH) & PM_PMC_MSK;
psel = event & PM_PMCSEL_MSK;
if (pmc >= 5)
return 0;
bit = -1;
if (psel < sizeof(direct_event_is_marked)) {
if (direct_event_is_marked[psel] & (1 << pmc))
return 1;
if (direct_event_is_marked[psel] & 0x80)
bit = 4;
else if (psel == 0x08)
bit = pmc - 1;
else if (psel == 0x10)
bit = 4 - pmc;
else if (psel == 0x1b && (pmc == 1 || pmc == 3))
bit = 4;
} else if ((psel & 0x58) == 0x40)
bit = psel & 7;
if (!(event & PM_BUSEVENT_MSK))
return 0;
byte = (event >> PM_BYTE_SH) & PM_BYTE_MSK;
unit = (event >> PM_UNIT_SH) & PM_UNIT_MSK;
if (unit == PM_LSU0) {
/* byte 1 bits 0-7, byte 2 bits 0,2-4,6 */
mask = 0x5dff00;
} else if (unit == PM_LSU1 && byte >= 4) {
byte -= 4;
/* byte 4 bits 1,3,5,7, byte 5 bits 6-7, byte 7 bits 0-4,6 */
mask = 0x5f00c0aa;
} else
return 0;
return (mask >> (byte * 8 + bit)) & 1;
}
perfcounters/powerpc: Add support for POWER5 processors This adds the back-end for the PMU on the POWER5 processor. This knows how to use the fixed-function PMC5 and PMC6 (instructions completed and run cycles). Unlike POWER6, PMC5/6 obey the freeze conditions and can generate interrupts, so their use doesn't impose any extra restrictions. POWER5+ is different and is not supported by this patch. Signed-off-by: Paul Mackerras <paulus@samba.org>
2009-02-24 00:33:56 +00:00
static int power5_compute_mmcr(unsigned int event[], int n_ev,
unsigned int hwc[], u64 mmcr[])
{
u64 mmcr1 = 0;
perf_counter: powerpc: set sample enable bit for marked instruction events Impact: enable access to hardware feature POWER processors have the ability to "mark" a subset of the instructions and provide more detailed information on what happens to the marked instructions as they flow through the pipeline. This marking is enabled by the "sample enable" bit in MMCRA, and there are synchronization requirements around setting and clearing the bit. This adds logic to the processor-specific back-ends so that they know which events relate to marked instructions and set the sampling enable bit if any event that we want to put on the PMU is a marked instruction event. It also adds logic to the generic powerpc code to do the necessary synchronization if that bit is set. Signed-off-by: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <18908.31930.1024.228867@cargo.ozlabs.ibm.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-08 10:30:18 +00:00
u64 mmcra = 0;
perfcounters/powerpc: Add support for POWER5 processors This adds the back-end for the PMU on the POWER5 processor. This knows how to use the fixed-function PMC5 and PMC6 (instructions completed and run cycles). Unlike POWER6, PMC5/6 obey the freeze conditions and can generate interrupts, so their use doesn't impose any extra restrictions. POWER5+ is different and is not supported by this patch. Signed-off-by: Paul Mackerras <paulus@samba.org>
2009-02-24 00:33:56 +00:00
unsigned int pmc, unit, byte, psel;
unsigned int ttm, grp;
int i, isbus, bit, grsel;
unsigned int pmc_inuse = 0;
unsigned int pmc_grp_use[2];
unsigned char busbyte[4];
unsigned char unituse[16];
int ttmuse;
if (n_ev > 6)
return -1;
/* First pass to count resource use */
pmc_grp_use[0] = pmc_grp_use[1] = 0;
memset(busbyte, 0, sizeof(busbyte));
memset(unituse, 0, sizeof(unituse));
for (i = 0; i < n_ev; ++i) {
pmc = (event[i] >> PM_PMC_SH) & PM_PMC_MSK;
if (pmc) {
if (pmc > 6)
return -1;
if (pmc_inuse & (1 << (pmc - 1)))
return -1;
pmc_inuse |= 1 << (pmc - 1);
/* count 1/2 vs 3/4 use */
if (pmc <= 4)
++pmc_grp_use[(pmc - 1) >> 1];
}
if (event[i] & PM_BUSEVENT_MSK) {
unit = (event[i] >> PM_UNIT_SH) & PM_UNIT_MSK;
byte = (event[i] >> PM_BYTE_SH) & PM_BYTE_MSK;
if (unit > PM_LASTUNIT)
return -1;
if (unit == PM_ISU0_ALT)
unit = PM_ISU0;
if (byte >= 4) {
if (unit != PM_LSU1)
return -1;
++unit;
byte &= 3;
}
if (!pmc)
++pmc_grp_use[byte & 1];
if (busbyte[byte] && busbyte[byte] != unit)
return -1;
busbyte[byte] = unit;
unituse[unit] = 1;
}
}
if (pmc_grp_use[0] > 2 || pmc_grp_use[1] > 2)
return -1;
/*
* Assign resources and set multiplexer selects.
*
* PM_ISU0 can go either on TTM0 or TTM1, but that's the only
* choice we have to deal with.
*/
if (unituse[PM_ISU0] &
(unituse[PM_FPU] | unituse[PM_IFU] | unituse[PM_ISU1])) {
unituse[PM_ISU0_ALT] = 1; /* move ISU to TTM1 */
unituse[PM_ISU0] = 0;
}
/* Set TTM[01]SEL fields. */
ttmuse = 0;
for (i = PM_FPU; i <= PM_ISU1; ++i) {
if (!unituse[i])
continue;
if (ttmuse++)
return -1;
mmcr1 |= (u64)i << MMCR1_TTM0SEL_SH;
}
ttmuse = 0;
for (; i <= PM_GRS; ++i) {
if (!unituse[i])
continue;
if (ttmuse++)
return -1;
mmcr1 |= (u64)(i & 3) << MMCR1_TTM1SEL_SH;
}
if (ttmuse > 1)
return -1;
/* Set byte lane select fields, TTM[23]SEL and GRS_*SEL. */
for (byte = 0; byte < 4; ++byte) {
unit = busbyte[byte];
if (!unit)
continue;
if (unit == PM_ISU0 && unituse[PM_ISU0_ALT]) {
/* get ISU0 through TTM1 rather than TTM0 */
unit = PM_ISU0_ALT;
} else if (unit == PM_LSU1 + 1) {
/* select lower word of LSU1 for this byte */
mmcr1 |= 1ull << (MMCR1_TTM3SEL_SH + 3 - byte);
}
ttm = unit >> 2;
mmcr1 |= (u64)ttm << (MMCR1_TD_CP_DBG0SEL_SH - 2 * byte);
}
/* Second pass: assign PMCs, set PMCxSEL and PMCx_ADDER_SEL fields */
for (i = 0; i < n_ev; ++i) {
pmc = (event[i] >> PM_PMC_SH) & PM_PMC_MSK;
unit = (event[i] >> PM_UNIT_SH) & PM_UNIT_MSK;
byte = (event[i] >> PM_BYTE_SH) & PM_BYTE_MSK;
psel = event[i] & PM_PMCSEL_MSK;
isbus = event[i] & PM_BUSEVENT_MSK;
if (!pmc) {
/* Bus event or any-PMC direct event */
for (pmc = 0; pmc < 4; ++pmc) {
if (pmc_inuse & (1 << pmc))
continue;
grp = (pmc >> 1) & 1;
if (isbus) {
if (grp == (byte & 1))
break;
} else if (pmc_grp_use[grp] < 2) {
++pmc_grp_use[grp];
break;
}
}
pmc_inuse |= 1 << pmc;
} else if (pmc <= 4) {
/* Direct event */
--pmc;
if ((psel == 8 || psel == 0x10) && isbus && (byte & 2))
/* add events on higher-numbered bus */
mmcr1 |= 1ull << (MMCR1_PMC1_ADDER_SEL_SH - pmc);
} else {
/* Instructions or run cycles on PMC5/6 */
--pmc;
}
if (isbus && unit == PM_GRS) {
bit = psel & 7;
grsel = (event[i] >> PM_GRS_SH) & PM_GRS_MSK;
mmcr1 |= (u64)grsel << grsel_shift[bit];
}
perf_counter: powerpc: set sample enable bit for marked instruction events Impact: enable access to hardware feature POWER processors have the ability to "mark" a subset of the instructions and provide more detailed information on what happens to the marked instructions as they flow through the pipeline. This marking is enabled by the "sample enable" bit in MMCRA, and there are synchronization requirements around setting and clearing the bit. This adds logic to the processor-specific back-ends so that they know which events relate to marked instructions and set the sampling enable bit if any event that we want to put on the PMU is a marked instruction event. It also adds logic to the generic powerpc code to do the necessary synchronization if that bit is set. Signed-off-by: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <18908.31930.1024.228867@cargo.ozlabs.ibm.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-08 10:30:18 +00:00
if (power5_marked_instr_event(event[i]))
mmcra |= MMCRA_SAMPLE_ENABLE;
perfcounters/powerpc: Add support for POWER5 processors This adds the back-end for the PMU on the POWER5 processor. This knows how to use the fixed-function PMC5 and PMC6 (instructions completed and run cycles). Unlike POWER6, PMC5/6 obey the freeze conditions and can generate interrupts, so their use doesn't impose any extra restrictions. POWER5+ is different and is not supported by this patch. Signed-off-by: Paul Mackerras <paulus@samba.org>
2009-02-24 00:33:56 +00:00
if (pmc <= 3)
mmcr1 |= psel << MMCR1_PMCSEL_SH(pmc);
hwc[i] = pmc;
}
/* Return MMCRx values */
mmcr[0] = 0;
if (pmc_inuse & 1)
mmcr[0] = MMCR0_PMC1CE;
if (pmc_inuse & 0x3e)
mmcr[0] |= MMCR0_PMCjCE;
mmcr[1] = mmcr1;
perf_counter: powerpc: set sample enable bit for marked instruction events Impact: enable access to hardware feature POWER processors have the ability to "mark" a subset of the instructions and provide more detailed information on what happens to the marked instructions as they flow through the pipeline. This marking is enabled by the "sample enable" bit in MMCRA, and there are synchronization requirements around setting and clearing the bit. This adds logic to the processor-specific back-ends so that they know which events relate to marked instructions and set the sampling enable bit if any event that we want to put on the PMU is a marked instruction event. It also adds logic to the generic powerpc code to do the necessary synchronization if that bit is set. Signed-off-by: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> LKML-Reference: <18908.31930.1024.228867@cargo.ozlabs.ibm.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-08 10:30:18 +00:00
mmcr[2] = mmcra;
perfcounters/powerpc: Add support for POWER5 processors This adds the back-end for the PMU on the POWER5 processor. This knows how to use the fixed-function PMC5 and PMC6 (instructions completed and run cycles). Unlike POWER6, PMC5/6 obey the freeze conditions and can generate interrupts, so their use doesn't impose any extra restrictions. POWER5+ is different and is not supported by this patch. Signed-off-by: Paul Mackerras <paulus@samba.org>
2009-02-24 00:33:56 +00:00
return 0;
}
static void power5_disable_pmc(unsigned int pmc, u64 mmcr[])
{
if (pmc <= 3)
mmcr[1] &= ~(0x7fUL << MMCR1_PMCSEL_SH(pmc));
}
static int power5_generic_events[] = {
[PERF_COUNT_CPU_CYCLES] = 0xf,
[PERF_COUNT_INSTRUCTIONS] = 0x100009,
[PERF_COUNT_CACHE_REFERENCES] = 0x4c1090, /* LD_REF_L1 */
[PERF_COUNT_CACHE_MISSES] = 0x3c1088, /* LD_MISS_L1 */
[PERF_COUNT_BRANCH_INSTRUCTIONS] = 0x230e4, /* BR_ISSUED */
[PERF_COUNT_BRANCH_MISSES] = 0x230e5, /* BR_MPRED_CR */
};
struct power_pmu power5_pmu = {
.n_counter = 6,
.max_alternatives = MAX_ALT,
.add_fields = 0x7000090000555ull,
.test_adder = 0x3000490000000ull,
.compute_mmcr = power5_compute_mmcr,
.get_constraint = power5_get_constraint,
.get_alternatives = power5_get_alternatives,
.disable_pmc = power5_disable_pmc,
.n_generic = ARRAY_SIZE(power5_generic_events),
.generic_events = power5_generic_events,
};
Reference in a new issue Copy permalink