.\" $OpenBSD: pctr.4,v 1.10 1999/06/05 13:18:35 aaron Exp $ .\" .\" Pentium performance counter driver for OpenBSD. .\" Copyright 1996 David Mazieres . .\" .\" Modification and redistribution in source and binary forms is .\" permitted provided that due credit is given to the author and the .\" OpenBSD project (for instance by leaving this copyright notice .\" intact). .\" .Dd August 15, 1996 .Dt PCTR 4 i386 .Os .Sh NAME .Nm pctr .Nd driver for cpu performance counters .Sh SYNOPSIS .Cd pseudo-device pctr .Sh DESCRIPTION The .Nm device provides access to the performance counters on Intel brand processors, and to the TSC on others. .Pp Intel processors have two 40-bit performance counters which can be programmed to count events such as cache misses, branch target buffer hits, TLB misses, dual-issues, interrupts, pipeline flushes, and more. .Pp There is one .Em ioctl call to read the status of all counters, and one .Em ioctl call to program the function of each counter. All require the following includes: .Bd -literal -offset indent #include #include #include .Ed The current state of all counters can be read with the .Dv PCIOCRD .Em ioctl , which takes an argument of type .Dv "struct pctrst" : .Bd -literal -offset indent #define PCTR_NUM 2 struct pctrst { u_int pctr_fn[PCTR_NUM]; pctrval pctr_tsc; pctrval pctr_hwc[PCTR_NUM]; pctrval pctr_idl; }; .Ed In this structure, .Dv ctr_fn contains the functions of the two counters, as previously set by the .Dv PCIOCS0 and .Dv PCIOCS1 ioctls (see bellow). .Dv pctr_hwc contains the actual value of the two hardware counters. .Dv pctr_tsc is a free-running, 64-bit cycle counter. Finally, .Dv pctr_idl is a 64-bit count of idle-loop iterations. The functions of the two counters can be programmed with ioctls .Dv PCIOCS0 and .Dv PCIOCS1 , which require a writeable file descriptor and take an argument of type .Dv "unsigned int" . \& The meaning of this integer is dependent on the particular CPU. .\" The .\" following procedure can be used to determine which counters are .\" available on a given cpu: .\" .Bd -literal -offset indent .\" ctrval id = __cpuid(); .\" if (__hasp5ctr(id)) { .\" /* The machine has Pentium counters */ .\" } else if (__hasp6ctr(id)) { .\" /* The machine has Pentium Pro counters */ .\" } else if (__hastsc(id)) { .\" /* The machine just has a time stamp counter */ .\" } else { .\" /* No counters at all */ .\"} .\" .Ed .Ss Time stamp counter The time stamp counter is available on all machines with Pentium and Pentium Pro counters, as well as on some 486s and non-intel CPUs. It is set to zero at boot time, and then increments with each cycle. Because the counter is 64-bits wide, it does not overflow. .Pp The time stamp counter can be read directly from user-mode using the .Fn rdtsc macro, which returns a 64-bit value of type .Dv pctrval . The following example illustrates a simple use of .Dv rdtsc to measure the execution time of a hypothetical subroutine called .Fn functionx : .Bd -literal -offset indent void time_functionx(void) { pctrval tsc; tsc = rdtsc(); functionx(); tsc = rdtsc() - tsc; printf ("Functionx took %qd cycles.\en", tsc); } .Ed .Pp The value of the time stamp counter is also returned by the .Dv PCIOCRD .Em ioctl , so that one can get an exact timestamp on readings of the hardware event counters. .Ss Pentium counters The Pentium counters are programmed with a 9 bit function. The top three bits contain the following flags: .Pp .Bl -tag -width P5CTR_C .It Dv P5CTR_K Enables counting of events that occur in kernel mode. .It Dv P5CTR_U Enables counting of events that occur in user mode. You must set at least one of .Dv P5CTR_U and .Dv P5CTR_K to count anything. .It Dv P5CTR_C When this flag is set, the counter attempts to count the number of cycles spent servicing a particular event, rather than simply the number of occurrences of that event. .El .Pp The bottom 6 bits set the particular event counted. Here is the event type of each permissible value for the bottom 6 bits of the counter function: .Pp .Bl -tag -width "0x00" -compact -offset indent .It 0x00 Data read .It 0x01 Data write .It 0x02 Data TLB miss .It 0x03 Data read miss .It 0x04 Data write miss .It 0x05 Write (hit) to M or E state lines .It 0x06 Data cache lines written back .It 0x07 Data cache snoops .It 0x08 Data cache snoop hits .It 0x09 Memory accesses in both pipes .It 0x0a Bank conflicts .It 0x0b Misaligned data memory references .It 0x0c Code read .It 0x0d Code TLB miss .It 0x0e Code cache miss .It 0x0f Any segment register load .It 0x12 Branches .It 0x13 BTB hits .It 0x14 Taken branch or BTB hit .It 0x15 Pipeline flushes .It 0x16 Instructions executed .It 0x17 Instructions executed in the V-pipe .It 0x18 Bus utilization (clocks) .It 0x19 Pipeline stalled by write backup .It 0x1a Pipeline stalled by data memory read .It 0x1b Pipeline stalled by write to E or M line .It 0x1c Locked bus cycle .It 0x1d I/O read or write cycle .It 0x1e Noncacheable memory references .It 0x1f AGI (Address Generation Interlock) .It 0x22 Floating-point operations .It 0x23 Breakpoint 0 match .It 0x24 Breakpoint 1 match .It 0x25 Breakpoint 2 match .It 0x26 Breakpoint 3 match .It 0x27 Hardware interrupts .It 0x28 Data read or data write .It 0x29 Data read miss or data write miss .El .Ss Pentium Pro counters The Pentium Pro counter functions contain several parts. The most significant byte (an 8-bit integer shifted left by .Dv P6CTR_CM_SHIFT ) contains a .Em "counter mask" . \& If non-zero, this sets a threshold for the number of times an event must occur in one cycle for the counter to be incremented. The .Em "counter mask" can therefore be used to count cycles in which an event occurs at least some number of times. The next byte contains several flags: .Pp .Bl -tag -width P6CTR_EN .It Dv P6CTR_U Enables counting of events that occur in user mode. .It Dv P6CTR_K Enables counting of events that occur in kernel mode. You must set at least one of .Dv P6CTR_K and .Dv P6CTR_U to count anything. .It Dv P6CTR_E Counts edges rather than cycles. For some functions this allows you to get an estimate of the number of events rather than the number of cycles occupied by those events. .It Dv P6CTR_EN Enable counters. This bit must be set in the function for counter 0 in order for either of the counters to be enabled. This bit should probably be set in counter 1 as well. .It Dv P6CTR_I Inverts the sense of the .Em "counter mask" . \& When this bit is set, the counter only increments on cycles in which there are no .Em more events than specified in the .Em "counter mask" . .El .Pp The next byte, also known as the .Em "unit mask" , contains flags specific to the event being counted. For events dealing with the L2 cache, the following flags are valid: .Pp .Bl -tag -width P6CTR_UM_M .It Dv P6CTR_UM_M Count events involving modified cache lines. .It Dv P6CTR_UM_E Count events involving exclusive cache lines. .It Dv P6CTR_UM_S Count events involving shared cache lines. .It Dv P6CTR_UM_I Count events involving invalid cache lines. .El To measure all L2 cache activity, all these bits should be set. They can be set with the macro .Dv P6CTR_UM_MESI which contains the bitwise or of all of the above. .Pp For event types dealing with bus transactions, there is another flag that can be set in the .Em "unit mask" : .Bl -tag -width P6CTR_UM_A .It Dv P6CTR_UM_A Count all appropriate bus events, not just those initiated by the processor. .El .Pp Finally, the least significant byte of the counter function is the event type to count. The following values are available: .Pp .Bl -tag -width 0x00 -compact .It 0x03 LD_BLOCKS Number of store buffer blocks. .It 0x04 SB_DRAINS Number of store buffer drain cycles. .It 0x05 MISALIGN_MEM_REF Number of misaligned data memory references. .It 0x06 SEGMENT_REG_LOADS Number of segment register loads. .It 0x10 FP_COMP_OPS_EXE (ctr0 only) Number of computational floating-point operations executed. .It 0x11 FP_ASSIST (ctr1 only) Number of floating-point exception cases handled by microcode. .It 0x12 MUL (ctr1 only) Number of multiplies. .It 0x13 DIV (ctr1 only) Number of divides. .It 0x14 CYCLES_DIV_BUSY (ctr0 only) Number of cycles during which the divider is busy. .It 0x21 L2_ADS Number of L2 address strobes. .It 0x22 L2_DBUS_BUSY Number of cycles durring which the data bus was busy. .It 0x23 L2_DBUS_BUSY_RD Number of cycles during which the data bus was busy transferring data from L2 to the processor. .It 0x24 L2_LINES_IN Number of lines allocated in the L2. .It 0x25 L2_M_LINES_INM Number of modified lines allocated in the L2. .It 0x26 L2_LINES_OUT Number of lines removed from the L2 for any reason. .It 0x27 L2_M_LINES_OUTM Number of modified lines removed from the L2 for any reason. .It 0x28 L2_IFETCH/mesi Number of L2 instruction fetches. .It 0x29 L2_LD/mesi Number of L2 data loads. .It 0x2a L2_ST/mesi Number of L2 data stores. .It 0x2e L2_RQSTS/mesi Number of L2 requests. .It 0x43 DATA_MEM_REFS All memory references, both cacheable and non-cacheable. .It 0x45 DCU_LINES_IN Total lines allocated in the DCU. .It 0x46 DCU_M_LINES_IN Number of M state lines allocated in the DCU. .It 0x47 DCU_M_LINES_OUT Number of M state lines evicted from the DCU. This includes evictions via snoop HITM, intervention or replacement .It 0x48 DCU_MISS_OUTSTANDING Weighted number of cycles while a DCU miss is outstanding. .It 0x60 BUS_REQ_OUTSTANDING Number of bus requests outstanding. .It 0x61 BUS_BNR_DRV Number of bus clock cycles during which the processor is driving the BNR pin. .It 0x62 BUS_DRDY_CLOCKS/a Number of clocks during which DRDY is asserted. .It 0x63 BUS_LOCK_CLOCKS/a Number of clocks during which LOCK is asserted. .It 0x64 BUS_DATA_RCV Number of bus clock cycles during which the processor is receiving data. .It 0x65 BUS_TRAN_BRD/a Number of burst read transactions. .It 0x66 BUS_TRAN_RFO/a Number of read for ownership transactions. .It 0x67 BUS_TRANS_WB/a Number of write back transactions. .It 0x68 BUS_TRAN_IFETCH/a Number of instruction fetch transactions. .It 0x69 BUS_TRAN_INVAL/a Number of invalidate transactions. .It 0x6a BUS_TRAN_PWR/a Number of partial write transactions. .It 0x6b BUS_TRANS_P/a Number of partial transactions. .It 0x6c BUS_TRANS_IO/a Number of I/O transactions. .It 0x6d BUS_TRAN_DEF/a Number of deferred transactions. .It 0x6e BUS_TRAN_BURST/a Number of burst transactions. .It 0x6f BUS_TRAN_MEM/a Number of memory transactions. .It 0x70 BUS_TRAN_ANY/a Number of all transactions. .It 0x79 CPU_CLK_UNHALTED Number of cycles during which the processor is not halted. .It 0x7a BUS_HIT_DRV Number of bus clock cycles during which the processor is driving the HIT pin. .It 0x7b BUS_HITM_DRV Number of bus clock cycles during which the processor is driving the HITM pin. .It 0x7e BUS_SNOOP_STALL Number of clock cycles during which the bus is snoop stalled. .It 0x80 IFU_IFETCH Number of instruction fetches, both cacheable and non-cacheable. .It 0x81 IFU_IFETCH_MISS Number of instruction fetch misses. .It 0x85 ITLB_MISS Number of ITLB misses. .It 0x86 IFU_MEM_STALL Number of cycles that the instruction fetch pipe stage is stalled, including cache mises, ITLB misses, ITLB faults, and victim cache evictions .It 0x87 ILD_STALL Number of cycles that the instruction length decoder is stalled .It 0xa2 RESOURCE_STALLS Number of cycles during which there are resource-related stalls. .It 0xc0 INST_RETIRED Number of instructions retired. .It 0xc1 FLOPS (ctr0 only) Number of computational floating-point operations retired. .It 0xc2 UOPS_RETIRED Number of UOPs retired. .It 0xc4 BR_INST_RETIRED Number of branch instructions retired. .It 0xc5 BR_MISS_PRED_RETIRED Number of mispredicted branches retired. .It 0xc6 CYCLES_INT_MASKED Number of processor cycles for which interrupts are disabled. .It 0xc7 CYCLES_INT_PENDING_AND_MASKED Number of processor cycles for which interrupts are disabled and interrupts are pending. .It 0xc8 HW_INT_RX Number of hardware interrupts received. .It 0xc9 BR_TAKEN_RETIRED Number of taken branches retired. .It 0xca BR_MISS_PRED_TAKEN_RET Number of taken mispredictioned branches retired. .It 0xd0 INST_DECODER Number of instructions decoded. .It 0xd2 PARTIAL_RAT_STALLS Number of cycles or events for partial stalls. .It 0xe0 BR_INST_DECODED Number of branch instructions decoded. .It 0xe2 BTB_MISSES Number of branches that miss the BTB. .It 0xe4 BR_BOGUS Number of bogus branches. .It 0xe6 BACLEARS Number of times BACLEAR is asserted. .El .Pp Events marked /mesi require the .Dv P6CTR_UM_[MESI] bits in the .Em "unit mask" . \& Events marked /a can take the .Dv P6CTR_UM_A bit. .Pp Unlike the Pentium counters, the Pentium Pro counters can be read directly from user-mode without need to invoke the kernel. The macro .Fn rdpmc ctr takes 0 or 1 as an argument to specify a counter, and returns that counter's 40-bit value (which will be of type .Dv pctrval ) . This is generally preferable to making a system call as it introduces less distortion in measurements. However, you should be aware of the possibility of an interrupt between invocations of .Fn rdpmc and/or .Fn rdtsc . .Sh SEE ALSO .Xr pctr 1 .Sh FILES .Pa /dev/pctr .Sh HISTORY A .Nm device first appeared in .Ox 2.0 . .Sh AUTHORS The .Nm device was written by David Mazieres .Aq dm@lcs.mit.edu . .Sh ERRORS .Bl -tag -width "[ENODEV]" .It Bq Er ENODEV An attempt was made to set the counter functions on a CPU that does not support counters. .It Bq Er EINVAL An invalid counter function was provided as an argument to the .Dv PCIOCS0 or .Dv PCIOCS1 .Em ioctl . .It Bq Er EPERM An attempt was made to set the counter functions, but the device was not open for writing. .El .Sh BUGS Not all counter functions are completely accurate. Some of the functions don't seem to make any sense at all.