— One Instruction/Clock Execution
— Core Clock Rate is 2x the Bus Clock
— Load/Store Programming Model
— Sixteen 32-Bit Global Registers
— Sixteen 32-Bit Local Registers (8 sets)
— Nine Addressing Modes
— User/Supervisor Protection Model
■ Two-Way Set Associative Instruction Cache
— 80960JD - 4 Kbyte
— Programmable Cache Locking
Mechanism
■ Direct Mapped Data Cache
— 80960JD - 2 Kbyte
— Write Through Operation
■ On-Chip Stack Frame Cache
— Seven Register Sets Can Be Saved
— Automatic Allocation on Call/Return
— 0-7 Frames Reserved for High-Priority
2.2 Burst Bus ............................................................................................................................................2
2.3 Timer Unit ...........................................................................................................................................3
2.5 Instruction Set Summary ....................................................................................................................3
2.6 Faults and Debugging .........................................................................................................................3
2.7 Low Power Operation .........................................................................................................................4
2.8 Test Features ......................................................................................................................................4
2.9 Memory-Mapped Control Registers ....................................................................................................4
2.10 Data Types and Memory Addressing Modes ....................................................................................4
3.0 PACKAGE INFORMATION ........................................................................................................................6
Figure 12.Output Delay Waveform for T
Figure 13.Output Float Waveform for T
Figure 14.Input Setup and Hold Waveform for T
Figure 15.Input Setup and Hold Waveform for T
Figure 16.Input Setup and Hold Waveform for T
Figure 17.Input Setup and Hold Waveform for T
Figure 18.Relative Timings Waveform for T
Figure 19.DT/R
and DEN Timings Waveform .......................................................................................... 38
Figure 21.Input Setup and Hold Waveforms for T
Figure 22.Output Delay and Output Float Waveform for T
Figure 23.Output Delay and Output Float Waveform for T
Figure 24.Input Setup and Hold Waveform for T
Figure 25.Non-Burst Read and Write Transactions Without Wait States, 32-Bit Bus ...............................42
Figure 26.Burst Read and Write Transactions Without Wait States, 32-Bit Bus ......................................43
Figure 27.Burst Write Transactions With 2,1,1,1 Wait States, 32-Bit Bus ................................................44
Figure 28.Burst Read and Write Transactions Without Wait States, 8-Bit Bus ........................................ 45
Figure 29.Burst Read and Write Transactions With 1, 0 Wait States
and Extra Tr State on Read, 16-Bit Bus ...................................................................................46
Figure 30.Bus Transactions Generated by Double Word Read Bus Request,
Misaligned One Byte From Quad Word Boundary, 32-Bit Bus, Little Endian ........................... 47
Figure 31.HOLD/HOLDA Waveform For Bus Arbitration .......................................................................... 48
This document contains advance information for the
80960JD microprocessor, including electrical
characteristics and package pinout information.
Detailed functional descriptions — other than
parametric performance — are published in the
®
Jx Microprocessor User’s Guide (272483).
i960
Throughout this data sheet, references to “80960Jx”
indicate features which apply to all of the following:
• 80960JD — 5V, 4 Kbyte instruction cache, 2 Kbyte
data cache and clock doubling
• 80L960JA — 3.3 V version of the 80960JA
• 80L960JF — 3.3 V version of the 80960JF
2.080960JD OVERVIEW
The 80960JD offers high performance to costsensitive 32-bit embedded applications. The
80960JD is object code compatible with the 80960
Core Architecture and is capable of sustained
execution at the rate of one instruction per clock.
This processor’s features include generous
instruction cache, data cache and data RAM. It also
boasts a fast interrupt mechanism, dual programmable timer units and new instructions.
The 80960JD’s clock doubler operates the processor
core at twice the bus clock rate to improve execution
performance without increasing the complexity of
board designs.
Memory subsystems for cost-sensitive embedded
applications often impose substantial wait state
penalties. The 80960JD integrates considerable
storage resources on-chip to decouple CPU
execution from the external bus.
The 80960JD rapidly allocates and deallocates local
register sets during context switches. The processor
needs to flush a register set to the stack only when it
saves more than seven sets to its local register
cache.
A 32-bit multiplexed burst bus provides a high-speed
interface to system memory and I/O. A full
complement of control signals simplifies the
connection of the 80960JD to external components.
The user programs physical and logical memory
attributes through memory-mapped control registers
(MMRs) — an extension not found on the i960 Kx,
Sx or Cx processors. Physical and logical configuration registers enable the processor to operate with
all combinations of bus width and data object
alignment. The processor supports a homogeneous
byte ordering model.
This processor integrates two important peripherals:
a timer unit and an interrupt controller. These and
other hardware resources are programmed through
memory-mapped control registers, an extension to
the familiar 80960 architecture.
The timer unit (TU) offers two independent 32-bit
timers for use as real-time system clocks and
general-purpose system timing. These operate in
either single-shot or auto-reload mode and can
generate interrupts.
The interrupt controller unit (ICU) provides a flexible
means for requesting interrupts. The ICU provides
full programmability of up to 240 interrupt sources
into 31 priority levels. The ICU takes advantage of a
cached priority table and optional routine caching to
minimize interrupt latency. Clock doubling reduces
interrupt latency by 40% compared to the
80960JA/JF. Local registers may be dedicated to
high-priority interrupts to further reduce latency.
Acting independently from the core, the ICU
compares the priorities of posted interrupts with the
current process priority, off-loading this task from the
core. The ICU also supports the integrated timer
interrupts.
The 80960JD features a Halt mode designed to
support applications where low power consumption
is critical. The halt instruction shuts down instruction
execution, resulting in a power savings of up to 90
percent.
The 80960JD’s testability features, including ONCE
(On-Circuit Emulation) mode and Boundary Scan
(JTAG), provide a powerful environment for design
debug and fault diagnosis.
The Solutions960® program features a wide variety
of development tools which support the i960
processor family. Many of these tools are developed
by partner companies; some are developed by Intel,
such as profile-driven optimizing compilers. For more
information on these products, contact your local
Intel representative.
PRELIMINARY
1
80960JDA
SRC1
SRC2
DEST
SRC1
SRC2
DEST
SRC1
DEST
CLKIN
TAP
Local Register Cache
PLL, Clocks,
Power Mgmt
Boundary Scan
5
8-Set
128
Global / Local
Register File
SRC2 DESTSRC1
Controller
3 Independent 32-Bit SRC1, SRC2, and DEST Buses
4 KByte Instruction Cache
Two-Way Set Associative
Constants
and
Address
Unit
effective
address
Control
Multiply
Divide
Unit
Instruction Sequencer
Execution
Generation
Figure 2. 80960JD Block Diagram
Memory
Interface
Unit
32-bit Address
32-bit Data
32-bit buses
address / data
Physical Region
Configuration
Bus
Control Unit
Bus Request
Queues
Two 32-Bit
Timers
Programmable
Interrupt Controller
Memory-Mapped
Register Interface
1 K byte
Data RAM
2 Kbyte Direct
Mapped Data
Cache
Control
21
Address/
Data Bus
32
Interrupt
Port
9
2.180960 Processor Core
The 80960Jx family is a scalar implementation of the
80960 Core Architecture. Intel designed this
processor core as a very high performance device
that is also cost-effective. Factors that contribute to
the core’s performance include:
• Core operates at twice the bus speed (80960JD
only)
• Register and resource scoreboarding allow
overlapped instruction execution
2
• 128-bit register bus speeds local register caching
• 4 Kbyte two-way set associative, integrated
instruction cache
• 2 Kbyte direct-mapped, integrated data cache
• 1 Kbyte integrated data RAM delivers zero wait
state program data
2.2Burst Bus
A 32-bit high-performance bus controller interfaces
the 80960JD to external memory and peripherals.
The BCU fetches instructions and transfers data at
the rate of up to four 32-bit words per six clock
cycles. The external address/data bus is multiplexed.
PRELIMINARY
A80960JD
Users may configure the 80960JD’s bus controller to
match an application’s fundamental memory organization. Physical bus width is register-programmed for
up to eight regions. Byte ordering and data caching
are programmed through a group of logical memory
templates and a defaults register.
The BCU’s features include:
• Multiplexed external bus to minimize pin count
• 32-, 16- and 8-bit bus widths to simplify I/O
interfaces
• External ready control for address-to-data, data-todata and data-to-next-address wait state types
• Support for big or little endian byte ordering to
facilitate the porting of existing program code
• Unaligned bus accesses performed transparently
• Three-deep load/store queue to decouple the bus
from the core
Upon reset, the 80960JD conducts an internal self
test. Then, before executing its first instruction, it
performs an external bus confidence test by
performing a checksum on the first words of the
initialization boot record (IBR).
The user may examine the contents of the caches at
any time by executing special cache control instructions.
2.3Timer Unit
The timer unit (TU) contains two independent 32-bit
timers which are capable of counting at several clock
rates and generating interrupts. Each is programmed
by use of the TU registers. These memory-mapped
registers are addressable on 32-bit boundaries. The
timers have a single-shot mode and auto-reload
capabilities for continuous operation. Each timer has
an independent interrupt request to the 80960JD’s
interrupt controller. The TU can generate a fault
when unauthorized writes from user mode are
detected. Clock prescaling is supported.
2.4Priority Interrupt Controller
A programmable interrupt controller manages up to
240 external sources through an 8-bit external
interrupt port. Alternatively, the interrupt inputs may
be configured for individual edge- or level-triggered
inputs. The interrupt unit (IU) also accepts interrupts
from the two on-chip timer channels and a single
Non-Maskable Interrupt (NMI
serviced according to their priority levels relative to
the current process priority.
Low interrupt latency is critical to many embedded
applications. As part of its highly flexible interrupt
mechanism, the 80960JD exploits several
techniques to minimize latency:
• Interrupt vectors and interrupt handler routines can
be reserved on-chip
• Register frames for high-priority interrupt handlers
can be cached on-chip
• The interrupt stack can be placed in cacheable
memory space
• Interrupt microcode executes at twice the bus
frequency
) pin. Interrupts are
2.5Instruction Set Summary
The 80960Jx adds several new instructions to the
i960 core architecture. The new instructions are:
• Conditional Move
• Conditional Add
• Conditional Subtract
• Byte Swap
• Halt
• Cache Control
• Interrupt Control
Table 1 identifies the instructions that the 80960Jx
supports. Refer to i960Guide (272483) for a detailed description of each
instruction.
®
Jx Microprocessor User’s
2.6Faults and Debugging
The 80960Jx employs a comprehensive fault model.
The processor responds to faults by making implicit
calls to a fault handling routine. Specific information
collected for each fault allows the fault handler to
diagnose exceptions and recover appropriately.
The processor also has built-in debug capabilities. In
software, the 80960Jx may be configured to detect
as many as seven different trace event types. Alter-
PRELIMINARY
3
80960JDA
natively, mark and fmark instructions can generate
trace events explicitly in the instruction stream.
Hardware breakpoint registers are also available to
trap on execution and data addresses.
2.7Low Power Operation
Intel fabricates the 80960Jx using an advanced submicron manufacturing process. The processor’s submicron topology provides the circuit density for
optimal cache size and high operating speeds while
dissipating modest power. The processor also uses
dynamic power management to turn off clocks to
unused circuits.
Users may program the 80960Jx to enter Halt mode
for maximum power savings. In Halt mode, the
processor core stops completely while the integrated
peripherals continue to function, reducing overall
power requirements up to 90 percent. Processor
execution resumes from internally or externally
generated interrupts.
2.8Test Features
The 80960Jx incorporates numerous features which
enhance the user’s ability to test both the processor
and the system to which it is attached. These
features include ONCE (On-Circuit Emulation) mode
and Boundary Scan (JTAG).
The 80960Jx provides testability features compatible
with IEEE Standard Test Access Port and Boundary
Scan Architecture (IEEE Std. 1149.1).
One of the boundary scan instructions, HIGHZ,
forces the processor to float all its output pins
(ONCE mode). ONCE mode can also be initiated at
reset without using the boundary scan mechanism.
ONCE mode is useful for board-level testing. This
feature allows a mounted 80960JD to electrically
“remove” itself from a circuit board. This allows for
system-level testing where a remote tester — such
as an in-circuit emulator — can exercise the
processor system.
The provided test logic does not interfere with
component or circuit board behavior and ensures
that components function correctly, connections
between various components are correct, and
various components interact correctly on the printed
circuit board.
The JTAG Boundary Scan feature is an attractive
alternative to conventional “bed-of-nails” testing. It
can examine connections which might otherwise be
inaccessible to a test system.
2.9Memory-Mapped Control
Registers
The 80960JD, though compliant with i960 series
processor core, has the added advantage of
memory-mapped, internal control registers not found
on the i960 Kx, Sx or Cx processors. These give
software the interface to easily read and modify
internal control registers.
Each of these registers is accessed as a memorymapped, 32-bit register. Access is accomplished
through regular memory-format instructions. The
processor ensures that these accesses do not
generate external bus cycles.
2.10 Data Types and Memory
Addressing Modes
As with all i960 family processors, the 80960Jx
instruction set supports several data types and
formats:
will be specified for operation at V
over a case temperature range of 0° to 100°C:
= 5.0 V ± 5%
cc
• NG80960JD-40 (40 MHz core, 20 MHz bus)
• NG80960JD-33 (33.33 MHz core, 16.67 MHz bus)
For complete package specifications and infor-
mation, refer to Intel’s Packaging Handbook
(240800).
3.1Pin Descriptions
This section describes the pins for the 80960JD in
the 132-pin ceramic Pin Grid Array (PGA) package
and 132-lead Plastic Quad Flatpack Package
(PQFP).
Section 3.1.1, Functional Pin Definitions
describes pin function; Section 3.1.2, 80960Jx 132Lead PGA Pinout and Section 3.1.3, 80960Jx
PQFP Pinout define the signal and pin locations for
the supported package types.
3.1.1 Functional Pin Definitions
Table 2 presents the legend for interpreting the pin
descriptions which follow. Pins associated with the
bus interface are described in Table 3. Pins
associated with basic control and test functions are
described in Table 4. Pins associated with the
Interrupt Unit are described in Table 5.
Table 2. Pin Description Nomenclature
SymbolDescription
IInput pin only.
OOutput pin only.
I/OPin can be either an input or output.
–Pin must be connected as described.
SSynchronous. Inputs must meet setup
and hold times relative to CLKIN for
proper operation.
R(1) is driven to V
R(0) is driven to V
R(Q) is a valid output
CC
SS
R(X) is driven to unknown state
R(H) is pulled up to V
CC
H (...)While the processor is in the hold state,
the pin:
H(1) is driven to V
H(0) is driven to V
H(Q) Maintains previous state or
CC
SS
continues to be a valid output
H(Z) Floats
P (...)While the processor is halted, the pin:
P(1) is driven to V
P(0) is driven to V
P(Q) Maintains previous state or
CC
SS
continues to be a valid output
6
PRELIMINARY
A80960JD
Table 3. Pin Description — External Bus Signals (Sheet 1 of 4)
NAMETYPEDESCRIPTION
AD31:0I/O
S(L)
R(X)
H(Z)
P(Q)
ADDRESS / DATA BUS carries 32-bit physical addresses and 8-, 16- or 32-bit data
T
to and from memory. During an address (
address (bits 0-1 indicate SIZE; see below). During a data (T
) cycle, bits 31:2 contain a physical word
a
data is present on one or more contiguous bytes, comprising AD31:24, AD23:16,
AD15:8 and AD7:0. During write operations, unused pins are driven to determinate
values.
SIZE, which comprises bits 0-1 of the AD lines during a
number of data transfers during the bus transaction.
When the processor enters Halt mode, if the previous bus operation was a:
• write — AD31:2 are driven with the last data value on the AD bus.
• read — AD31:4 are driven with the last address value on the AD bus; AD3:2 are
driven with the value of A3:2 from the last data cycle.
Typically, AD1:0 reflect the SIZE information of the last bus transaction (either
instruction fetch or load/store) that was executed before entering Halt mode.
ALEO
R(0)
H(Z)
ADDRESS LATCH ENABLE indicates the transfer of a physical address. ALE is
T
asserted during a
active HIGH and floats to a high impedance state during a hold cycle (T
cycle and deasserted before the beginning of the Td state. It is
a
P(0)
ADDRESS LATCH ENABLE indicates the transfer of a physical address. ALE
ALE
O
R(1)
H(Z)
inverted version of ALE. This signal gives the 80960JD a high degree of compatibility
with existing 80960Kx systems.
P(1)
ADDRESS STROBE indicates a valid address and the start of a new bus access.
ADS
O
R(1)
H(Z)
The processor asserts ADS
samples ADS
at the end of the cycle.
for the entire Ta cycle. External bus control logic typically
P(1)
A3:2O
R(X)
H(Z)
P(Q)
ADDRESS3:2 comprise a partial demultiplexed address bus.
32-bit memory accesses: the processor asserts address bits A3:2 during
partial word address increments with each assertion of RDYRCV
16-bit memory accesses: the processor asserts address bits A3:1 during
driven on the
of RDYRCV
BE1 pin. The partial short word address increments with each assertion
during a burst.
8-bit memory accesses: the processor asserts address bits A3:0 during
driven on BE1:0
RDYRCV
. The partial byte address increments with each assertion of
during a burst.
) cycle, read or write
d
T
cycle, specifies the
a
h
during a burst.
T
).
is the
T
. The
a
T
with A1
a
, with A1:0
a
PRELIMINARY
7
80960JDA
Table 3. Pin Description — External Bus Signals (Sheet 2 of 4)
NAMETYPEDESCRIPTION
BE3:0O
R(1)
H(Z)
P(1)
WIDTH/
HLTD1:0
R(0)
H(Z)
P(1)
D/C
R(X)
H(Z)
P(Q)
W/R
R(0)
H(Z)
P(Q)
BYTE ENABLES select which of up to four data bytes on the bus participate in the
current bus access. Byte enable encoding is dependent on the bus width of the
memory region accessed:
32-bit bus:
enables data on AD31:24
BE3
enables data on AD23:16
BE2
enables data on AD15:8
BE1
enables data on AD7:0
BE0
16-bit bus:
becomes Byte High Enable (enables data on AD15:8)
BE3
is not used (state is high)
BE2
becomes Address Bit 1 (A1)
BE1
becomes Byte Low Enable (enables data on AD7:0)
BE0
8-bit bus:
is not used (state is high)
BE3
is not used (state is high)
BE2
becomes Address Bit 1 (A1)
BE1
becomes Address Bit 0 (A0)
BE0
The processor asserts byte enables, byte high enable and byte low enable during
Since unaligned bus requests are split into separate bus transactions, these signals
do not toggle during a burst. They remain active through the last T
cycle.
d
For accesses to 8- and 16-bit memory, the processor asserts the address bits in
conjunction with A3:2 described above.
WIDTH/HALTED signals denote the physical memory attributes for a bus transaction:
The processor floats the WIDTH/HLTD pins whenever it relinquishes the bus in
response to a HOLD request, regardless of prior operating state.
DATA/CODE indicates that a bus access is a data access (1) or an instruction access
O
has the same timing as W/R.
(0). D/C
0 = instruction access
1 = data access
WRITE/READ specifies, during a
O
(0). It is latched on-chip and remains valid during T
T
cycle, whether the operation is a write (1) or read
a
cycles.
d
0 = read
1 = write
T
.
a
8
PRELIMINARY
A80960JD
Table 3. Pin Description — External Bus Signals (Sheet 3 of 4)
NAMETYPEDESCRIPTION
DT/RO
DEN
BLAST
RDYRCV
/
LOCK
ONCE
R(0)
H(Z)
P(Q)
R(1)
H(Z)
P(1)
R(1)
H(Z)
P(1)
S(L)
S(L)
R(H)
H(Z)
P(1)
DATA TRANSMIT / RECEIVE indicates the direction of data transfer to and from the
address/data bus. It is low during T
/Td cycles for a write. DT/R never changes state when DEN is asserted.
and T
w
0 = receive
1 = transmit
DATA ENABLE indicates data transfer cycles during a bus access. DEN
O
at the start of the first data cycle in a bus access and deasserted at the end of the last
data cycle. DEN
to the data bus.
0 = data cycle
1 = not data cycle
BURST LAST indicates the last transfer in a bus access. BLAST
O
last data transfer of burst and non-burst accesses. BLAST
wait states are inserted via the RDYRCV
data transfer in a bus cycle.
0 = last data transfer
1 = not last data transfer
I
READY/RECOVER indicates that data on AD lines can be sampled or removed. If
RDYRCV
by inserting a wait state (T
0 = sample data
1 = don’t sample data
The RDYRCV
continues to insert additional recovery states until it samples the pin HIGH. This
function gives slow external devices more time to float their buffers before the
processor begins to drive address again.
0 = insert wait states
1 = recovery complete
BUS LOCK indicates that an atomic read-modify-write operation is in progress. The
I/O
output is asserted in the first clock of an atomic operation and deasserted in
LOCK
the last data transfer of the sequence. The processor does not grant HOLDA while it
is asserting LOCK
semaphore operations.
0 = Atomic read-modify-write in progress
1 = Atomic read-modify-write not in progress
ONCE MODE: The processor samples the ONCE
LOW at the end of reset, the processor enters ONCE mode. In ONCE mode, the
processor stops all clocks and floats all output pins. The pin has a weak internal
pullup which is active during reset to ensure normal operation when the pin is left
unconnected.
0 = ONCE mode enabled
1 = ONCE mode not enabled
is used with DT/R to provide control for data transceivers connected
is not asserted during a Td cycle, the Td cycle is extended to the next cycle
w
pin has another function during the recovery (Tr) state. The processor
. This prevents external agents from accessing memory involved in
and Tw/Td cycles for a read; it is high during Ta
a
is asserted in the
pin. BLAST becomes inactive after the final
).
remains active as long as
input during reset. If it is asserted
is asserted
PRELIMINARY
9
80960JDA
Table 3. Pin Description — External Bus Signals (Sheet 4 of 4)
NAMETYPEDESCRIPTION
HOLDI
S(L)
HOLDAO
R(Q)
H(1)
P(Q)
BSTATO
R(0)
H(Q)
P(0)
Table 4. Pin Description — Processor Control Signals, Test Signals and Power (Sheet 1 of 2)
NAMETYPEDESCRIPTION
CLKINICLOCK INPUT provides the processor’s fundamental time base; both the processor
RESET
A(L)
STESTI
S(L)
HOLD: A request from an external bus master to acquire the bus. When the
processor receives HOLD and grants bus control to another master, it asserts
HOLDA, floats the address/data and control lines and enters the T
HOLD is deasserted, the processor deasserts HOLDA and enters either the T
state. When
h
state, resuming control of the address/data and control lines.
0 = no hold request
1 = hold request
HOLD ACKNOWLEDGE indicates to an external bus master that the processor has
relinquished control of the bus. The processor can grant HOLD requests and enter
state during reset and while halted as well as during regular operation.
the T
h
0 = hold not acknowledged
1 = hold acknowledged
BUS STATUS indicates that the processor may soon stall unless it has sufficient
access to the bus; see i960
logic can examine this signal to determine when an external bus master should
acquire/relinquish the bus.
0 = no potential stall
1 = potential stall
core and the external bus run at the CLKIN rate. All input and output timings are
specified relative to a rising CLKIN edge.
I
RESET initializes the processor and clears its internal logic. During reset, the
processor places the address/data bus and control output pins in their idle (inactive)
states.
During reset, the input pins are ignored with the exception of LOCK
/ONCE, STEST
and HOLD.
The RESET
ization during power up, RESET
cycles with V
a minimum of 15 cycles.
pin has an internal synchronizer. To ensure predictable processor initial-
must be asserted a minimum of 10,000 CLKIN
and CLKIN stable. On a warm reset, RESET should be asserted for
CC
SELF TEST enables or disables the processor’s internal self-test feature at initialization. STEST is examined at the end of reset. When STEST is asserted, the
processor performs its internal self-test and the external bus confidence test. When
STEST is deasserted, the processor performs only the external bus confidence test.
0 = self test disabled
1 = self test enabled
or Ta
i
10
PRELIMINARY
A80960JD
Table 4. Pin Description — Processor Control Signals, Test Signals and Power (Sheet 2 of 2)
NAMETYPEDESCRIPTION
FAILO
TCKITEST CLOCK is a CPU input which provides the clocking function for IEEE 1149.1
TDII
TDOO
TRST
TMSI
V
CC
V
CCPLL
V
SS
NC–NO CONNECT pins. Do not make any system connections to these pins.
R(0)
H(Q)
P(1)
S(L)
R(Q)
HQ)
P(Q)
A(L)
S(L)
FAIL indicates a failure of the processor’s built-in self-test performed during initialization. FAIL
indicate the status of individual tests:
• When self-test passes, the processor deasserts FAIL
user code.
• When self-test fails, the processor asserts FAIL
0 = self test failed
1 = self test passed
Boundary Scan Testing (JTAG). State information and data are clocked into the
processor on the rising edge; data is clocked out of the processor on the falling edge.
TEST DATA INPUT is the serial input pin for JTAG. TDI is sampled on the rising edge
of TCK, during the SHIFT-IR and SHIFT-DR states of the Test Access Port.
TEST DATA OUTPUT is the serial output pin for JTAG. TDO is driven on the falling
edge of TCK during the SHIFT-IR and SHIFT-DR states of the Test Access Port. At
other times, TDO floats. TDO does not float during ONCE mode.
I
TEST RESET asynchronously resets the Test Access Port (TAP) controller function
of IEEE 1149.1 Boundary Scan testing (JTAG). When using the Boundary Scan
feature, connect a pulldown resistor between this pin and V
pin must be connected to V
Connection Recommendations (pg. 24).
TEST MODE SELECT is sampled at the rising edge of TCK to select the operation
of the test logic for IEEE 1149.1 Boundary Scan testing.
–POWER pins intended for external connection to a VCC board plane.
–PLL POWER is a separate VCC supply pin for the phase lock loop clock generator. It
is intended for external connection to the V
add a simple bypass filter circuit to reduce noise-induced clock jitter and its effects on
timing relationships.
–GROUND pins intended for external connection to a VSS board plane.
is asserted immediately upon reset and toggles during self-test to
and begins operation from
and then stops executing.
. If TAP is not used, this
; however, no resistor is required. See Section 4.3,
SS
board plane. In noisy environments,
CC
SS
PRELIMINARY
11
80960JDA
Table 5. Pin Description — Interrupt Unit Signals
NAMETYPEDESCRIPTION
I
XINT7:0
A(E/L)
NMI
A(E)
EXTERNAL INTERRUPT pins are used to request interrupt service. The XINT7:0
pins can be configured in three modes:
Dedicated Mode: Each pin is assigned a dedicated interrupt level. Dedicated inputs
can be programmed to be level (low) or edge (falling) sensitive.
Expanded Mode: All eight pins act as a vectored interrupt source. The interrupt pins
are level sensitive in this mode.
Mixed Mode: The XINT7:5
pins act as dedicated sources and the XINT4:0 pins
act as the five most significant bits of a vectored source. The least
significant bits of the vectored source are set to 010
Unused external interrupt pins should be connected to V
I
NON-MASKABLE INTERRUPT causes a non-maskable interrupt event to occur.
is the highest priority interrupt source and is falling edge-triggered. If NMI is