Class uvm_pkg::uvm_mem
Collaboration Diagram of uvm_mem
Name |
Type |
Description |
|---|---|---|
mam |
variable mam Memory allocation manager Memory allocation manager for the memory corresponding to this abstraction class instance. Can be used to allocate regions of consecutive addresses of specific sizes, such as DMA buffers, or to locate virtual register array. |
Constructors
- new(string name, longint unsigned size, int unsigned n_bits, string access = "RW", int has_coverage = UVM_NO_COVERAGE)
Function
new
Create a new instance and type-specific configuration
Creates an instance of a memory abstraction class with the specified name.
size specifies the total number of memory locations. n_bits specifies the total number of bits in each memory location. access specifies the access policy of this memory and may be one of "RW for RAMs and "RO" for ROMs.
has_coverage specifies which functional coverage models are present in the extension of the register abstraction class. Multiple functional coverage models may be specified by adding their symbolic names, as defined by the uvm_coverage_model_e type. New
- Parameters:
name (string)
size (longint unsigned)
n_bits (int unsigned)
access (string)
has_coverage (int)
Enums
- init_e
- Enum Items:
UNKNOWNS
ZEROES
ONES
ADDRESS
VALUE
INCR
DECR
Functions
- configure(uvm_reg_block parent, string hdl_path = "")
Function
configure
Instance-specific configuration
Specify the parent block of this memory.
If this memory is implemented in a single HDL variable, its name is specified as the hdl_path . Otherwise, if the memory is implemented as a concatenation of variables (usually one per bank), then the HDL path must be specified using the add_hdl_path() or add_hdl_path_slice() method. Configure
- Parameters:
parent (uvm_reg_block)
hdl_path (string)
- set_offset(uvm_reg_map map, uvm_reg_addr_t offset, bit unmapped = 0)
Function
set_offset
Modify the offset of the memory
The offset of a memory within an address map is set using the uvm_reg_map::add_mem() method. This method is used to modify that offset dynamically.
Note
Modifying the offset of a memory will make the abstract model
diverge from the specification that was used to create it. Set_offset
- Parameters:
map (uvm_reg_map)
offset (uvm_reg_addr_t)
unmapped (bit)
- set_parent(uvm_reg_block parent)
Set_parent
- Parameters:
parent (uvm_reg_block) --
Local
- add_map(uvm_reg_map map)
Add_map
- Parameters:
map (uvm_reg_map) --
Local
- Xlock_modelX()
Xlock_modelXlocal
- Xadd_vregX(uvm_vreg vreg)
Xadd_vregX
- Parameters:
vreg (uvm_vreg) --
Local
- Xdelete_vregX(uvm_vreg vreg)
Xdelete_vregX
- Parameters:
vreg (uvm_vreg) --
Local
- get_full_name()
Function
get_full_name
Get the hierarchical name
Return the hierarchal name of this memory. The base of the hierarchical name is the root block. Get_full_name
- get_parent()
Function
get_parent
Get the parent block. Get_parent
- Return type:
- get_block()
Get_block
- Return type:
- get_n_maps()
Function
get_n_maps
Returns the number of address maps this memory is mapped in. Get_n_maps
- is_in_map(uvm_reg_map map)
Function
is_in_map
Return TRUE if this memory is in the specified address map . Is_in_map
- Parameters:
map (uvm_reg_map)
- get_maps(uvm_reg_map maps)
Function
get_maps
Returns all of the address maps where this memory is mapped. Get_maps
- Parameters:
maps (uvm_reg_map)
- get_local_map(uvm_reg_map map, string caller = "")
Get_local_map
- Parameters:
map (uvm_reg_map) --
Local
caller (string)
- Return type:
- get_default_map(string caller = "")
Get_default_map
- Parameters:
caller (string) --
Local
- Return type:
- get_rights(uvm_reg_map map = null)
Function
get_rights
Returns the access rights of this memory.
Returns "RW", "RO" or "WO". The access rights of a memory is always "RW", unless it is a shared memory with access restriction in a particular address map.
If no address map is specified and the memory is mapped in only one address map, that address map is used. If the memory is mapped in more than one address map, the default address map of the parent block is used.
If an address map is specified and the memory is not mapped in the specified address map, an error message is issued and "RW" is returned. Get_rights
- Parameters:
map (uvm_reg_map)
- get_access(uvm_reg_map map = null)
Function
get_access
Returns the access policy of the memory when written and read via an address map.
If the memory is mapped in more than one address map, an address map must be specified. If access restrictions are present when accessing a memory through the specified address map, the access mode returned takes the access restrictions into account. For example, a read-write memory accessed through a domain with read-only restrictions would return "RO". Get_access
- Parameters:
map (uvm_reg_map)
- get_size()
Function
get_size
Returns the number of unique memory locations in this memory. Get_size
- get_n_bytes()
Function
get_n_bytes
Return the width, in number of bytes, of each memory location. Get_n_bytes
- get_n_bits()
Function
get_n_bits
Returns the width, in number of bits, of each memory location. Get_n_bits
- get_max_size()
Function
get_max_size
Returns the maximum width, in number of bits, of all memories. Get_max_size
- get_virtual_registers(uvm_vreg regs)
Function
get_virtual_registers
Return the virtual registers in this memory
Fills the specified array with the abstraction class for all of the virtual registers implemented in this memory. The order in which the virtual registers are located in the array is not specified. Get_virtual_registers
- Parameters:
regs (uvm_vreg)
- get_virtual_fields(uvm_vreg_field fields)
Function
get_virtual_fields
Return the virtual fields in the memory
Fills the specified dynamic array with the abstraction class for all of the virtual fields implemented in this memory. The order in which the virtual fields are located in the array is not specified. Get_virtual_fields
- Parameters:
fields (uvm_vreg_field)
- get_vreg_by_name(string name)
Function
get_vreg_by_name
Find the named virtual register
Finds a virtual register with the specified name implemented in this memory and returns its abstraction class instance. If no virtual register with the specified name is found, returns null . Get_vreg_by_name
- Parameters:
name (string)
- Return type:
- get_vfield_by_name(string name)
Function
get_vfield_by_name
Find the named virtual field
Finds a virtual field with the specified name implemented in this memory and returns its abstraction class instance. If no virtual field with the specified name is found, returns null . Get_vfield_by_name
- Parameters:
name (string)
- Return type:
- get_vreg_by_offset(uvm_reg_addr_t offset, uvm_reg_map map = null)
Function
get_vreg_by_offset
Find the virtual register implemented at the specified offset
Finds the virtual register implemented in this memory at the specified offset in the specified address map and returns its abstraction class instance. If no virtual register at the offset is found, returns null . Get_vreg_by_offset
- Parameters:
offset (uvm_reg_addr_t)
map (uvm_reg_map)
- Return type:
- get_offset(uvm_reg_addr_t offset = 0, uvm_reg_map map = null)
Function
get_offset
Returns the base offset of a memory location
Returns the base offset of the specified location in this memory in an address map .
If no address map is specified and the memory is mapped in only one address map, that address map is used. If the memory is mapped in more than one address map, the default address map of the parent block is used.
If an address map is specified and the memory is not mapped in the specified address map, an error message is issued. Get_offset
- Parameters:
offset (uvm_reg_addr_t)
map (uvm_reg_map)
- Return type:
- get_address(uvm_reg_addr_t offset = 0, uvm_reg_map map = null)
Function
get_address
Returns the base external physical address of a memory location
Returns the base external physical address of the specified location in this memory if accessed through the specified address map .
If no address map is specified and the memory is mapped in only one address map, that address map is used. If the memory is mapped in more than one address map, the default address map of the parent block is used.
If an address map is specified and the memory is not mapped in the specified address map, an error message is issued. Get_address
- Parameters:
offset (uvm_reg_addr_t)
map (uvm_reg_map)
- Return type:
- get_addresses(uvm_reg_addr_t offset = 0, uvm_reg_map map = null, uvm_reg_addr_t addr)
Function
get_addresses
Identifies the external physical address(es) of a memory location
Computes all of the external physical addresses that must be accessed to completely read or write the specified location in this memory. The addressed are specified in little endian order. Returns the number of bytes transferred on each access.
If no address map is specified and the memory is mapped in only one address map, that address map is used. If the memory is mapped in more than one address map, the default address map of the parent block is used.
If an address map is specified and the memory is not mapped in the specified address map, an error message is issued. Get_addresses
- Parameters:
offset (uvm_reg_addr_t)
map (uvm_reg_map)
addr (uvm_reg_addr_t)
- set_frontdoor(uvm_reg_frontdoor ftdr, uvm_reg_map map = null, string fname = "", int lineno = 0)
Function
set_frontdoor
Set a user-defined frontdoor for this memory
By default, memories are mapped linearly into the address space of the address maps that instantiate them. If memories are accessed using a different mechanism, a user-defined access mechanism must be defined and associated with the corresponding memory abstraction class
If the memory is mapped in multiple address maps, an address map must be specified. Set_frontdoor
- Parameters:
ftdr (uvm_reg_frontdoor)
map (uvm_reg_map)
fname (string)
lineno (int)
- get_frontdoor(uvm_reg_map map = null)
Function
get_frontdoor
Returns the user-defined frontdoor for this memory
If null , no user-defined frontdoor has been defined. A user-defined frontdoor is defined by using the uvm_mem::set_frontdoor() method.
If the memory is mapped in multiple address maps, an address map must be specified. Get_frontdoor
- Parameters:
map (uvm_reg_map)
- Return type:
- set_backdoor(uvm_reg_backdoor bkdr, string fname = "", int lineno = 0)
Function
set_backdoor
Set a user-defined backdoor for this memory
By default, memories are accessed via the built-in string-based DPI routines if an HDL path has been specified using the uvm_mem::configure() or uvm_mem::add_hdl_path() method. If this default mechanism is not suitable (e.g. because the memory is not implemented in pure SystemVerilog) a user-defined access mechanism must be defined and associated with the corresponding memory abstraction class. Set_backdoor
- Parameters:
bkdr (uvm_reg_backdoor)
fname (string)
lineno (int)
- get_backdoor(bit inherited = 1)
Function
get_backdoor
Returns the user-defined backdoor for this memory
If null , no user-defined backdoor has been defined. A user-defined backdoor is defined by using the uvm_reg::set_backdoor() method.
If inherit is TRUE, returns the backdoor of the parent block if none have been specified for this memory. Get_backdoor
- Parameters:
inherited (bit)
- Return type:
- clear_hdl_path(string kind = "RTL")
Function
clear_hdl_path
Delete HDL paths
Remove any previously specified HDL path to the memory instance for the specified design abstraction. Clear_hdl_path
- Parameters:
kind (string)
- add_hdl_path(uvm_hdl_path_slice slices, string kind = "RTL")
Function
add_hdl_path
Add an HDL path
Add the specified HDL path to the memory instance for the specified design abstraction. This method may be called more than once for the same design abstraction if the memory is physically duplicated in the design abstraction. Add_hdl_path
- Parameters:
slices (uvm_hdl_path_slice)
kind (string)
- add_hdl_path_slice(string name, int offset, int size, bit first = 0, string kind = "RTL")
Function
add_hdl_path_slice
Add the specified HDL slice to the HDL path for the specified design abstraction. If first is TRUE, starts the specification of a duplicate HDL implementation of the memory. Add_hdl_path_slice
- Parameters:
name (string)
offset (int)
size (int)
first (bit)
kind (string)
- has_hdl_path(string kind = "")
Function
has_hdl_path
Check if a HDL path is specified
Returns TRUE if the memory instance has a HDL path defined for the specified design abstraction. If no design abstraction is specified, uses the default design abstraction specified for the parent block. Has_hdl_path
- Parameters:
kind (string)
- get_hdl_path(uvm_hdl_path_concat paths, string kind = "")
Function
get_hdl_path
Get the incremental HDL path(s)
Returns the HDL path(s) defined for the specified design abstraction in the memory instance. Returns only the component of the HDL paths that corresponds to the memory, not a full hierarchical path
If no design abstraction is specified, the default design abstraction for the parent block is used. Get_hdl_path
- Parameters:
paths (uvm_hdl_path_concat)
kind (string)
- get_full_hdl_path(uvm_hdl_path_concat paths, string kind = "", string separator = ".")
Function
get_full_hdl_path
Get the full hierarchical HDL path(s)
Returns the full hierarchical HDL path(s) defined for the specified design abstraction in the memory instance. There may be more than one path returned even if only one path was defined for the memory instance, if any of the parent components have more than one path defined for the same design abstraction
If no design abstraction is specified, the default design abstraction for each ancestor block is used to get each incremental path. Get_full_hdl_path
- Parameters:
paths (uvm_hdl_path_concat)
kind (string)
separator (string)
- get_hdl_path_kinds(string kinds)
Function
get_hdl_path_kinds
Get design abstractions for which HDL paths have been defined. Get_hdl_path_kinds
- Parameters:
kinds (string)
- backdoor_read_func(uvm_reg_item rw)
Function
backdoor_read_func
User-defined backdoor read access
Override the default string-based DPI backdoor access read for this memory type. Backdoor_read_func
- Parameters:
rw (uvm_reg_item)
- Return type:
- has_coverage(uvm_reg_cvr_t models)
Function
has_coverage
Check if memory has coverage model(s)
Returns TRUE if the memory abstraction class contains a coverage model for all of the models specified. Models are specified by adding the symbolic value of individual coverage model as defined in uvm_coverage_model_e. Has_coverage
- Parameters:
models (uvm_reg_cvr_t)
- set_coverage(uvm_reg_cvr_t is_on)
Function
set_coverage
Turns on coverage measurement.
Turns the collection of functional coverage measurements on or off for this memory. The functional coverage measurement is turned on for every coverage model specified using uvm_coverage_model_e symbolic identifiers. Multiple functional coverage models can be specified by adding the functional coverage model identifiers. All other functional coverage models are turned off. Returns the sum of all functional coverage models whose measurements were previously on.
This method can only control the measurement of functional coverage models that are present in the memory abstraction classes, then enabled during construction. See the uvm_mem::has_coverage() method to identify the available functional coverage models. Set_coverage
- Parameters:
is_on (uvm_reg_cvr_t)
- Return type:
- get_coverage(uvm_reg_cvr_t is_on)
Function
get_coverage
Check if coverage measurement is on.
Returns TRUE if measurement for all of the specified functional coverage models are currently on. Multiple functional coverage models can be specified by adding the functional coverage model identifiers.
See uvm_mem::set_coverage() for more details. Get_coverage
- Parameters:
is_on (uvm_reg_cvr_t)
- XsampleX(uvm_reg_addr_t addr, bit is_read, uvm_reg_map map)
- Parameters:
addr (uvm_reg_addr_t) --
Local
is_read (bit)
map (uvm_reg_map)
- do_print(uvm_printer printer)
Core ovm_object operations. Do_print
- Parameters:
printer (uvm_printer)
- convert2string()
Convert2string
- clone()
Clone
- Return type:
- do_copy(uvm_object rhs)
Do_copy
- Parameters:
rhs (uvm_object)
- do_compare(uvm_object rhs, uvm_comparer comparer)
Do_compare
- Parameters:
rhs (uvm_object)
comparer (uvm_comparer)
- do_pack(uvm_packer packer)
Do_pack
- Parameters:
packer (uvm_packer)
- do_unpack(uvm_packer packer)
Do_unpack
- Parameters:
packer (uvm_packer)
Tasks
- write(uvm_status_e status, uvm_reg_addr_t offset, uvm_reg_data_t value, uvm_path_e path = UVM_DEFAULT_PATH, uvm_reg_map map = null, uvm_sequence_base parent = null, int prior = -1, uvm_object extension = null, string fname = "", int lineno = 0)
Task
write
Write the specified value in a memory location
Write value in the memory location that corresponds to this abstraction class instance at the specified offset using the specified access path . If the memory is mapped in more than one address map, an address map must be specified if a physical access is used (front-door access). If a back-door access path is used, the effect of writing the register through a physical access is mimicked. For example, a read-only memory will not be written. Write
- Parameters:
status (uvm_status_e)
offset (uvm_reg_addr_t)
value (uvm_reg_data_t)
path (uvm_path_e)
map (uvm_reg_map)
parent (uvm_sequence_base)
prior (int)
extension (uvm_object)
fname (string)
lineno (int)
- read(uvm_status_e status, uvm_reg_addr_t offset, uvm_reg_data_t value, uvm_path_e path = UVM_DEFAULT_PATH, uvm_reg_map map = null, uvm_sequence_base parent = null, int prior = -1, uvm_object extension = null, string fname = "", int lineno = 0)
Task
read
Read the current value from a memory location
Read and return value from the memory location that corresponds to this abstraction class instance at the specified offset using the specified access path . If the register is mapped in more than one address map, an address map must be specified if a physical access is used (front-door access). Read
- Parameters:
status (uvm_status_e)
offset (uvm_reg_addr_t)
value (uvm_reg_data_t)
path (uvm_path_e)
map (uvm_reg_map)
parent (uvm_sequence_base)
prior (int)
extension (uvm_object)
fname (string)
lineno (int)
- burst_write(uvm_status_e status, uvm_reg_addr_t offset, uvm_reg_data_t value, uvm_path_e path = UVM_DEFAULT_PATH, uvm_reg_map map = null, uvm_sequence_base parent = null, int prior = -1, uvm_object extension = null, string fname = "", int lineno = 0)
Task
burst_write
Write the specified values in memory locations
Burst-write the specified values in the memory locations beginning at the specified offset . If the memory is mapped in more than one address map, an address map must be specified if not using the backdoor. If a back-door access path is used, the effect of writing the register through a physical access is mimicked. For example, a read-only memory will not be written. Burst_write
- Parameters:
status (uvm_status_e)
offset (uvm_reg_addr_t)
value (uvm_reg_data_t)
path (uvm_path_e)
map (uvm_reg_map)
parent (uvm_sequence_base)
prior (int)
extension (uvm_object)
fname (string)
lineno (int)
- burst_read(uvm_status_e status, uvm_reg_addr_t offset, uvm_reg_data_t value, uvm_path_e path = UVM_DEFAULT_PATH, uvm_reg_map map = null, uvm_sequence_base parent = null, int prior = -1, uvm_object extension = null, string fname = "", int lineno = 0)
Task
burst_read
Read values from memory locations
Burst-read into values the data the memory locations beginning at the specified offset . If the memory is mapped in more than one address map, an address map must be specified if not using the backdoor. If a back-door access path is used, the effect of writing the register through a physical access is mimicked. For example, a read-only memory will not be written. Burst_read
- Parameters:
status (uvm_status_e)
offset (uvm_reg_addr_t)
value (uvm_reg_data_t)
path (uvm_path_e)
map (uvm_reg_map)
parent (uvm_sequence_base)
prior (int)
extension (uvm_object)
fname (string)
lineno (int)
- poke(uvm_status_e status, uvm_reg_addr_t offset, uvm_reg_data_t value, string kind = "", uvm_sequence_base parent = null, uvm_object extension = null, string fname = "", int lineno = 0)
Task
poke
Deposit the specified value in a memory location
Deposit the value in the DUT memory location corresponding to this abstraction class instance at the specified offset , as-is, using a back-door access.
Uses the HDL path for the design abstraction specified by kind . Poke
- Parameters:
status (uvm_status_e)
offset (uvm_reg_addr_t)
value (uvm_reg_data_t)
kind (string)
parent (uvm_sequence_base)
extension (uvm_object)
fname (string)
lineno (int)
- peek(uvm_status_e status, uvm_reg_addr_t offset, uvm_reg_data_t value, string kind = "", uvm_sequence_base parent = null, uvm_object extension = null, string fname = "", int lineno = 0)
Task
peek
Read the current value from a memory location
Sample the value in the DUT memory location corresponding to this abstraction class instance at the specified offset using a back-door access. The memory location value is sampled, not modified.
Uses the HDL path for the design abstraction specified by kind . Peek
- Parameters:
status (uvm_status_e)
offset (uvm_reg_addr_t)
value (uvm_reg_data_t)
kind (string)
parent (uvm_sequence_base)
extension (uvm_object)
fname (string)
lineno (int)
- do_write(uvm_reg_item rw)
Do_write
- Parameters:
rw (uvm_reg_item)
- do_read(uvm_reg_item rw)
Do_read
- Parameters:
rw (uvm_reg_item)
- backdoor_write(uvm_reg_item rw)
Function
backdoor_write
User-defined backdoor read access
Override the default string-based DPI backdoor access write for this memory type. Backdoor_write
- Parameters:
rw (uvm_reg_item)
- pre_write(uvm_reg_item rw)
Task
pre_write
Called before memory write.
If the offset , value , access path , or address map are modified, the updated offset, data value, access path or address map will be used to perform the memory operation. If the status is modified to anything other than
, the operation is aborted. The registered callback methods are invoked after the invocation of this method.
- Parameters:
rw (uvm_reg_item)
- post_write(uvm_reg_item rw)
Task
post_write
Called after memory write.
If the status is modified, the updated status will be returned by the memory operation.
The registered callback methods are invoked before the invocation of this method.
- Parameters:
rw (uvm_reg_item)
- pre_read(uvm_reg_item rw)
Task
pre_read
Called before memory read.
If the offset , access path or address map are modified, the updated offset, access path or address map will be used to perform the memory operation. If the status is modified to anything other than
, the operation is aborted. The registered callback methods are invoked after the invocation of this method.
- Parameters:
rw (uvm_reg_item)
- post_read(uvm_reg_item rw)
Task
post_read
Called after memory read.
If the readback data or status is modified, the updated readback //data or status will be returned by the memory operation.
The registered callback methods are invoked before the invocation of this method.
- Parameters:
rw (uvm_reg_item)
CLASS
uvm_mem
Memory abstraction base class
A memory is a collection of contiguous locations. A memory may be accessible via more than one address map.
Unlike registers, memories are not mirrored because of the potentially large data space: tests that walk the entire memory space would negate any benefit from sparse memory modelling techniques. Rather than relying on a mirror, it is recommended that backdoor access be used instead.