In the vvp code generator, when zero/sign extending a sub-expression,
the sub-expression is first evaluated and stored in one location in
the thread vector store, then is copied and extended into a second
location. The storage used for the initial sub-expression evaluation
is never deallocated.
This patch fixes a number of problems related to the divide and
modulus operators.
The net version (CA) of modulus did not support a signed version.
Division or modulus of a value wider than the machine word did
not correctly check for division by zero and return 'bx.
Fixed a problem in procedural modulus. The sign of the result is
only dependent on the L-value.
Division or modulus of a signed value that was the same width as
the machine word was creating an incorrect sign mask.
Division of a signed value that would fit into a single machine
word was not checking for division by zero.
Division or modulus of a wide value was always being done as
unsigned.
Added a negative operator for vvp_vector2_t. This made
implementing the signed wide division and modulus easier.
Support arrays of realtime variable arrays and net arrays. This
involved a simple fix to the ivl core parser, proper support in
the code generator, and rework the runtime support in vvp.
Even on 64 bit machines immediate values should be limited to
32 bits so that the a.out file will run correctly on a 32 bit
machine. This patch fixes a number of places where the code
generator was not checking for/observing this.
This patch adds non-blocking event control for array words.
It also fixes a problem where the word used to put the
calculated delay for a non-blocking array assignment was
not being released. It also fixes the non-blocking array
assignments to correctly handle off the end/beginning part
selects.
Since some event control assignments can be skipped we need an
event control clear so that future %evctl statements do not fail
their assert. This patch adds %evctl/c and uses it in the compiler
as appropriate to keep the event control information in sync.
This patch adds full event control for vectors and parts of a
vector. It also fixes the other non-blocking part select code
to correctly handle a negative offset ([1:-2] of a [4:0] will
have an offset of -2).
When evaluating an expression into an index register bit 4 is used
to determine if the expression had any X/Z bits. In the case of a
constant immediate value this can not happen so make sure the bit
is cleared.
We need consistent pointer output (%p) and under MinGW we need to
us _snprintf instead of snprintf to get this. The recent event
control addition missed this. This has been reported to the MinGW
developers and they are working on a solution.
This patch pushes the non-blocking event control information to
the code generator. It adds the %evctl statements that are used
to put the event control information into the special thread
event control registers. The signed version (%evctl/s) required
the implementation of %ix/getv/s to load a signed value into
an index register. It then adds %assign/wr/e event control based
non-blocking assignment for real values. It also fixes the other
non-blocking real assignments to use Transport instead of inertial
delays.
When the -V flag is passed to the iverilog command, we can easily
print the version information for the driver itself, but it is also
valuable to probe all the components that would have been used for
a real compile. So the driver executes the preprocessor and the ivl
core to have them print version information.
The ivl core program also tries to load the target code generator
and get version information to print. For this to work, create a new
optional entry point "target_query" that takes a query key string as
an argument and returns a const string as the result. Use this with
the key "version" to get version information out of the target.
Target selection is done by the DLL target code generator, so there
is no value having a layer of target selection ahead of it. Remove
all that redundant code and simplify the target config files to reflect
this.
It is possible for signe-extend to have a delay attached to it. (Same
for repeat.) Handle it like other LPM devices, by stuffing a .delay
device into the output path of the device, if appropriate.
Nothing to do with tab width! Eliminates useless
trailing spaces and tabs, and nearly all <space><tab>
pairings. No change to derived files (e.g., .vvp),
non-master files (e.g., lxt2_write.c) or the new tgt-vhdl
directory.
Low priority, simple entropy reduction. Please apply
unless it deletes some steganographic content you want
to keep.
This patch adds blocking repeat event controls and also makes the
base repeat statement sign aware. If the argument to repeat is
negative (it must be a signed variable) then this is treated just
like an argument of 0 (there is no looping). Doing this allows us
to model the repeat event control as follows.
lhs = repeat(count) @(event) rhs;
is translated to:
begin
temp = rhs;
repeat (count) @(event);
lhs = temp;
end
This patch also pushes the non-blocking event control
information to the elaboration phase where it will report they
are not currently supported.
Be more sophisticated with the code generated for constant values.
When values are large, use an optimal mix of %movi and %mov
instructions to get the desired value, no matter what the content.
This patch pushes the automatic property for both tasks and
functions to the code generators. The vvp back end does not
currently support this so it will error out during code
generation. The VHDL back end should be able to use this
property and tgt-stub prints the property. Having this will
also make it easier when we do adding this to the runtime.
This patch fixes a bug in the VVP code generator that causes syntactically
incorrect code to be generated if an event expression contains a memory or
array port.
This patch adds .cast/int and updates .cast/real to act as a local
(temporary) net and to support either a signed or unsigned input.
The vvp_vector4_t class not can convert an arbitrarily sized double
to a vector value. This removes the restriction of lround().
Also document the new statements.
This handles the general case of a non-real operand to a real-valued
division. This can turn up if only 1 operand of a divide is real. In
this case the division as a whole is real and the other operand must
be cast to real.
This method creates an extra node, but it should be a very compact
node and this node does no evaluation tricks so in the run time should
be no more expensive then folding the cast into the .arith/div.r itself.
The load-and-add for vectors %load/vp0/s can be combined with the
load-and-add for array words, and the %load/avp0/s added to round
out the combinations. This can make for fewer instructions when
words are padded in arithmetic expressions.
The %load/vp0 instruction adds a signed value to the signal value being
loaded, but it doesn't allow for a signed source vector. Add the
%load/vp0/s instruction that pads the loaded vector, and add the code
generator details to properly use it.
Actually, the immediate value handling is a little chaotic and should
be cleaned up. This patch opens the door for allowing signed immediate
values, and uses them in a few places where they are explicitly handled.
We must go through the opcodes that can take immediate values and make
explicit whether they are signed/unsigned/etc, and what their size
limits are.
The part select of a vector is converted by the compiler during
elaboration to a 0-based canonical address. But since it is legal
to address bits below the LSB, the canonical address can be negative.
So make the part select base for selecting from signals work with
signed arithmetic and make the code generator generate negative
indices when needed.
This is not a solution to all the problems, but is a better catch-all
then what is currently there. Allow the index field to be a T<> that
accesses the thread to get the address index.
Note that the lexor.lex currently returns the T<> as a T_SYMBOL, and the
users of T_SYMBOL objects need to interpret the meaning. This is
probably not the best idea, in light of all the other *<> formats that
now exist.
Fold the bi-directional part select into the pass switch (tran) support
so that it can be really bi-directional. This involves adding a new
tranvp device that does part select in tran islands, and reworking the
tran island resolution to handle non-identical nodes. This will be needed
for resistive tran devices anyhow.
The draw_net_input function is modified to account for nexus that is
a port of an island. Draw the ports (and the islands if necessary)
to the island and use the port output for the nexus instead of the
port input. This allows the bi-directional behavior of the port to
interpose itself in the data flow.
In this process of these changes, the draw_net_input function was
reorganized, and all the considerable amount of code for it was
moved to a file of its own. (vvp_scope.c is pretty unruly.)
The handling of immediate add used to do 16bits at a time. When it went
up to 32bits, the need to work in chunks vanished, but the chunk handling
was still there, this time shifting by 32, which causes problems on 32bit
machines. Simplify the %addi handling to avoid this.
Clarify that operands are typically 32bits, and have the code generator
make better use of this.
Also improve the %movi implementation to work well with marger vectors.
Add the %andi instruction to use immediate operands.
The MinGW system() implementation appears to return the straight
return value instead of the waitpid() like result that more
normal systems return. Because of this just return the system()
result without processing for MinGW compilations.
Older version of the MinGW runtime (pre 3.14) just used the
underlying vsnprintf(). Which has some problems. The 3.14 version
has some nice improvements, but it has a sever bug when processing
"%*.*f", -1, -1, <some_real_value>. Because of this we need to use
the underlying version without the enhancements for now.
snprintf prints %p differently than the other printf routines
so use _snprintf to get consistent results.
Only build the PDF files if both man and ps2pdf exist.
MinGW does not know about the z modifier for %d, %u, etc.
Add some missing Makefile check targets.
Array words don't have a vpiHandle with a label, so the %vpi_call
needs a special syntac for arguments that reference array words.
This syntax creates an array word reference that persists and can
be used at a VPI object by system tasks.
Memory words may have part selects assigned, but the code messed up
the testing for the validity of the part select base. This fixes do
detect constant bases so that base validity tests are handled at
compile time.
The draw_input_from_net function was being used to access words of
a var array, which doesn't work. Have the draw_input_from_net punt
on that case, and by the way the mux inputs don't need to use that
function, instead they should use the general draw_net_input function
to get the input labels.
It is possible for an array to have 1 word in it, so using the array
count to detect an array is incorrect. Use the ivl_signal_dimensions
function, which is there exactly for that purpose.
Save tons of space per memory word by not creating a vpi handle for
each and every word of a variable array. (Net arrays still get a
vpiHandle for every word.) The consequence of this is that all
accesses to a variable array need to go through the indexing.
This commit handles the most common places where this impacts, but
there are still problems.
It is a quirk of the $signed() system function that the argument
is converted to signed, but the operation that is performed is
not changed. So arithmetic operators on unsigned arguments inside
a $signed() expression still perform unsigned arithmetic.
The abs() function needs to be able to turn -0.0 into 0.0. This proved
to be too clunky (and perhaps impossible) to do with tests and jumps,
so add an %abs/wr opcode to do it using fabs().
The min/max functions need to take special care with the handling
of NaN operands. These matter, so generate the extra code to handle
them.
This patch adds file and line information for parameters and
local parameters. It also adds file/line stubs for signals in
the tgt-* files. It adds the pform code needed to eventually
do genvar checks and passing of genvar file/line information.
It verifies that a genvar does not have the same name as a
parameter/local parameter.
Implement in behavioral the abs/min/max operators for real values.
The parser treats these builtin functions as operators, unary or
binary, and elaborates them appropriately.
Also add enough code generator support to handle real valued expressions
in thread context.
This patch adds code to push the file and line information
for scope objects (modules, functions, tasks, etc.) to the
runtime. For modules, this includes the definition fields.
This patch adds ifnone functionality. It does not produce an
error when both an ifnone and an unconditional simple module
path are given. For this case the ifnone delays are ignored.
When generating a real expression you can have bits of the
expression that use vector only operands. When this happens
evaluate that part of the expression as a vector and then
convert it to a real value.
Detect thread bit allocation failures and fail gracefully. Print an
error message that points at the expression in question, and return
with an error code so that the compiler exits with an error.
This patch adds the ability to assign/deassign a bit or part select.
It also cleans up the code and fixes some problem in the forcing of
strength aware nets.
This patch adds a %assign/av/d opcode. This is a version of %assign/av
that allows a delay expression. Ultimately this allows a dynamically
indexed array to have a delay expression (non-constant delay value).
This patch removes the overly optimistic lookaside save for a signal
that has been set (%set/v). This is incorrect because if a force or
assign are active the value will not be set as expected.
vvp did not have the ability to handle real parameters.
This patch fixes that omission. Parameters are only used
by vpi calls to get compile time information.
The dup_expr() function for NetESignal was not copying the word_
expression. This would cause an individual array access to turn
into an entire array access. The file and line information was
also not being set correctly.
An assert() in tgt-vvp/eval_expr.c was also replaced with an
exit() since the appropriate information was already being
printed.
This patch adds functionality to do a bit or part select release
when a constant value is forced to the net/register. It also adds an
error message when the user tries to force a signal to a bit/part
select. This is not currently handled by the run time, so is now
caught in the compiler (tgt-vvp). Where when this functionality is
needed, it will be easy to know what to do instead of trying to track
down some odd runtime functionality.
What this all means is that you can force a signal to an entire
signal or you can force a constant to any part of a signal (bit,
part or entire) and release any of the above. Technically the
release of a constant value does not have to match the force.
The runtime verifies that if you are releasing a signal driver
it is being done as a full release. I don't see an easy way to
check this in the compiler.
To fix the signal deficiencies we need to rework the force_link
code to allow multiple drivers and partial unlinking. Much of
this is in the runtime, but the %force/link operator may also
need to be changed like I did to the %release opcode.
The IVL_EX_NONE type was not displaying any message and default did not
use the new ivl_expr_{file,lineno} functions to get the source location.
This patch remedies these problems.
This patch reworks much of the ternary code to short circuit when
possible and supports real values better. It adds a blend operator
for real values that returns 0.0 when the values differ and the value
when they match. This deviates slightly from the standard which
specifies that the value for reals is always 0.0 when the conditional
is 'bx. There are also a couple bug fixes.
These fixes have not been ported to continuous assignments yet.
Ternary operators used at compile time and in procedural assignments
should be complete (short circuit and support real values).
This patch adds support for delaying constants at time zero. It also
cleans up the code in elab_net.cc to use this capability instead of
building it with an extra BUFZ to carry the delay information.
This patch adds a real error message instead of an assert when
someone attempts to release a part or bit select. I plan to work
on the real functionality soon.
This patch adds the power operator for signed bit based values
in a continuous assignment. It also fixes a few other power
expression width problems. The expression width is still not
calculated correctly, since the correct method can produce huge
possible bit widths. The result is currently limited to the width
of the native long. This is because lround() is used to convert
from a double to an integer. A check in the code generator protects
the runtime from this limitation.
This patch adds the power operator for unsigned bit based values
in a continuous assignment. It also refactors the power code for
normal expressions and continuous assignments.
User functions called in a continuous assignment were not getting
their scope mangled. This is needed to handle arrayed instances and
possibly other things.
This patch adds bit based power support to normal expressions.
It also pushes the constant unsigned bit based calculation to
the runtime until the bit based method can be copied to the
compiler. Continuous assignments also need to use this type
of calculation.
This patch adds a check and prints a warning message when the power
operator is used with unsigned bit based values. It also fixes a couple
of typos and adds an asserts if the above power operator happens to
get to the tgt-vvp back end.
This patch add delays in continuous assignments for the shift operators,
concatenations and replications. It also reports an error if the user
attempts to take the modulus of a real value.
This patch uncovers a couple other problems in the system. I am trying
to fix one of them. I will report the other problem shortly.
This patch fixes some of the delays in bit based continuous
assignments (unary -, unary +constants, string constants,
reduction operators and user functions).
Use draw_lpm_output_delay() in all functions that need it.
When used in continuous assignments, user defined functions may have
delayed output. Handle that by generating the proper .delay node when
needed to delay the output of a .ufunc node.
Previously only the logical operators (~, &, |, ^, etc.) supported a
delayed value in a continuous assignment. This patch should extend this
to all operators. An extensive check of real values was done. The same
will be done shortly for bit based nets.
Checks for constructs currently unsupported in continuous assignments
provide a more explicit message (** operator, real user functions,
{!, && and ||} operators with a real argument).
Rework the encoding of a real value in the Cr<> label to be similar to
the format used by the %loadi/wr instruction. This mantissa-expoment
format better carries all the bits of the desired real value with
plenty of fidelity and range.
The comment that shows the intended value for real constants was
being printed with a %f. This is wrong for large or small values.
This patch changes the %f to a %g to get more consistent results.
A right shift may generate extra bits to preserve the proper shift
characteristic. This patch replaces the assert that was forcing the
input vector to not be greater than the input port width with code
to only select the required lower bits from the vector if it is
larger than the input port.
This patch adds support for the following statement:
<vector>[<i>] <= #<delay> <value>.
It is a copy with slight modification of code from the next
else/if clause.
This patch adds a new opcode %load/avp0 that is used to load a
word from an array and add a value to it. %load/vp0 was
changed/fixed to do the summation at the result width not the
vector width. This allows small vectors to index large arrays with
an offset. A few errors in the opcodes.txt file were also fixed.
tgt-vvp/eval_expr.c uninitialized variables
vpi/sys_display.c uninitialized variables
vvp/vpi_priv.cc deprecated string constant usage
vvp/vpi_vthr_vector.cc deprecated string constant usage
the last entry invokes vpip_name_string() and uses const char *
in the same way as the other 9 callers in vvp/*.cc, the only
difference is that the argument is static instead of computed.
Rework the ivl_file_table_* interface to be more generic and easier
to use. Also all the vvp examples except for memory.vvp have been
fixed to run correctly with the current vvp. Someone with a bit more
experience will need to fix memory.vvp.
Add the vpiFile and vpiLineNo properties to system functions.
Most other objects have stubs that return "N/A"/0. Interactive
functions (called from the debugger) use <interactive> for the
file name.
With this change, local symbols are not emitted in the vvp target,
but are marked as local. When thus marked, the vvp run time does not
offfer any VPI access and the signals (net or var) are effectively
invisible.
Zero-based part selects are a special case that is handled with
optimized code. Properly handle the part select width to get the
desired padding.
Signed-off-by: Stephen Williams <steve@icarus.com>
When user defined function calls are withing a $signed, the result
needs to be properly padded. To make this work, handle padding of
a function result exactly type the padding of a signal vector. this
is natural in that the function return value *is* in a signal vector.
This fixes pr1841300.
When loading a part select from the least significant bits, it is
OK to use the %load/v instruction to strip the high bits off. This
allows the zero-based part select to work in one step, without
loading excess bits.
Some patch caused a variable that held the constant base of a part
select to become unassigned. This led to unexplained failures to part
select a signal. Fix by properly detecting the case of a constant base
of 0 for the part select.
The %ix/getv instruction loads an integer register directly from
a signal vector. This is an optimization that an index register
is loaded from an expression is only a signal. It avoids the thread
vector space.
Where and expression is an immediate value added to a signal value,
it is possible to optimize them to a single instruction that combines
the load with an add at the same time.
String values are known to be 2-value bits, so they are natural
candidates for using the movi instruction to load the value in
far fewer instructions. With this patch, up to 16bits (two bytes)
at a time can be loaded per instruction.
Signed-off-by: Stephen Williams <steve@icarus.com>
tgt-vvp was taking strings and processing them as is, but the strings
could contain octal constants. These constants needed to be converted
to binary form before they could be turned into a bit pattern.
CMOS gates like NMOS and PMOS have a strength aware output.
They were missed when the NMOS and PMOS were fixed. "cmos.v"
will now pass in the test suite.
This patch fixes/enhances the array part select code. It also
verifies that any lval array index in a continuous assignment
is a constant value. Also, %set/av now uses index register 1 as
described in the documentation (as a bit offset).
Modpaths need to be in the proper scope in order to be available to
vpi functions. The code generator manages that by writing the modpath
records with scope switch functions. Batch the modpath records at the
end of the structural stuff so that the scope switching doesn't cause
trouble.
Also seperate the modpath code into its own source file.
Signed-off-by: Stephen Williams <steve@icarus.com>
The << operator may not be defined when the RHS is greater than or
equal to the LHS width. To work around the problem the mask is
incrementally generated.
There have been reports in the field of number-to-real conversions
doing bad things with the calculated mantissa. This patch eliminates
a opssible portability problem by reworking the negating of negative
mantissa values.
Signed-off-by: Stephen Williams <steve@icarus.com>
Constant propagation incorrectly elided an entire constant net node
if only the LSB of the driven vector was HiZ. This caused the entire
vector to look like HiZ. Also, the code generator for writing the
constant values missed bits.
Signed-off-by: Stephen Williams <steve@icarus.com>
Implement compare-immediate instructions and generate code to use
these new instructions to improve runtime performance.
Signed-off-by: Stephen Williams <steve@icarus.com>
This patch adds ivl_scope_time_precision() to the compiler which can
be used to extract the local scope precision. tgt-stub and tgt-vvp
have been modified to use this new function and output a value that
is appropriate. The vvp runtime has been altered to use this new
data which is accessed with the vpip_time_precision_from_handle()
function. vpiTimePrecision uses this function to return the correct
precision.
The outputs to bufif and nmos/pmos devices are always strength-aware,
so should have their outputs connected to a net8 device.
Signed-off-by: Stephen Williams <steve@icarus.com>
Real value are vector width of 1, fix real literal to reflect this.
fix leaking real registers in code generation for function arguments.
Load of signal should handle conversion from real to vector. Function
arguments, type vector passed a real value, are an example where this
comes up.
Signed-off-by: Stephen Williams <steve@icarus.com>
Arrays of nets that have all their words collapsed together can become
a collapsed array as a whole. Add support for this case in the vvp code
generator and runtime.
more general concept of arrays. The NetMemory and NetEMemory
classes are removed from the ivl core program, and the IVL_LPM_RAM
lpm type is removed from the ivl_target API.
Cary R
Martin Whitaker
Stephen Williams
Larry Doolittle
Martin Whitaker
steve