This increases compatibility with certain older parsers in some cases
that worked before commit 15fae357 but do not work with the current
compat-int mode
Before this commit, every initial assignment to a memory generated
two wires and four assigns in a process. For unknown reasons (I did
not investigate), large amounts of assigns cause quadratic slowdown
later in the AST frontend, in processAst/removeSignalFromCaseTree.
As a consequence, common and reasonable Verilog code, such as:
reg [`WIDTH:0] mem [0:`DEPTH];
integer i; initial for (i = 0; i <= `DEPTH; i++) mem[i] = 0;
took extremely long time to be processed; around 80 s for a 8-wide,
8192-deep memory.
After this commit, initial assignments where address and/or data are
constant (after `generate`) do not incur the cost of intermediate
wires; expressions like `mem[i+1]=i^(i<<1)` are considered constant.
This results in speedups of orders of magnitude for common memory
sizes; it now takes merely 0.4 s to process a 8-wide, 8192-deep
memory, and only 5.8 s to process a 8-wide, 131072-deep one.
As a bonus, this change also results in nontrivial speedups later
in the synthesis pipeline, since pass sequencing issues meant that
all of these intermediate wires were subject to transformations such
as width reduction, even though they existed solely to be constant
folded away in `memory_collect`.
This commit tries to carefully follow the documented behavior of LSE
and Synplify. It will use `syn_ramstyle` attribute if there are any
write ports, and `syn_romstyle` attribute otherwise.
* LSE supports both `syn_ramstyle` and `syn_romstyle`.
* Synplify only supports `syn_ramstyle`, with same values as LSE.
* Synplify also supports `syn_rw_conflict_logic`, which is not
documented as supported for LSE.
Limitations of the Yosys implementation:
* LSE/Synplify support `syn_ramstyle="block_ram,no_rw_check"`
syntax to turn off insertion of transparency logic. There is
currently no way to support multiple valued attributes in
memory_bram. It is also not clear if that is a good idea, since
it can cause sim/synth mismatches.
* LSE/Synplify/1364.1 support block ROM inference from full case
statements. Yosys does not currently perform this transformation.
* LSE/Synplify propagate `syn_ramstyle`/`syn_romstyle` attributes
from the module to the inner memories. There is currently no way
to do this in Yosys (attrmvcp only works on cells and wires).
This commit tries to carefully follow the documented behavior of LSE
and Synplify. It will use `syn_ramstyle` attribute if there are any
write ports, and `syn_romstyle` attribute otherwise.
* LSE supports both `syn_ramstyle` and `syn_romstyle`.
* Synplify only supports `syn_ramstyle`, with same values as LSE.
* Synplify also supports `syn_rw_conflict_logic`, which is not
documented as supported for LSE.
Limitations of the Yosys implementation:
* LSE/Synplify appear to interpret attribute values insensitive
to case. There is currently no way to do this in Yosys (attrmap
can only change case of attribute names).
* LSE/Synplify support `syn_ramstyle="block_ram,no_rw_check"`
syntax to turn off insertion of transparency logic. There is
currently no way to support multiple valued attributes in
memory_bram. It is also not clear if that is a good idea, since
it can cause sim/synth mismatches.
* LSE/Synplify/1364.1 support block ROM inference from full case
statements. Yosys does not currently perform this transformation.
* LSE/Synplify propagate `syn_ramstyle`/`syn_romstyle` attributes
from the module to the inner memories. There is currently no way
to do this in Yosys (attrmvcp only works on cells and wires).
whitequark
R. Ou
Eddie Hung
Marcin Kościelnicki
Diego H
Rodrigo Alejandro Melo
Stefan Biereigel
David Shah