Commit Graph

184 Commits

Author SHA1 Message Date
David Shah 9f241c9a42 nexus: DSP inference support
Signed-off-by: David Shah <dave@ds0.me>
2020-11-20 08:45:55 +00:00
Xiretza 86e0440da9
Update nexus arch tests to new harness 2020-10-29 14:42:07 +01:00
Marcelina Kościelnicka d3b6b7fe98 xilinx: Fix attributes_test.ys
This test pretty much passes by accident — the `prep` command runs
memory_collect without memory_dff first, which prevents merging read
register into the memory, and thus blocks block RAM inference for a
reason completely unrelated to the attribute.

The attribute setting didn't actually work because it was set on the
containing module instead of the actual memory.
2020-10-24 23:52:37 +02:00
Marcelina Kościelnicka eb76d35e80 memory_dff: Fix needlessly duplicating enable bits.
When the register being merged into the EN signal happens to be a $sdff,
the current code creates a new $mux for every bit, even if they happen
to be identical (as is usually the case), preventing proper grouping
further down the flow.  Fix this by adding a simple cache.

Fixes #2409.
2020-10-22 13:03:42 +02:00
Miodrag Milanović ac0bd2ffc4
Merge pull request #2397 from daveshah1/nexus
synth_nexus: Initial implementation
2020-10-19 11:20:56 +02:00
David Shah 4d584d9319 synth_nexus: Initial implementation
Signed-off-by: David Shah <dave@ds0.me>
2020-10-15 08:52:15 +01:00
clairexen 2412e75495
Merge pull request #2380 from Xiretza/parallel-tests
Clean up and parallelize testsuite
2020-10-01 18:12:31 +02:00
Eddie Hung de79978372
xilinx: do not make DSP48E1 a whitebox for ABC9 by default (#2325)
* xilinx: eliminate SCCs from DSP48E1 model

* xilinx: add SCC test for DSP48E1

* Update techlibs/xilinx/cells_sim.v

* xilinx: Gate DSP48E1 being a whitebox behind ALLOW_WHITEBOX_DSP48E1

Have a test that checks it works through ABC9 when enabled
2020-09-23 09:15:24 -07:00
Xiretza acd47bbd52
tests: Centralize test collection and Makefile generation 2020-09-21 15:07:02 +02:00
Dan Ravensloft 1a07b330f8 intel_alm: Add multiply signedness to cells
Quartus assumes unsigned multiplication by default, breaking signed
multiplies, so add an input signedness parameter to the MISTRAL_MUL*
cells to propagate to Quartus' <family>_mac cells.
2020-08-26 22:50:16 +02:00
Marcelina Kościelnicka 50d532f01c techmap/shift_shiftx: Remove the "shiftx2mux" special path.
Our techmap rules for $shift and $shiftx cells contained a special path
that aimed to decompose the shift LSB-first instead of MSB-first in
select cases that come up in pmux lowering.  This path was needlessly
overcomplicated and contained bugs.

Instead of doing that, just switch over the main path to iterate
LSB-first (except for the specially-handled MSB for signed shifts
and overflow handling).  This also makes the code consistent with
shl/shr/sshl/sshr cells, which are already decomposed LSB-first.

Fixes #2346.
2020-08-20 12:44:09 +02:00
Marcelina Kościelnicka 9a4f420b4b Replace opt_rmdff with opt_dff. 2020-08-07 13:21:03 +02:00
Marcelina Kościelnicka 6cd135a5eb opt_expr: Remove -clkinv option, make it the default.
Adds -noclkinv option just in case the old behavior was actually useful
to someone.
2020-07-31 00:08:15 +02:00
Marcelina Kościelnicka cf60699884 synth_ice40: Use opt_dff.
The main part is converting ice40_dsp to recognize the new FF types
created in opt_dff instead of trying to recognize the mux patterns on
its own.

The fsm call has been moved upwards because the passes cannot deal with
$dffe/$sdff*, and other optimizations don't help it much anyway.
2020-07-30 22:26:20 +02:00
Marcelina Kościelnicka 8501342fc5 synth_xilinx: Use opt_dff.
The main part is converting xilinx_dsp to recognize the new FF types
created in opt_dff instead of trying to recognize the patterns on its
own.

The fsm call has been moved upwards because the passes cannot deal with
$dffe/$sdff*, and other optimizations don't help it much anyway.
2020-07-30 22:26:09 +02:00
Dan Ravensloft a2fb84fd0c intel_alm: direct M10K instantiation
This reverts commit a3a90f6377.
2020-07-27 15:39:06 +02:00
Dan Ravensloft 62311b7ec0 intel_alm: increase abc9 -W 2020-07-26 23:56:54 +02:00
Marcelina Kościelnicka 0c6d0d4b5d satgen: Add support for dffe, sdff, sdffe, sdffce cells. 2020-07-24 03:19:21 +02:00
Dan Ravensloft 4d9d90079c intel_alm: add additional ABC9 timings 2020-07-23 11:57:07 +01:00
Miodrag Milanović 910f421324
Merge pull request #2238 from YosysHQ/mwk/dfflegalize-anlogic
anlogic: Use dfflegalize.
2020-07-16 18:07:58 +02:00
Marcelina Kościelnicka 3050454d6e anlogic: Use dfflegalize. 2020-07-14 05:02:50 +02:00
Lofty a3a90f6377 Revert "intel_alm: direct M10K instantiation"
This reverts commit 09ecb9b2cf.
2020-07-13 18:05:38 +02:00
Marcelina Kościelnicka 347dd01c2f xilinx: Fix srl regression.
Of standard yosys cells, xilinx_srl only works on $_DFF_?_ and
$_DFFE_?P_, which get upgraded to $_SDFFE_?P?P_ by dfflegalize at the
point where xilinx_srl is called for non-abc9.  Fix this by running
ff_map.v first, resulting in FDRE cells, which are handled correctly.
2020-07-12 23:41:27 +02:00
Marcelina Kościelnicka c73ebeb90e gowin: Use dfflegalize. 2020-07-06 12:27:46 +02:00
Dan Ravensloft 09ecb9b2cf intel_alm: direct M10K instantiation 2020-07-05 23:28:59 +02:00
Dan Ravensloft 7f45cab27a synth_gowin: ABC9 support
This adds ABC9 support for synth_gowin; drastically improving
synthesis quality.
2020-07-05 22:07:17 +02:00
Dan Ravensloft 0d4c2f0a65 intel_alm: add Cyclone 10 GX tests 2020-07-05 21:36:38 +02:00
Dan Ravensloft b004f09018 intel_alm: DSP inference 2020-07-05 05:39:20 +02:00
Marcelina Kościelnicka 3ca2de0f77 synth_intel_alm: Use dfflegalize. 2020-07-04 22:56:16 +02:00
Dan Ravensloft c6765443fd Improve MISTRAL_FF specify rules
Co-authored-by: Eddie Hung <eddie@fpgeh.com>
2020-07-04 19:45:10 +02:00
Eddie Hung 52fbaeca07 tests: update fsm.ys resource count
Suspect it is to do with map/set ordering in techmap; should
be fixed by #1862?
2020-07-04 19:45:10 +02:00
Dan Ravensloft 8b4eb78849 intel_alm: fix DFFE matching 2020-06-11 19:55:51 +02:00
Claire Wolf 7112f187cd Add missing .gitignore file
Signed-off-by: Claire Wolf <claire@symbioticeda.com>
2020-06-04 22:25:47 +02:00
Eddie Hung 69850204c4
Merge pull request #2077 from YosysHQ/eddie/abc9_dff_improve
abc9: -dff improvements
2020-06-04 08:15:25 -07:00
Eddie Hung 45cd323055
Merge pull request #2082 from YosysHQ/eddie/abc9_scc_fixes
abc9: fixes around handling combinatorial loops
2020-06-03 17:35:46 -07:00
Eddie Hung 8a11019d38 tests: tidy up testcase 2020-06-03 08:41:55 -07:00
Eddie Hung 46ed0db2ec
Merge pull request #2080 from YosysHQ/eddie/fix_test_warnings
tests: reduce test warnings
2020-06-03 08:37:07 -07:00
Miodrag Milanovic 0a88f002e5 allow range for mux test 2020-06-01 13:48:19 +02:00
Eddie Hung ea4374a223 abc9_ops: update messaging (credit to @Xiretza for spotting) 2020-05-30 08:57:48 -07:00
Eddie Hung 1dce798dc5 tests: add ecp5 latch testcase with -abc9 2020-05-25 16:39:16 -07:00
Eddie Hung 08221edbc1 tests: xilinx macc test to have initval, shorten BMC depth for runtime 2020-05-25 10:09:05 -07:00
Eddie Hung 60aa804915 tests: fix some test warnings 2020-05-25 10:07:58 -07:00
Eddie Hung 9c6d216a06 tests: add test for abc9 -dff removing a redundant flop entirely 2020-05-25 08:43:33 -07:00
Eddie Hung 8dd93e389e tests: add testcase for abc9 -dff preserving flop names 2020-05-25 08:43:33 -07:00
Marcelina Kościelnicka aee439360b Add force_downto and force_upto wire attributes.
Fixes #2058.
2020-05-19 01:42:40 +02:00
Eddie Hung 7cd3f4a79b abc9_ops: add -prep_bypass for auto bypass boxes; refactor
Eliminate need for abc9_{,un}map.v in xilinx
-prep_dff_{hier,unmap} -> -prep_hier
2020-05-14 10:33:56 -07:00
Eddie Hung 8d7b3c06b2 abc9: suppress warnings when no compatible + used flop boxes formed 2020-05-14 10:33:56 -07:00
Eddie Hung cdd250ef16 xilinx: update abc9_dff tests 2020-05-14 10:33:56 -07:00
Eddie Hung 762b6ad74a xilinx: remove no-longer-relevant test 2020-05-14 10:33:56 -07:00
Dan Ravensloft 5b779f7f4e intel_alm: direct LUTRAM cell instantiation
By instantiating the LUTRAM cell directly, we avoid a trip through
altsyncram, which speeds up Quartus synthesis time. This also gives
a little more flexibility, as Yosys can build RAMs out of individual
32x1 LUTRAM cells.

While working on this, I discovered that the mem_init0 parameter of
<family>_mlab_cell gets ignored by Quartus.
2020-05-07 21:03:13 +02:00