Skip to content

ci(fpga): place gf6/gf8/gf16-mul-clean and gf24-sub with HeAP, not SA #864

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
gHashTag merged 1 commit into from
Oct 4, 2026
Merged

ci(fpga): place gf6/gf8/gf16-mul-clean and gf24-sub with HeAP, not SA #864

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 6 additions & 1 deletion .github/workflows/ax7203-gf16-mul-clean.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,11 @@ jobs:
docker run --rm -v "$PWD:/work" -w /work regymm/openxc7 bash -c "cd /nextpnr-xilinx && python3 xilinx/python/bbaexport.py --device xc7a200tfbg484-2 --bba /work/build/gf16_mul/chipdb/xc7a200tfbg484-2.bba && bbasm -l /work/build/gf16_mul/chipdb/xc7a200tfbg484-2.bba /work/build/gf16_mul/chipdb/xc7a200tfbg484-2.bin"
- name: nextpnr (seed search — placer heap + per-seed timeout, lessons from gf20)
run: |
# --placer heap, not sa (#862). The step name already said heap; the flag did not. SA leaves invalid
# tiles on this design ("post-placement validity check failed for Bel 'SLICE_.../A5FF'",
# openXC7/nextpnr-xilinx#162) and placed 4 of seeds 1-16 (6, 11, 12, 15; 35-62 s each, Fmax 27.9-37.2 MHz),
# so this loop took 98 s to reach seed 6. HeAP placed 16 of 16, 8-10 s each, Fmax 60.6-66.5 MHz
# (regymm/openxc7 sha256:35c91073, run 37215621928).
CLEAN=0; for s in 1 2 3 4 5 6 7 8; do
rm -f build/gf16_mul/gf16_mul_ax7203.fasm
timeout 20m docker run --rm -v "$PWD:/work" -w /work regymm/openxc7 timeout --signal=KILL 1800 nextpnr-xilinx \
Expand All @@ -45,7 +50,7 @@ jobs:
--json /work/build/gf16_mul/gf16_mul_ax7203.json \
--write /work/build/gf16_mul/gf16_mul_ax7203_routed.json \
--fasm /work/build/gf16_mul/gf16_mul_ax7203.fasm \
--freq 50.0 --seed $s --placer sa --router router1 --timing-allow-fail > /tmp/np_$s.log 2>&1 || true
--freq 50.0 --seed $s --placer heap --router router1 --timing-allow-fail > /tmp/np_$s.log 2>&1 || true
if grep -q "Failed to find a route" /tmp/np_$s.log || [ ! -f build/gf16_mul/gf16_mul_ax7203.fasm ]; then continue; fi
cp /tmp/np_$s.log /tmp/np.log; CLEAN=$s; break
done; [ "$CLEAN" = "0" ] && { echo "::error::no clean seed"; exit 1; }; echo "seed=$CLEAN"
Expand Down
7 changes: 6 additions & 1 deletion .github/workflows/ax7203-gf24-sub.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,11 @@ jobs:
docker run --rm -v "$PWD:/work" -w /work regymm/openxc7 bash -c "cd /nextpnr-xilinx && python3 xilinx/python/bbaexport.py --device xc7a200tfbg484-2 --bba /work/build/gf24_sub/chipdb/xc7a200tfbg484-2.bba && bbasm -l /work/build/gf24_sub/chipdb/xc7a200tfbg484-2.bba /work/build/gf24_sub/chipdb/xc7a200tfbg484-2.bin"
- name: nextpnr (seed search)
run: |
# --placer heap, not sa (#862). SA leaves invalid tiles on this design ("post-placement validity
# check failed for Bel 'SLICE_.../A5FF'", openXC7/nextpnr-xilinx#162) and placed 3 of seeds 1-16
# (6, 8, 13; 56-68 s each, Fmax 15.6-16.2 MHz), so this loop took 215 s to reach seed 6. HeAP placed
# 16 of 16, 6-9 s each, Fmax 26.3-28.1 MHz; both are under the 50 MHz target, which --timing-allow-fail
# already accepts (regymm/openxc7 sha256:35c91073, run 37215621928).
CLEAN=0; for s in 1 2 3 4 5 6 7 8; do
rm -f build/gf24_sub/gf24_sub_ax7203.fasm
docker run --rm -v "$PWD:/work" -w /work regymm/openxc7 timeout --signal=KILL 1800 nextpnr-xilinx \
Expand All @@ -45,7 +50,7 @@ jobs:
--json /work/build/gf24_sub/gf24_sub_ax7203.json \
--write /work/build/gf24_sub/gf24_sub_ax7203_routed.json \
--fasm /work/build/gf24_sub/gf24_sub_ax7203.fasm \
--freq 50.0 --seed $s --placer sa --router router1 --timing-allow-fail > /tmp/np_$s.log 2>&1 || true
--freq 50.0 --seed $s --placer heap --router router1 --timing-allow-fail > /tmp/np_$s.log 2>&1 || true
if grep -q "Failed to find a route" /tmp/np_$s.log || [ ! -f build/gf24_sub/gf24_sub_ax7203.fasm ]; then continue; fi
cp /tmp/np_$s.log /tmp/np.log; CLEAN=$s; break
done; [ "$CLEAN" = "0" ] && { echo "::error::no clean seed"; exit 1; }; echo "seed=$CLEAN"
Expand Down
7 changes: 6 additions & 1 deletion .github/workflows/ax7203-gf6-mul-clean.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,11 @@ jobs:
docker run --rm -v "$PWD:/work" -w /work regymm/openxc7 bash -c "cd /nextpnr-xilinx && python3 xilinx/python/bbaexport.py --device xc7a200tfbg484-2 --bba /work/build/gf6_mul/chipdb/xc7a200tfbg484-2.bba && bbasm -l /work/build/gf6_mul/chipdb/xc7a200tfbg484-2.bba /work/build/gf6_mul/chipdb/xc7a200tfbg484-2.bin"
- name: nextpnr (seed search — placer heap + per-seed timeout, lessons from gf20)
run: |
# --placer heap, not sa (#862). The step name already said heap; the flag did not. SA leaves invalid
# tiles on this design ("post-placement validity check failed for Bel 'SLICE_.../A5FF'",
# openXC7/nextpnr-xilinx#162) and placed 4 of seeds 1-16 (8, 10, 11, 15; 43-108 s each, Fmax 42.4-59.1 MHz),
# so this loop took 194 s to reach seed 8, the last one it tries. HeAP placed 16 of 16, 5-6 s each,
# Fmax 146.6-159.2 MHz (regymm/openxc7 sha256:35c91073, run 37215621928).
CLEAN=0; for s in 1 2 3 4 5 6 7 8; do
rm -f build/gf6_mul/gf6_mul_ax7203.fasm
timeout 20m docker run --rm -v "$PWD:/work" -w /work regymm/openxc7 timeout --signal=KILL 1800 nextpnr-xilinx \
Expand All @@ -45,7 +50,7 @@ jobs:
--json /work/build/gf6_mul/gf6_mul_ax7203.json \
--write /work/build/gf6_mul/gf6_mul_ax7203_routed.json \
--fasm /work/build/gf6_mul/gf6_mul_ax7203.fasm \
--freq 50.0 --seed $s --placer sa --router router1 --timing-allow-fail > /tmp/np_$s.log 2>&1 || true
--freq 50.0 --seed $s --placer heap --router router1 --timing-allow-fail > /tmp/np_$s.log 2>&1 || true
if grep -q "Failed to find a route" /tmp/np_$s.log || [ ! -f build/gf6_mul/gf6_mul_ax7203.fasm ]; then continue; fi
cp /tmp/np_$s.log /tmp/np.log; CLEAN=$s; break
done; [ "$CLEAN" = "0" ] && { echo "::error::no clean seed"; exit 1; }; echo "seed=$CLEAN"
Expand Down
7 changes: 6 additions & 1 deletion .github/workflows/ax7203-gf8-mul-clean.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,11 @@ jobs:
docker run --rm -v "$PWD:/work" -w /work regymm/openxc7 bash -c "cd /nextpnr-xilinx && python3 xilinx/python/bbaexport.py --device xc7a200tfbg484-2 --bba /work/build/gf8_mul/chipdb/xc7a200tfbg484-2.bba && bbasm -l /work/build/gf8_mul/chipdb/xc7a200tfbg484-2.bba /work/build/gf8_mul/chipdb/xc7a200tfbg484-2.bin"
- name: nextpnr (seed search — placer heap + per-seed timeout, lessons from gf20)
run: |
# --placer heap, not sa (#862). The step name already said heap; the flag did not. SA leaves invalid
# tiles on this design ("post-placement validity check failed for Bel 'SLICE_.../A5FF'",
# openXC7/nextpnr-xilinx#162) and placed 4 of seeds 1-16 (4, 10, 11, 12; 32-38 s each, Fmax 42.3-51.4 MHz),
# so this loop took 80 s to reach seed 4. HeAP placed 16 of 16, 5-7 s each, Fmax 102.2-116.0 MHz
# (regymm/openxc7 sha256:35c91073, run 37215621928).
CLEAN=0; for s in 1 2 3 4 5 6 7 8; do
rm -f build/gf8_mul/gf8_mul_ax7203.fasm
timeout 20m docker run --rm -v "$PWD:/work" -w /work regymm/openxc7 timeout --signal=KILL 1800 nextpnr-xilinx \
Expand All @@ -45,7 +50,7 @@ jobs:
--json /work/build/gf8_mul/gf8_mul_ax7203.json \
--write /work/build/gf8_mul/gf8_mul_ax7203_routed.json \
--fasm /work/build/gf8_mul/gf8_mul_ax7203.fasm \
--freq 50.0 --seed $s --placer sa --router router1 --timing-allow-fail > /tmp/np_$s.log 2>&1 || true
--freq 50.0 --seed $s --placer heap --router router1 --timing-allow-fail > /tmp/np_$s.log 2>&1 || true
if grep -q "Failed to find a route" /tmp/np_$s.log || [ ! -f build/gf8_mul/gf8_mul_ax7203.fasm ]; then continue; fi
cp /tmp/np_$s.log /tmp/np.log; CLEAN=$s; break
done; [ "$CLEAN" = "0" ] && { echo "::error::no clean seed"; exit 1; }; echo "seed=$CLEAN"
Expand Down
Loading