add rdna3/rdna4/cdna4 to testamd (#14778)

* add rdna3/rdna4/cdna4 to testamd

* test simplify

* ci cleanups

* mergable

* skip slow
This commit is contained in:
George Hotz
2026-02-16 09:45:16 +08:00
committed by GitHub
parent ac079e43d7
commit bd18217f32

View File

@@ -615,56 +615,6 @@ jobs:
- name: Run process replay tests
uses: ./.github/actions/process-replay
testamd:
strategy:
fail-fast: false
matrix:
backend: [amd, amdllvm]
name: Linux (${{ matrix.backend }})
runs-on: ubuntu-22.04
timeout-minutes: 20
env:
AMD: 1
MOCKGPU: 1
FORWARD_ONLY: 1
AMD_LLVM: ${{ matrix.backend == 'amdllvm' && '1' || matrix.backend != 'amdllvm' && '0' }}
steps:
- name: Checkout Code
uses: actions/checkout@v4
- name: Setup Environment
uses: ./.github/actions/setup-tinygrad
with:
key: ${{ matrix.backend }}-minimal
deps: testing_unit
amd: 'true'
llvm: ${{ matrix.backend == 'amdllvm' && 'true' }}
- name: Check Device.DEFAULT and print some source
run: |
python3 -c "from tinygrad import Device; assert Device.DEFAULT in ['AMD'], Device.DEFAULT"
DEBUG=5 FORWARD_ONLY=1 python3 test/test_tiny.py TestTiny.test_plus
- name: Run LLVM test
if: matrix.backend=='amdllvm'
run: python test/device/test_amd_llvm.py
- name: Run pytest (amd)
run: python -m pytest -n=auto test/backend/test_ops.py test/backend/test_dtype.py test/backend/test_dtype_alu.py test/backend/test_linearizer.py test/backend/test_randomness.py test/backend/test_jit.py test/backend/test_graph.py test/backend/test_multitensor.py test/device/test_hcq.py test/testextra/test_cfg_viz.py --durations=20
- name: Run pytest (amd)
run: python -m pytest test/external/external_test_am.py --durations=20
- name: Run TRANSCENDENTAL math
run: TRANSCENDENTAL=2 python -m pytest -n=auto test/backend/test_ops.py::TestOps::test_sin test/backend/test_ops.py::TestOps::test_cos test/backend/test_ops.py::TestOps::test_tan test/backend/test_ops.py::TestOps::test_exp test/backend/test_ops.py::TestOps::test_log --durations=20
- name: Run TestOps.test_add with SQTT
run: |
VIZ=-2 DEBUG=5 python3 test/backend/test_ops.py TestOps.test_add
extra/sqtt/rgptool.py create "/tmp/profile.pkl.$USER" -o /tmp/gpu0.rgp
- name: Run AMD emulated tests on NULL backend
env:
AMD: 0
run: |
PYTHONPATH=. NULL=1 EMULATE=AMD python extra/mmapeak/mmapeak.py
PYTHONPATH=. NULL=1 EMULATE=AMD_CDNA4 python3 -m pytest -n=auto test/testextra/test_tk.py
- name: Run process replay tests
uses: ./.github/actions/process-replay
testamdasm:
name: AMD ASM IDE
runs-on: ubuntu-24.04
@@ -693,25 +643,62 @@ jobs:
echo "deb http://apt.llvm.org/$(lsb_release -cs)/ llvm-toolchain-$(lsb_release -cs)-21 main" | sudo tee /etc/apt/sources.list.d/llvm.list
sudo apt-get update
sudo apt-get install llvm-21 llvm-21-tools cloc
- name: RDNA3 Line Count
run: cloc --by-file tinygrad/renderer/amd/*.py
- name: Install rocprof-trace-decoder
run: sudo PYTHONPATH="." ./extra/sqtt/install_sqtt_decoder.py
- name: Run RDNA3 emulator tests
- name: Run AMD renderer tests
run: AMD_LLVM=0 python -m pytest -n=auto test/amd/ --durations 20
- name: Run RDNA3 emulator tests (AMD_LLVM=1)
- name: Run AMD renderer tests (AMD_LLVM=1)
run: AMD_LLVM=1 python -m pytest -n=auto test/amd/ --durations 20
- name: Run RDNA3 dtype tests
run: AMD_LLVM=0 pytest -n=auto test/backend/test_dtype_alu.py test/backend/test_dtype.py --durations 20
- name: Run RDNA3 dtype tests (AMD_LLVM=1)
run: AMD_LLVM=1 pytest -n=auto test/backend/test_dtype_alu.py test/backend/test_dtype.py --durations 20
# TODO: run all once emulator is faster
- name: Run RDNA3 ops tests
run: SKIP_SLOW_TEST=1 AMD_LLVM=0 pytest -n=auto test/backend/test_ops.py -k "test_sparse_categorical_crossentropy or test_tril or test_nonzero or test_softmax_argmax" --durations 20
- name: Run RDNA4 emulator tests
run: MOCKGPU_ARCH=rdna4 python -m pytest test/test_tiny.py -v --durations 20
- name: Run CDNA4 emulator tests
run: AMD_LLVM=1 MOCKGPU_ARCH=cdna4 python -m pytest test/test_tiny.py -v --durations 20
- name: Run TestOps.test_add with SQTT
run: |
VIZ=-2 DEBUG=5 python3 test/backend/test_ops.py TestOps.test_add
extra/sqtt/rgptool.py create "/tmp/profile.pkl.$USER" -o /tmp/gpu0.rgp
- name: Run AMD emulated tests on NULL backend
env:
AMD: 0
run: |
PYTHONPATH=. NULL=1 EMULATE=AMD python extra/mmapeak/mmapeak.py
PYTHONPATH=. NULL=1 EMULATE=AMD_CDNA4 python3 -m pytest -n=auto test/testextra/test_tk.py
- name: Run LLVM test
run: AMD_LLVM=1 python test/device/test_amd_llvm.py
testamd:
strategy:
fail-fast: false
matrix:
backend: [amd, amdllvm]
arch: [rdna3]
#arch: [rdna3, rdna4, cdna4]
name: Linux (${{ matrix.backend }} ${{ matrix.arch }})
runs-on: ubuntu-22.04
timeout-minutes: 15
env:
AMD: 1
MOCKGPU: 1
MOCKGPU_ARCH: ${{ matrix.arch }}
SKIP_SLOW_TEST: 1
AMD_LLVM: ${{ matrix.backend == 'amdllvm' && '1' || matrix.backend != 'amdllvm' && '0' }}
steps:
- name: Checkout Code
uses: actions/checkout@v4
- name: Setup Environment
uses: ./.github/actions/setup-tinygrad
with:
key: ${{ matrix.backend }}-minimal
deps: testing_unit
amd: 'true'
llvm: ${{ matrix.backend == 'amdllvm' && 'true' }}
- name: Check Device.DEFAULT and print some source
run: |
python3 -c "from tinygrad import Device; assert Device.DEFAULT in ['AMD'], Device.DEFAULT"
DEBUG=5 FORWARD_ONLY=1 python3 test/test_tiny.py TestTiny.test_plus
- name: Run pytest (amd)
run: python -m pytest -n=auto test/backend/test_ops.py test/backend/test_dtype.py test/backend/test_dtype_alu.py test/backend/test_linearizer.py test/backend/test_randomness.py test/backend/test_jit.py test/backend/test_graph.py test/backend/test_multitensor.py test/device/test_hcq.py test/testextra/test_cfg_viz.py test/external/external_test_am.py --durations=20
- name: Run TRANSCENDENTAL math
run: TRANSCENDENTAL=2 python -m pytest -n=auto test/backend/test_ops.py::TestOps::test_sin test/backend/test_ops.py::TestOps::test_cos test/backend/test_ops.py::TestOps::test_tan test/backend/test_ops.py::TestOps::test_exp test/backend/test_ops.py::TestOps::test_log --durations=20
- name: Run process replay tests
uses: ./.github/actions/process-replay
testnvidia:
strategy: