github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-01-08 22:48:25 -05:00
Code Issues Packages Projects Releases Wiki Activity

set .amdhsa_kernarg_size in asm test (#13826)

Browse Source
This commit is contained in:
qazal
2025-12-25 13:08:14 +09:00
committed by GitHub
parent da1cb6a9ec
commit a1c1684b91
2 changed files with 6 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
.github/workflows/test.yml vendored
View File

@@ -642,7 +642,7 @@ jobs:
if: matrix.backend=='amdllvm'
run: python test/device/test_amd_llvm.py
- name: Run pytest (amd)
run: python -m pytest -n=auto test/test_ops.py test/test_dtype.py test/test_dtype_alu.py test/test_linearizer.py test/test_randomness.py test/test_jit.py test/test_graph.py test/test_multitensor.py test/device/test_hcq.py --durations=20
run: python -m pytest -n=auto test/test_ops.py test/test_dtype.py test/test_dtype_alu.py test/test_linearizer.py test/test_randomness.py test/test_jit.py test/test_graph.py test/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

8
test/testextra/test_cfg_viz.py
View File

@@ -4,7 +4,7 @@ import textwrap
from tinygrad import Device, Tensor
from tinygrad.uop.ops import UOp, Ops, track_rewrites
from tinygrad.renderer import ProgramSpec
from tinygrad.helpers import TracingKey
from tinygrad.helpers import TracingKey, getenv
from tinygrad.engine.realize import ExecItem, CompiledRunner
# TODO: use the RDNA3 renderer when it's in master
@@ -18,6 +18,7 @@ fn_name:
.rodata
.p2align 6
.amdhsa_kernel fn_name
.amdhsa_kernarg_size 8
.amdhsa_user_sgpr_kernarg_segment_ptr 1
.amdhsa_next_free_vgpr .amdgcn.next_free_vgpr
.amdhsa_next_free_sgpr .amdgcn.next_free_sgpr
@@ -53,12 +54,13 @@ amdhsa.kernels:
@track_rewrites(name=lambda *args,ret,**kwargs: TracingKey(ret.name, ret=ret))
def run_asm(name:str, src:str) -> ProgramSpec:
prg = ProgramSpec(name, template.replace("fn_name", name).replace("INSTRUCTION", textwrap.dedent(src)), Device.DEFAULT, UOp(Ops.SINK))
prg = ProgramSpec(name, template.replace("fn_name", name).replace("INSTRUCTION", textwrap.dedent(src)), Device.DEFAULT, UOp(Ops.SINK),
global_size=[1, 1, 1], local_size=[1, 1, 1], globals=[0])
ei = ExecItem(UOp(Ops.SINK), [Tensor.empty(1).uop.buffer.ensure_allocated()], prg=CompiledRunner(prg))
ei.run()
return prg
@unittest.skipUnless(Device.DEFAULT == "AMD", "only on AMD")
@unittest.skipUnless(Device.DEFAULT == "AMD" and not getenv("AMD_LLVM"), "only on AMD with comgr")
class TestCfg(unittest.TestCase):
def setUp(self):
arch = Device["AMD"].arch