tinygrad 0.10.0__py3-none-any.whl → 0.10.1__py3-none-any.whl
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- tinygrad/codegen/kernel.py +114 -172
- tinygrad/codegen/linearize.py +211 -81
- tinygrad/codegen/lowerer.py +30 -35
- tinygrad/codegen/{uopgraph.py → rewriter.py} +69 -59
- tinygrad/codegen/transcendental.py +12 -13
- tinygrad/device.py +170 -47
- tinygrad/dtype.py +28 -26
- tinygrad/engine/jit.py +80 -63
- tinygrad/engine/memory.py +4 -5
- tinygrad/engine/multi.py +162 -0
- tinygrad/engine/realize.py +58 -107
- tinygrad/engine/schedule.py +381 -314
- tinygrad/engine/search.py +40 -44
- tinygrad/gradient.py +70 -0
- tinygrad/helpers.py +77 -58
- tinygrad/nn/__init__.py +30 -32
- tinygrad/nn/datasets.py +1 -2
- tinygrad/nn/optim.py +22 -26
- tinygrad/nn/state.py +89 -64
- tinygrad/ops.py +562 -446
- tinygrad/renderer/__init__.py +79 -36
- tinygrad/renderer/cstyle.py +70 -84
- tinygrad/renderer/llvmir.py +32 -20
- tinygrad/renderer/ptx.py +79 -99
- tinygrad/renderer/wgsl.py +87 -0
- tinygrad/runtime/autogen/amd_gpu.py +39507 -12
- tinygrad/runtime/autogen/comgr.py +2 -0
- tinygrad/runtime/autogen/kfd.py +4 -3
- tinygrad/runtime/autogen/kgsl.py +1 -1
- tinygrad/runtime/autogen/libpciaccess.py +2023 -0
- tinygrad/runtime/autogen/llvm.py +11379 -0
- tinygrad/runtime/autogen/vfio.py +891 -0
- tinygrad/runtime/graph/cuda.py +8 -9
- tinygrad/runtime/graph/hcq.py +84 -79
- tinygrad/runtime/graph/metal.py +19 -21
- tinygrad/runtime/ops_amd.py +488 -327
- tinygrad/runtime/ops_clang.py +15 -28
- tinygrad/runtime/ops_cloud.py +34 -34
- tinygrad/runtime/ops_cuda.py +30 -27
- tinygrad/runtime/ops_disk.py +62 -63
- tinygrad/runtime/ops_dsp.py +129 -38
- tinygrad/runtime/ops_gpu.py +30 -30
- tinygrad/runtime/ops_hip.py +29 -31
- tinygrad/runtime/ops_llvm.py +45 -40
- tinygrad/runtime/ops_metal.py +93 -73
- tinygrad/runtime/ops_npy.py +2 -2
- tinygrad/runtime/ops_nv.py +232 -270
- tinygrad/runtime/ops_python.py +51 -46
- tinygrad/runtime/ops_qcom.py +129 -157
- tinygrad/runtime/ops_webgpu.py +63 -0
- tinygrad/runtime/support/allocator.py +94 -0
- tinygrad/runtime/support/am/__init__.py +0 -0
- tinygrad/runtime/support/am/amdev.py +384 -0
- tinygrad/runtime/support/am/ip.py +463 -0
- tinygrad/runtime/support/compiler_cuda.py +4 -2
- tinygrad/runtime/support/elf.py +26 -4
- tinygrad/runtime/support/hcq.py +254 -324
- tinygrad/runtime/support/llvm.py +32 -0
- tinygrad/shape/shapetracker.py +84 -53
- tinygrad/shape/view.py +103 -138
- tinygrad/spec.py +154 -0
- tinygrad/tensor.py +744 -496
- {tinygrad-0.10.0.dist-info → tinygrad-0.10.1.dist-info}/METADATA +32 -21
- tinygrad-0.10.1.dist-info/RECORD +86 -0
- {tinygrad-0.10.0.dist-info → tinygrad-0.10.1.dist-info}/WHEEL +1 -1
- tinygrad/engine/lazy.py +0 -228
- tinygrad/function.py +0 -212
- tinygrad/multi.py +0 -177
- tinygrad/runtime/graph/clang.py +0 -39
- tinygrad-0.10.0.dist-info/RECORD +0 -77
- {tinygrad-0.10.0.dist-info → tinygrad-0.10.1.dist-info}/LICENSE +0 -0
- {tinygrad-0.10.0.dist-info → tinygrad-0.10.1.dist-info}/top_level.txt +0 -0
@@ -10,6 +10,8 @@ import ctypes, ctypes.util, os
|
|
10
10
|
PATHS_TO_TRY = [
|
11
11
|
'/opt/rocm/lib/libamd_comgr.so',
|
12
12
|
os.getenv('ROCM_PATH', '')+'/lib/libamd_comgr.so',
|
13
|
+
'/usr/local/lib/libamd_comgr.dylib',
|
14
|
+
'/opt/homebrew/lib/libamd_comgr.dylib',
|
13
15
|
]
|
14
16
|
def _try_dlopen_amd_comgr():
|
15
17
|
library = ctypes.util.find_library("amd_comgr")
|
tinygrad/runtime/autogen/kfd.py
CHANGED
@@ -10,10 +10,11 @@ import ctypes, os
|
|
10
10
|
|
11
11
|
|
12
12
|
|
13
|
-
import
|
13
|
+
import functools
|
14
|
+
from tinygrad.runtime.support.hcq import HWInterface
|
14
15
|
|
15
|
-
def _do_ioctl(__idir, __base, __nr, __user_struct, __fd, **kwargs):
|
16
|
-
ret =
|
16
|
+
def _do_ioctl(__idir, __base, __nr, __user_struct, __fd:HWInterface, **kwargs):
|
17
|
+
ret = __fd.ioctl((__idir<<30) | (ctypes.sizeof(made := __user_struct(**kwargs))<<16) | (__base<<8) | __nr, made)
|
17
18
|
if ret != 0: raise RuntimeError(f"ioctl returned {ret}")
|
18
19
|
return made
|
19
20
|
|
tinygrad/runtime/autogen/kgsl.py
CHANGED
@@ -129,7 +129,7 @@ else:
|
|
129
129
|
import fcntl, functools
|
130
130
|
|
131
131
|
def _do_ioctl(__idir, __base, __nr, __user_struct, __fd, __payload=None, **kwargs):
|
132
|
-
ret =
|
132
|
+
ret = __fd.ioctl((__idir<<30) | (ctypes.sizeof(made := (__payload or __user_struct(**kwargs)))<<16) | (__base<<8) | __nr, made)
|
133
133
|
if ret != 0: raise RuntimeError(f"ioctl returned {ret}")
|
134
134
|
return made
|
135
135
|
|