tinygrad 0.10.2__py3-none-any.whl → 0.11.0__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/__init__.py +1 -1
- tinygrad/apps/llm.py +206 -0
- tinygrad/codegen/__init__.py +116 -0
- tinygrad/codegen/devectorizer.py +315 -172
- tinygrad/codegen/expander.py +8 -16
- tinygrad/codegen/gpudims.py +89 -0
- tinygrad/codegen/linearize.py +205 -203
- tinygrad/codegen/lowerer.py +92 -139
- tinygrad/codegen/opt/__init__.py +38 -0
- tinygrad/codegen/opt/heuristic.py +125 -0
- tinygrad/codegen/opt/kernel.py +510 -0
- tinygrad/{engine → codegen/opt}/search.py +51 -35
- tinygrad/codegen/opt/swizzler.py +134 -0
- tinygrad/codegen/opt/tc.py +127 -0
- tinygrad/codegen/quantize.py +67 -0
- tinygrad/device.py +122 -132
- tinygrad/dtype.py +152 -35
- tinygrad/engine/jit.py +81 -54
- tinygrad/engine/memory.py +46 -27
- tinygrad/engine/realize.py +82 -41
- tinygrad/engine/schedule.py +70 -445
- tinygrad/frontend/__init__.py +0 -0
- tinygrad/frontend/onnx.py +1253 -0
- tinygrad/frontend/torch.py +5 -0
- tinygrad/gradient.py +19 -27
- tinygrad/helpers.py +95 -47
- tinygrad/nn/__init__.py +7 -8
- tinygrad/nn/optim.py +72 -41
- tinygrad/nn/state.py +37 -23
- tinygrad/renderer/__init__.py +40 -60
- tinygrad/renderer/cstyle.py +143 -128
- tinygrad/renderer/llvmir.py +113 -62
- tinygrad/renderer/ptx.py +50 -32
- tinygrad/renderer/wgsl.py +27 -23
- tinygrad/runtime/autogen/am/am.py +5861 -0
- tinygrad/runtime/autogen/am/pm4_nv.py +962 -0
- tinygrad/runtime/autogen/am/pm4_soc15.py +931 -0
- tinygrad/runtime/autogen/am/sdma_4_0_0.py +5209 -0
- tinygrad/runtime/autogen/am/sdma_4_4_2.py +5209 -0
- tinygrad/runtime/autogen/am/sdma_5_0_0.py +7103 -0
- tinygrad/runtime/autogen/am/sdma_6_0_0.py +8085 -0
- tinygrad/runtime/autogen/am/smu_v13_0_0.py +3068 -0
- tinygrad/runtime/autogen/am/smu_v14_0_2.py +3605 -0
- tinygrad/runtime/autogen/amd_gpu.py +1433 -67197
- tinygrad/runtime/autogen/comgr.py +35 -9
- tinygrad/runtime/autogen/comgr_3.py +906 -0
- tinygrad/runtime/autogen/cuda.py +2419 -494
- tinygrad/runtime/autogen/hsa.py +57 -16
- tinygrad/runtime/autogen/ib.py +7171 -0
- tinygrad/runtime/autogen/io_uring.py +917 -118
- tinygrad/runtime/autogen/kfd.py +748 -26
- tinygrad/runtime/autogen/libc.py +613 -218
- tinygrad/runtime/autogen/libusb.py +1643 -0
- tinygrad/runtime/autogen/nv/nv.py +8602 -0
- tinygrad/runtime/autogen/nv_gpu.py +7218 -2072
- tinygrad/runtime/autogen/opencl.py +2 -4
- tinygrad/runtime/autogen/sqtt.py +1789 -0
- tinygrad/runtime/autogen/vfio.py +3 -3
- tinygrad/runtime/autogen/webgpu.py +273 -264
- tinygrad/runtime/graph/cuda.py +3 -3
- tinygrad/runtime/graph/hcq.py +68 -29
- tinygrad/runtime/graph/metal.py +29 -13
- tinygrad/runtime/graph/remote.py +114 -0
- tinygrad/runtime/ops_amd.py +537 -320
- tinygrad/runtime/ops_cpu.py +108 -7
- tinygrad/runtime/ops_cuda.py +12 -14
- tinygrad/runtime/ops_disk.py +13 -10
- tinygrad/runtime/ops_dsp.py +47 -40
- tinygrad/runtime/ops_gpu.py +13 -11
- tinygrad/runtime/ops_hip.py +6 -9
- tinygrad/runtime/ops_llvm.py +35 -15
- tinygrad/runtime/ops_metal.py +29 -19
- tinygrad/runtime/ops_npy.py +5 -3
- tinygrad/runtime/ops_null.py +28 -0
- tinygrad/runtime/ops_nv.py +306 -234
- tinygrad/runtime/ops_python.py +62 -52
- tinygrad/runtime/ops_qcom.py +28 -39
- tinygrad/runtime/ops_remote.py +482 -0
- tinygrad/runtime/ops_webgpu.py +28 -28
- tinygrad/runtime/support/am/amdev.py +114 -249
- tinygrad/runtime/support/am/ip.py +211 -172
- tinygrad/runtime/support/amd.py +138 -0
- tinygrad/runtime/support/{compiler_hip.py → compiler_amd.py} +40 -8
- tinygrad/runtime/support/compiler_cuda.py +8 -11
- tinygrad/runtime/support/elf.py +2 -1
- tinygrad/runtime/support/hcq.py +184 -97
- tinygrad/runtime/support/ib.py +172 -0
- tinygrad/runtime/support/llvm.py +3 -4
- tinygrad/runtime/support/memory.py +251 -0
- tinygrad/runtime/support/nv/__init__.py +0 -0
- tinygrad/runtime/support/nv/ip.py +581 -0
- tinygrad/runtime/support/nv/nvdev.py +183 -0
- tinygrad/runtime/support/system.py +170 -0
- tinygrad/runtime/support/usb.py +268 -0
- tinygrad/runtime/support/webgpu.py +18 -0
- tinygrad/schedule/__init__.py +0 -0
- tinygrad/schedule/grouper.py +119 -0
- tinygrad/schedule/kernelize.py +368 -0
- tinygrad/schedule/multi.py +231 -0
- tinygrad/shape/shapetracker.py +40 -46
- tinygrad/shape/view.py +88 -52
- tinygrad/tensor.py +968 -542
- tinygrad/uop/__init__.py +117 -0
- tinygrad/{codegen/transcendental.py → uop/decompositions.py} +125 -38
- tinygrad/uop/mathtraits.py +169 -0
- tinygrad/uop/ops.py +1021 -0
- tinygrad/uop/spec.py +228 -0
- tinygrad/{codegen → uop}/symbolic.py +239 -216
- tinygrad/uop/upat.py +163 -0
- tinygrad/viz/assets/cdnjs.cloudflare.com/ajax/libs/highlight.js/11.10.0/languages/x86asm.min.js +19 -0
- tinygrad/viz/assets/d3js.org/d3.v7.min.js +2 -0
- tinygrad/viz/assets/dagrejs.github.io/project/dagre/latest/dagre.min.js +801 -0
- tinygrad/viz/index.html +203 -403
- tinygrad/viz/js/index.js +718 -0
- tinygrad/viz/js/worker.js +29 -0
- tinygrad/viz/serve.py +224 -102
- {tinygrad-0.10.2.dist-info → tinygrad-0.11.0.dist-info}/METADATA +24 -16
- tinygrad-0.11.0.dist-info/RECORD +141 -0
- {tinygrad-0.10.2.dist-info → tinygrad-0.11.0.dist-info}/WHEEL +1 -1
- tinygrad/codegen/kernel.py +0 -693
- tinygrad/engine/multi.py +0 -161
- tinygrad/ops.py +0 -1003
- tinygrad/runtime/ops_cloud.py +0 -220
- tinygrad/runtime/support/allocator.py +0 -94
- tinygrad/spec.py +0 -155
- tinygrad/viz/assets/d3js.org/d3.v5.min.js +0 -2
- tinygrad/viz/assets/dagrejs.github.io/project/dagre-d3/latest/dagre-d3.min.js +0 -4816
- tinygrad/viz/perfetto.html +0 -178
- tinygrad-0.10.2.dist-info/RECORD +0 -99
- {tinygrad-0.10.2.dist-info → tinygrad-0.11.0.dist-info/licenses}/LICENSE +0 -0
- {tinygrad-0.10.2.dist-info → tinygrad-0.11.0.dist-info}/top_level.txt +0 -0
@@ -19,7 +19,7 @@ def _try_dlopen_amd_comgr():
|
|
19
19
|
for candidate in PATHS_TO_TRY:
|
20
20
|
try: return ctypes.CDLL(candidate)
|
21
21
|
except OSError: pass
|
22
|
-
|
22
|
+
return None
|
23
23
|
|
24
24
|
|
25
25
|
def string_cast(char_pointer, encoding='utf-8', errors='strict'):
|
@@ -226,7 +226,8 @@ amd_comgr_data_kind_s__enumvalues = {
|
|
226
226
|
17: 'AMD_COMGR_DATA_KIND_AR',
|
227
227
|
18: 'AMD_COMGR_DATA_KIND_BC_BUNDLE',
|
228
228
|
19: 'AMD_COMGR_DATA_KIND_AR_BUNDLE',
|
229
|
-
|
229
|
+
20: 'AMD_COMGR_DATA_KIND_OBJ_BUNDLE',
|
230
|
+
20: 'AMD_COMGR_DATA_KIND_LAST',
|
230
231
|
}
|
231
232
|
AMD_COMGR_DATA_KIND_UNDEF = 0
|
232
233
|
AMD_COMGR_DATA_KIND_SOURCE = 1
|
@@ -242,7 +243,8 @@ AMD_COMGR_DATA_KIND_FATBIN = 16
|
|
242
243
|
AMD_COMGR_DATA_KIND_AR = 17
|
243
244
|
AMD_COMGR_DATA_KIND_BC_BUNDLE = 18
|
244
245
|
AMD_COMGR_DATA_KIND_AR_BUNDLE = 19
|
245
|
-
|
246
|
+
AMD_COMGR_DATA_KIND_OBJ_BUNDLE = 20
|
247
|
+
AMD_COMGR_DATA_KIND_LAST = 20
|
246
248
|
amd_comgr_data_kind_s = ctypes.c_uint32 # enum
|
247
249
|
amd_comgr_data_kind_t = amd_comgr_data_kind_s
|
248
250
|
amd_comgr_data_kind_t__enumvalues = amd_comgr_data_kind_s__enumvalues
|
@@ -515,6 +517,24 @@ try:
|
|
515
517
|
amd_comgr_action_info_get_option_list_item.argtypes = [amd_comgr_action_info_t, size_t, ctypes.POINTER(ctypes.c_uint64), ctypes.POINTER(ctypes.c_char)]
|
516
518
|
except AttributeError:
|
517
519
|
pass
|
520
|
+
try:
|
521
|
+
amd_comgr_action_info_set_bundle_entry_ids = _libraries['libamd_comgr.so'].amd_comgr_action_info_set_bundle_entry_ids
|
522
|
+
amd_comgr_action_info_set_bundle_entry_ids.restype = amd_comgr_status_t
|
523
|
+
amd_comgr_action_info_set_bundle_entry_ids.argtypes = [amd_comgr_action_info_t, ctypes.POINTER(ctypes.c_char) * 0, size_t]
|
524
|
+
except AttributeError:
|
525
|
+
pass
|
526
|
+
try:
|
527
|
+
amd_comgr_action_info_get_bundle_entry_id_count = _libraries['libamd_comgr.so'].amd_comgr_action_info_get_bundle_entry_id_count
|
528
|
+
amd_comgr_action_info_get_bundle_entry_id_count.restype = amd_comgr_status_t
|
529
|
+
amd_comgr_action_info_get_bundle_entry_id_count.argtypes = [amd_comgr_action_info_t, ctypes.POINTER(ctypes.c_uint64)]
|
530
|
+
except AttributeError:
|
531
|
+
pass
|
532
|
+
try:
|
533
|
+
amd_comgr_action_info_get_bundle_entry_id = _libraries['libamd_comgr.so'].amd_comgr_action_info_get_bundle_entry_id
|
534
|
+
amd_comgr_action_info_get_bundle_entry_id.restype = amd_comgr_status_t
|
535
|
+
amd_comgr_action_info_get_bundle_entry_id.argtypes = [amd_comgr_action_info_t, size_t, ctypes.POINTER(ctypes.c_uint64), ctypes.POINTER(ctypes.c_char)]
|
536
|
+
except AttributeError:
|
537
|
+
pass
|
518
538
|
try:
|
519
539
|
amd_comgr_action_info_set_working_directory_path = _libraries['libamd_comgr.so'].amd_comgr_action_info_set_working_directory_path
|
520
540
|
amd_comgr_action_info_set_working_directory_path.restype = amd_comgr_status_t
|
@@ -560,7 +580,8 @@ amd_comgr_action_kind_s__enumvalues = {
|
|
560
580
|
15: 'AMD_COMGR_ACTION_COMPILE_SOURCE_WITH_DEVICE_LIBS_TO_BC',
|
561
581
|
16: 'AMD_COMGR_ACTION_COMPILE_SOURCE_TO_RELOCATABLE',
|
562
582
|
17: 'AMD_COMGR_ACTION_COMPILE_SOURCE_TO_EXECUTABLE',
|
563
|
-
|
583
|
+
18: 'AMD_COMGR_ACTION_UNBUNDLE',
|
584
|
+
18: 'AMD_COMGR_ACTION_LAST',
|
564
585
|
}
|
565
586
|
AMD_COMGR_ACTION_SOURCE_TO_PREPROCESSOR = 0
|
566
587
|
AMD_COMGR_ACTION_ADD_PRECOMPILED_HEADERS = 1
|
@@ -580,7 +601,8 @@ AMD_COMGR_ACTION_COMPILE_SOURCE_TO_FATBIN = 14
|
|
580
601
|
AMD_COMGR_ACTION_COMPILE_SOURCE_WITH_DEVICE_LIBS_TO_BC = 15
|
581
602
|
AMD_COMGR_ACTION_COMPILE_SOURCE_TO_RELOCATABLE = 16
|
582
603
|
AMD_COMGR_ACTION_COMPILE_SOURCE_TO_EXECUTABLE = 17
|
583
|
-
|
604
|
+
AMD_COMGR_ACTION_UNBUNDLE = 18
|
605
|
+
AMD_COMGR_ACTION_LAST = 18
|
584
606
|
amd_comgr_action_kind_s = ctypes.c_uint32 # enum
|
585
607
|
amd_comgr_action_kind_t = amd_comgr_action_kind_s
|
586
608
|
amd_comgr_action_kind_t__enumvalues = amd_comgr_action_kind_s__enumvalues
|
@@ -801,12 +823,13 @@ __all__ = \
|
|
801
823
|
'AMD_COMGR_ACTION_LINK_RELOCATABLE_TO_RELOCATABLE',
|
802
824
|
'AMD_COMGR_ACTION_OPTIMIZE_BC_TO_BC',
|
803
825
|
'AMD_COMGR_ACTION_SOURCE_TO_PREPROCESSOR',
|
804
|
-
'
|
805
|
-
'
|
806
|
-
'
|
826
|
+
'AMD_COMGR_ACTION_UNBUNDLE', 'AMD_COMGR_DATA_KIND_AR',
|
827
|
+
'AMD_COMGR_DATA_KIND_AR_BUNDLE', 'AMD_COMGR_DATA_KIND_BC',
|
828
|
+
'AMD_COMGR_DATA_KIND_BC_BUNDLE', 'AMD_COMGR_DATA_KIND_BYTES',
|
829
|
+
'AMD_COMGR_DATA_KIND_DIAGNOSTIC',
|
807
830
|
'AMD_COMGR_DATA_KIND_EXECUTABLE', 'AMD_COMGR_DATA_KIND_FATBIN',
|
808
831
|
'AMD_COMGR_DATA_KIND_INCLUDE', 'AMD_COMGR_DATA_KIND_LAST',
|
809
|
-
'AMD_COMGR_DATA_KIND_LOG',
|
832
|
+
'AMD_COMGR_DATA_KIND_LOG', 'AMD_COMGR_DATA_KIND_OBJ_BUNDLE',
|
810
833
|
'AMD_COMGR_DATA_KIND_PRECOMPILED_HEADER',
|
811
834
|
'AMD_COMGR_DATA_KIND_RELOCATABLE', 'AMD_COMGR_DATA_KIND_SOURCE',
|
812
835
|
'AMD_COMGR_DATA_KIND_UNDEF', 'AMD_COMGR_LANGUAGE_HC',
|
@@ -828,6 +851,8 @@ __all__ = \
|
|
828
851
|
'AMD_COMGR_SYMBOL_TYPE_OBJECT', 'AMD_COMGR_SYMBOL_TYPE_SECTION',
|
829
852
|
'AMD_COMGR_SYMBOL_TYPE_UNKNOWN', 'amd_comgr_action_data_count',
|
830
853
|
'amd_comgr_action_data_get_data',
|
854
|
+
'amd_comgr_action_info_get_bundle_entry_id',
|
855
|
+
'amd_comgr_action_info_get_bundle_entry_id_count',
|
831
856
|
'amd_comgr_action_info_get_isa_name',
|
832
857
|
'amd_comgr_action_info_get_language',
|
833
858
|
'amd_comgr_action_info_get_logging',
|
@@ -835,6 +860,7 @@ __all__ = \
|
|
835
860
|
'amd_comgr_action_info_get_option_list_item',
|
836
861
|
'amd_comgr_action_info_get_options',
|
837
862
|
'amd_comgr_action_info_get_working_directory_path',
|
863
|
+
'amd_comgr_action_info_set_bundle_entry_ids',
|
838
864
|
'amd_comgr_action_info_set_isa_name',
|
839
865
|
'amd_comgr_action_info_set_language',
|
840
866
|
'amd_comgr_action_info_set_logging',
|