TVM 解读#
- 项目配置
- Relax
- Relay 前端
tvm.arith
解读tvm.arith.analyzer
解读- 解读
StructuralEqual
- 解读
ExprDeepEqual
canonical_simplify()
canonical_simplify()
mul+sumcanonical_simplify()
plit_indexcanonical_simplify()
divcanonical_simplify()
fp16_const_foldtvm.arith.Analyzer.canonical_simplify()
floormodcanonical_simplify()
canonical_mixedcanonical_simplify()
reduce_combinercanonical_simplify()
reducecanonical_simplify()
if_then_elsecanonical_simplify()
complex_casescanonical_simplify()
castcanonical_simplify()
normalize_min_value_exprcanonical_simplify()
proddivcanonical_simplify()
floormod_twocanonical_simplify()
le
tvm.arith.analyzer.ConstIntBound
simplify()
rewrite_simplify()
- 解读
- 解读
tvm.target
- 解读
tvm.ir
- 解读
tvm/script
- 运行时(C++)
- TVM FFI
- 解读张量表达式
- 解读
tvm.topi
- tvm.tir 解读
- TVM op
- 数据布局
- 属性
- TVM 表达式
- TVM 节点反射
- TVM 变换
- 源码映射机制
NameSupply
- 内存计划
- UMA 开发
- Relay
- 解读
DFPatternCallback
- 解读 tvm.relay.quantize
- 类型推断
ToMixedPrecision
解读- 解读
_test_type_solver
FForwardRewrite
tvm::relay::transformForwardRewrite
tvm::relay::RewriteAnnotatedOps()
- Relay 后端
- 构建 Relay 模型
- graph executor
- cpp-build-module
- graph-partitioning
- analysis-get-calibration_data
- memory-passes
- AnnotateUsedMemory
- 解读
- QNN
- 测试模块
- 解读代码生成
tvm.contrib