From 0ecdfa610187a882015beda6bf67edef6902d898 Mon Sep 17 00:00:00 2001
From: Hashim Sharif <hsharif3@tyler.cs.illinois.edu>
Date: Mon, 7 Dec 2020 13:54:20 -0600
Subject: [PATCH] Adding list of porting changes required (docs)

---
 llvm/projects/hpvm-tensor-rt/docs/Changes.md | 63 ++++++++++++++++++++
 1 file changed, 63 insertions(+)
 create mode 100644 llvm/projects/hpvm-tensor-rt/docs/Changes.md

diff --git a/llvm/projects/hpvm-tensor-rt/docs/Changes.md b/llvm/projects/hpvm-tensor-rt/docs/Changes.md
new file mode 100644
index 0000000000..4247938b6c
--- /dev/null
+++ b/llvm/projects/hpvm-tensor-rt/docs/Changes.md
@@ -0,0 +1,63 @@
+
+# Header Files Needing Modification
+
+- llvm/include/IR/IntrisicsVISC.td
+  * Add intrinsics here
+
+- bechmarks/common/include/visc.h - Benchmarks header
+  * Include tensor intrinsic declarations
+
+- Header files for "PROMISE_TARGET" and "CUDNN_TARGET" Hints
+* /include/SupportVISC/DFG2LLVM.h
+* /include/SupportVISC/VISCUtils.h
+* /include/SupportVISC/VISCHint.h
+* 
+
+
+
+# Passes Needing Modification
+
+- GenVISC
+  *  Handle Tensor Intrinsics
+  * Add Declarations for Intrinsic Functions called (at top)
+  * Modify runOnModule to include rules for functions -> intrinsics
+
+
+- ClearDFG
+  * Incorporate new changes for visc.node.id
+
+
+# Passes to Move
+
+- lib/Transforms/DFG2LLVM_CUDNN
+- lib/Transforms/DFG2LLVM_WrapperAPI
+- lib/Transforms/FusedNodesHPVM
+
+** Changes needed:
+- Modify paths to tensor_runtime.ll
+- Use `hpvm' instead of `visc' intrinsics
+
+
+# Projects to Move
+
+- hpvm-tensor-rt
+  * Make sure to not move 'tuner_results' and 'model_params' subdirectories
+
+- gpu_profiler
+
+- soc_simlator
+
+- Keras (wait on it)
+
+
+
+# Build System
+
+- Add rules to /lib/Transforms/CMakeLists.txt
+- Add rules to /llvm/projects/CmakeLists.txt
+- Automate the generation of 'tensor_runtime.ll`
+- Add CUDNN, CUDA paths to the template environment setup file
+- Move to using Cmake-3.17 (earlier 3.15)
+
+
+
-- 
GitLab