Improvements to README.md.

Author: npfet

README.md

@@ -29,12 +29,30 @@ which were used for this project:
  - iverilog (v12)
  - python3 (v3.11.0)
 
+**Please note that iverilog v13 (development release) is not supported.**
+
+To install Python 3.11:
+```
+$ conda create -n codex python=3.11
+$ conda activate codex
+```
+
+Install [ICARUS Verilog](https://github.com/steveicarus/iverilog):
+```
+$ git clone https://github.com/steveicarus/iverilog.git && cd iverilog \
+        && git checkout v12-branch \
+        && sh ./autoconf.sh && ./configure && make -j4\
+        && make install
+```
+
 You will also need the following Python packages:
 
 ```
  % pip install langchain langchain-openai langchain-nvidia-ai-endpoints
 ```
 
+We plan to provide a Dockerfile and backwards compatibility mode with a prebuilt jsonl soon.
+
 ### Usage 
 
 The evalution harness is run using make and various evaluation parameters can be set as below:
@@ -45,6 +63,8 @@ mkdir -p build/
 make
 ```
 
+Evaluation can be sped up by providing the `-j` flag to make, such as `-j4` to run 4 worker processes.
+
 Available tasks are `code-complete-iccad2023` and `spec-to-rtl` with each referencing their corresponding `dataset_$task` directory containig the problems. Problem themselves are identical between the two datasets and only the task format changes.
 
 Valid models are listed at the top of `scripts/sv-generate`. The number of in-context learning examples can be between 0-4, and given with `--with-examples`. Samples to collect per problem are given by `--with-samples`. Finally, model temperature and top_p can be set to --with-temperature and --with-top-p, respectively.