update doc

.github/workflows/stale.yaml

@@ -2,7 +2,7 @@ name: 'Close stale issues and PRs'
 
 on:
   schedule:
-    - cron: '30 1 1 * *'
+    - cron: '30 1 1 1 *'
   workflow_dispatch:
     inputs:
       logLevel:

README.md

@@ -60,6 +60,8 @@ Check [spconv 2.x algorithm introduction](docs/spconv2_algo.pdf) to understand s
 
 Use spconv >= cu114 if possible. cuda 11.4 can compile greatly faster kernel in some situation.
 
+Update Spconv: you **MUST UNINSTALL** all spconv/cumm/spconv-cuxxx/cumm-cuxxx first, use ```pip list | grep spconv``` and ```pip list | grep cumm``` to check all installed package. then use pip to install new spconv.
+
 ## NEWS
 
 * spconv 2.2: ampere feature support (by [EvernightAurora](https://github.com/EvernightAurora)), pure c++ code generation, nvrtc, drop python 3.6
@@ -67,7 +69,7 @@ Use spconv >= cu114 if possible. cuda 11.4 can compile greatly faster kernel in
 ## Spconv 2.2 vs Spconv 2.1
 
 * faster fp16 conv kernels (~5-30%) in ampere GPUs (tested in RTX 3090)
-* greatly faster int8 conv kernels (~1.2x~2.7x) in ampere GPUs (tested in RTX 3090)
+* greatly faster int8 conv kernels (~1.2x-2.7x) in ampere GPUs (tested in RTX 3090)
 * drop python 3.6 support
 * nvrtc support: kernel in old GPUs will be compiled in runtime.
 * [libspconv](docs/PURE_CPP_BUILD.md): pure c++ build of all spconv ops. see [example](example/libspconv/run_build.sh)

docs/BENCHMARK.md

@@ -18,25 +18,29 @@
 
 ### Network Benchmark without batchnorm (TF32/F16) in Different GPUs
 
-Basic: ```python -m spconv.benchmark bench_basic f16``` and ```python -m spconv.benchmark bench_basic tf32```
+Basic (120k voxels): ```python -m spconv.benchmark bench_basic f16``` and ```python -m spconv.benchmark bench_basic tf32```
 
 | GPUs | F16-Forward | F16-Backward | TF32-Forward  | TF32-Backward |
 | -------------- |:---------------------:|---------------------:|---------------------:| ---------------------:|
 | T4 | 18.74     | 25.51    | N/A      | N/A      |
 | RTX 3080 Laptop (150W) | 8.2    | 11.51    | 15.04      | 26.90      |
 | A100 | 13.02    | 12.43    | 12.35      | 14.93      |
-| RTX3090 | 11.84    | 11.84    | 13.23      | 15.79      |
+| RTX 3090 | 11.84    | 11.84    | 13.23      | 15.79      |
 | RTX A6000 | 11.11    | 8.97    | 12.30      | 12.79      |
+| TESLA V100-32G | 15.55    | 14.90    | N/A       | N/A      |
+| TESLA V100-16G | 10.61    | 13.91    | N/A       | N/A      |
 
-Large: ```python -m spconv.benchmark bench_large f16``` and ```python -m spconv.benchmark bench_large tf32```
+Large (900k voxels): ```python -m spconv.benchmark bench_large f16``` and ```python -m spconv.benchmark bench_large tf32```
 
 | GPUs | F16-Forward | F16-Backward | TF32-Forward  | TF32-Backward |
 | -------------- |:---------------------:|---------------------:|---------------------:| ---------------------:|
 | T4 | 128.7     | 203.3    | N/A      | N/A      |
 | RTX 3080 Laptop (150W) | 43.15    | 74.57    | 84.65      | 165.19      |
 | A100 | 19.85    | 31.24    | 29.58      | 55.63      |
-| RTX3090 | 27.83    | 40.45    | 44.51      | 73.17      |
+| RTX 3090 | 27.83    | 40.45    | 44.51      | 73.17      |
 | RTX A6000 | 28.62    | 39.86    | 45.43      | 74.11      |
+| TESLA V100-32G | 50.37    | 72.99   | N/A      | N/A      |
+| TESLA V100-16G | 38.65    | 61.47    | N/A       | N/A      |
 
 
 **NOTE** 

example/libspconv/README.md

 ## libspconv Example
 
-run ```run_build.sh``` to get ```libspconv.so```.
+run ```run_build.sh``` to get ```libspconv.so```.
\ No newline at end of file
+
+## libspconv API
+
+currently not available, but you can check python code to understand how to use C++ apis, spconv python and libspconv use same c++ code.
\ No newline at end of file