diff --git a/kernels/flash_attention/compile_flash.sh b/kernels/flash_attention/compile_flash.sh
index 5808508c..a2209047 100755
--- a/kernels/flash_attention/compile_flash.sh
+++ b/kernels/flash_attention/compile_flash.sh
@@ -27,6 +27,7 @@ ln -sf input.c.rand.fp32.seqlen1024headdim64.row.bin input.c.bin
 
 for arch in "${archs[@]}"; do
     git checkout ae-flash-$arch
+    git pull
 
     # re-compile libvortexrt.a
     # FIXME after restructure
diff --git a/kernels/sgemm_gemmini_dma/compile_virgo.sh b/kernels/sgemm_gemmini_dma/compile_virgo.sh
index f2deb04c..de9b4b38 100755
--- a/kernels/sgemm_gemmini_dma/compile_virgo.sh
+++ b/kernels/sgemm_gemmini_dma/compile_virgo.sh
@@ -1,5 +1,9 @@
 #!/bin/sh
 
+# hopper and virgo has the same SIMT configurations
+git checkout ae-hopper
+git pull
+
 if [ ! -f input.a.rand01.fp16.m256n256k256.row.bin ]; then
     echo "input binaries not found, generating operands"
     python3 generate_operands.py
diff --git a/kernels/sgemm_gemmini_dma/kernel.cpp b/kernels/sgemm_gemmini_dma/kernel.cpp
index 8ae36bd8..5b6c3f71 100644
--- a/kernels/sgemm_gemmini_dma/kernel.cpp
+++ b/kernels/sgemm_gemmini_dma/kernel.cpp
@@ -53,7 +53,7 @@
 #define PRINTF(...) sprintf(PRINT_BUF, __VA_ARGS__)
 // #define PRINTF(...) vx_printf(__VA_ARGS__)
 #define SWISH(beta, x) ((x) / (1 + exp(-(beta) * (x))))
-// #define POWER
+#define POWER
 
 typedef uint16_t smem_elem_t;
 // typedef float smem_elem_t;
diff --git a/kernels/sgemm_tcore/compile_tcore.sh b/kernels/sgemm_tcore/compile_tcore.sh
index 128eb175..d338521b 100755
--- a/kernels/sgemm_tcore/compile_tcore.sh
+++ b/kernels/sgemm_tcore/compile_tcore.sh
@@ -53,6 +53,7 @@ done
 
 for arch in "${archs[@]}"; do
     git checkout ae-$arch
+    git pull
 
     # re-compile libvortexrt.a
     # FIXME after restructure