Merge branch 'ae' into ae-hopper

kernels/flash_attention/compile_flash.sh

@@ -30,7 +30,6 @@ for arch in "${archs[@]}"; do
     git pull
 
     # re-compile libvortexrt.a
-    # FIXME after restructure
     pushd ../../lib
     make
     popd
@@ -40,6 +39,7 @@ for arch in "${archs[@]}"; do
     # touch source file to force re-building, as the Makefile does not track
     # binary changes
     touch kernel.cpp
+    touch kernel.gemmini.cpp
 
     make CONFIG=flash.$arch.seqlen1024.headdim64
 done

kernels/sgemm_gemmini_dma/compile_virgo.sh

@@ -4,6 +4,11 @@
 git checkout ae-hopper
 git pull
 
+# re-compile libvortexrt.a
+pushd ../../lib
+make
+popd
+
 if [ ! -f input.a.rand01.fp16.m256n256k256.row.bin ]; then
     echo "input binaries not found, generating operands"
     python3 generate_operands.py

kernels/sgemm_tcore/compile_tcore.sh

@@ -56,7 +56,6 @@ for arch in "${archs[@]}"; do
     git pull
 
     # re-compile libvortexrt.a
-    # FIXME after restructure
     pushd ../../lib
     make
     popd