Switch from aeabi_lmul to custom assembly.

Spotted a sizeable performance regression on an app with a hot long
multiply. aebi_lmul is noticably slower than the previous custom Dalvik
assembly so switch to using that.

Change-Id: Id829de69193506ba7ac8fdc8b78009351631b0fc
2 files changed
tree: 90c732aa100677ab2cacd68036ca283171a4de94
  1. build/
  2. jdwpspy/
  3. src/
  4. test/
  5. tools/
  6. .gitignore
  7. Android.mk