Here's an old 2001 article on potential VFP performance:
ARM adds floating-point unit to processor cores
Looks like a fully pipelined unit with a single-cycle FMAC, but with no SSE or Altivec like SIMD ops except short vector load/store.
In XCode, put a floating point intensive compute function in a small file, set device for the target, turn off Thumb code generation in build options, and pick the the "Show Assembly" menu item to see some VFP code generation for that file.
.
ARM adds floating-point unit to processor cores
Looks like a fully pipelined unit with a single-cycle FMAC, but with no SSE or Altivec like SIMD ops except short vector load/store.
In XCode, put a floating point intensive compute function in a small file, set device for the target, turn off Thumb code generation in build options, and pick the the "Show Assembly" menu item to see some VFP code generation for that file.
.