Test Performance with Latest HW Drop

[aroelke]

Current state of functionality:

32-bit	Positive	Negative
rwx	works	works
heap	- stanford_int_treesort_fixed fails with violation - everything else works	works
stack	works	works
cfi	- inline_asm_works_1 fails with violation - everything else works	works
threeClass	works	works

64-bit	Positive	Negative
rwx	works	works
heap	- stanford_int_treesort_fixed fails with violation - everything else works	- offset_free_fails_1 hangs - everything else works
stack	works	works
cfi	- inline_asm_works_1 fails with violation - everything else works	works
threeClass	works	works

[aroelke]

64-bit heap/offset_free_fails_1: misaligned load in pvPortFree, specifically with ld a1,8(a1) at PC 0xc00248c4 attempting to load 8 bytes from address 0xc008238c. This causes an exception in the AP for which there is no handler, so it jumps to PC 0x0 and hangs there (looping on exceptions, since that's also illegal).

This is due to an explicit addition of 4 to the pointer to the allocated memory, which on a 64-bit system causes it to point to the middle of a word, which maps to the same tag as the word-aligned original value (preventing detection of the illegal free), but confusing our heap allocation library.

Addressed by #127.

[aroelke]

The cfi policy appears to require a "Target" tag for all instructions that can be jumped to (whether using direct jumps like jal or indirect ones like beq), but the branch targets of the code in the inline assembly of inline_asm_works_1 do not have this tag.

This is a compiler bug and may not be fixed.