1 Target Independent Opportunities:
3 ===-------------------------------------------------------------------------===
5 FreeBench/mason contains code like this:
7 static p_type m0u(p_type p) {
8 int m[]={0, 8, 1, 2, 16, 5, 13, 7, 14, 9, 3, 4, 11, 12, 15, 10, 17, 6};
16 We currently compile this into a memcpy from a static array into 'm', then
17 a bunch of loads from m. It would be better to avoid the memcpy and just do
18 loads from the static array.
20 ===-------------------------------------------------------------------------===
22 Get the C front-end to expand hypot(x,y) -> llvm.sqrt(x*x+y*y) when errno and
23 precision don't matter (ffastmath). Misc/mandel will like this. :)
25 ===-------------------------------------------------------------------------===
27 For all targets, not just X86:
28 When llvm.memcpy, llvm.memset, or llvm.memmove are lowered, they should be
29 optimized to a few store instructions if the source is constant and the length
30 is smallish (< 8). This will greatly help some tests like Shootout/strcat.c
33 //===---------------------------------------------------------------------===//
35 Solve this DAG isel folding deficiency:
53 The problem is the store's chain operand is not the load X but rather
54 a TokenFactor of the load X and load Y, which prevents the folding.
56 There are two ways to fix this:
58 1. The dag combiner can start using alias analysis to realize that y/x
59 don't alias, making the store to X not dependent on the load from Y.
60 2. The generated isel could be made smarter in the case it can't
61 disambiguate the pointers.
63 Number 1 is the preferred solution.
65 //===---------------------------------------------------------------------===//