diff options
Diffstat (limited to 'lib/Target/ARM/README.txt')
| -rw-r--r-- | lib/Target/ARM/README.txt | 31 |
1 files changed, 11 insertions, 20 deletions
diff --git a/lib/Target/ARM/README.txt b/lib/Target/ARM/README.txt index 8fb1da30088f9..fb64d9fb4ce5b 100644 --- a/lib/Target/ARM/README.txt +++ b/lib/Target/ARM/README.txt @@ -8,12 +8,8 @@ Reimplement 'select' in terms of 'SEL'. add doesn't need to overflow between the two 16-bit chunks. * Implement pre/post increment support. (e.g. PR935) -* Coalesce stack slots! * Implement smarter constant generation for binops with large immediates. -* Consider materializing FP constants like 0.0f and 1.0f using integer - immediate instructions then copy to FPU. Slower than load into FPU? - //===---------------------------------------------------------------------===// Crazy idea: Consider code that uses lots of 8-bit or 16-bit values. By the @@ -422,14 +418,6 @@ are not remembered when the same two values are compared twice. //===---------------------------------------------------------------------===// -More register scavenging work: - -1. Use the register scavenger to track frame index materialized into registers - (those that do not fit in addressing modes) to allow reuse in the same BB. -2. Finish scavenging for Thumb. - -//===---------------------------------------------------------------------===// - More LSR enhancements possible: 1. Teach LSR about pre- and post- indexed ops to allow iv increment be merged @@ -540,10 +528,6 @@ while ARMConstantIslandPass only need to worry about LDR (literal). //===---------------------------------------------------------------------===// -We need to fix constant isel for ARMv6t2 to use MOVT. - -//===---------------------------------------------------------------------===// - Constant island pass should make use of full range SoImm values for LEApcrel. Be careful though as the last attempt caused infinite looping on lencod. @@ -593,10 +577,17 @@ it saves an instruction and a register. //===---------------------------------------------------------------------===// -add/sub/and/or + i32 imm can be simplified by folding part of the immediate -into the operation. +It might be profitable to cse MOVi16 if there are lots of 32-bit immediates +with the same bottom half. //===---------------------------------------------------------------------===// -It might be profitable to cse MOVi16 if there are lots of 32-bit immediates -with the same bottom half. +Robert Muth started working on an alternate jump table implementation that +does not put the tables in-line in the text. This is more like the llvm +default jump table implementation. This might be useful sometime. Several +revisions of patches are on the mailing list, beginning at: +http://lists.cs.uiuc.edu/pipermail/llvmdev/2009-June/022763.html + +//===---------------------------------------------------------------------===// + +Make use of the "rbit" instruction. |