mirror of
https://github.com/RPCSX/llvm.git
synced 2024-12-25 21:45:21 +00:00
0aa1d5e02f
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@146 91177308-0d34-0410-b5e6-96231b3b80d8
78 lines
4.0 KiB
Plaintext
78 lines
4.0 KiB
Plaintext
* Remove Neg instruction. It is equilivent to sub 0, <v>
|
|
* Fix DCE to elminate br <c>, %L1, %L1 so that it can optimize the main of
|
|
fib.ll better. Currently I have to do this to get best results:
|
|
as < fib.ll | opt -inline -sccp -dce -sccp -dce |dis
|
|
* Fix DCE to work better, so that SCCP can show it's true value.
|
|
* Implement ADCE
|
|
* Fix the const pool printer to print out constants in some sort of "sorted"
|
|
order. Then enable TestOptimizer.sh to diff -sccp output. Currently it
|
|
doesn't work because the diff fails because of ordering of the constant
|
|
pool. :(
|
|
* Enable DoConstantPoolMerging to do trivial DCE of constant values.
|
|
* Should provide "castTerminator, castPHI, etc" functions in Instruction, and
|
|
similar functions in other classes, that effectively do dynamic casts. This
|
|
would allow code like this:
|
|
if (I->isTerminator()) {
|
|
TerminatorInst *TI = (TerminatorInst*)I;
|
|
...
|
|
}
|
|
to be written as:
|
|
if (TerminatorInst *TI = I->castTerminatorInst()) {
|
|
...
|
|
}
|
|
* Think about whether edge split SSA form would be useful to do.
|
|
* Inlining should attempt to give block names the same name in the inlined
|
|
method (using SymbolTable::getUniqueName)
|
|
* The dropAllReferences code can be a noop when NDEBUG!!!
|
|
* Finish xvcg output
|
|
* pred/succ iterators on basic blocks don't handle switch statements correctly
|
|
* Enhance BB to make predecessor handling easier (to update PHI nodes)
|
|
* Provide a pass that eliminates critical edges from the CFG
|
|
* Provide a print pass to print out xvcg format files for vis
|
|
* I need to provide an option to the bytecode loader to ignore memory
|
|
dependance edges. Instead, the VM would just treat memory operations
|
|
(load, store, getfield, putfield, call) as pinned instructions.
|
|
* I need to have a way to prevent taking the address of a constant pool
|
|
reference. You should only be able to take the address of a variable.
|
|
Maybe taking the address of a constant copies it? What about virtual
|
|
function tables? Maybe a const pointer would be better...
|
|
* Structures should be accessed something like this: ubyte is ok. Limits
|
|
structure size to 256 members. This can be fixed later by either:
|
|
1. adding varient that takes ushort
|
|
2. Splitting structures into nested structures each of half size
|
|
<float> %f = load *{int, {float}} Str, 1, 0
|
|
store float %f, *{int, {float}} Str, 1, 0
|
|
* I'm noticing me writing a lot of code that looks like this (dtor material here):
|
|
ConstPool.dropAllReferences();
|
|
ConstPool.delete_all();
|
|
ConstPool.setParent(0);
|
|
~ConstPool
|
|
|
|
* Need a way to attach bytecode block info at various levels of asm code.
|
|
* Rename "ConstantPool" to "ConstPool"
|
|
* Maybe ConstantPool objects should keep themselves sorted as things are
|
|
inserted.
|
|
* Need to be able to inflate recursive types. %x = { *%x }, %x = %x ()
|
|
* Recognize and save comments in assembly and bytecode format
|
|
* Encode line number table in bytecode (like #line), optional table
|
|
|
|
* Encode negative relative offsets in the bytecode file
|
|
|
|
* Implement switch to switch on a constant pool array of type:
|
|
[{ label, int }] or [label] (lookup vs index switch)
|
|
* Apparently bison has a %pure_parser option. Maybe useful for Assembly/Parser
|
|
|
|
* Implement a header file that can read either assembly or bytecode, implement
|
|
a writer that can output either based on what is read with this reader..
|
|
* Implement the following derived types:
|
|
* structure/record { int %foo, int %bar} or { %foo = int, int }
|
|
* pointer int *
|
|
* "packed format", like this: [4 x sbyte]: Packed SIMD datatype
|
|
* Maybe 'tailcall' also?
|
|
* Include a method level bytecode block that defines a mapping between values
|
|
and registers that defines a minimally register allocated code. This can
|
|
make me finally address how to encode extensions in assembly.
|
|
* Bytecode reader should use extensions that may or may not be linked into the
|
|
application to read blocks. Thus an easy way to ignore symbol table info
|
|
would be to not link in that reader into the app.
|