mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2024-08-19 19:29:31 +00:00
add some performance notes from looking at sgefa
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@25788 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
parent
6a28456e18
commit
5164a313e0
@ -248,3 +248,37 @@ This should be a high-priority to fix. With the fp-stack, this is a single
|
||||
instruction. With SSE it could be far better than this. Why is the sequence
|
||||
above using 'setp'? It shouldn't care about nan's.
|
||||
|
||||
//===---------------------------------------------------------------------===//
|
||||
|
||||
Is there a better way to implement Y = -X (fneg) than the literal code:
|
||||
|
||||
float %test(float %X) {
|
||||
%Y = sub float -0.0, %X
|
||||
ret float %Y
|
||||
}
|
||||
|
||||
movss LCPI1_0, %xmm0 ;; load -0.0
|
||||
subss 8(%esp), %xmm0 ;; subtract
|
||||
|
||||
//===---------------------------------------------------------------------===//
|
||||
|
||||
None of the SSE instructions are handled in X86RegisterInfo::foldMemoryOperand,
|
||||
which prevents the spiller from folding spill code into the instructions.
|
||||
|
||||
This leads to code like this:
|
||||
|
||||
mov %eax, 8(%esp)
|
||||
cvtsi2sd %eax, %xmm0
|
||||
instead of:
|
||||
cvtsi2sd 8(%esp), %xmm0
|
||||
|
||||
//===---------------------------------------------------------------------===//
|
||||
|
||||
This instruction selector selects 'int X = 0' as 'mov Reg, 0' not 'xor Reg,Reg'
|
||||
This is bigger and slower.
|
||||
|
||||
//===---------------------------------------------------------------------===//
|
||||
|
||||
LSR should be turned on for the X86 backend and tuned to take advantage of its
|
||||
addressing modes.
|
||||
|
||||
|
Loading…
Reference in New Issue
Block a user