llvm/TailCallElim at 33fe993f2ec3a146e5954db14aa9a751141677f0 - llvm

RPCSX/llvm

mirror of https://github.com/RPCSX/llvm.git synced 2025-02-28 00:50:37 +00:00

History

Nick Lewycky 05da4dd998 Improve 'tail' call marking in TRE. A bootstrap of clang goes from 375k calls marked tail in the IR to 470k, however this improvement does not carry into an improvement of the call/jmp ratio on x86. The most common pattern is a tail call + br to a block with nothing but a 'ret'.

The number of tail call to loop conversions remains the same (1618 by my count).

The new algorithm does a local scan over the use-def chains to identify local "alloca-derived" values, as well as points where the alloca could escape. Then, a visit over the CFG marks blocks as being before or after the allocas have escaped, and annotates the calls accordingly.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@208017 91177308-0d34-0410-b5e6-96231b3b80d8

2014-05-05 23:59:03 +00:00

2010-06-26-MultipleReturnValues.ll

Catch more CHECK that can be converted to CHECK-LABEL in Transforms for easier debugging. No functionality change.

2013-07-14 01:50:49 +00:00

accum_recursion.ll

Catch more CHECK that can be converted to CHECK-LABEL in Transforms for easier debugging. No functionality change.

2013-07-14 01:50:49 +00:00

ackermann.ll

Revert the test moves from 176733. Use "REQUIRES: asserts" instead.

2013-03-12 16:27:52 +00:00

basic.ll

Improve 'tail' call marking in TRE. A bootstrap of clang goes from 375k calls marked tail in the IR to 470k, however this improvement does not carry into an improvement of the call/jmp ratio on x86. The most common pattern is a tail call + br to a block with nothing but a 'ret'.