<html><head><meta http-equiv="Content-Type" content="text/html charset=us-ascii"><base href="x-msg://1411/"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space; "><br><div><div>On Mar 13, 2013, at 4:37 PM, Yin Ma <<a href="mailto:yinma@codeaurora.org">yinma@codeaurora.org</a>> wrote:</div><br class="Apple-interchange-newline"><blockquote type="cite"><div lang="EN-US" link="blue" vlink="purple" style="font-family: Helvetica; font-size: medium; font-style: normal; font-variant: normal; font-weight: normal; letter-spacing: normal; line-height: normal; orphans: 2; text-align: -webkit-auto; text-indent: 0px; text-transform: none; white-space: normal; widows: 2; word-spacing: 0px; -webkit-text-size-adjust: auto; -webkit-text-stroke-width: 0px; "><div class="WordSection1" style="page: WordSection1; "><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">Hi All,<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; "><o:p> </o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">In the target I am working, we comes cross a situation that the loop strength reduction<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">could deliver a better result but currently not, because<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt 0.5in; font-size: 11pt; font-family: Calibri, sans-serif; text-indent: -0.25in; "><span>1.<span style="font-style: normal; font-variant: normal; font-weight: normal; font-size: 7pt; line-height: normal; font-family: 'Times New Roman'; "> <span class="Apple-converted-space"> </span></span></span>the algorithm narrows search space by winner registers without considering<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt 0.5in; font-size: 11pt; font-family: Calibri, sans-serif; ">the target preferred format. (NarrowSearchSpaceByPickingWinnerRegs)<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt 0.5in; font-size: 11pt; font-family: Calibri, sans-serif; text-indent: -0.25in; "><span>2.<span style="font-style: normal; font-variant: normal; font-weight: normal; font-size: 7pt; line-height: normal; font-family: 'Times New Roman'; "> <span class="Apple-converted-space"> </span></span></span>Cost comparison solely favors the number register without considering other<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt 0.5in; font-size: 11pt; font-family: Calibri, sans-serif; ">Impacts.<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; "><o:p> </o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">For the case one,<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">NarrowSearchSpaceByPickingWinnerRegs filters by most occurred registers.<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">ld(basereg, immediate) is a target preferred addressing mode. However, it may<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">be deleted because basereg is very likely not to be the most occurred register<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">because the less opportunity in a combination.<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; "><o:p> </o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">For the case two, by observing the cost comparison equation<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">bool Cost::operator<(const Cost &Other) const {<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; "> if (NumRegs != Other.NumRegs) return NumRegs < Other.NumRegs;<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; "> if (AddRecCost != Other.AddRecCost) return AddRecCost < Other.AddRecCost;<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; "> if (NumIVMuls != Other.NumIVMuls) return NumIVMuls < Other.NumIVMuls;<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; "> if (NumBaseAdds != Other.NumBaseAdds) return NumBaseAdds < Other.NumBaseAdds;<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; "> if (ImmCost != Other.ImmCost) return ImmCost < Other.ImmCost;<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; "> if (SetupCost != Other.SetupCost) return SetupCost < Other.SetupCost;<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; "> return false;<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">}<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; "><o:p> </o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">If we have a case to compare<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">Cost at 5 regs, with addrec cost 1, plus 15 base adds, plus 1 imm cost, plus 4 setup cost.<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">Cost at 4 regs, with addrec cost 1, plus 28 base adds, plus 1 imm cost, plus 2 setup cost.<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">The current mode will select 4 regs case even there are 14 more base adds. And base<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">Adds matters in our targets.<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; "><o:p> </o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">So I think the current LSR should be pushing more decision into target dependent backend.<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">Like calling new functions in TargetTransformInfo. At least, in narrow search space and cost<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">comparison phase, or more in cost rating phase. LSR can be tightly cooped with the target<o:p></o:p></div><div style="margin: 0in 0in 0.0001pt; font-size: 11pt; font-family: Calibri, sans-serif; ">attributes in order to get the most beneficial result.</div></div></div></blockquote><br></div><div>Yes. LSR decisions are tightly coupled with the target architecture and potentially the subtarget microarcthitecture. As you figured out, the way to handle it is to communicate more information to LSR through TTI. It's easy to do this to solve individual benchmarks on your target. It's hard to know if you have a general solution that works across targets. But if you can add hooks in a way that preserves existing behavior on other targets it shouldn't be a problem. We want to design general hooks, but leave it up to someone doing the benchmarking to tune them for a particular target.</div><div><br></div><div>-Andy</div></body></html>