llvm dev - Dec 2012 - [LLVMdev] vmlx forwarding an cortex A9 question

Hi all,

On following code when I use llc targeting ARM Cortex-A9 as follows, if
vmlx-forwarding is turned off then 'vmla' instructions are generated. It
seems that -mcpu=cortex-a9 enables it by default and thus less 'vmla'
instructions are generated. On this specific example it doesn't make any
difference in term of performance, but on a more complex example disabling
vmlx-forwarding leads to significant win in performance wise (10% speed-up).
Does anyone can explain me what is the exact purpose of vmlx-forwarding option,
and why it is enabled by default on cortex-A9.

On attached file:
llc fun2.ll -march=arm -mcpu=cortex-a9 -mattr=+neon -float-abi=soft -o fun2.s
=> vmla instruction not generated
On attached file:
llc fun2.ll -march=arm -mcpu=cortex-a9 -mattr=+neon,-vmlx-forwarding
-float-abi=soft -o fun2.s => vmla instruction generated

Thanks for your answers.
Best Regards
Seb

-------------- next part --------------
A non-text attachment was scrubbed...
Name: fun2.ll
Type: application/octet-stream
Size: 2731 bytes
Desc: fun2.ll
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20121220/ce4615b5/attachment.obj>

I tried llc fun2.ll -march=arm -mcpu=cortex-a9 -mattr=+neon -float-abi=soft -o
fun2.s and I see vmla is generated:
  vmla.f32        q13, q13, q13

Besides, I don't think HasVMLxForwarding() is being used now. 

Thanks,
Weiming


-----Original Message-----
From: llvmdev-bounces at cs.uiuc.edu [mailto:llvmdev-bounces at cs.uiuc.edu] On
Behalf Of Sebastien DELDON-GNB
Sent: Thursday, December 20, 2012 2:26 AM
To: LLVMdev at cs.uiuc.edu
Subject: [LLVMdev] vmlx forwarding an cortex A9 question

Hi all,

On following code when I use llc targeting ARM Cortex-A9 as follows, if
vmlx-forwarding is turned off then 'vmla' instructions are generated. It
seems that -mcpu=cortex-a9 enables it by default and thus less 'vmla'
instructions are generated. On this specific example it doesn't make any
difference in term of performance, but on a more complex example disabling
vmlx-forwarding leads to significant win in performance wise (10% speed-up).
Does anyone can explain me what is the exact purpose of vmlx-forwarding option,
and why it is enabled by default on cortex-A9.

On attached file:
llc fun2.ll -march=arm -mcpu=cortex-a9 -mattr=+neon -float-abi=soft -o fun2.s
=> vmla instruction not generated On attached file:
llc fun2.ll -march=arm -mcpu=cortex-a9 -mattr=+neon,-vmlx-forwarding
-float-abi=soft -o fun2.s => vmla instruction generated

Thanks for your answers.
Best Regards
Seb