[llvm-bugs] [Bug 35397] New: [AMDGPU][MC][GFX8][GFX9] Use of bound_ctrl is incompatible with SP3
via llvm-bugs
llvm-bugs at lists.llvm.org
Thu Nov 23 03:50:14 PST 2017
https://bugs.llvm.org/show_bug.cgi?id=35397
Bug ID: 35397
Summary: [AMDGPU][MC][GFX8][GFX9] Use of bound_ctrl is
incompatible with SP3
Product: libraries
Version: trunk
Hardware: PC
OS: All
Status: NEW
Severity: enhancement
Priority: P
Component: Backend: AMDGPU
Assignee: unassignedbugs at nondot.org
Reporter: dpreobrazhensky at luxoft.com
CC: llvm-bugs at lists.llvm.org
SP3 handles bound_ctrl differently in gfx8 and gfx9.
This causes incompatibility and inconsistency issues.
---------------------------------------------------------------------
Modifier SP3 meaning decoded by SP3 as
gfx8 gfx9 gfx8 gfx9
---------------------------------------------------------------------
bound_ctrl=0 set bit clr bit bound_ctrl=0 <omitted>
bound_ctrl=1 set bit set bit bound_ctrl=0 bound_ctrl=1
<omitted> clr bit clr bit <omitted> <omitted>
---------------------------------------------------------------------
Our assembler is incompatible with SP3 implementation:
-------------------------------------------------------------
Modifier LLVM Asm meaning decoded by LLVM Asm as
-------------------------------------------------------------
bound_ctrl=0 set bit bound_ctrl=0
bound_ctrl=1 <not supported> N/A
<omitted> clr bit <omitted>
-------------------------------------------------------------
I believe we should improve current implementation:
- make it more compatible with SP3;
- make it more consistent.
I see the following options:
a) Do not change anything;
b) Strictly follow SP3 implementation;
c) Mimic SP3 gfx8 implementation for both gfx8 and gfx9;
d) Mimic SP3 gfx9 implementation for both gfx8 and gfx9.
Artem and Sam suggested option c) while I tend to prefer option d).
Matt, what is your opinion?
--
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20171123/bcec3144/attachment.html>
More information about the llvm-bugs
mailing list