[llvm-bugs] [Bug 33597] New: RegCoalescer introduces "Virtual register defs don't dominate all uses" error
via llvm-bugs
llvm-bugs at lists.llvm.org
Mon Jun 26 11:19:37 PDT 2017
https://bugs.llvm.org/show_bug.cgi?id=33597
Bug ID: 33597
Summary: RegCoalescer introduces "Virtual register defs don't
dominate all uses" error
Product: libraries
Version: trunk
Hardware: PC
OS: All
Status: NEW
Severity: enhancement
Priority: P
Component: Register Allocator
Assignee: unassignedbugs at nondot.org
Reporter: Matthew.Arsenault at amd.com
CC: llvm-bugs at lists.llvm.org
Created attachment 18715
--> https://bugs.llvm.org/attachment.cgi?id=18715&action=edit
Testcase
Renaming independent subregister live ranges in vreg_does_not_dominate
%vreg19: Found 4 equivalence classes.
%vreg19: Splitting into newly created: %vreg20 %vreg21 %vreg22
# Machine code for function vreg_does_not_dominate: NoPHIs, TracksLiveness
Function Live Ins: %VGPR0 in %vreg2, %SGPR30_SGPR31 in %vreg15
BB#0:
Live Ins: %VGPR0 %SGPR30_SGPR31 %SGPR5
%vreg15<def> = COPY %SGPR30_SGPR31; SReg_64:%vreg15
%vreg2<def> = COPY %VGPR0; VGPR_32:%vreg2
%vreg20:sub1<def,read-undef,tied3> = V_MAC_F32_e32 %vreg7<undef>,
%vreg6<undef>, %vreg20:sub1<undef,tied0>, %EXEC<imp-use>; VReg_128:%vreg20
VGPR_32:%vreg7,%vreg6
%vreg21:sub0<def> = V_MOV_B32_e32 0, %EXEC<imp-use>; VReg_128:%vreg21
%vreg19:sub2<def,read-undef> = COPY %vreg19:sub0; VReg_128:%vreg19
S_CBRANCH_VCCNZ <BB#2>, %VCC<imp-use,undef>
S_BRANCH <BB#1>
Successors according to CFG: BB#2(0x40000000 / 0x80000000 = 50.00%)
BB#1(0x40000000 / 0x80000000 = 50.00%)
BB#1:
Predecessors according to CFG: BB#0
%vreg22:sub3<def> = V_ADD_F32_e32 %vreg9<undef>, %vreg9<undef>,
%EXEC<imp-use>; VReg_128:%vreg22 VGPR_32:%vreg9
%vreg21:sub0<def> = V_ADD_F32_e64 0, 0, 0, 0, 0, 0, %EXEC<imp-use>;
VReg_128:%vreg21
%vreg20:sub1<def> = V_ADD_F32_e32 %vreg19:sub1, %vreg19:sub1,
%EXEC<imp-use>; VReg_128:%vreg20,%vreg19,%vreg19
%vreg19:sub2<def,read-undef> = COPY %vreg19:sub0; VReg_128:%vreg19
Successors according to CFG: BB#2(0x80000000 / 0x80000000 = 100.00%)
BB#2:
Predecessors according to CFG: BB#0 BB#1
BUFFER_STORE_DWORD_OFFEN %vreg19:sub3, %vreg2,
%SGPR0_SGPR1_SGPR2_SGPR3, %SGPR4, 12, 0, 0, 0, %EXEC<imp-use>; VReg_128:%vreg19
VGPR_32:%vreg2
BUFFER_STORE_DWORD_OFFEN %vreg19:sub2, %vreg2,
%SGPR0_SGPR1_SGPR2_SGPR3, %SGPR4, 8, 0, 0, 0, %EXEC<imp-use>; VReg_128:%vreg19
VGPR_32:%vreg2
BUFFER_STORE_DWORD_OFFEN %vreg19:sub1, %vreg2,
%SGPR0_SGPR1_SGPR2_SGPR3, %SGPR4, 4, 0, 0, 0, %EXEC<imp-use>; VReg_128:%vreg19
VGPR_32:%vreg2
BUFFER_STORE_DWORD_OFFEN %vreg19:sub0, %vreg2,
%SGPR0_SGPR1_SGPR2_SGPR3, %SGPR4, 0, 0, 0, 0, %EXEC<imp-use>; VReg_128:%vreg19
VGPR_32:%vreg2
%SGPR30_SGPR31<def> = COPY %vreg15; SReg_64:%vreg15
%SGPR5<def> = COPY %SGPR5
S_SETPC_B64_return %SGPR30_SGPR31, %SGPR5<imp-use>
# End machine code for function vreg_does_not_dominate.
*** Bad machine code: Virtual register defs don't dominate all uses. ***
- function: vreg_does_not_dominate
- v. register: %vreg22
LLVM ERROR: Found 1 machine code errors.
--
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/20170626/f47bb3b6/attachment.html>
More information about the llvm-bugs
mailing list