yao via llvm-dev
2018-Apr-09 08:30 UTC
[llvm-dev] How to get the case value from Machine Instruction
Hi, guys
I am interesting about how to get the switch case value form the Machine
Instruction.
I know the switch will be converted to jump-table in the Machine Instruction.
And in the phase CodeGen , the case-value of SwitchInst can get esasly.
but it seems no case -value in Machine Instruction.
The MI as follows:
Frame Objects:
fi#0: size=1, align=0, at location [SP]
fi#1: size=4, align=4, at location [SP+8]
fi#2: size=4, align=4, at location [SP+4]
fi#3: size=4, align=4, at location [SP]
Jump Tables:
%jump-table.0: %bb.2 %bb.3 %bb.4 %bb.5
%bb.0: derived from LLVM BB %0
%r0 = MOVi 0, 14, %noreg, %noreg
STRi12 %r0, %stack.1, 14, %noreg
%r0 = MOVi 4, 14, %noreg, %noreg
STRi12 %r0, %stack.2, 14, %noreg
%r0 = LDRi12 %stack.2, 14, %noreg
%r0 = SUBri %r0, 1, 14, %noreg, %noreg
CMPri %r0, 3, 14, %noreg, implicit-def %cpsr
STRi12 %r0, %stack.3, 14, %noreg
Bcc %bb.6, 8, %cpsr
Successors according to CFG: %bb.6 %bb.1
%bb.1: derived from LLVM BB %1
Predecessors according to CFG: %bb.0
%1:gprnopc = LEApcrelJT %jump-table.0, 14, %noreg
%2:gprnopc = LDRrs killed %0:gprnopc, %1, 0, 14, %noreg; mem:LD4[JumpTable]
BR_JTr killed %2, %jump-table.0
Successors according to CFG: %bb.2(?%) %bb.3(?%) %bb.4(?%) %bb.5(?%)
%bb.2: derived from LLVM BB %2
Predecessors according to CFG: %bb.1
%r0 = LDRi12 %stack.2, 14, %noreg
%r0 = ADDri %r0, 11, 14, %noreg, %noreg
STRi12 %r0, %stack.2, 14, %noreg
B %bb.6
Successors according to CFG: %bb.6
%bb.3: derived from LLVM BB %3
Predecessors according to CFG: %bb.1
%r0 = LDRi12 %stack.2, 14, %noreg
%r0 = ADDri %r0, 12, 14, %noreg, %noreg
STRi12 %r0, %stack.2, 14, %noreg
B %bb.6
Successors according to CFG: %bb.6
%bb.4: derived from LLVM BB %4
Predecessors according to CFG: %bb.1
%r0 = LDRi12 %stack.2, 14, %noreg
%r0 = ADDri %r0, 13, 14, %noreg, %noreg
STRi12 %r0, %stack.2, 14, %noreg
B %bb.6
Successors according to CFG: %bb.6
%bb.5: derived from LLVM BB %5
Predecessors according to CFG: %bb.1
%r0 = LDRi12 %stack.2, 14, %noreg
%r0 = ADDri %r0, 14, 14, %noreg, %noreg
STRi12 %r0, %stack.2, 14, %noreg
Successors according to CFG: %bb.6
%bb.6: derived from LLVM BB %6
Predecessors according to CFG: %bb.0 %bb.2 %bb.3 %bb.4 %bb.5
%r0 = LDRi12 %stack.2, 14, %noreg
BX_RET 14, %noreg
I hope to get the Case value from the MI,
Thanks,
yaoxiao
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180409/882c5b12/attachment-0001.html>
Martin J. O'Riordan via llvm-dev
2018-Apr-09 16:10 UTC
[llvm-dev] How to get the case value from Machine Instruction
Some glitch in the emailer? I have received this message 3 times in a row!?
I think that by the time it gets as far as MI-level there is no reversible
method of determining the 'case' label at all. The reason I say this,
is
that I have often seen optimisations that coalesce groups of values into
interesting logical tests and jump-tables are completely avoided. For
example, a simple range like:
switch(n) {
case 8:
case 9:
case 10:
case 11:
case 12:
case 13:
case 14:
case 15:
...action...
}
is quite likely to become (equivalent to):
if ((n & 0xFFFFFF8) == 0x00000008) ...action...
so the 'case' logic of the original expression is lost; and there are
other
logical reductions a lot less obvious than this. I expect that at IR-level,
especially pre-SCEV that the 'case' specific information might still be
available but at MI-level it has already been lost. The debug information
might still retain some back-association, but I am not particularly familiar
with Dwarf.
MartinO
From: llvm-dev [mailto:llvm-dev-bounces at lists.llvm.org] On Behalf Of yao via
llvm-dev
Sent: 09 April 2018 09:30
To: llvmdev <llvmdev at cs.uiuc.edu>
Subject: [llvm-dev] How to get the case value from Machine Instruction
Hi, guys
I am interesting about how to get the switch case value form the Machine
Instruction.
I know the switch will be converted to jump-table in the Machine
Instruction.
And in the phase CodeGen , the case-value of SwitchInst can get esasly.
but it seems no case -value in Machine Instruction.
The MI as follows:
Frame Objects:
fi#0: size=1, align=0, at location [SP]
fi#1: size=4, align=4, at location [SP+8]
fi#2: size=4, align=4, at location [SP+4]
fi#3: size=4, align=4, at location [SP]
Jump Tables:
%jump-table.0: %bb.2 %bb.3 %bb.4 %bb.5
%bb.0: derived from LLVM BB %0
%r0 = MOVi 0, 14, %noreg, %noreg
STRi12 %r0, %stack.1, 14, %noreg
%r0 = MOVi 4, 14, %noreg, %noreg
STRi12 %r0, %stack.2, 14, %noreg
%r0 = LDRi12 %stack.2, 14, %noreg
%r0 = SUBri %r0, 1, 14, %noreg, %noreg
CMPri %r0, 3, 14, %noreg, implicit-def %cpsr
STRi12 %r0, %stack.3, 14, %noreg
Bcc %bb.6, 8, %cpsr
Successors according to CFG: %bb.6 %bb.1
%bb.1: derived from LLVM BB %1
Predecessors according to CFG: %bb.0
%1:gprnopc = LEApcrelJT %jump-table.0, 14, %noreg
%2:gprnopc = LDRrs killed %0:gprnopc, %1, 0, 14, %noreg; mem:LD4[JumpTable]
BR_JTr killed %2, %jump-table.0
Successors according to CFG: %bb.2(?%) %bb.3(?%) %bb.4(?%) %bb.5(?%)
%bb.2: derived from LLVM BB %2
Predecessors according to CFG: %bb.1
%r0 = LDRi12 %stack.2, 14, %noreg
%r0 = ADDri %r0, 11, 14, %noreg, %noreg
STRi12 %r0, %stack.2, 14, %noreg
B %bb.6
Successors according to CFG: %bb.6
%bb.3: derived from LLVM BB %3
Predecessors according to CFG: %bb.1
%r0 = LDRi12 %stack.2, 14, %noreg
%r0 = ADDri %r0, 12, 14, %noreg, %noreg
STRi12 %r0, %stack.2, 14, %noreg
B %bb.6
Successors according to CFG: %bb.6
%bb.4: derived from LLVM BB %4
Predecessors according to CFG: %bb.1
%r0 = LDRi12 %stack.2, 14, %noreg
%r0 = ADDri %r0, 13, 14, %noreg, %noreg
STRi12 %r0, %stack.2, 14, %noreg
B %bb.6
Successors according to CFG: %bb.6
%bb.5: derived from LLVM BB %5
Predecessors according to CFG: %bb.1
%r0 = LDRi12 %stack.2, 14, %noreg
%r0 = ADDri %r0, 14, 14, %noreg, %noreg
STRi12 %r0, %stack.2, 14, %noreg
Successors according to CFG: %bb.6
%bb.6: derived from LLVM BB %6
Predecessors according to CFG: %bb.0 %bb.2 %bb.3 %bb.4 %bb.5
%r0 = LDRi12 %stack.2, 14, %noreg
BX_RET 14, %noreg
I hope to get the Case value from the MI,
Thanks,
yaoxiao
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180409/c2c56a4f/attachment-0001.html>
yao via llvm-dev
2018-Apr-10 02:35 UTC
[llvm-dev] How to get the case value from Machine Instruction
Thanks for your help.
Is there possible I can get the realily case value form the MI?
For the case in https://bugs.llvm.org/show_bug.cgi?id=34902.
as follows.
#############################
* GCC v7.1 generated assembly
#############################
** Options: -Os -marm -march=armv7-a
foo:
@ args = 0, pretend = 0, frame = 0
@ frame_needed = 0, uses_anonymous_args = 0
sub r0, r0, #15
push {r4, lr}
cmp r0, #5
ldrls pc, [pc, r0, asl #2]
b .L1
.L4:
.word .L3
.word .L5
.word .L6
.word .L7
.word .L8
.word .L9
.L3:
mov r0, #5
bl func
.L5:
mov r0, #59
bl func
.L6:
movw r0, #515
bl func
.L7:
mov r0, #65
bl func
.L8:
mov r0, #8
bl func
.L9:
mov r0, #15
pop {r4, lr}
b func
.L1:
pop {r4, pc}
the assembly shows the value was replaced by ldrls pc, [pc, r0, asl #2], I had
tried to convert the MI to DAG , but seems the jumptable seems only contain the
case address,
Could I calculate the value by the offset or other key method ?
Thanks,
yaoxiao
At 2018-04-10 00:10:31, "Martin J. O'Riordan" <MartinO at
theheart.ie> wrote:
Some glitch in the emailer? I have received this message 3 times in a row!?
I think that by the time it gets as far as MI-level there is no reversible
method of determining the ‘case’ label at all. The reason I say this, is that I
have often seen optimisations that coalesce groups of values into interesting
logical tests and jump-tables are completely avoided. For example, a simple
range like:
switch(n) {
case 8:
case 9:
case 10:
case 11:
case 12:
case 13:
case 14:
case 15:
...action...
}
is quite likely to become (equivalent to):
if ((n & 0xFFFFFF8) == 0x00000008) ...action...
so the ‘case’ logic of the original expression is lost; and there are other
logical reductions a lot less obvious than this. I expect that at IR-level,
especially pre-SCEV that the ‘case’ specific information might still be
available but at MI-level it has already been lost. The debug information might
still retain some back-association, but I am not particularly familiar with
Dwarf.
MartinO
From: llvm-dev [mailto:llvm-dev-bounces at lists.llvm.org] On Behalf Of yao via
llvm-dev
Sent: 09 April 2018 09:30
To: llvmdev <llvmdev at cs.uiuc.edu>
Subject: [llvm-dev] How to get the case value from Machine Instruction
Hi, guys
I am interesting about how to get the switch case value form the Machine
Instruction.
I know the switch will be converted to jump-table in the Machine Instruction.
And in the phase CodeGen , the case-value of SwitchInst can get esasly.
but it seems no case -value in Machine Instruction.
The MI as follows:
Frame Objects:
fi#0: size=1, align=0, at location [SP]
fi#1: size=4, align=4, at location [SP+8]
fi#2: size=4, align=4, at location [SP+4]
fi#3: size=4, align=4, at location [SP]
Jump Tables:
%jump-table.0: %bb.2 %bb.3 %bb.4 %bb.5
%bb.0: derived from LLVM BB %0
%r0 = MOVi 0, 14, %noreg, %noreg
STRi12 %r0, %stack.1, 14, %noreg
%r0 = MOVi 4, 14, %noreg, %noreg
STRi12 %r0, %stack.2, 14, %noreg
%r0 = LDRi12 %stack.2, 14, %noreg
%r0 = SUBri %r0, 1, 14, %noreg, %noreg
CMPri %r0, 3, 14, %noreg, implicit-def %cpsr
STRi12 %r0, %stack.3, 14, %noreg
Bcc %bb.6, 8, %cpsr
Successors according to CFG: %bb.6 %bb.1
%bb.1: derived from LLVM BB %1
Predecessors according to CFG: %bb.0
%1:gprnopc = LEApcrelJT %jump-table.0, 14, %noreg
%2:gprnopc = LDRrs killed %0:gprnopc, %1, 0, 14, %noreg; mem:LD4[JumpTable]
BR_JTr killed %2, %jump-table.0
Successors according to CFG: %bb.2(?%) %bb.3(?%) %bb.4(?%) %bb.5(?%)
%bb.2: derived from LLVM BB %2
Predecessors according to CFG: %bb.1
%r0 = LDRi12 %stack.2, 14, %noreg
%r0 = ADDri %r0, 11, 14, %noreg, %noreg
STRi12 %r0, %stack.2, 14, %noreg
B %bb.6
Successors according to CFG: %bb.6
%bb.3: derived from LLVM BB %3
Predecessors according to CFG: %bb.1
%r0 = LDRi12 %stack.2, 14, %noreg
%r0 = ADDri %r0, 12, 14, %noreg, %noreg
STRi12 %r0, %stack.2, 14, %noreg
B %bb.6
Successors according to CFG: %bb.6
%bb.4: derived from LLVM BB %4
Predecessors according to CFG: %bb.1
%r0 = LDRi12 %stack.2, 14, %noreg
%r0 = ADDri %r0, 13, 14, %noreg, %noreg
STRi12 %r0, %stack.2, 14, %noreg
B %bb.6
Successors according to CFG: %bb.6
%bb.5: derived from LLVM BB %5
Predecessors according to CFG: %bb.1
%r0 = LDRi12 %stack.2, 14, %noreg
%r0 = ADDri %r0, 14, 14, %noreg, %noreg
STRi12 %r0, %stack.2, 14, %noreg
Successors according to CFG: %bb.6
%bb.6: derived from LLVM BB %6
Predecessors according to CFG: %bb.0 %bb.2 %bb.3 %bb.4 %bb.5
%r0 = LDRi12 %stack.2, 14, %noreg
BX_RET 14, %noreg
I hope to get the Case value from the MI,
Thanks,
yaoxiao
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20180410/b6d4ba6a/attachment.html>