Skip to content

[PowerPC] Implement vector unpack instructions #151004

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

[PowerPC] Implement vector unpack instructions #151004

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
9b78cca
add unpack high/low nibble
lei137 Jul 28, 2025
ce4c27c
add remaining unpacked instruction implementation
lei137 Jul 28, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
78 changes: 78 additions & 0 deletions llvm/lib/Target/PowerPC/PPCInstrFuture.td
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,6 +45,67 @@ multiclass XOForm_RTAB5_L1r<bits<6> opcode, bits<9> xo, dag OOL, dag IOL,
}
}

class VXForm_VRTB5<bits<11> xo, bits<5> R, dag OOL, dag IOL, string asmstr,
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

no opcode(4) in the definition in the class. ?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

that's specified in the class it's inheriting from. See line 49.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

thanks, I saw it.

list<dag> pattern> : I<4, OOL, IOL, asmstr, NoItinerary> {
bits<5> VRT;
bits<5> VRB;

let Pattern = pattern;

let Inst{6 -10} = VRT;
let Inst{11 -15} = R;
let Inst{16 -20} = VRB;
let Inst{21 -31} = xo;
}

class VXForm_VRTB5_UIM2<bits<11> xo, bits<3> R, dag OOL, dag IOL, string asmstr,
list<dag> pattern>
: I<4, OOL, IOL, asmstr, NoItinerary> {
bits<5> VRT;
bits<5> VRB;
bits<2> UIM;

let Pattern = pattern;

let Inst{6 -10} = VRT;
let Inst{11 -13} = R;
let Inst{14 -15} = UIM;
let Inst{16 -20} = VRB;
let Inst{21 -31} = xo;
}

class VXForm_VRTB5_UIM1<bits<11> xo, bits<4> R, dag OOL, dag IOL, string asmstr,
list<dag> pattern>
: I<4, OOL, IOL, asmstr, NoItinerary> {
bits<5> VRT;
bits<5> VRB;
bits<1> UIM;

let Pattern = pattern;

let Inst{6 -10} = VRT;
let Inst{11 -14} = R;
let Inst{15} = UIM;
let Inst{16 -20} = VRB;
let Inst{21 -31} = xo;
}

class VXForm_VRTB5_UIM3<bits<11> xo, bits<2> R, dag OOL, dag IOL, string asmstr,
list<dag> pattern>
: I<4, OOL, IOL, asmstr, NoItinerary> {
bits<5> VRT;
bits<5> VRB;
bits<3> UIM;

let Pattern = pattern;

let Inst{6 -10} = VRT;
let Inst{11 -12} = R;
let Inst{13 -15} = UIM;
let Inst{16 -20} = VRB;
let Inst{21 -31} = xo;
}

Copy link
Contributor

@diggerlin diggerlin Jul 29, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

not sure whether to a base class for above 4 class is a good idea or not ?

since there has same

bits<5> VRT;
 bits<5> VRB;
 
  let Inst{6 -10} = VRT;
 let Inst{16 -20} = VRB;
 let Inst{21 -31} = xo;

and

I<4, OOL, IOL, asmstr, NoItinerary>

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

My preference is to keep them separate as the middle bits {11-16} are different combinations of fields for all of them.

let Predicates = [IsISAFuture] in {
defm SUBFUS : XOForm_RTAB5_L1r<31, 72, (outs g8rc:$RT),
(ins g8rc:$RA, g8rc:$RB, u1imm:$L),
Expand Down Expand Up @@ -81,4 +142,21 @@ let Predicates = [HasVSX, IsISAFuture] in {
(ins vsrprc:$XTp, memr:$RA, g8rc:$RB),
"stxvprll $XTp, $RA, $RB", IIC_LdStLFD, []>;
}

def VUPKHSNTOB : VXForm_VRTB5<387, 0, (outs vrrc:$VRT), (ins vrrc:$VRB),
"vupkhsntob $VRT, $VRB", []>;
def VUPKLSNTOB : VXForm_VRTB5<387, 1, (outs vrrc:$VRT), (ins vrrc:$VRB),
"vupklsntob $VRT, $VRB", []>;
def VUPKINT4TOBF16
: VXForm_VRTB5_UIM2<387, 2, (outs vrrc:$VRT), (ins vrrc:$VRB, u2imm:$UIM),
"vupkint4tobf16 $VRT, $VRB, $UIM", []>;
def VUPKINT8TOBF16
: VXForm_VRTB5_UIM1<387, 1, (outs vrrc:$VRT), (ins vrrc:$VRB, u1imm:$UIM),
"vupkint8tobf16 $VRT, $VRB, $UIM", []>;
def VUPKINT8TOFP32
: VXForm_VRTB5_UIM2<387, 3, (outs vrrc:$VRT), (ins vrrc:$VRB, u2imm:$UIM),
"vupkint8tofp32 $VRT, $VRB, $UIM", []>;
def VUPKINT4TOFP32
: VXForm_VRTB5_UIM3<387, 2, (outs vrrc:$VRT), (ins vrrc:$VRB, u3imm:$UIM),
"vupkint4tofp32 $VRT, $VRB, $UIM", []>;
}
18 changes: 18 additions & 0 deletions llvm/test/MC/Disassembler/PowerPC/ppc-encoding-ISAFuture.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -195,3 +195,21 @@

#CHECK: dmxxsha224256pad 0, 1
0xf0,0x18,0x0e,0x94

#CHECK: vupkhsntob 2, 4
0x10,0x40,0x21,0x83

#CHECK: vupklsntob 2, 4
0x10,0x41,0x21,0x83

#CHECK: vupkint4tobf16 2, 4, 3
0x10,0x4b,0x21,0x83

#CHECK: vupkint8tobf16 1, 3, 1
0x10,0x23,0x19,0x83

#CHECK: vupkint4tofp32 3, 5, 2
0x10,0x72,0x29,0x83

#CHECK: vupkint8tofp32 3, 5, 2
0x10,0x6e,0x29,0x83
18 changes: 18 additions & 0 deletions llvm/test/MC/Disassembler/PowerPC/ppc64le-encoding-ISAFuture.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -189,3 +189,21 @@

#CHECK: dmxxsha224256pad 0, 1
0x94,0x0e,0x18,0xf0

#CHECK: vupkhsntob 2, 4
0x83,0x21,0x40,0x10

#CHECK: vupklsntob 2, 4
0x83,0x21,0x41,0x10

#CHECK: vupkint4tobf16 2, 4, 3
0x83,0x21,0x4b,0x10

#CHECK: vupkint8tobf16 1, 3, 1
0x83,0x19,0x23,0x10

#CHECK: vupkint4tofp32 3, 5, 2
0x83,0x29,0x72,0x10

#CHECK: vupkint8tofp32 3, 5, 2
0x83,0x29,0x6e,0x10
24 changes: 24 additions & 0 deletions llvm/test/MC/PowerPC/ppc-encoding-ISAFuture.s
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -282,3 +282,27 @@
dmxxsha224256pad 0, 1
#CHECK-BE: dmxxsha224256pad 0, 1 # encoding: [0xf0,0x18,0x0e,0x94]
#CHECK-LE: dmxxsha224256pad 0, 1 # encoding: [0x94,0x0e,0x18,0xf0]

vupkhsntob 2, 4
#CHECK-BE: vupkhsntob 2, 4 # encoding: [0x10,0x40,0x21,0x83]
#CHECK-LE: vupkhsntob 2, 4 # encoding: [0x83,0x21,0x40,0x10]

vupklsntob 2, 4
#CHECK-BE: vupklsntob 2, 4 # encoding: [0x10,0x41,0x21,0x83]
#CHECK-LE: vupklsntob 2, 4 # encoding: [0x83,0x21,0x41,0x10]

vupkint4tobf16 2, 4, 3
#CHECK-BE: vupkint4tobf16 2, 4, 3 # encoding: [0x10,0x4b,0x21,0x83]
#CHECK-LE: vupkint4tobf16 2, 4, 3 # encoding: [0x83,0x21,0x4b,0x10]

vupkint8tobf16 1, 3, 1
#CHECK-BE: vupkint8tobf16 1, 3, 1 # encoding: [0x10,0x23,0x19,0x83]
#CHECK-LE: vupkint8tobf16 1, 3, 1 # encoding: [0x83,0x19,0x23,0x10]

vupkint4tofp32 3, 5, 2
#CHECK-BE: vupkint4tofp32 3, 5, 2 # encoding: [0x10,0x72,0x29,0x83]
#CHECK-LE: vupkint4tofp32 3, 5, 2 # encoding: [0x83,0x29,0x72,0x10]

vupkint8tofp32 3, 5, 2
#CHECK-BE: vupkint8tofp32 3, 5, 2 # encoding: [0x10,0x6e,0x29,0x83]
#CHECK-LE: vupkint8tofp32 3, 5, 2 # encoding: [0x83,0x29,0x6e,0x10]