Skip to content

[AMDGPU][True16][MC] Generate op_sel operands for VOPC instructions #125561

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

Merged
merged 2 commits into from
Feb 5, 2025
+270 −258
Merged

[AMDGPU][True16][MC] Generate op_sel operands for VOPC instructions #125561

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
7b833a4
udpate vopc asm converter
broxigarchen Feb 3, 2025
6c250bc
Update llvm/lib/Target/AMDGPU/VOPCInstructions.td
broxigarchen Feb 4, 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
20 changes: 16 additions & 4 deletions llvm/lib/Target/AMDGPU/VOPCInstructions.td
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -389,10 +389,10 @@ multiclass VOPC_Pseudos <string opName,
let SchedRW = P.Schedule;
let isCompare = 1;
let isCommutable = 1;
let AsmMatchConverter =
!if (P.HasOpSel, "cvtVOP3OpSel",
!if (!or(P.HasModifiers, P.HasOMod, P.HasIntClamp), "cvtVOP3",
""));
let AsmMatchConverter = !cond(
P.HasOpSel : "cvtVOP3OpSel",
!or(P.HasModifiers, P.HasOMod, P.HasIntClamp) : "cvtVOP3",
1 : "");
}

if P.HasExtSDWA then
Expand Down Expand Up @@ -454,6 +454,10 @@ multiclass VOPCX_Pseudos <string opName,
let isCommutable = 1;
let SubtargetPredicate = HasNoSdstCMPX;
let IsVCMPX = 1;
let AsmMatchConverter = !cond(
P_NoSDst.HasOpSel : "cvtVOP3OpSel",
!or(P_NoSDst.HasModifiers, P_NoSDst.HasOMod, P_NoSDst.HasIntClamp) : "cvtVOP3",
1 : "");
}

if P_NoSDst.HasExtSDWA then
Expand Down Expand Up @@ -1079,6 +1083,10 @@ multiclass VOPC_Class_Pseudos <string opName, VOPC_Profile p, bit DefExec,
VCMPXNoSDstTable<1, opName#"_e64"> {
let Defs = !if(DefExec, [EXEC], []);
let SchedRW = p.Schedule;
let AsmMatchConverter = !cond(
p.HasOpSel : "cvtVOP3OpSel",
!or(p.HasModifiers, p.HasOMod, p.HasIntClamp) : "cvtVOP3",
1 : "");
}

if p.HasExtSDWA then
Expand Down Expand Up @@ -1127,6 +1135,10 @@ multiclass VOPCX_Class_Pseudos <string opName,
let Defs = [EXEC];
let SchedRW = P_NoSDst.Schedule;
let SubtargetPredicate = HasNoSdstCMPX;
let AsmMatchConverter = !cond(
P_NoSDst.HasOpSel : "cvtVOP3OpSel",
!or(P_NoSDst.HasModifiers, P_NoSDst.HasOMod, P_NoSDst.HasIntClamp) : "cvtVOP3",
1 : "");
}

if P_NoSDst.HasExtSDWA then
Expand Down
8 changes: 4 additions & 4 deletions llvm/test/MC/AMDGPU/gfx11_asm_vop3_from_vopc.s
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -128,19 +128,19 @@ v_cmp_class_f16_e64 vcc_lo, 0.5, m0
// W64-ERR: :[[@LINE-2]]:21: error: invalid operand for instruction

v_cmp_class_f16_e64 s5, v255.h, v2.l
// W32: v_cmp_class_f16_e64 s5, v255.h, v2.l ; encoding: [0x05,0x08,0x7d,0xd4,0xff,0x05,0x02,0x00]
// W32: v_cmp_class_f16_e64 s5, v255.h, v2.l op_sel:[1,0,0] ; encoding: [0x05,0x08,0x7d,0xd4,0xff,0x05,0x02,0x00]
// W64-ERR: :[[@LINE-2]]:21: error: invalid operand for instruction

v_cmp_class_f16_e64 s5, s105, v255.h
// W32: v_cmp_class_f16_e64 s5, s105, v255.h ; encoding: [0x05,0x10,0x7d,0xd4,0x69,0xfe,0x03,0x00]
// W32: v_cmp_class_f16_e64 s5, s105, v255.h op_sel:[0,1,0] ; encoding: [0x05,0x10,0x7d,0xd4,0x69,0xfe,0x03,0x00]
// W64-ERR: :[[@LINE-2]]:21: error: invalid operand for instruction

v_cmp_class_f16_e64 s[10:11], v255.h, v2.l
// W64: v_cmp_class_f16_e64 s[10:11], v255.h, v2.l ; encoding: [0x0a,0x08,0x7d,0xd4,0xff,0x05,0x02,0x00]
// W64: v_cmp_class_f16_e64 s[10:11], v255.h, v2.l op_sel:[1,0,0] ; encoding: [0x0a,0x08,0x7d,0xd4,0xff,0x05,0x02,0x00]
// W32-ERR: :[[@LINE-2]]:21: error: invalid operand for instruction

v_cmp_class_f16_e64 s[10:11], s105, v255.h
// W64: v_cmp_class_f16_e64 s[10:11], s105, v255.h ; encoding: [0x0a,0x10,0x7d,0xd4,0x69,0xfe,0x03,0x00]
// W64: v_cmp_class_f16_e64 s[10:11], s105, v255.h op_sel:[0,1,0] ; encoding: [0x0a,0x10,0x7d,0xd4,0x69,0xfe,0x03,0x00]
// W32-ERR: :[[@LINE-2]]:21: error: invalid operand for instruction

v_cmp_class_f32_e64 s5, v1, v2
Expand Down
118 changes: 59 additions & 59 deletions llvm/test/MC/AMDGPU/gfx11_asm_vop3_from_vopcx.s
View file
Open in desktop

Large diffs are not rendered by default.

18 changes: 9 additions & 9 deletions llvm/test/MC/AMDGPU/gfx11_asm_vopc_t16_promote.s
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@
// RUN: llvm-mc -triple=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize64,+real-true16 -show-encoding %s 2>&1 | FileCheck --check-prefix=GFX11 %s

v_cmp_class_f16 vcc, v1.h, v255.h
// GFX11: v_cmp_class_f16_e64 vcc, v1.h, v255.h ; encoding: [0x6a,0x18,0x7d,0xd4,0x01,0xff,0x03,0x00]
// GFX11: v_cmp_class_f16_e64 vcc, v1.h, v255.h op_sel:[1,1,0] ; encoding: [0x6a,0x18,0x7d,0xd4,0x01,0xff,0x03,0x00]

v_cmp_class_f16 vcc, v1.h, v255.h dpp8:[7,6,5,4,3,2,1,0]
// GFX11: v_cmp_class_f16_e64_dpp vcc, v1.h, v255.h op_sel:[1,1] dpp8:[7,6,5,4,3,2,1,0] ; encoding: [0x6a,0x18,0x7d,0xd4,0xe9,0xfe,0x03,0x00,0x01,0x77,0x39,0x05]
Expand All @@ -20,10 +20,10 @@ v_cmp_class_f16 vcc, v1.l, v255.l quad_perm:[3,2,1,0]
// GFX11: v_cmp_class_f16_e64_dpp vcc, v1.l, v255.l quad_perm:[3,2,1,0] row_mask:0xf bank_mask:0xf ; encoding: [0x6a,0x00,0x7d,0xd4,0xfa,0xfe,0x03,0x00,0x01,0x1b,0x00,0xff]

v_cmp_class_f16 vcc, v127.h, v255.h
// GFX11: v_cmp_class_f16_e64 vcc, v127.h, v255.h ; encoding: [0x6a,0x18,0x7d,0xd4,0x7f,0xff,0x03,0x00]
// GFX11: v_cmp_class_f16_e64 vcc, v127.h, v255.h op_sel:[1,1,0] ; encoding: [0x6a,0x18,0x7d,0xd4,0x7f,0xff,0x03,0x00]

v_cmp_class_f16 vcc, v127.h, v255.h
// GFX11: v_cmp_class_f16_e64 vcc, v127.h, v255.h ; encoding: [0x6a,0x18,0x7d,0xd4,0x7f,0xff,0x03,0x00]
// GFX11: v_cmp_class_f16_e64 vcc, v127.h, v255.h op_sel:[1,1,0] ; encoding: [0x6a,0x18,0x7d,0xd4,0x7f,0xff,0x03,0x00]

v_cmp_class_f16 vcc, v127.h, v255.h dpp8:[7,6,5,4,3,2,1,0]
// GFX11: v_cmp_class_f16_e64_dpp vcc, v127.h, v255.h op_sel:[1,1] dpp8:[7,6,5,4,3,2,1,0] ; encoding: [0x6a,0x18,0x7d,0xd4,0xe9,0xfe,0x03,0x00,0x7f,0x77,0x39,0x05]
Expand Down Expand Up @@ -56,10 +56,10 @@ v_cmp_class_f16 vcc, v127.l, v255.l quad_perm:[3,2,1,0]
// GFX11: v_cmp_class_f16_e64_dpp vcc, v127.l, v255.l quad_perm:[3,2,1,0] row_mask:0xf bank_mask:0xf ; encoding: [0x6a,0x00,0x7d,0xd4,0xfa,0xfe,0x03,0x00,0x7f,0x1b,0x00,0xff]

v_cmp_class_f16 vcc, v128.h, v2.h
// GFX11: v_cmp_class_f16_e64 vcc, v128.h, v2.h ; encoding: [0x6a,0x18,0x7d,0xd4,0x80,0x05,0x02,0x00]
// GFX11: v_cmp_class_f16_e64 vcc, v128.h, v2.h op_sel:[1,1,0] ; encoding: [0x6a,0x18,0x7d,0xd4,0x80,0x05,0x02,0x00]

v_cmp_class_f16 vcc, v128.h, v2.h
// GFX11: v_cmp_class_f16_e64 vcc, v128.h, v2.h ; encoding: [0x6a,0x18,0x7d,0xd4,0x80,0x05,0x02,0x00]
// GFX11: v_cmp_class_f16_e64 vcc, v128.h, v2.h op_sel:[1,1,0] ; encoding: [0x6a,0x18,0x7d,0xd4,0x80,0x05,0x02,0x00]

v_cmp_class_f16 vcc, v128.h, v2.h dpp8:[7,6,5,4,3,2,1,0]
// GFX11: v_cmp_class_f16_e64_dpp vcc, v128.h, v2.h op_sel:[1,1] dpp8:[7,6,5,4,3,2,1,0] ; encoding: [0x6a,0x18,0x7d,0xd4,0xe9,0x04,0x02,0x00,0x80,0x77,0x39,0x05]
Expand Down Expand Up @@ -92,10 +92,10 @@ v_cmp_class_f16 vcc, v128.l, v2.l quad_perm:[3,2,1,0]
// GFX11: v_cmp_class_f16_e64_dpp vcc, v128.l, v2.l quad_perm:[3,2,1,0] row_mask:0xf bank_mask:0xf ; encoding: [0x6a,0x00,0x7d,0xd4,0xfa,0x04,0x02,0x00,0x80,0x1b,0x00,0xff]

v_cmp_class_f16 vcc, vcc_hi, v255.h
// GFX11: v_cmp_class_f16_e64 vcc, vcc_hi, v255.h ; encoding: [0x6a,0x10,0x7d,0xd4,0x6b,0xfe,0x03,0x00]
// GFX11: v_cmp_class_f16_e64 vcc, vcc_hi, v255.h op_sel:[0,1,0] ; encoding: [0x6a,0x10,0x7d,0xd4,0x6b,0xfe,0x03,0x00]

v_cmp_class_f16 vcc, vcc_hi, v255.h
// GFX11: v_cmp_class_f16_e64 vcc, vcc_hi, v255.h ; encoding: [0x6a,0x10,0x7d,0xd4,0x6b,0xfe,0x03,0x00]
// GFX11: v_cmp_class_f16_e64 vcc, vcc_hi, v255.h op_sel:[0,1,0] ; encoding: [0x6a,0x10,0x7d,0xd4,0x6b,0xfe,0x03,0x00]

v_cmp_class_f16 vcc, vcc_hi, v255.l
// GFX11: v_cmp_class_f16_e64 vcc, vcc_hi, v255.l ; encoding: [0x6a,0x00,0x7d,0xd4,0x6b,0xfe,0x03,0x00]
Expand All @@ -104,10 +104,10 @@ v_cmp_class_f16 vcc, vcc_hi, v255.l
// GFX11: v_cmp_class_f16_e64 vcc, vcc_hi, v255.l ; encoding: [0x6a,0x00,0x7d,0xd4,0x6b,0xfe,0x03,0x00]

v_cmp_class_f16 vcc, vcc_lo, v255.h
// GFX11: v_cmp_class_f16_e64 vcc, vcc_lo, v255.h ; encoding: [0x6a,0x10,0x7d,0xd4,0x6a,0xfe,0x03,0x00]
// GFX11: v_cmp_class_f16_e64 vcc, vcc_lo, v255.h op_sel:[0,1,0] ; encoding: [0x6a,0x10,0x7d,0xd4,0x6a,0xfe,0x03,0x00]

v_cmp_class_f16 vcc, vcc_lo, v255.h
// GFX11: v_cmp_class_f16_e64 vcc, vcc_lo, v255.h ; encoding: [0x6a,0x10,0x7d,0xd4,0x6a,0xfe,0x03,0x00]
// GFX11: v_cmp_class_f16_e64 vcc, vcc_lo, v255.h op_sel:[0,1,0] ; encoding: [0x6a,0x10,0x7d,0xd4,0x6a,0xfe,0x03,0x00]

v_cmp_class_f16 vcc, vcc_lo, v255.l
// GFX11: v_cmp_class_f16_e64 vcc, vcc_lo, v255.l ; encoding: [0x6a,0x00,0x7d,0xd4,0x6a,0xfe,0x03,0x00]
Expand Down
Loading