Skip to content

Commit cbe8f3a

Browse files
authored
[AMDGPU][True16][MC] fix fmac_f16_t16 vop3 format (llvm#135464)
add fmac_f16_t16_e64 to isfmac check to fix the vop3 format of fmac_f16_t16 instruction
1 parent 425ccd5 commit cbe8f3a

9 files changed

+228
-109
lines changed

llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp

+2
Original file line numberDiff line numberDiff line change
@@ -624,7 +624,9 @@ bool isMAC(unsigned Opc) {
624624
Opc == AMDGPU::V_FMAC_LEGACY_F32_e64_gfx10 ||
625625
Opc == AMDGPU::V_FMAC_DX9_ZERO_F32_e64_gfx11 ||
626626
Opc == AMDGPU::V_FMAC_F16_e64_gfx10 ||
627+
Opc == AMDGPU::V_FMAC_F16_t16_e64_gfx11 ||
627628
Opc == AMDGPU::V_FMAC_F16_fake16_e64_gfx11 ||
629+
Opc == AMDGPU::V_FMAC_F16_t16_e64_gfx12 ||
628630
Opc == AMDGPU::V_FMAC_F16_fake16_e64_gfx12 ||
629631
Opc == AMDGPU::V_DOT2C_F32_F16_e64_vi ||
630632
Opc == AMDGPU::V_DOT2C_F32_BF16_e64_vi ||

llvm/test/MC/AMDGPU/gfx11_asm_vop2_t16_promote.s

+16-7
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,4 @@
1-
// NOTE: Assertions have been autogenerated by utils/update_mc_test_checks.py UTC_ARGS: --sort --version 5
1+
// NOTE: Assertions have been autogenerated by utils/update_mc_test_checks.py UTC_ARGS: --unique --sort --version 5
22
// RUN: llvm-mc -triple=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,+real-true16 -show-encoding %s 2>&1 | FileCheck --check-prefix=GFX11 --implicit-check-not=_e32 %s
33
// RUN: llvm-mc -triple=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize64,+real-true16 -show-encoding %s 2>&1 | FileCheck --check-prefix=GFX11 --implicit-check-not=_e32 %s
44

@@ -56,14 +56,23 @@ v_add_f16 v5.l, v255.l, v2.l dpp8:[7,6,5,4,3,2,1,0]
5656
v_add_f16 v5.l, v255.l, v2.l quad_perm:[3,2,1,0]
5757
// GFX11: v_add_f16_e64_dpp v5.l, v255.l, v2.l quad_perm:[3,2,1,0] row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x32,0xd5,0xfa,0x04,0x02,0x00,0xff,0x1b,0x00,0xff]
5858

59-
v_fmac_f16 v255, v1, v2
60-
// GFX11: v_fmac_f16_e64 v255, v1, v2 ; encoding: [0xff,0x00,0x36,0xd5,0x01,0x05,0x02,0x00]
59+
v_fmac_f16 v255.h, v1.h, v2.h
60+
// GFX11: v_fmac_f16_e64 v255.h, v1.h, v2.h op_sel:[1,1,1,1] ; encoding: [0xff,0x58,0x36,0xd5,0x01,0x05,0x02,0x00]
6161

62-
v_fmac_f16 v5, v1, v255
63-
// GFX11: v_fmac_f16_e64 v5, v1, v255 ; encoding: [0x05,0x00,0x36,0xd5,0x01,0xff,0x03,0x00]
62+
v_fmac_f16 v255.l, v1.l, v2.l
63+
// GFX11: v_fmac_f16_e64 v255.l, v1.l, v2.l ; encoding: [0xff,0x00,0x36,0xd5,0x01,0x05,0x02,0x00]
6464

65-
v_fmac_f16 v5, v255, v2
66-
// GFX11: v_fmac_f16_e64 v5, v255, v2 ; encoding: [0x05,0x00,0x36,0xd5,0xff,0x05,0x02,0x00]
65+
v_fmac_f16 v5.h, v1.h, v255.h
66+
// GFX11: v_fmac_f16_e64 v5.h, v1.h, v255.h op_sel:[1,1,1,1] ; encoding: [0x05,0x58,0x36,0xd5,0x01,0xff,0x03,0x00]
67+
68+
v_fmac_f16 v5.h, v255.h, v2.h
69+
// GFX11: v_fmac_f16_e64 v5.h, v255.h, v2.h op_sel:[1,1,1,1] ; encoding: [0x05,0x58,0x36,0xd5,0xff,0x05,0x02,0x00]
70+
71+
v_fmac_f16 v5.l, v1.l, v255.l
72+
// GFX11: v_fmac_f16_e64 v5.l, v1.l, v255.l ; encoding: [0x05,0x00,0x36,0xd5,0x01,0xff,0x03,0x00]
73+
74+
v_fmac_f16 v5.l, v255.l, v2.l
75+
// GFX11: v_fmac_f16_e64 v5.l, v255.l, v2.l ; encoding: [0x05,0x00,0x36,0xd5,0xff,0x05,0x02,0x00]
6776

6877
v_ldexp_f16 v255.h, v1.h, v2.h
6978
// GFX11: v_ldexp_f16_e64 v255.h, v1.h, v2.h op_sel:[1,1,1] ; encoding: [0xff,0x58,0x3b,0xd5,0x01,0x05,0x02,0x00]

llvm/test/MC/AMDGPU/gfx11_asm_vop3_dpp16_from_vop2.s

+40-28
Original file line numberDiff line numberDiff line change
@@ -524,47 +524,59 @@ v_cvt_pkrtz_f16_f32_e64_dpp v5, -v1, |v2| row_xmask:0 row_mask:0x1 bank_mask:0x3
524524
v_cvt_pkrtz_f16_f32_e64_dpp v255, -|v255|, -|v255| clamp row_xmask:15 row_mask:0x3 bank_mask:0x0 bound_ctrl:0 fi:1
525525
// GFX11: v_cvt_pk_rtz_f16_f32_e64_dpp v255, -|v255|, -|v255| clamp row_xmask:15 row_mask:0x3 bank_mask:0x0 fi:1 ; encoding: [0xff,0x83,0x2f,0xd5,0xfa,0xfe,0x03,0x60,0xff,0x6f,0x05,0x30]
526526

527-
v_fmac_f16_e64_dpp v5, v1, v2 quad_perm:[3,2,1,0]
528-
// GFX11: v_fmac_f16_e64_dpp v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x1b,0x00,0xff]
527+
v_fmac_f16_e64_dpp v5.l, v1.l, v2.l quad_perm:[3,2,1,0]
528+
// GFX11: v_fmac_f16_e64_dpp v5.l, v1.l, v2.l quad_perm:[3,2,1,0] row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x1b,0x00,0xff]
529529

530-
v_fmac_f16_e64_dpp v5, v1, v2 quad_perm:[0,1,2,3]
531-
// GFX11: v_fmac_f16_e64_dpp v5, v1, v2 quad_perm:[0,1,2,3] row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0xe4,0x00,0xff]
530+
v_fmac_f16_e64_dpp v5.l, v1.l, v2.l quad_perm:[0,1,2,3]
531+
// GFX11: v_fmac_f16_e64_dpp v5.l, v1.l, v2.l quad_perm:[0,1,2,3] row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0xe4,0x00,0xff]
532532

533-
v_fmac_f16_e64_dpp v5, v1, v2 row_mirror
534-
// GFX11: v_fmac_f16_e64_dpp v5, v1, v2 row_mirror row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x40,0x01,0xff]
533+
v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_mirror
534+
// GFX11: v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_mirror row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x40,0x01,0xff]
535535

536-
v_fmac_f16_e64_dpp v5, v1, v2 row_half_mirror
537-
// GFX11: v_fmac_f16_e64_dpp v5, v1, v2 row_half_mirror row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x41,0x01,0xff]
536+
v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_half_mirror
537+
// GFX11: v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_half_mirror row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x41,0x01,0xff]
538538

539-
v_fmac_f16_e64_dpp v5, v1, v2 row_shl:1
540-
// GFX11: v_fmac_f16_e64_dpp v5, v1, v2 row_shl:1 row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x01,0x01,0xff]
539+
v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_shl:1
540+
// GFX11: v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_shl:1 row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x01,0x01,0xff]
541541

542-
v_fmac_f16_e64_dpp v5, v1, v2 row_shl:15
543-
// GFX11: v_fmac_f16_e64_dpp v5, v1, v2 row_shl:15 row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x0f,0x01,0xff]
542+
v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_shl:15
543+
// GFX11: v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_shl:15 row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x0f,0x01,0xff]
544544

545-
v_fmac_f16_e64_dpp v5, v1, v2 row_shr:1
546-
// GFX11: v_fmac_f16_e64_dpp v5, v1, v2 row_shr:1 row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x11,0x01,0xff]
545+
v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_shr:1
546+
// GFX11: v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_shr:1 row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x11,0x01,0xff]
547547

548-
v_fmac_f16_e64_dpp v5, v1, v2 row_shr:15
549-
// GFX11: v_fmac_f16_e64_dpp v5, v1, v2 row_shr:15 row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x1f,0x01,0xff]
548+
v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_shr:15
549+
// GFX11: v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_shr:15 row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x1f,0x01,0xff]
550550

551-
v_fmac_f16_e64_dpp v5, v1, v2 row_ror:1
552-
// GFX11: v_fmac_f16_e64_dpp v5, v1, v2 row_ror:1 row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x21,0x01,0xff]
551+
v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_ror:1
552+
// GFX11: v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_ror:1 row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x21,0x01,0xff]
553553

554-
v_fmac_f16_e64_dpp v5, v1, v2 row_ror:15
555-
// GFX11: v_fmac_f16_e64_dpp v5, v1, v2 row_ror:15 row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x2f,0x01,0xff]
554+
v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_ror:15
555+
// GFX11: v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_ror:15 row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x2f,0x01,0xff]
556556

557-
v_fmac_f16_e64_dpp v5, v1, v2 row_share:0 row_mask:0xf bank_mask:0xf
558-
// GFX11: v_fmac_f16_e64_dpp v5, v1, v2 row_share:0 row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x50,0x01,0xff]
557+
v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_share:0 row_mask:0xf bank_mask:0xf
558+
// GFX11: v_fmac_f16_e64_dpp v5.l, v1.l, v2.l row_share:0 row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x50,0x01,0xff]
559559

560-
v_fmac_f16_e64_dpp v5, |v1|, -v2 mul:2 row_share:15 row_mask:0x0 bank_mask:0x1
561-
// GFX11: v_fmac_f16_e64_dpp v5, |v1|, -v2 mul:2 row_share:15 row_mask:0x0 bank_mask:0x1 ; encoding: [0x05,0x01,0x36,0xd5,0xfa,0x04,0x02,0x48,0x01,0x5f,0x01,0x01]
560+
v_fmac_f16_e64_dpp v5.l, |v1.l|, -v2.l mul:2 row_share:15 row_mask:0x0 bank_mask:0x1
561+
// GFX11: v_fmac_f16_e64_dpp v5.l, |v1.l|, -v2.l mul:2 row_share:15 row_mask:0x0 bank_mask:0x1 ; encoding: [0x05,0x01,0x36,0xd5,0xfa,0x04,0x02,0x48,0x01,0x5f,0x01,0x01]
562562

563-
v_fmac_f16_e64_dpp v5, -v1, |v2| mul:4 row_xmask:0 row_mask:0x1 bank_mask:0x3 bound_ctrl:1 fi:0
564-
// GFX11: v_fmac_f16_e64_dpp v5, -v1, |v2| mul:4 row_xmask:0 row_mask:0x1 bank_mask:0x3 bound_ctrl:1 ; encoding: [0x05,0x02,0x36,0xd5,0xfa,0x04,0x02,0x30,0x01,0x60,0x09,0x13]
563+
v_fmac_f16_e64_dpp v5.l, -v1.l, |v2.l| mul:4 row_xmask:0 row_mask:0x1 bank_mask:0x3 bound_ctrl:1 fi:0
564+
// GFX11: v_fmac_f16_e64_dpp v5.l, -v1.l, |v2.l| mul:4 row_xmask:0 row_mask:0x1 bank_mask:0x3 bound_ctrl:1 ; encoding: [0x05,0x02,0x36,0xd5,0xfa,0x04,0x02,0x30,0x01,0x60,0x09,0x13]
565565

566-
v_fmac_f16_e64_dpp v255, -|v255|, -|v255| clamp div:2 row_xmask:15 row_mask:0x3 bank_mask:0x0 bound_ctrl:0 fi:1
567-
// GFX11: v_fmac_f16_e64_dpp v255, -|v255|, -|v255| clamp div:2 row_xmask:15 row_mask:0x3 bank_mask:0x0 fi:1 ; encoding: [0xff,0x83,0x36,0xd5,0xfa,0xfe,0x03,0x78,0xff,0x6f,0x05,0x30]
566+
v_fmac_f16_e64_dpp v255.l, -|v255.l|, -|v255.l| clamp div:2 row_xmask:15 row_mask:0x3 bank_mask:0x0 bound_ctrl:0 fi:1
567+
// GFX11: v_fmac_f16_e64_dpp v255.l, -|v255.l|, -|v255.l| clamp div:2 row_xmask:15 row_mask:0x3 bank_mask:0x0 fi:1 ; encoding: [0xff,0x83,0x36,0xd5,0xfa,0xfe,0x03,0x78,0xff,0x6f,0x05,0x30]
568+
569+
v_fmac_f16_e64_dpp v5.h, v1.h, v2.h row_share:0 row_mask:0xf bank_mask:0xf
570+
// GFX11: v_fmac_f16_e64_dpp v5.h, v1.h, v2.h op_sel:[1,1,1,1] row_share:0 row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x58,0x36,0xd5,0xfa,0x04,0x02,0x00,0x01,0x50,0x01,0xff]
571+
572+
v_fmac_f16_e64_dpp v5.l, |v1.h|, -v2.l mul:2 row_share:15 row_mask:0x0 bank_mask:0x1
573+
// GFX11: v_fmac_f16_e64_dpp v5.l, |v1.h|, -v2.l op_sel:[1,0,0,0] mul:2 row_share:15 row_mask:0x0 bank_mask:0x1 ; encoding: [0x05,0x09,0x36,0xd5,0xfa,0x04,0x02,0x48,0x01,0x5f,0x01,0x01]
574+
575+
v_fmac_f16_e64_dpp v5.l, -v1.l, |v2.h| mul:4 row_xmask:0 row_mask:0x1 bank_mask:0x3 bound_ctrl:1 fi:0
576+
// GFX11: v_fmac_f16_e64_dpp v5.l, -v1.l, |v2.h| op_sel:[0,1,0,0] mul:4 row_xmask:0 row_mask:0x1 bank_mask:0x3 bound_ctrl:1 ; encoding: [0x05,0x12,0x36,0xd5,0xfa,0x04,0x02,0x30,0x01,0x60,0x09,0x13]
577+
578+
v_fmac_f16_e64_dpp v255.h, -|v255.l|, -|v255.l| clamp div:2 row_xmask:15 row_mask:0x3 bank_mask:0x0 bound_ctrl:0 fi:1
579+
// GFX11: v_fmac_f16_e64_dpp v255.h, -|v255.l|, -|v255.l| op_sel:[0,0,1,1] clamp div:2 row_xmask:15 row_mask:0x3 bank_mask:0x0 fi:1 ; encoding: [0xff,0xc3,0x36,0xd5,0xfa,0xfe,0x03,0x78,0xff,0x6f,0x05,0x30]
568580

569581
v_fmac_f32_e64_dpp v5, v1, v2 quad_perm:[3,2,1,0]
570582
// GFX11: v_fmac_f32_e64_dpp v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0xf bank_mask:0xf ; encoding: [0x05,0x00,0x2b,0xd5,0xfa,0x04,0x02,0x00,0x01,0x1b,0x00,0xff]

llvm/test/MC/AMDGPU/gfx11_asm_vop3_dpp8_from_vop2.s

+20-8
Original file line numberDiff line numberDiff line change
@@ -169,17 +169,29 @@ v_cvt_pkrtz_f16_f32_e64_dpp v5, -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
169169
v_cvt_pkrtz_f16_f32_e64_dpp v255, -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
170170
// GFX11: v_cvt_pk_rtz_f16_f32_e64_dpp v255, -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] ; encoding: [0xff,0x83,0x2f,0xd5,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
171171

172-
v_fmac_f16_e64_dpp v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0]
173-
// GFX11: v_fmac_f16_e64_dpp v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0] ; encoding: [0x05,0x00,0x36,0xd5,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
172+
v_fmac_f16_e64_dpp v5.l, v1.l, v2.l dpp8:[7,6,5,4,3,2,1,0]
173+
// GFX11: v_fmac_f16_e64_dpp v5.l, v1.l, v2.l dpp8:[7,6,5,4,3,2,1,0] ; encoding: [0x05,0x00,0x36,0xd5,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
174174

175-
v_fmac_f16_e64_dpp v5, |v1|, -v2 mul:2 dpp8:[7,6,5,4,3,2,1,0]
176-
// GFX11: v_fmac_f16_e64_dpp v5, |v1|, -v2 mul:2 dpp8:[7,6,5,4,3,2,1,0] ; encoding: [0x05,0x01,0x36,0xd5,0xe9,0x04,0x02,0x48,0x01,0x77,0x39,0x05]
175+
v_fmac_f16_e64_dpp v5.l, |v1.l|, -v2.l mul:2 dpp8:[7,6,5,4,3,2,1,0]
176+
// GFX11: v_fmac_f16_e64_dpp v5.l, |v1.l|, -v2.l mul:2 dpp8:[7,6,5,4,3,2,1,0] ; encoding: [0x05,0x01,0x36,0xd5,0xe9,0x04,0x02,0x48,0x01,0x77,0x39,0x05]
177177

178-
v_fmac_f16_e64_dpp v5, -v1, |v2| mul:4 dpp8:[7,6,5,4,3,2,1,0] fi:1
179-
// GFX11: v_fmac_f16_e64_dpp v5, -v1, |v2| mul:4 dpp8:[7,6,5,4,3,2,1,0] fi:1 ; encoding: [0x05,0x02,0x36,0xd5,0xea,0x04,0x02,0x30,0x01,0x77,0x39,0x05]
178+
v_fmac_f16_e64_dpp v5.l, -v1.l, |v2.l| mul:4 dpp8:[7,6,5,4,3,2,1,0] fi:1
179+
// GFX11: v_fmac_f16_e64_dpp v5.l, -v1.l, |v2.l| mul:4 dpp8:[7,6,5,4,3,2,1,0] fi:1 ; encoding: [0x05,0x02,0x36,0xd5,0xea,0x04,0x02,0x30,0x01,0x77,0x39,0x05]
180180

181-
v_fmac_f16_e64_dpp v255, -|v255|, -|v255| clamp div:2 dpp8:[0,0,0,0,0,0,0,0] fi:0
182-
// GFX11: v_fmac_f16_e64_dpp v255, -|v255|, -|v255| clamp div:2 dpp8:[0,0,0,0,0,0,0,0] ; encoding: [0xff,0x83,0x36,0xd5,0xe9,0xfe,0x03,0x78,0xff,0x00,0x00,0x00]
181+
v_fmac_f16_e64_dpp v255.l, -|v255.l|, -|v255.l| clamp div:2 dpp8:[0,0,0,0,0,0,0,0] fi:0
182+
// GFX11: v_fmac_f16_e64_dpp v255.l, -|v255.l|, -|v255.l| clamp div:2 dpp8:[0,0,0,0,0,0,0,0] ; encoding: [0xff,0x83,0x36,0xd5,0xe9,0xfe,0x03,0x78,0xff,0x00,0x00,0x00]
183+
184+
v_fmac_f16_e64_dpp v5.h, v1.h, v2.h dpp8:[7,6,5,4,3,2,1,0]
185+
// GFX11: v_fmac_f16_e64_dpp v5.h, v1.h, v2.h op_sel:[1,1,1,1] dpp8:[7,6,5,4,3,2,1,0] ; encoding: [0x05,0x58,0x36,0xd5,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
186+
187+
v_fmac_f16_e64_dpp v5.l, |v1.h|, -v2.l mul:2 dpp8:[7,6,5,4,3,2,1,0]
188+
// GFX11: v_fmac_f16_e64_dpp v5.l, |v1.h|, -v2.l op_sel:[1,0,0,0] mul:2 dpp8:[7,6,5,4,3,2,1,0] ; encoding: [0x05,0x09,0x36,0xd5,0xe9,0x04,0x02,0x48,0x01,0x77,0x39,0x05]
189+
190+
v_fmac_f16_e64_dpp v5.l, -v1.l, |v2.h| mul:4 dpp8:[7,6,5,4,3,2,1,0] fi:1
191+
// GFX11: v_fmac_f16_e64_dpp v5.l, -v1.l, |v2.h| op_sel:[0,1,0,0] mul:4 dpp8:[7,6,5,4,3,2,1,0] fi:1 ; encoding: [0x05,0x12,0x36,0xd5,0xea,0x04,0x02,0x30,0x01,0x77,0x39,0x05]
192+
193+
v_fmac_f16_e64_dpp v255.h, -|v255.l|, -|v255.l| clamp div:2 dpp8:[0,0,0,0,0,0,0,0] fi:0
194+
// GFX11: v_fmac_f16_e64_dpp v255.h, -|v255.l|, -|v255.l| op_sel:[0,0,1,1] clamp div:2 dpp8:[0,0,0,0,0,0,0,0] ; encoding: [0xff,0xc3,0x36,0xd5,0xe9,0xfe,0x03,0x78,0xff,0x00,0x00,0x00]
183195

184196
v_fmac_f32_e64_dpp v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0]
185197
// GFX11: v_fmac_f32_e64_dpp v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0] ; encoding: [0x05,0x00,0x2b,0xd5,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]

llvm/test/MC/AMDGPU/gfx11_asm_vop3_from_vop2.s

+39-30
Original file line numberDiff line numberDiff line change
@@ -587,50 +587,59 @@ v_fmac_dx9_zero_f32_e64 v5, -src_scc, |vcc_lo| mul:4
587587
v_fmac_dx9_zero_f32_e64 v255, -|0xaf123456|, -|vcc_hi| clamp div:2
588588
// GFX11: v_fmac_dx9_zero_f32_e64 v255, -|0xaf123456|, -|vcc_hi| clamp div:2 ; encoding: [0xff,0x83,0x06,0xd5,0xff,0xd6,0x00,0x78,0x56,0x34,0x12,0xaf]
589589

590-
v_fmac_f16_e64 v5, v1, v2
591-
// GFX11: v_fmac_f16_e64 v5, v1, v2 ; encoding: [0x05,0x00,0x36,0xd5,0x01,0x05,0x02,0x00]
590+
v_fmac_f16_e64 v5.l, v1.l, v2.l
591+
// GFX11: v_fmac_f16_e64 v5.l, v1.l, v2.l ; encoding: [0x05,0x00,0x36,0xd5,0x01,0x05,0x02,0x00]
592592

593-
v_fmac_f16_e64 v5, v255, v255
594-
// GFX11: v_fmac_f16_e64 v5, v255, v255 ; encoding: [0x05,0x00,0x36,0xd5,0xff,0xff,0x03,0x00]
593+
v_fmac_f16_e64 v5.l, v255.l, v255.l
594+
// GFX11: v_fmac_f16_e64 v5.l, v255.l, v255.l ; encoding: [0x05,0x00,0x36,0xd5,0xff,0xff,0x03,0x00]
595595

596-
v_fmac_f16_e64 v5, s1, s2
597-
// GFX11: v_fmac_f16_e64 v5, s1, s2 ; encoding: [0x05,0x00,0x36,0xd5,0x01,0x04,0x00,0x00]
596+
v_fmac_f16_e64 v5.l, s1, s2
597+
// GFX11: v_fmac_f16_e64 v5.l, s1, s2 ; encoding: [0x05,0x00,0x36,0xd5,0x01,0x04,0x00,0x00]
598598

599-
v_fmac_f16_e64 v5, s105, s105
600-
// GFX11: v_fmac_f16_e64 v5, s105, s105 ; encoding: [0x05,0x00,0x36,0xd5,0x69,0xd2,0x00,0x00]
599+
v_fmac_f16_e64 v5.l, s105, s105
600+
// GFX11: v_fmac_f16_e64 v5.l, s105, s105 ; encoding: [0x05,0x00,0x36,0xd5,0x69,0xd2,0x00,0x00]
601601

602-
v_fmac_f16_e64 v5, vcc_lo, ttmp15
603-
// GFX11: v_fmac_f16_e64 v5, vcc_lo, ttmp15 ; encoding: [0x05,0x00,0x36,0xd5,0x6a,0xf6,0x00,0x00]
602+
v_fmac_f16_e64 v5.l, vcc_lo, ttmp15
603+
// GFX11: v_fmac_f16_e64 v5.l, vcc_lo, ttmp15 ; encoding: [0x05,0x00,0x36,0xd5,0x6a,0xf6,0x00,0x00]
604604

605-
v_fmac_f16_e64 v5, vcc_hi, 0xfe0b
606-
// GFX11: v_fmac_f16_e64 v5, vcc_hi, 0xfe0b ; encoding: [0x05,0x00,0x36,0xd5,0x6b,0xfe,0x01,0x00,0x0b,0xfe,0x00,0x00]
605+
v_fmac_f16_e64 v5.l, vcc_hi, 0xfe0b
606+
// GFX11: v_fmac_f16_e64 v5.l, vcc_hi, 0xfe0b ; encoding: [0x05,0x00,0x36,0xd5,0x6b,0xfe,0x01,0x00,0x0b,0xfe,0x00,0x00]
607607

608-
v_fmac_f16_e64 v5, ttmp15, src_scc
609-
// GFX11: v_fmac_f16_e64 v5, ttmp15, src_scc ; encoding: [0x05,0x00,0x36,0xd5,0x7b,0xfa,0x01,0x00]
608+
v_fmac_f16_e64 v5.l, ttmp15, src_scc
609+
// GFX11: v_fmac_f16_e64 v5.l, ttmp15, src_scc ; encoding: [0x05,0x00,0x36,0xd5,0x7b,0xfa,0x01,0x00]
610610

611-
v_fmac_f16_e64 v5, m0, 0.5
612-
// GFX11: v_fmac_f16_e64 v5, m0, 0.5 ; encoding: [0x05,0x00,0x36,0xd5,0x7d,0xe0,0x01,0x00]
611+
v_fmac_f16_e64 v5.l, m0, 0.5
612+
// GFX11: v_fmac_f16_e64 v5.l, m0, 0.5 ; encoding: [0x05,0x00,0x36,0xd5,0x7d,0xe0,0x01,0x00]
613613

614-
v_fmac_f16_e64 v5, exec_lo, -1
615-
// GFX11: v_fmac_f16_e64 v5, exec_lo, -1 ; encoding: [0x05,0x00,0x36,0xd5,0x7e,0x82,0x01,0x00]
614+
v_fmac_f16_e64 v5.l, exec_lo, -1
615+
// GFX11: v_fmac_f16_e64 v5.l, exec_lo, -1 ; encoding: [0x05,0x00,0x36,0xd5,0x7e,0x82,0x01,0x00]
616616

617-
v_fmac_f16_e64 v5, |exec_hi|, null
618-
// GFX11: v_fmac_f16_e64 v5, |exec_hi|, null ; encoding: [0x05,0x01,0x36,0xd5,0x7f,0xf8,0x00,0x00]
617+
v_fmac_f16_e64 v5.l, |exec_hi|, null
618+
// GFX11: v_fmac_f16_e64 v5.l, |exec_hi|, null ; encoding: [0x05,0x01,0x36,0xd5,0x7f,0xf8,0x00,0x00]
619619

620-
v_fmac_f16_e64 v5, null, exec_lo
621-
// GFX11: v_fmac_f16_e64 v5, null, exec_lo ; encoding: [0x05,0x00,0x36,0xd5,0x7c,0xfc,0x00,0x00]
620+
v_fmac_f16_e64 v5.l, null, exec_lo
621+
// GFX11: v_fmac_f16_e64 v5.l, null, exec_lo ; encoding: [0x05,0x00,0x36,0xd5,0x7c,0xfc,0x00,0x00]
622622

623-
v_fmac_f16_e64 v5, -1, exec_hi
624-
// GFX11: v_fmac_f16_e64 v5, -1, exec_hi ; encoding: [0x05,0x00,0x36,0xd5,0xc1,0xfe,0x00,0x00]
623+
v_fmac_f16_e64 v5.l, -1, exec_hi
624+
// GFX11: v_fmac_f16_e64 v5.l, -1, exec_hi ; encoding: [0x05,0x00,0x36,0xd5,0xc1,0xfe,0x00,0x00]
625625

626-
v_fmac_f16_e64 v5, 0.5, -m0 mul:2
627-
// GFX11: v_fmac_f16_e64 v5, 0.5, -m0 mul:2 ; encoding: [0x05,0x00,0x36,0xd5,0xf0,0xfa,0x00,0x48]
626+
v_fmac_f16_e64 v5.l, 0.5, -m0 mul:2
627+
// GFX11: v_fmac_f16_e64 v5.l, 0.5, -m0 mul:2 ; encoding: [0x05,0x00,0x36,0xd5,0xf0,0xfa,0x00,0x48]
628628

629-
v_fmac_f16_e64 v5, -src_scc, |vcc_lo| mul:4
630-
// GFX11: v_fmac_f16_e64 v5, -src_scc, |vcc_lo| mul:4 ; encoding: [0x05,0x02,0x36,0xd5,0xfd,0xd4,0x00,0x30]
629+
v_fmac_f16_e64 v5.l, -src_scc, |vcc_lo| mul:4
630+
// GFX11: v_fmac_f16_e64 v5.l, -src_scc, |vcc_lo| mul:4 ; encoding: [0x05,0x02,0x36,0xd5,0xfd,0xd4,0x00,0x30]
631631

632-
v_fmac_f16_e64 v255, -|0xfe0b|, -|vcc_hi| clamp div:2
633-
// GFX11: v_fmac_f16_e64 v255, -|0xfe0b|, -|vcc_hi| clamp div:2 ; encoding: [0xff,0x83,0x36,0xd5,0xff,0xd6,0x00,0x78,0x0b,0xfe,0x00,0x00]
632+
v_fmac_f16_e64 v255.l, -|0xfe0b|, -|vcc_hi| clamp div:2
633+
// GFX11: v_fmac_f16_e64 v255.l, -|0xfe0b|, -|vcc_hi| clamp div:2 ; encoding: [0xff,0x83,0x36,0xd5,0xff,0xd6,0x00,0x78,0x0b,0xfe,0x00,0x00]
634+
635+
v_fmac_f16_e64 v5.l, v1.h, v2.l
636+
// GFX11: v_fmac_f16_e64 v5.l, v1.h, v2.l op_sel:[1,0,0,0] ; encoding: [0x05,0x08,0x36,0xd5,0x01,0x05,0x02,0x00]
637+
638+
v_fmac_f16_e64 v5.l, v255.l, v255.h
639+
// GFX11: v_fmac_f16_e64 v5.l, v255.l, v255.h op_sel:[0,1,0,0] ; encoding: [0x05,0x10,0x36,0xd5,0xff,0xff,0x03,0x00]
640+
641+
v_fmac_f16_e64 v255.h, -|0xfe0b|, -|vcc_hi| clamp div:2
642+
// GFX11: v_fmac_f16_e64 v255.h, -|0xfe0b|, -|vcc_hi| op_sel:[0,0,1,1] clamp div:2 ; encoding: [0xff,0xc3,0x36,0xd5,0xff,0xd6,0x00,0x78,0x0b,0xfe,0x00,0x00]
634643

635644
v_fmac_f32_e64 v5, v1, v2
636645
// GFX11: v_fmac_f32_e64 v5, v1, v2 ; encoding: [0x05,0x00,0x2b,0xd5,0x01,0x05,0x02,0x00]

0 commit comments

Comments
 (0)