maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[AMDGPU][MC][GFX11][NFC] Consolidate VOP tests by encoding 

Differential Revision: https://reviews.llvm.org/D131559
This commit is contained in:
Dmitry Preobrazhensky 2022-08-11 15:00:40 +03:00
parent 7309e8cfbe
commit c0de0e0c64
22 changed files with 16627 additions and 20743 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

223
llvm/test/MC/AMDGPU/gfx11_asm_dpp.s
View File

@ -1,223 +0,0 @@
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11,W32 %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11,W64 %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 %s 2>&1 | FileCheck --check-prefix=W32-ERR --implicit-check-not=error: %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 %s 2>&1 | FileCheck --check-prefix=W64-ERR --implicit-check-not=error: %s
; DPP8
; VOP1->3
v_cvt_f16_f32_e64_dpp v5, v1 div:2 dpp8:[0,2,1,3,4,5,6,7]
// GFX11: encoding: [0x05,0x00,0x8a,0xd5,0xe9,0x00,0x00,0x18,0x01,0x50,0xc6,0xfa]
v_ffbh_i32_e64_dpp v5, v1 dpp8:[0,0,2,3,4,4,6,7]
// GFX11: encoding: [0x05,0x00,0xbb,0xd5,0xe9,0x00,0x00,0x00,0x01,0x80,0x46,0xfa]
v_ctz_i32_b32_e64_dpp v199, v1 dpp8:[0,0,2,3,4,4,6,7] fi:1
// GFX11: encoding: [0xc7,0x00,0xba,0xd5,0xea,0x00,0x00,0x00,0x01,0x80,0x46,0xfa]
; VOP2->3
v_xnor_b32_e64_dpp v8, v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: encoding: [0x08,0x00,0x1e,0xd5,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_add_nc_u32_e64_dpp v60, v61, v62 dpp8:[7,6,5,3,4,2,1,0] fi:1
// GFX11: encoding: [0x3c,0x00,0x25,0xd5,0xea,0x7c,0x02,0x00,0x3d,0x77,0x47,0x05]
v_mul_f32_e64_dpp v0, v1, v2 dpp8:[0,1,2,3,1,5,6,7]
// GFX11: encoding: [0x00,0x00,0x08,0xd5,0xe9,0x04,0x02,0x00,0x01,0x88,0x96,0xfa]
v_cndmask_b32_e64_dpp v202, v1, v2, s[4:5] dpp8:[7,1,7,2,7,3,7,4]
// W64: encoding: [0xca,0x00,0x01,0xd5,0xe9,0x04,0x12,0x00,0x01,0xcf,0xf5,0x9d]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
; VOP2B -> VOP3B
v_add_co_ci_u32_e64_dpp v0, s2, v1, v2, s1 clamp dpp8:[7,6,5,3,4,2,1,0] fi:1
// W32: encoding: [0x00,0x82,0x20,0xd5,0xea,0x04,0x06,0x00,0x01,0x77,0x47,0x05]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_add_co_ci_u32_e64_dpp v0, s[2:3], v1, v2, s[2:3] clamp dpp8:[7,6,5,3,4,2,1,0] fi:1
// W64: encoding: [0x00,0x82,0x20,0xd5,0xea,0x04,0x0a,0x00,0x01,0x77,0x47,0x05]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_sub_co_ci_u32_e64_dpp v0, vcc, v1, v2, vcc clamp dpp8:[7,6,5,3,4,2,1,0]
// W64: encoding: [0x00,0xea,0x21,0xd5,0xe9,0x04,0xaa,0x01,0x01,0x77,0x47,0x05]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_sub_co_ci_u32_e64_dpp v0, vcc, v1, v2, s[2:3] dpp8:[7,6,5,3,4,2,1,0] fi:1
// W64: encoding: [0x00,0x6a,0x21,0xd5,0xea,0x04,0x0a,0x00,0x01,0x77,0x47,0x05]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
; VOP3
v_bfe_u32_e64_dpp v0, v1, v2, v3 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0x00,0x00,0x10,0xd6,0xe9,0x04,0x0e,0x04,0x01,0x88,0xc6,0xfa]
v_maxmin_f32_e64_dpp v0, v1, v2, v3 clamp dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0x00,0x80,0x5e,0xd6,0xe9,0x04,0x0e,0x04,0x01,0x88,0xc6,0xfa]
v_maxmin_f32_e64_dpp v0, v1, v2, v3 div:2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0x00,0x00,0x5e,0xd6,0xea,0x04,0x0e,0x1c,0x01,0x88,0xc6,0xfa]
v_minmax_f32_e64_dpp v0, -v1, -v2, -v3 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0x00,0x00,0x5f,0xd6,0xe9,0x04,0x0e,0xe4,0x01,0x88,0xc6,0xfa]
v_minmax_f32_e64_dpp v0, abs(v1), v2, v3 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0x00,0x01,0x5f,0xd6,0xe9,0x04,0x0e,0x04,0x01,0x88,0xc6,0xfa]
v_fma_f32_e64_dpp v80, v81, v82, v81 dpp8:[0,1,6,3,4,5,6,7]
// GFX11: encoding: [0x50,0x00,0x13,0xd6,0xe9,0xa4,0x46,0x05,0x51,0x88,0xc7,0xfa]
v_fma_f32_e64_dpp v80, v81, abs(v82), v81 dpp8:[0,1,6,3,4,5,6,7]
// GFX11: encoding: [0x50,0x02,0x13,0xd6,0xe9,0xa4,0x46,0x05,0x51,0x88,0xc7,0xfa]
v_max3_f32_e64_dpp v5, v1, v2, s3 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0x05,0x00,0x1c,0xd6,0xe9,0x04,0x0e,0x00,0x01,0x77,0x39,0x05]
v_add3_u32_e64_dpp v5, v1, v2, -1 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0x05,0x00,0x55,0xd6,0xe9,0x04,0x06,0x03,0x01,0x77,0x39,0x05]
v_add3_u32_e64_dpp v5, v1, v2, s4 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0x05,0x00,0x55,0xd6,0xe9,0x04,0x12,0x00,0x01,0x77,0x39,0x05]
;VOP3B
v_subrev_co_u32_e64_dpp v5, null, v1, v2 dpp8:[2,2,2,2,4,4,4,4]
// GFX11: encoding: [0x05,0x7c,0x02,0xd7,0xe9,0x04,0x02,0x00,0x01,0x92,0x44,0x92]
v_subrev_co_u32_e64_dpp v5, vcc_lo, v1, v2 dpp8:[2,2,2,2,4,4,4,4]
// W32: encoding: [0x05,0x6a,0x02,0xd7,0xe9,0x04,0x02,0x00,0x01,0x92,0x44,0x92]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
; VOP3P
v_fma_mix_f32 v0, v1, v2, v3 dpp8:[2,2,2,2,4,4,4,4]
// GFX11: encoding: [0x00,0x00,0x20,0xcc,0xe9,0x04,0x0e,0x04,0x01,0x92,0x44,0x92]
v_fma_mix_f32 v0, v1, v2, v3 clamp dpp8:[2,2,2,2,4,4,4,4] fi:1
// GFX11: encoding: [0x00,0x80,0x20,0xcc,0xea,0x04,0x0e,0x04,0x01,0x92,0x44,0x92]
v_fma_mixlo_f16 v0, abs(v1), -v2, abs(v3) dpp8:[2,2,2,2,4,4,4,4]
// GFX11: encoding: [0x00,0x05,0x21,0xcc,0xe9,0x04,0x0e,0x44,0x01,0x92,0x44,0x92]
; For test purpose only. OP_SEL has to be set to all 0 and OP_SEL_HI has to be set to all 1
v_fma_mixlo_f16 v0, abs(v1), -v2, abs(v3) op_sel:[1,0,0] op_sel_hi:[1,0,0] dpp8:[2,2,2,2,4,4,4,4]
// GFX11: encoding: [0x00,0x0d,0x21,0xcc,0xe9,0x04,0x0e,0x4c,0x01,0x92,0x44,0x92]
v_dot2_f32_f16_e64_dpp v0, v1, v2, v3 neg_lo:[0,1,1] neg_hi:[1,0,1] dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0x00,0x05,0x13,0xcc,0xe9,0x04,0x0e,0xc4,0x01,0x77,0x39,0x05]
; DPP
; VOP1->3
v_floor_f32_e64_dpp v5, v1 clamp row_shl:0x7 row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0x05,0x80,0xa4,0xd5,0xfa,0x00,0x00,0x00,0x01,0x07,0x05,0x00]
v_fract_f32_e64_dpp v5, v1 mul:2 quad_perm:[1,3,1,0] row_mask:0x7
// GFX11: encoding: [0x05,0x00,0xa0,0xd5,0xfa,0x00,0x00,0x08,0x01,0x1d,0x00,0x7f]
v_sat_pk_u8_i16_e64_dpp v0, v2 row_mirror bank_mask:0x2 fi:1
// GFX11: encoding: [0x00,0x00,0xe2,0xd5,0xfa,0x00,0x00,0x00,0x02,0x40,0x05,0xf2]
v_cvt_i32_f32_e64_dpp v5, v1 clamp row_shl:0x7 row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0x05,0x80,0x88,0xd5,0xfa,0x00,0x00,0x00,0x01,0x07,0x01,0x00]
; VOP2->3
v_mul_f16_e64_dpp v0, v2, v4 row_share:0xa bound_ctrl:0
// GFX11: encoding: [0x00,0x00,0x35,0xd5,0xfa,0x08,0x02,0x00,0x02,0x5a,0x09,0xff]
v_xnor_b32_e64_dpp v8, v5, v2 quad_perm:[1,0,2,3] row_mask:0x1 bank_mask:0x0
// GFX11: encoding: [0x08,0x00,0x1e,0xd5,0xfa,0x04,0x02,0x00,0x05,0xe1,0x00,0x10]
v_mbcnt_lo_u32_b32_e64_dpp v5, v126, v2 row_half_mirror bound_ctrl:0
// GFX11: encoding: [0x05,0x00,0x1f,0xd7,0xfa,0x04,0x02,0x00,0x7e,0x41,0x09,0xff]
v_mul_i32_i24_e64_dpp v208, v101, v4 clamp row_shr:0xe row_mask:0x3 bank_mask:0xa bound_ctrl:0
// GFX11: encoding: [0xd0,0x80,0x09,0xd5,0xfa,0x08,0x02,0x00,0x65,0x1e,0x09,0x3a]
v_cndmask_b16_e64_dpp v0, v1, v2, null quad_perm:[2,3,0,0]
// GFX11: encoding: [0x00,0x00,0x5d,0xd6,0xfa,0x04,0xf2,0x01,0x01,0x0e,0x00,0xff]
v_cndmask_b32_e64_dpp v202, v1, v2, s1 quad_perm:[2,3,0,0] fi:1
// W32: encoding: [0xca,0x00,0x01,0xd5,0xfa,0x04,0x06,0x00,0x01,0x0e,0x04,0xff]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
; VOP2
v_dot2c_f32_f16_dpp v5, v1, v2 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x04,0x01,0xe4,0x00,0x00]
v_dot2c_f32_f16_dpp v5, v1, v2 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x04,0x01,0xe4,0x04,0x00]
v_dot2c_f32_f16_dpp v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0xe9,0x04,0x0a,0x04,0x01,0x77,0x39,0x05]
v_dot2c_f32_f16_dpp v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x04,0x01,0x77,0x39,0x05]
v_dot2acc_f32_f16_dpp v5, v1, v2 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x04,0x01,0xe4,0x00,0x00]
v_dot2acc_f32_f16_dpp v5, v1, v2 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x04,0x01,0xe4,0x04,0x00]
v_dot2acc_f32_f16_dpp v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0xe9,0x04,0x0a,0x04,0x01,0x77,0x39,0x05]
v_dot2acc_f32_f16_dpp v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x04,0x01,0x77,0x39,0x05]
; VOP2B -> VOP3B
v_subrev_co_ci_u32_e64_dpp v0, vcc_lo, v1, v2, vcc_lo quad_perm:[2,2,3,1]
// W32: encoding: [0x00,0x6a,0x22,0xd5,0xfa,0x04,0xaa,0x01,0x01,0x7a,0x00,0xff]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_add_co_ci_u32_e64_dpp v0, s5, v1, v2, vcc_hi quad_perm:[0,2,3,1] row_mask:0x0
// W32: encoding: [0x00,0x05,0x20,0xd5,0xfa,0x04,0xae,0x01,0x01,0x78,0x00,0x0f]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_add_co_ci_u32_e64_dpp v0, vcc, v1, v2, s[0:1] quad_perm:[0,2,3,1] row_mask:0x0
// W64: encoding: [0x00,0x6a,0x20,0xd5,0xfa,0x04,0x02,0x00,0x01,0x78,0x00,0x0f]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
; VOP3
v_bfe_u32_e64_dpp v0, v1, v2, v3 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0x00,0x00,0x10,0xd6,0xfa,0x04,0x0e,0x04,0x01,0x1b,0x00,0x00]
v_fma_f32_e64_dpp v93, abs(v94), v95, v94 quad_perm:[3,2,1,0] bank_mask:0xe
// GFX11: encoding: [0x5d,0x01,0x13,0xd6,0xfa,0xbe,0x7a,0x05,0x5e,0x1b,0x00,0xfe]
v_sub_nc_i32_e64_dpp v93, v94, v95 row_ror:7 bank_mask:0x1 bound_ctrl:0
// GFX11: encoding: [0x5d,0x00,0x25,0xd7,0xfa,0xbe,0x02,0x00,0x5e,0x27,0x09,0xf1]
v_lshl_or_b32_e64_dpp v255, v5, v0, vcc_hi row_xmask:0x6 row_mask:0x0 fi:1
// GFX11: encoding: [0xff,0x00,0x56,0xd6,0xfa,0x00,0xae,0x01,0x05,0x66,0x05,0x0f]
v_cubesc_f32_e64_dpp v5, v1, v2, 1 row_shr:4 row_mask:0xf bank_mask:0xf
// GFX11: encoding: [0x05,0x00,0x0d,0xd6,0xfa,0x04,0x06,0x02,0x01,0x14,0x01,0xff]
v_cubesc_f32_e64_dpp v5, v1, v2, s2 row_shr:4 row_mask:0xf bank_mask:0xf
// GFX11: encoding: [0x05,0x00,0x0d,0xd6,0xfa,0x04,0x0a,0x00,0x01,0x14,0x01,0xff]
; VOP3B
v_add_co_u32_e64_dpp v5, s4, v1, v2 clamp quad_perm:[2,2,3,1] bound_ctrl:0 fi:1
// W32: encoding: [0x05,0x84,0x00,0xd7,0xfa,0x04,0x02,0x00,0x01,0x7a,0x0c,0xff]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_add_co_u32_e64_dpp v243, vcc, v243, v2 clamp quad_perm:[1,2,3,1] bank_mask: 0x5 fi:1
// W64: encoding: [0xf3,0xea,0x00,0xd7,0xfa,0x04,0x02,0x00,0xf3,0x79,0x04,0xf5]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
; VOP3P
v_dot2_f32_f16 v0, v1, v2, v3 neg_lo:[0,0,0] neg_hi:[0,0,0] quad_perm:[2,2,3,1] bound_ctrl:0 fi:1
// GFX11: v_dot2_f32_f16_e64_dpp v0, v1, v2, v3 quad_perm:[2,2,3,1] row_mask:0xf bank_mask:0xf bound_ctrl:1 fi:1 ; encoding: [0x00,0x00,0x13,0xcc,0xfa,0x04,0x0e,0x04,0x01,0x7a,0x0c,0xff]
v_dot2_f32_f16 v0, v1, v2, v3 neg_lo:[1,1,0] neg_hi:[1,0,1] quad_perm:[3,2,1,0] bank_mask:0xe
// GFX11: v_dot2_f32_f16_e64_dpp v0, v1, v2, v3 neg_lo:[1,1,0] neg_hi:[1,0,1] quad_perm:[3,2,1,0] row_mask:0xf bank_mask:0xe ; encoding: [0x00,0x05,0x13,0xcc,0xfa,0x04,0x0e,0x64,0x01,0x1b,0x00,0xfe]
v_fma_mix_f32 v0, v1, v2, v3 op_sel:[0,0,0] row_ror:7 bank_mask:0x1 bound_ctrl:0
// GFX11: v_fma_mix_f32_e64_dpp v0, v1, v2, v3 row_ror:7 row_mask:0xf bank_mask:0x1 bound_ctrl:1 ; encoding: [0x00,0x00,0x20,0xcc,0xfa,0x04,0x0e,0x04,0x01,0x27,0x09,0xf1]
v_fma_mixhi_f16 v0, v1, v2, v3 op_sel_hi:[1,1,1] clamp quad_perm:[0,2,3,1] row_mask:0x0
// GFX11: v_fma_mixhi_f16_e64_dpp v0, v1, v2, v3 op_sel_hi:[1,1,1] clamp quad_perm:[0,2,3,1] row_mask:0x0 bank_mask:0xf ; encoding: [0x00,0xc0,0x22,0xcc,0xfa,0x04,0x0e,0x1c,0x01,0x78,0x00,0x0f]

271
llvm/test/MC/AMDGPU/gfx11_asm_dpp16.s → llvm/test/MC/AMDGPU/gfx11_asm_vop1_dpp16.s
View File

@ -1,5 +1,5 @@
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s // RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -show-encoding %s 2>&1 | FileCheck --check-prefixes=GFX11-ERR --implicit-check-not=error %s // RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
v_mov_b32 v5, v1 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 v_mov_b32 v5, v1 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x02,0x0a,0x7e,0x01,0x1b,0x00,0x00] // GFX11: encoding: [0xfa,0x02,0x0a,0x7e,0x01,0x1b,0x00,0x00]
@ -229,114 +229,6 @@ v_cvt_norm_i16_f16 v5, v1 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
v_cvt_norm_u16_f16 v5, v1 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 v_cvt_norm_u16_f16 v5, v1 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0xc8,0x0a,0x7e,0x01,0x1b,0x00,0x00] // GFX11: encoding: [0xfa,0xc8,0x0a,0x7e,0x01,0x1b,0x00,0x00]
v_add_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x06,0x01,0x1b,0x00,0x00]
v_add_f32 v5, -v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x06,0x01,0x1b,0x10,0x00]
v_add_f32 v5, v1, -v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x06,0x01,0x1b,0x40,0x00]
v_sub_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x08,0x01,0x1b,0x00,0x00]
v_subrev_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x0a,0x01,0x1b,0x00,0x00]
v_mul_legacy_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x0e,0x01,0x1b,0x00,0x00]
v_mul_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x10,0x01,0x1b,0x00,0x00]
v_mul_i32_i24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x12,0x01,0x1b,0x00,0x00]
v_mul_hi_i32_i24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x14,0x01,0x1b,0x00,0x00]
v_mul_u32_u24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x16,0x01,0x1b,0x00,0x00]
v_mul_hi_u32_u24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x18,0x01,0x1b,0x00,0x00]
v_min_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x1e,0x01,0x1b,0x00,0x00]
v_max_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x20,0x01,0x1b,0x00,0x00]
v_min_i32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x22,0x01,0x1b,0x00,0x00]
v_max_i32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x24,0x01,0x1b,0x00,0x00]
v_min_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x26,0x01,0x1b,0x00,0x00]
v_max_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x28,0x01,0x1b,0x00,0x00]
v_lshrrev_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x32,0x01,0x1b,0x00,0x00]
v_ashrrev_i32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x34,0x01,0x1b,0x00,0x00]
v_lshlrev_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x30,0x01,0x1b,0x00,0x00]
v_and_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x36,0x01,0x1b,0x00,0x00]
v_or_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x38,0x01,0x1b,0x00,0x00]
v_xor_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x3a,0x01,0x1b,0x00,0x00]
v_xnor_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x3c,0x01,0x1b,0x00,0x00]
v_add_co_ci_u32 v5, vcc_lo, v1, v2, vcc_lo quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x40,0x01,0x1b,0x00,0x00]
v_sub_co_ci_u32 v5, vcc_lo, v1, v2, vcc_lo quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x42,0x01,0x1b,0x00,0x00]
v_subrev_co_ci_u32 v5, vcc_lo, v1, v2, vcc_lo quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x44,0x01,0x1b,0x00,0x00]
v_fmac_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x56,0x01,0x1b,0x00,0x00]
v_add_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x64,0x01,0x1b,0x00,0x00]
v_sub_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x66,0x01,0x1b,0x00,0x00]
v_subrev_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x68,0x01,0x1b,0x00,0x00]
v_mul_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x6a,0x01,0x1b,0x00,0x00]
v_fmac_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x6c,0x01,0x1b,0x00,0x00]
v_max_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x72,0x01,0x1b,0x00,0x00]
v_min_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x74,0x01,0x1b,0x00,0x00]
v_ldexp_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x76,0x01,0x1b,0x00,0x00]
v_mov_b32 v5, v1 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:0 v_mov_b32 v5, v1 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:0
// GFX11: encoding: [0xfa,0x02,0x0a,0x7e,0x01,0x1b,0x00,0x00] // GFX11: encoding: [0xfa,0x02,0x0a,0x7e,0x01,0x1b,0x00,0x00]
@ -505,123 +397,6 @@ v_cvt_norm_i16_f16 v5, v1 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
v_cvt_norm_u16_f16 v5, v1 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1 v_cvt_norm_u16_f16 v5, v1 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0xc8,0x0a,0x7e,0x01,0x1b,0x04,0x00] // GFX11: encoding: [0xfa,0xc8,0x0a,0x7e,0x01,0x1b,0x04,0x00]
v_add_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x06,0x01,0x1b,0x04,0x00]
v_sub_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x08,0x01,0x1b,0x04,0x00]
v_subrev_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x0a,0x01,0x1b,0x04,0x00]
v_mul_legacy_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x0e,0x01,0x1b,0x04,0x00]
v_mul_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x10,0x01,0x1b,0x04,0x00]
v_mul_i32_i24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x12,0x01,0x1b,0x04,0x00]
v_mul_hi_i32_i24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x14,0x01,0x1b,0x04,0x00]
v_mul_u32_u24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x16,0x01,0x1b,0x04,0x00]
v_mul_hi_u32_u24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x18,0x01,0x1b,0x04,0x00]
v_min_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x1e,0x01,0x1b,0x04,0x00]
v_max_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x20,0x01,0x1b,0x04,0x00]
v_min_i32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x22,0x01,0x1b,0x04,0x00]
v_max_i32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x24,0x01,0x1b,0x04,0x00]
v_min_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x26,0x01,0x1b,0x04,0x00]
v_max_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x28,0x01,0x1b,0x04,0x00]
v_lshrrev_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x32,0x01,0x1b,0x04,0x00]
v_ashrrev_i32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x34,0x01,0x1b,0x04,0x00]
v_lshlrev_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x30,0x01,0x1b,0x04,0x00]
v_and_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x36,0x01,0x1b,0x04,0x00]
v_or_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x38,0x01,0x1b,0x04,0x00]
v_xor_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x3a,0x01,0x1b,0x04,0x00]
v_xnor_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x3c,0x01,0x1b,0x04,0x00]
v_fmac_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x56,0x01,0x1b,0x04,0x00]
v_add_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x64,0x01,0x1b,0x04,0x00]
v_sub_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x66,0x01,0x1b,0x04,0x00]
v_subrev_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x68,0x01,0x1b,0x04,0x00]
v_mul_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x6a,0x01,0x1b,0x04,0x00]
v_fmac_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x6c,0x01,0x1b,0x04,0x00]
v_max_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x72,0x01,0x1b,0x04,0x00]
v_min_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x74,0x01,0x1b,0x04,0x00]
v_ldexp_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x76,0x01,0x1b,0x04,0x00]
v_add_nc_u32 v5, v1, v2 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x4a,0x01,0xe4,0x00,0x00]
v_add_nc_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x4a,0x01,0x1b,0x04,0x00]
v_sub_nc_u32 v5, v1, v2 row_mirror row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x4c,0x01,0x40,0x01,0x00]
v_sub_nc_u32 v5, v1, v2 row_half_mirror row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x4c,0x01,0x41,0x01,0x00]
v_sub_nc_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x4c,0x01,0x1b,0x04,0x00]
v_subrev_nc_u32 v5, v1, v2 row_xmask:15 row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x4e,0x01,0x6f,0x01,0x00]
v_subrev_nc_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x1 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x4e,0x01,0x1b,0x00,0x10]
v_subrev_nc_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x4e,0x01,0x1b,0x04,0x00]
v_movreld_b32 v1, v0 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 v_movreld_b32 v1, v0 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x84,0x02,0x7e,0x00,0x1b,0x00,0x00] // GFX11: encoding: [0xfa,0x84,0x02,0x7e,0x00,0x1b,0x00,0x00]
@ -633,45 +408,3 @@ v_movrelsd_2_b32 v0, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
v_movrelsd_b32 v0, v255 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 v_movrelsd_b32 v0, v255 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x88,0x00,0x7e,0xff,0x1b,0x00,0x00] // GFX11: encoding: [0xfa,0x88,0x00,0x7e,0xff,0x1b,0x00,0x00]
v_dot2_f16_f16_e64_dpp v0, v1, v2, v3 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0x00,0x00,0x66,0xd6,0xfa,0x04,0x0e,0x04,0x01,0xe4,0x04,0x00]
v_dot2_f16_f16_e64_dpp v0, v1, v2, v3 op_sel:[1,1,0,0] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid op_sel operand
v_dot2_f16_f16_e64_dpp v0, s1, v2, v3 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
v_dot2_f16_f16_e64_dpp v0, v1, s2, v3 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
v_dot2_f16_f16_e64_dpp v0, v1, v2, v3 op_sel:[0,0,1,1] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0x00,0x60,0x66,0xd6,0xfa,0x04,0x0e,0x04,0x01,0xe4,0x04,0x00]
v_dot2_f16_f16_e64_dpp v0, |v1|, -v2, -|s3| op_sel:[0,0,1,1] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0x00,0x65,0x66,0xd6,0xfa,0x04,0x0e,0xc0,0x01,0xe4,0x04,0x00]
v_dot2_f16_f16_e64_dpp v5, v1, v2, 0.5 quad_perm:[3,2,1,0] row_mask:0xf bank_mask:0xf
// GFX11: encoding: [0x05,0x00,0x66,0xd6,0xfa,0x04,0xc2,0x03,0x01,0x1b,0x00,0xff]
v_dot2_bf16_bf16_e64_dpp v0, v1, v2, v3 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0x00,0x00,0x67,0xd6,0xfa,0x04,0x0e,0x04,0x01,0xe4,0x04,0x00]
v_dot2_bf16_bf16_e64_dpp v0, v1, v2, v3 op_sel:[1,1,0,0] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid op_sel operand
v_dot2_bf16_bf16_e64_dpp v0, s1, v2, v3 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
// GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
v_dot2_bf16_bf16_e64_dpp v0, v1, s2, v3 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
// GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
v_dot2_bf16_bf16_e64_dpp v0, v1, v2, v3 op_sel:[0,0,1,1] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0x00,0x60,0x67,0xd6,0xfa,0x04,0x0e,0x04,0x01,0xe4,0x04,0x00]
v_dot2_bf16_bf16_e64_dpp v0, |v1|, -v2, -|s3| op_sel:[0,0,1,1] quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0x00,0x65,0x67,0xd6,0xfa,0x04,0x0e,0xc0,0x01,0xe4,0x04,0x00]
v_dot2_bf16_bf16_e64_dpp v5, v1, v2, 0 quad_perm:[3,2,1,0] row_mask:0xf bank_mask:0xf
// GFX11: encoding: [0x05,0x00,0x67,0xd6,0xfa,0x04,0x02,0x02,0x01,0x1b,0x00,0xff]

238
llvm/test/MC/AMDGPU/gfx11_asm_dpp8.s → llvm/test/MC/AMDGPU/gfx11_asm_vop1_dpp8.s
View File

@ -1,5 +1,5 @@
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s // RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -show-encoding %s 2>&1 | FileCheck --check-prefixes=GFX11-ERR --implicit-check-not=error %s // RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
v_mov_b32 v5, v1 dpp8:[0,1,2,3,4,5,6,7] v_mov_b32 v5, v1 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x02,0x0a,0x7e,0x01,0x88,0xc6,0xfa] // GFX11: encoding: [0xe9,0x02,0x0a,0x7e,0x01,0x88,0xc6,0xfa]
@ -163,93 +163,6 @@ v_cvt_norm_i16_f16 v5, v1 dpp8:[0,1,2,3,4,5,6,7]
v_cvt_norm_u16_f16 v5, v1 dpp8:[0,1,2,3,4,5,6,7] v_cvt_norm_u16_f16 v5, v1 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0xc8,0x0a,0x7e,0x01,0x88,0xc6,0xfa] // GFX11: encoding: [0xe9,0xc8,0x0a,0x7e,0x01,0x88,0xc6,0xfa]
v_add_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x06,0x01,0x88,0xc6,0xfa]
v_sub_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x08,0x01,0x88,0xc6,0xfa]
v_subrev_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x0a,0x01,0x88,0xc6,0xfa]
v_mul_legacy_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x0e,0x01,0x88,0xc6,0xfa]
v_mul_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x10,0x01,0x88,0xc6,0xfa]
v_mul_i32_i24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x12,0x01,0x88,0xc6,0xfa]
v_mul_hi_i32_i24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x14,0x01,0x88,0xc6,0xfa]
v_mul_u32_u24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x16,0x01,0x88,0xc6,0xfa]
v_mul_hi_u32_u24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x18,0x01,0x88,0xc6,0xfa]
v_min_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x1e,0x01,0x88,0xc6,0xfa]
v_max_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x20,0x01,0x88,0xc6,0xfa]
v_min_i32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x22,0x01,0x88,0xc6,0xfa]
v_max_i32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x24,0x01,0x88,0xc6,0xfa]
v_min_u32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x26,0x01,0x88,0xc6,0xfa]
v_max_u32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x28,0x01,0x88,0xc6,0xfa]
v_lshrrev_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x32,0x01,0x88,0xc6,0xfa]
v_ashrrev_i32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x34,0x01,0x88,0xc6,0xfa]
v_lshlrev_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x30,0x01,0x88,0xc6,0xfa]
v_and_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x36,0x01,0x88,0xc6,0xfa]
v_or_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x38,0x01,0x88,0xc6,0xfa]
v_xor_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x3a,0x01,0x88,0xc6,0xfa]
v_xnor_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x3c,0x01,0x88,0xc6,0xfa]
v_add_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x64,0x01,0x88,0xc6,0xfa]
v_sub_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x66,0x01,0x88,0xc6,0xfa]
v_subrev_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x68,0x01,0x88,0xc6,0xfa]
v_mul_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x6a,0x01,0x88,0xc6,0xfa]
v_max_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x72,0x01,0x88,0xc6,0xfa]
v_min_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x74,0x01,0x88,0xc6,0xfa]
v_ldexp_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x76,0x01,0x88,0xc6,0xfa]
v_mov_b32 v5, v1 dpp8:[0,1,2,3,4,5,6,7] fi:0 v_mov_b32 v5, v1 dpp8:[0,1,2,3,4,5,6,7] fi:0
// GFX11: encoding: [0xe9,0x02,0x0a,0x7e,0x01,0x88,0xc6,0xfa] // GFX11: encoding: [0xe9,0x02,0x0a,0x7e,0x01,0x88,0xc6,0xfa]
@ -414,150 +327,3 @@ v_cvt_norm_i16_f16 v5, v1 dpp8:[0,1,2,3,4,5,6,7] fi:1
v_cvt_norm_u16_f16 v5, v1 dpp8:[0,1,2,3,4,5,6,7] fi:1 v_cvt_norm_u16_f16 v5, v1 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0xc8,0x0a,0x7e,0x01,0x88,0xc6,0xfa] // GFX11: encoding: [0xea,0xc8,0x0a,0x7e,0x01,0x88,0xc6,0xfa]
v_add_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x06,0x01,0x88,0xc6,0xfa]
v_sub_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x08,0x01,0x88,0xc6,0xfa]
v_subrev_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x0a,0x01,0x88,0xc6,0xfa]
v_mul_legacy_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x0e,0x01,0x88,0xc6,0xfa]
v_mul_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x10,0x01,0x88,0xc6,0xfa]
v_mul_i32_i24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x12,0x01,0x88,0xc6,0xfa]
v_mul_hi_i32_i24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x14,0x01,0x88,0xc6,0xfa]
v_mul_u32_u24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x16,0x01,0x88,0xc6,0xfa]
v_mul_hi_u32_u24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x18,0x01,0x88,0xc6,0xfa]
v_min_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x1e,0x01,0x88,0xc6,0xfa]
v_max_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x20,0x01,0x88,0xc6,0xfa]
v_min_i32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x22,0x01,0x88,0xc6,0xfa]
v_max_i32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x24,0x01,0x88,0xc6,0xfa]
v_min_u32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x26,0x01,0x88,0xc6,0xfa]
v_max_u32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x28,0x01,0x88,0xc6,0xfa]
v_lshrrev_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x32,0x01,0x88,0xc6,0xfa]
v_ashrrev_i32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x34,0x01,0x88,0xc6,0xfa]
v_lshlrev_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x30,0x01,0x88,0xc6,0xfa]
v_and_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x36,0x01,0x88,0xc6,0xfa]
v_or_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x38,0x01,0x88,0xc6,0xfa]
v_xor_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x3a,0x01,0x88,0xc6,0xfa]
v_xnor_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x3c,0x01,0x88,0xc6,0xfa]
v_add_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x64,0x01,0x88,0xc6,0xfa]
v_sub_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x66,0x01,0x88,0xc6,0xfa]
v_subrev_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x68,0x01,0x88,0xc6,0xfa]
v_mul_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x6a,0x01,0x88,0xc6,0xfa]
v_max_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x72,0x01,0x88,0xc6,0xfa]
v_min_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x74,0x01,0x88,0xc6,0xfa]
v_ldexp_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x76,0x01,0x88,0xc6,0xfa]
v_add_nc_u32 v5, v1, v255 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0xe9,0xfe,0x0b,0x4a,0x01,0x77,0x39,0x05]
v_add_nc_u32 v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x4a,0x01,0x77,0x39,0x05]
v_sub_nc_u32 v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0xe9,0x04,0x0a,0x4c,0x01,0x77,0x39,0x05]
v_sub_nc_u32 v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x4c,0x01,0x77,0x39,0x05]
v_subrev_nc_u32 v5, v1, v255 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0xe9,0xfe,0x0b,0x4e,0x01,0x77,0x39,0x05]
v_subrev_nc_u32 v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x4e,0x01,0x77,0x39,0x05]
v_dot2_f16_f16_e64_dpp v0, v1, v2, v3 dpp8:[0,1,2,3,4,4,4,4]
// GFX11: encoding: [0x00,0x00,0x66,0xd6,0xe9,0x04,0x0e,0x04,0x01,0x88,0x46,0x92]
v_dot2_f16_f16_e64_dpp v0, v1, v2, v3 op_sel:[1,1,0,0] dpp8:[0,1,2,3,4,4,4,4]
// GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid op_sel operand
v_dot2_f16_f16_e64_dpp v0, s1, v2, v3 dpp8:[0,1,2,3,4,4,4,4]
// GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
v_dot2_f16_f16_e64_dpp v0, v1, s2, v3 dpp8:[0,1,2,3,4,4,4,4]
// GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
v_dot2_f16_f16_e64_dpp v0, v1, v2, v3 op_sel:[0,0,1,1] dpp8:[0,1,2,3,4,4,4,4]
// GFX11: encoding: [0x00,0x60,0x66,0xd6,0xe9,0x04,0x0e,0x04,0x01,0x88,0x46,0x92]
v_dot2_f16_f16_e64_dpp v0, |v1|, -v2, -|s3| op_sel:[0,0,1,1] dpp8:[0,1,2,3,4,4,4,4]
// GFX11: encoding: [0x00,0x65,0x66,0xd6,0xe9,0x04,0x0e,0xc0,0x01,0x88,0x46,0x92]
v_dot2_f16_f16_e64_dpp v5, v1, v2, 0.5 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0x05,0x00,0x66,0xd6,0xe9,0x04,0xc2,0x03,0x01,0x77,0x39,0x05]
v_dot2_bf16_bf16_e64_dpp v0, v1, v2, v3 dpp8:[0,1,2,3,4,4,4,4]
// GFX11: encoding: [0x00,0x00,0x67,0xd6,0xe9,0x04,0x0e,0x04,0x01,0x88,0x46,0x92]
v_dot2_bf16_bf16_e64_dpp v0, v1, v2, v3 op_sel:[1,1,0,0] dpp8:[0,1,2,3,4,4,4,4]
// GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid op_sel operand
v_dot2_bf16_bf16_e64_dpp v0, s1, v2, v3 dpp8:[0,1,2,3,4,4,4,4]
// GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
v_dot2_bf16_bf16_e64_dpp v0, v1, s2, v3 dpp8:[0,1,2,3,4,4,4,4]
// GFX11-ERR: :[[@LINE-1]]:{{[0-9]+}}: error: invalid operand for instruction
v_dot2_bf16_bf16_e64_dpp v0, v1, v2, v3 op_sel:[0,0,1,1] dpp8:[0,1,2,3,4,4,4,4]
// GFX11: encoding: [0x00,0x60,0x67,0xd6,0xe9,0x04,0x0e,0x04,0x01,0x88,0x46,0x92]
v_dot2_bf16_bf16_e64_dpp v0, |v1|, -v2, -|s3| op_sel:[0,0,1,1] dpp8:[0,1,2,3,4,4,4,4]
// GFX11: encoding: [0x00,0x65,0x67,0xd6,0xe9,0x04,0x0e,0xc0,0x01,0x88,0x46,0x92]
v_dot2_bf16_bf16_e64_dpp v5, v1, v2, 0 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0x05,0x00,0x67,0xd6,0xe9,0x04,0x02,0x02,0x01,0x77,0x39,0x05]

256
llvm/test/MC/AMDGPU/gfx11_asm_vop2_dpp16.s Normal file
View File

@ -0,0 +1,256 @@
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11,W32 %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11,W64 %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 %s 2>&1 | FileCheck --check-prefix=W32-ERR --implicit-check-not=error: %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 %s 2>&1 | FileCheck --check-prefix=W64-ERR --implicit-check-not=error: %s
v_dot2c_f32_f16_dpp v5, v1, v2 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x04,0x01,0xe4,0x00,0x00]
v_dot2c_f32_f16_dpp v5, v1, v2 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x04,0x01,0xe4,0x04,0x00]
v_dot2acc_f32_f16_dpp v5, v1, v2 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x04,0x01,0xe4,0x00,0x00]
v_dot2acc_f32_f16_dpp v5, v1, v2 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x04,0x01,0xe4,0x04,0x00]
v_add_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x06,0x01,0x1b,0x00,0x00]
v_add_f32 v5, -v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x06,0x01,0x1b,0x10,0x00]
v_add_f32 v5, v1, -v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x06,0x01,0x1b,0x40,0x00]
v_sub_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x08,0x01,0x1b,0x00,0x00]
v_subrev_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x0a,0x01,0x1b,0x00,0x00]
v_mul_legacy_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x0e,0x01,0x1b,0x00,0x00]
v_mul_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x10,0x01,0x1b,0x00,0x00]
v_mul_i32_i24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x12,0x01,0x1b,0x00,0x00]
v_mul_hi_i32_i24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x14,0x01,0x1b,0x00,0x00]
v_mul_u32_u24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x16,0x01,0x1b,0x00,0x00]
v_mul_hi_u32_u24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x18,0x01,0x1b,0x00,0x00]
v_min_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x1e,0x01,0x1b,0x00,0x00]
v_max_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x20,0x01,0x1b,0x00,0x00]
v_min_i32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x22,0x01,0x1b,0x00,0x00]
v_max_i32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x24,0x01,0x1b,0x00,0x00]
v_min_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x26,0x01,0x1b,0x00,0x00]
v_max_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x28,0x01,0x1b,0x00,0x00]
v_lshrrev_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x32,0x01,0x1b,0x00,0x00]
v_ashrrev_i32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x34,0x01,0x1b,0x00,0x00]
v_lshlrev_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x30,0x01,0x1b,0x00,0x00]
v_and_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x36,0x01,0x1b,0x00,0x00]
v_or_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x38,0x01,0x1b,0x00,0x00]
v_xor_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x3a,0x01,0x1b,0x00,0x00]
v_xnor_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x3c,0x01,0x1b,0x00,0x00]
v_add_co_ci_u32 v5, vcc_lo, v1, v2, vcc_lo quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// W32: encoding: [0xfa,0x04,0x0a,0x40,0x01,0x1b,0x00,0x00]
// W64-ERR: error: operands are not valid for this GPU or mode
v_add_co_ci_u32 v5, vcc, v1, v2, vcc quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// W64: encoding: [0xfa,0x04,0x0a,0x40,0x01,0x1b,0x00,0x00]
// W32-ERR: error: operands are not valid for this GPU or mode
v_sub_co_ci_u32 v5, vcc_lo, v1, v2, vcc_lo quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// W32: encoding: [0xfa,0x04,0x0a,0x42,0x01,0x1b,0x00,0x00]
// W64-ERR: error: operands are not valid for this GPU or mode
v_sub_co_ci_u32 v5, vcc, v1, v2, vcc quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// W64: encoding: [0xfa,0x04,0x0a,0x42,0x01,0x1b,0x00,0x00]
// W32-ERR: error: operands are not valid for this GPU or mode
v_subrev_co_ci_u32 v5, vcc_lo, v1, v2, vcc_lo quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// W32: encoding: [0xfa,0x04,0x0a,0x44,0x01,0x1b,0x00,0x00]
// W64-ERR: error: operands are not valid for this GPU or mode
v_subrev_co_ci_u32 v5, vcc, v1, v2, vcc quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// W64: encoding: [0xfa,0x04,0x0a,0x44,0x01,0x1b,0x00,0x00]
// W32-ERR: error: operands are not valid for this GPU or mode
v_fmac_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x56,0x01,0x1b,0x00,0x00]
v_add_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x64,0x01,0x1b,0x00,0x00]
v_sub_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x66,0x01,0x1b,0x00,0x00]
v_subrev_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x68,0x01,0x1b,0x00,0x00]
v_mul_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x6a,0x01,0x1b,0x00,0x00]
v_fmac_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x6c,0x01,0x1b,0x00,0x00]
v_max_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x72,0x01,0x1b,0x00,0x00]
v_min_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x74,0x01,0x1b,0x00,0x00]
v_ldexp_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x76,0x01,0x1b,0x00,0x00]
v_add_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x06,0x01,0x1b,0x04,0x00]
v_sub_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x08,0x01,0x1b,0x04,0x00]
v_subrev_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x0a,0x01,0x1b,0x04,0x00]
v_mul_legacy_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x0e,0x01,0x1b,0x04,0x00]
v_mul_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x10,0x01,0x1b,0x04,0x00]
v_mul_i32_i24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x12,0x01,0x1b,0x04,0x00]
v_mul_hi_i32_i24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x14,0x01,0x1b,0x04,0x00]
v_mul_u32_u24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x16,0x01,0x1b,0x04,0x00]
v_mul_hi_u32_u24 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x18,0x01,0x1b,0x04,0x00]
v_min_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x1e,0x01,0x1b,0x04,0x00]
v_max_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x20,0x01,0x1b,0x04,0x00]
v_min_i32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x22,0x01,0x1b,0x04,0x00]
v_max_i32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x24,0x01,0x1b,0x04,0x00]
v_min_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x26,0x01,0x1b,0x04,0x00]
v_max_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x28,0x01,0x1b,0x04,0x00]
v_lshrrev_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x32,0x01,0x1b,0x04,0x00]
v_ashrrev_i32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x34,0x01,0x1b,0x04,0x00]
v_lshlrev_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x30,0x01,0x1b,0x04,0x00]
v_and_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x36,0x01,0x1b,0x04,0x00]
v_or_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x38,0x01,0x1b,0x04,0x00]
v_xor_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x3a,0x01,0x1b,0x04,0x00]
v_xnor_b32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x3c,0x01,0x1b,0x04,0x00]
v_fmac_f32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x56,0x01,0x1b,0x04,0x00]
v_add_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x64,0x01,0x1b,0x04,0x00]
v_sub_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x66,0x01,0x1b,0x04,0x00]
v_subrev_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x68,0x01,0x1b,0x04,0x00]
v_mul_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x6a,0x01,0x1b,0x04,0x00]
v_fmac_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x6c,0x01,0x1b,0x04,0x00]
v_max_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x72,0x01,0x1b,0x04,0x00]
v_min_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x74,0x01,0x1b,0x04,0x00]
v_ldexp_f16 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x76,0x01,0x1b,0x04,0x00]
v_add_nc_u32 v5, v1, v2 quad_perm:[0,1,2,3] row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x4a,0x01,0xe4,0x00,0x00]
v_add_nc_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x4a,0x01,0x1b,0x04,0x00]
v_sub_nc_u32 v5, v1, v2 row_mirror row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x4c,0x01,0x40,0x01,0x00]
v_sub_nc_u32 v5, v1, v2 row_half_mirror row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x4c,0x01,0x41,0x01,0x00]
v_sub_nc_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x4c,0x01,0x1b,0x04,0x00]
v_subrev_nc_u32 v5, v1, v2 row_xmask:15 row_mask:0x0 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x4e,0x01,0x6f,0x01,0x00]
v_subrev_nc_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x1 bank_mask:0x0
// GFX11: encoding: [0xfa,0x04,0x0a,0x4e,0x01,0x1b,0x00,0x10]
v_subrev_nc_u32 v5, v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x0a,0x4e,0x01,0x1b,0x04,0x00]

206
llvm/test/MC/AMDGPU/gfx11_asm_vop2_dpp8.s Normal file
View File

@ -0,0 +1,206 @@
// RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
// RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
v_dot2c_f32_f16_dpp v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0xe9,0x04,0x0a,0x04,0x01,0x77,0x39,0x05]
v_dot2c_f32_f16_dpp v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x04,0x01,0x77,0x39,0x05]
v_dot2acc_f32_f16_dpp v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0xe9,0x04,0x0a,0x04,0x01,0x77,0x39,0x05]
v_dot2acc_f32_f16_dpp v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x04,0x01,0x77,0x39,0x05]
v_add_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x06,0x01,0x88,0xc6,0xfa]
v_sub_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x08,0x01,0x88,0xc6,0xfa]
v_subrev_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x0a,0x01,0x88,0xc6,0xfa]
v_mul_legacy_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x0e,0x01,0x88,0xc6,0xfa]
v_mul_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x10,0x01,0x88,0xc6,0xfa]
v_mul_i32_i24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x12,0x01,0x88,0xc6,0xfa]
v_mul_hi_i32_i24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x14,0x01,0x88,0xc6,0xfa]
v_mul_u32_u24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x16,0x01,0x88,0xc6,0xfa]
v_mul_hi_u32_u24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x18,0x01,0x88,0xc6,0xfa]
v_min_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x1e,0x01,0x88,0xc6,0xfa]
v_max_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x20,0x01,0x88,0xc6,0xfa]
v_min_i32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x22,0x01,0x88,0xc6,0xfa]
v_max_i32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x24,0x01,0x88,0xc6,0xfa]
v_min_u32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x26,0x01,0x88,0xc6,0xfa]
v_max_u32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x28,0x01,0x88,0xc6,0xfa]
v_lshrrev_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x32,0x01,0x88,0xc6,0xfa]
v_ashrrev_i32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x34,0x01,0x88,0xc6,0xfa]
v_lshlrev_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x30,0x01,0x88,0xc6,0xfa]
v_and_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x36,0x01,0x88,0xc6,0xfa]
v_or_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x38,0x01,0x88,0xc6,0xfa]
v_xor_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x3a,0x01,0x88,0xc6,0xfa]
v_xnor_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x3c,0x01,0x88,0xc6,0xfa]
v_add_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x64,0x01,0x88,0xc6,0xfa]
v_sub_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x66,0x01,0x88,0xc6,0xfa]
v_subrev_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x68,0x01,0x88,0xc6,0xfa]
v_mul_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x6a,0x01,0x88,0xc6,0xfa]
v_max_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x72,0x01,0x88,0xc6,0xfa]
v_min_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x74,0x01,0x88,0xc6,0xfa]
v_ldexp_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7]
// GFX11: encoding: [0xe9,0x04,0x0a,0x76,0x01,0x88,0xc6,0xfa]
v_add_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x06,0x01,0x88,0xc6,0xfa]
v_sub_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x08,0x01,0x88,0xc6,0xfa]
v_subrev_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x0a,0x01,0x88,0xc6,0xfa]
v_mul_legacy_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x0e,0x01,0x88,0xc6,0xfa]
v_mul_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x10,0x01,0x88,0xc6,0xfa]
v_mul_i32_i24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x12,0x01,0x88,0xc6,0xfa]
v_mul_hi_i32_i24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x14,0x01,0x88,0xc6,0xfa]
v_mul_u32_u24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x16,0x01,0x88,0xc6,0xfa]
v_mul_hi_u32_u24 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x18,0x01,0x88,0xc6,0xfa]
v_min_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x1e,0x01,0x88,0xc6,0xfa]
v_max_f32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x20,0x01,0x88,0xc6,0xfa]
v_min_i32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x22,0x01,0x88,0xc6,0xfa]
v_max_i32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x24,0x01,0x88,0xc6,0xfa]
v_min_u32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x26,0x01,0x88,0xc6,0xfa]
v_max_u32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x28,0x01,0x88,0xc6,0xfa]
v_lshrrev_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x32,0x01,0x88,0xc6,0xfa]
v_ashrrev_i32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x34,0x01,0x88,0xc6,0xfa]
v_lshlrev_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x30,0x01,0x88,0xc6,0xfa]
v_and_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x36,0x01,0x88,0xc6,0xfa]
v_or_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x38,0x01,0x88,0xc6,0xfa]
v_xor_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x3a,0x01,0x88,0xc6,0xfa]
v_xnor_b32 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x3c,0x01,0x88,0xc6,0xfa]
v_add_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x64,0x01,0x88,0xc6,0xfa]
v_sub_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x66,0x01,0x88,0xc6,0xfa]
v_subrev_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x68,0x01,0x88,0xc6,0xfa]
v_mul_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x6a,0x01,0x88,0xc6,0xfa]
v_max_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x72,0x01,0x88,0xc6,0xfa]
v_min_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x74,0x01,0x88,0xc6,0xfa]
v_ldexp_f16 v5, v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x76,0x01,0x88,0xc6,0xfa]
v_add_nc_u32 v5, v1, v255 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0xe9,0xfe,0x0b,0x4a,0x01,0x77,0x39,0x05]
v_add_nc_u32 v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x4a,0x01,0x77,0x39,0x05]
v_sub_nc_u32 v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0xe9,0x04,0x0a,0x4c,0x01,0x77,0x39,0x05]
v_sub_nc_u32 v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x4c,0x01,0x77,0x39,0x05]
v_subrev_nc_u32 v5, v1, v255 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0xe9,0xfe,0x0b,0x4e,0x01,0x77,0x39,0x05]
v_subrev_nc_u32 v5, v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: encoding: [0xea,0x04,0x0a,0x4e,0x01,0x77,0x39,0x05]

4
llvm/test/MC/AMDGPU/gfx11_asm_vop3.s
View File

@ -11790,10 +11790,6 @@ v_cndmask_b16 v5, v1, v2, vcc
// W64: encoding: [0x05,0x00,0x5d,0xd6,0x01,0x05,0xaa,0x01] // W64: encoding: [0x05,0x00,0x5d,0xd6,0x01,0x05,0xaa,0x01]
// W32-ERR: error: invalid operand for instruction // W32-ERR: error: invalid operand for instruction
v_cndmask_b16 v5, v1, v2, vcc
// W64: encoding: [0x05,0x00,0x5d,0xd6,0x01,0x05,0xaa,0x01]
// W32-ERR: error: invalid operand for instruction
v_cndmask_b16 v5, -v1, v2, s[4:5] v_cndmask_b16 v5, -v1, v2, s[4:5]
// W64: encoding: [0x05,0x00,0x5d,0xd6,0x01,0x05,0x12,0x20] // W64: encoding: [0x05,0x00,0x5d,0xd6,0x01,0x05,0x12,0x20]
// W32-ERR: error: invalid operand for instruction // W32-ERR: error: invalid operand for instruction

9582
llvm/test/MC/AMDGPU/gfx11_asm_vop3_dpp16.s
View File

File diff suppressed because it is too large Load Diff

3220
llvm/test/MC/AMDGPU/gfx11_asm_vop3_dpp8.s
View File

File diff suppressed because it is too large Load Diff

568
llvm/test/MC/AMDGPU/gfx11_asm_vopc_e64.s → llvm/test/MC/AMDGPU/gfx11_asm_vop3c.s
View File

File diff suppressed because it is too large Load Diff

6852
llvm/test/MC/AMDGPU/gfx11_asm_vop3c_dpp16.s Normal file
View File

File diff suppressed because it is too large Load Diff

2500
llvm/test/MC/AMDGPU/gfx11_asm_vop3c_dpp8.s Normal file
View File

File diff suppressed because it is too large Load Diff

3233
llvm/test/MC/AMDGPU/gfx11_asm_vop3cx.s Normal file
View File

File diff suppressed because it is too large Load Diff

2691
llvm/test/MC/AMDGPU/gfx11_asm_vop3cx_dpp16.s Normal file
View File

File diff suppressed because it is too large Load Diff

680
llvm/test/MC/AMDGPU/gfx11_asm_vop3cx_dpp8.s Normal file
View File

@ -0,0 +1,680 @@
// RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
// RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
v_cmpx_class_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xfd,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_class_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xfd,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_class_f16_e64_dpp -|v255|, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x01,0xfd,0xd4,0xe9,0xfe,0x03,0x20,0xff,0x00,0x00,0x00]
v_cmpx_class_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xfe,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_class_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xfe,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_class_f32_e64_dpp -|v255|, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x01,0xfe,0xd4,0xe9,0xfe,0x03,0x20,0xff,0x00,0x00,0x00]
v_cmpx_eq_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x82,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_eq_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x82,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_eq_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x82,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_eq_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x82,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_eq_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x92,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_eq_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x92,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_eq_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x92,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_eq_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x92,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_eq_i16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xb2,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_eq_i16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xb2,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_eq_i16_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xb2,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_eq_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xc2,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_eq_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xc2,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_eq_i32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xc2,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_eq_u16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xba,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_eq_u16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xba,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_eq_u16_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xba,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_eq_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xca,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_eq_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xca,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_eq_u32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xca,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_f_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x80,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_f_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x80,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_f_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x80,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_f_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x80,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_f_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x90,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_f_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x90,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_f_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x90,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_f_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x90,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_f_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xc0,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_f_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xc0,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_f_i32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xc0,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_f_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xc8,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_f_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xc8,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_f_u32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xc8,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_ge_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x86,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ge_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x86,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_ge_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x86,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_ge_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x86,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_ge_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x96,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ge_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x96,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_ge_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x96,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_ge_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x96,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_ge_i16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xb6,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ge_i16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xb6,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ge_i16_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xb6,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_ge_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xc6,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ge_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xc6,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ge_i32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xc6,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_ge_u16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xbe,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ge_u16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xbe,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ge_u16_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xbe,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_ge_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xce,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ge_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xce,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ge_u32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xce,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_gt_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x84,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_gt_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x84,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_gt_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x84,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_gt_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x84,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_gt_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x94,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_gt_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x94,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_gt_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x94,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_gt_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x94,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_gt_i16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xb4,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_gt_i16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xb4,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_gt_i16_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xb4,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_gt_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xc4,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_gt_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xc4,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_gt_i32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xc4,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_gt_u16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xbc,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_gt_u16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xbc,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_gt_u16_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xbc,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_gt_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xcc,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_gt_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xcc,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_gt_u32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xcc,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_le_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x83,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_le_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x83,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_le_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x83,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_le_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x83,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_le_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x93,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_le_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x93,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_le_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x93,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_le_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x93,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_le_i16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xb3,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_le_i16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xb3,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_le_i16_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xb3,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_le_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xc3,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_le_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xc3,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_le_i32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xc3,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_le_u16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xbb,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_le_u16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xbb,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_le_u16_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xbb,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_le_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xcb,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_le_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xcb,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_le_u32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xcb,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_lg_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x85,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_lg_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x85,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_lg_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x85,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_lg_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x85,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_lg_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x95,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_lg_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x95,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_lg_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x95,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_lg_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x95,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_lt_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x81,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_lt_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x81,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_lt_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x81,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_lt_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x81,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_lt_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x91,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_lt_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x91,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_lt_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x91,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_lt_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x91,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_lt_i16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xb1,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_lt_i16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xb1,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_lt_i16_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xb1,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_lt_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xc1,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_lt_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xc1,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_lt_i32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xc1,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_lt_u16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xb9,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_lt_u16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xb9,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_lt_u16_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xb9,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_lt_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xc9,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_lt_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xc9,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_lt_u32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xc9,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_ne_i16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xb5,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ne_i16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xb5,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ne_i16_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xb5,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_ne_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xc5,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ne_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xc5,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ne_i32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xc5,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_ne_u16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xbd,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ne_u16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xbd,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ne_u16_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xbd,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_ne_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xcd,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ne_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xcd,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ne_u32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xcd,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_neq_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x8d,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_neq_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x8d,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_neq_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x8d,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_neq_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x8d,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_neq_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x9d,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_neq_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x9d,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_neq_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x9d,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_neq_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x9d,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_nge_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x89,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_nge_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x89,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_nge_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x89,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_nge_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x89,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_nge_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x99,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_nge_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x99,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_nge_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x99,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_nge_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x99,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_ngt_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x8b,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ngt_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x8b,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_ngt_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x8b,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_ngt_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x8b,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_ngt_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x9b,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_ngt_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x9b,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_ngt_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x9b,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_ngt_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x9b,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_nle_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x8c,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_nle_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x8c,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_nle_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x8c,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_nle_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x8c,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_nle_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x9c,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_nle_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x9c,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_nle_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x9c,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_nle_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x9c,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_nlg_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x8a,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_nlg_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x8a,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_nlg_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x8a,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_nlg_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x8a,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_nlg_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x9a,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_nlg_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x9a,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_nlg_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x9a,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_nlg_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x9a,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_nlt_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x8e,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_nlt_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x8e,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_nlt_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x8e,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_nlt_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x8e,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_nlt_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x9e,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_nlt_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x9e,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_nlt_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x9e,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_nlt_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x9e,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_o_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x87,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_o_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x87,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_o_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x87,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_o_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x87,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_o_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x97,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_o_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x97,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_o_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x97,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_o_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x97,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_t_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x8f,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_t_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x8f,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_t_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x8f,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_t_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x8f,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_t_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x9f,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_t_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x9f,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_t_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x9f,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_t_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x9f,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_t_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xc7,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_t_i32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xc7,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_t_i32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xc7,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_t_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0xcf,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_t_u32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x00,0xcf,0xd4,0xea,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_t_u32_e64_dpp v255, v255 dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x00,0xcf,0xd4,0xe9,0xfe,0x03,0x00,0xff,0x00,0x00,0x00]
v_cmpx_tru_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x8f,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_tru_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x8f,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_tru_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x8f,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_tru_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x8f,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_tru_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x9f,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_tru_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x9f,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_tru_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x9f,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_tru_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x9f,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_u_f16_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x88,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_u_f16_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x88,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_u_f16_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x88,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_u_f16_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x88,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]
v_cmpx_u_f32_e64_dpp v1, v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x00,0x98,0xd4,0xe9,0x04,0x02,0x00,0x01,0x77,0x39,0x05]
v_cmpx_u_f32_e64_dpp |v1|, -v2 dpp8:[7,6,5,4,3,2,1,0]
// GFX11: [0x7e,0x01,0x98,0xd4,0xe9,0x04,0x02,0x40,0x01,0x77,0x39,0x05]
v_cmpx_u_f32_e64_dpp -v1, |v2| dpp8:[7,6,5,4,3,2,1,0] fi:1
// GFX11: [0x7e,0x02,0x98,0xd4,0xea,0x04,0x02,0x20,0x01,0x77,0x39,0x05]
v_cmpx_u_f32_e64_dpp -|v255|, -|v255| clamp dpp8:[0,0,0,0,0,0,0,0] fi:0
// GFX11: [0x7e,0x83,0x98,0xd4,0xe9,0xfe,0x03,0x60,0xff,0x00,0x00,0x00]

14
llvm/test/MC/AMDGPU/gfx11_asm_vop3p_dpp16.s Normal file
View File

@ -0,0 +1,14 @@
// RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
// RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
v_dot2_f32_f16 v0, v1, v2, v3 neg_lo:[0,0,0] neg_hi:[0,0,0] quad_perm:[2,2,3,1] bound_ctrl:0 fi:1
// GFX11: v_dot2_f32_f16_e64_dpp v0, v1, v2, v3 quad_perm:[2,2,3,1] row_mask:0xf bank_mask:0xf bound_ctrl:1 fi:1 ; encoding: [0x00,0x00,0x13,0xcc,0xfa,0x04,0x0e,0x04,0x01,0x7a,0x0c,0xff]
v_dot2_f32_f16 v0, v1, v2, v3 neg_lo:[1,1,0] neg_hi:[1,0,1] quad_perm:[3,2,1,0] bank_mask:0xe
// GFX11: v_dot2_f32_f16_e64_dpp v0, v1, v2, v3 neg_lo:[1,1,0] neg_hi:[1,0,1] quad_perm:[3,2,1,0] row_mask:0xf bank_mask:0xe ; encoding: [0x00,0x05,0x13,0xcc,0xfa,0x04,0x0e,0x64,0x01,0x1b,0x00,0xfe]
v_fma_mix_f32 v0, v1, v2, v3 op_sel:[0,0,0] row_ror:7 bank_mask:0x1 bound_ctrl:0
// GFX11: v_fma_mix_f32_e64_dpp v0, v1, v2, v3 row_ror:7 row_mask:0xf bank_mask:0x1 bound_ctrl:1 ; encoding: [0x00,0x00,0x20,0xcc,0xfa,0x04,0x0e,0x04,0x01,0x27,0x09,0xf1]
v_fma_mixhi_f16 v0, v1, v2, v3 op_sel_hi:[1,1,1] clamp quad_perm:[0,2,3,1] row_mask:0x0
// GFX11: v_fma_mixhi_f16_e64_dpp v0, v1, v2, v3 op_sel_hi:[1,1,1] clamp quad_perm:[0,2,3,1] row_mask:0x0 bank_mask:0xf ; encoding: [0x00,0xc0,0x22,0xcc,0xfa,0x04,0x0e,0x1c,0x01,0x78,0x00,0x0f]

18
llvm/test/MC/AMDGPU/gfx11_asm_vop3p_dpp8.s Normal file
View File

@ -0,0 +1,18 @@
// RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
// RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
v_fma_mix_f32 v0, v1, v2, v3 dpp8:[2,2,2,2,4,4,4,4]
// GFX11: encoding: [0x00,0x00,0x20,0xcc,0xe9,0x04,0x0e,0x04,0x01,0x92,0x44,0x92]
v_fma_mix_f32 v0, v1, v2, v3 clamp dpp8:[2,2,2,2,4,4,4,4] fi:1
// GFX11: encoding: [0x00,0x80,0x20,0xcc,0xea,0x04,0x0e,0x04,0x01,0x92,0x44,0x92]
v_fma_mixlo_f16 v0, abs(v1), -v2, abs(v3) dpp8:[2,2,2,2,4,4,4,4]
// GFX11: encoding: [0x00,0x05,0x21,0xcc,0xe9,0x04,0x0e,0x44,0x01,0x92,0x44,0x92]
// For test purpose only. OP_SEL has to be set to all 0 and OP_SEL_HI has to be set to all 1
v_fma_mixlo_f16 v0, abs(v1), -v2, abs(v3) op_sel:[1,0,0] op_sel_hi:[1,0,0] dpp8:[2,2,2,2,4,4,4,4]
// GFX11: encoding: [0x00,0x0d,0x21,0xcc,0xe9,0x04,0x0e,0x4c,0x01,0x92,0x44,0x92]
v_dot2_f32_f16_e64_dpp v0, v1, v2, v3 neg_lo:[0,1,1] neg_hi:[1,0,1] dpp8:[7,6,5,4,3,2,1,0]
// GFX11: encoding: [0x00,0x05,0x13,0xcc,0xe9,0x04,0x0e,0xc4,0x01,0x77,0x39,0x05]

233
llvm/test/MC/AMDGPU/gfx11_asm_vopc_dpp.s
View File

@ -1,233 +0,0 @@
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11,W32 %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11,W64 %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 %s 2>&1 | FileCheck --check-prefix=W32-ERR --implicit-check-not=error: %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 %s 2>&1 | FileCheck --check-prefix=W64-ERR --implicit-check-not=error: %s
// =========================================================
// dpp8
// =========================================================
// 32 bit
v_cmp_le_u16_dpp v1, v2 dpp8:[7,7,7,3,4,4,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x76,0x7c,0x01,0xff,0x47,0xfa]
v_cmp_le_i16_dpp v1, v2 dpp8:[7,7,7,3,4,4,6,7]
// GFX11: encoding: [0xe9,0x04,0x66,0x7c,0x01,0xff,0x47,0xfa]
// w32
v_cmp_le_i32_dpp vcc_lo, v1, v255 dpp8:[0,2,1,3,4,5,6,7]
// W32: encoding: [0xe9,0xfe,0x87,0x7c,0x01,0x50,0xc6,0xfa]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_cmp_tru_f32_dpp vcc_lo, v1, v2 dpp8:[0,2,1,3,4,5,6,7]
// W32: encoding: [0xe9,0x04,0x3e,0x7c,0x01,0x50,0xc6,0xfa]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// check vcc/vcc_lo have been added
v_cmp_lt_f32_dpp v1, v2 dpp8:[2,3,4,1,3,3,3,3]
// W32: v_cmp_lt_f32 vcc_lo, v1, v2 dpp8:[2,3,4,1,3,3,3,3] ; encoding: [0xe9,0x04,0x22,0x7c,0x01,0x1a,0xb3,0x6d]
// W64: v_cmp_lt_f32 vcc, v1, v2 dpp8:[2,3,4,1,3,3,3,3] ; encoding: [0xe9,0x04,0x22,0x7c,0x01,0x1a,0xb3,0x6d]
// w64
v_cmp_lt_u16_dpp vcc, v1, v2 dpp8:[7,6,5,3,4,2,1,0] fi:1
// W64: encoding: [0xea,0x04,0x72,0x7c,0x01,0x77,0x47,0x05]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// cmp class
v_cmp_class_f16_dpp vcc, v1, v2 dpp8:[7,6,5,3,4,2,1,0] fi:1
// W64: encoding: [0xea,0x04,0xfa,0x7c,0x01,0x77,0x47,0x05]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// cmpx
v_cmpx_f_i32_dpp v0, v2 dpp8:[7,6,5,3,4,2,1,0] fi:1
// GFX11: encoding: [0xea,0x04,0x80,0x7d,0x00,0x77,0x47,0x05]
v_cmpx_t_f32_dpp v255, v2 dpp8:[7,6,5,3,4,2,1,0]
// GFX11: encoding: [0xe9,0x04,0x3e,0x7d,0xff,0x77,0x47,0x05]
// cmpx class
v_cmpx_class_f16_dpp v12, v101 dpp8:[7,6,5,3,4,2,1,0]
// GFX11: encoding: [0xe9,0xca,0xfa,0x7d,0x0c,0x77,0x47,0x05]
// =========================================================
// 64 bit
v_cmp_lt_i16_e64_dpp s[50:51], v1, v22 dpp8:[7,6,5,3,4,2,1,0]
// W64: encoding: [0x32,0x00,0x31,0xd4,0xe9,0x2c,0x02,0x00,0x01,0x77,0x47,0x05]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_cmp_lt_i16_e64_dpp s10, v10, v2 dpp8:[7,6,5,3,4,2,1,0]
// W32: encoding: [0x0a,0x00,0x31,0xd4,0xe9,0x04,0x02,0x00,0x0a,0x77,0x47,0x05]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// w32
v_cmp_gt_i32_e64_dpp s10, v1, v50 dpp8:[0,1,2,3,4,5,6,7] fi:1
// W32: encoding: [0x0a,0x00,0x44,0xd4,0xea,0x64,0x02,0x00,0x01,0x88,0xc6,0xfa]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_cmp_ngt_f32_e64_dpp s10, -v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// W32: encoding: [0x0a,0x00,0x1b,0xd4,0xea,0x04,0x02,0x20,0x01,0x88,0xc6,0xfa]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// w64
v_cmp_f_f32_e64_dpp s[10:11], v1, v2 dpp8:[0,1,2,3,4,5,6,7] fi:1
// W64: encoding: [0x0a,0x00,0x10,0xd4,0xea,0x04,0x02,0x00,0x01,0x88,0xc6,0xfa]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_cmp_nle_f32_e64_dpp s[10:11], v2, v201 dpp8:[0,1,6,3,4,5,6,7]
// W64: encoding: [0x0a,0x00,0x1c,0xd4,0xe9,0x92,0x03,0x00,0x02,0x88,0xc7,0xfa]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_cmp_lt_f16_e64_dpp s[10:11], v240, -v2 dpp8:[0,1,6,3,4,5,6,7]
// W64: encoding: [0x0a,0x00,0x01,0xd4,0xe9,0x04,0x02,0x40,0xf0,0x88,0xc7,0xfa]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// cmp class
// w32
v_cmp_class_f32_e64_dpp s10, v10, v2 dpp8:[0,1,6,3,4,5,6,7]
// W32: encoding: [0x0a,0x00,0x7e,0xd4,0xe9,0x04,0x02,0x00,0x0a,0x88,0xc7,0xfa]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// w64
v_cmp_class_f16_e64_dpp vcc, v1, v2 dpp8:[0,1,6,3,4,5,6,7]
// W64: encoding: [0x6a,0x00,0x7d,0xd4,0xe9,0x04,0x02,0x00,0x01,0x88,0xc7,0xfa]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// cmpx
v_cmpx_eq_i16_e64_dpp v5, v25 dpp8:[0,1,6,3,4,5,6,7] fi:1
// GFX11: encoding: [0x7e,0x00,0xb2,0xd4,0xea,0x32,0x02,0x00,0x05,0x88,0xc7,0xfa]
v_cmpx_ge_i32_e64_dpp v0, v3 dpp8:[0,1,6,3,4,5,6,7] fi:1
// GFX11: encoding: [0x7e,0x00,0xc6,0xd4,0xea,0x06,0x02,0x00,0x00,0x88,0xc7,0xfa]
// cmpx class
v_cmpx_class_f32_e64_dpp v1, v20 dpp8:[0,1,6,3,4,5,6,7] fi:1
// GFX11: encoding: [0x7e,0x00,0xfe,0xd4,0xea,0x28,0x02,0x00,0x01,0x88,0xc7,0xfa]
// ================================================================
// dpp
// ================================================================
// 32 bit
v_cmp_gt_u16_dpp v1, v2 row_shl:0x7 row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x78,0x7c,0x01,0x07,0x05,0x00]
v_cmp_gt_i16_dpp v1, v2 quad_perm:[1,3,1,0] row_mask:0x7
// GFX11: encoding: [0xfa,0x04,0x68,0x7c,0x01,0x1d,0x00,0x7f]
v_cmp_lt_f32 v1, -v2 quad_perm:[0,1,2,2]
// GFX11: encoding: [0xfa,0x04,0x22,0x7c,0x01,0xa4,0x40,0xff]
// w32
v_cmp_gt_i32_dpp vcc_lo, v1, v255 row_mirror bank_mask:0x2 fi:1
// W32: encoding: [0xfa,0xfe,0x89,0x7c,0x01,0x40,0x05,0xf2]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_cmp_f_f32_dpp vcc_lo, v1, v2 row_shl:0x7 row_mask:0x0 bank_mask:0x0
// W32: encoding: [0xfa,0x04,0x20,0x7c,0x01,0x07,0x01,0x00]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// w64
v_cmp_ge_u16_dpp vcc, v1, v2 row_share:0xa bound_ctrl:0
// W64: encoding: [0xfa,0x04,0x7c,0x7c,0x01,0x5a,0x09,0xff]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// cmp class
v_cmp_class_f16_dpp vcc, v1, v2 row_half_mirror bound_ctrl:0
// W64: encoding: [0xfa,0x04,0xfa,0x7c,0x01,0x41,0x09,0xff]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// cmpx
v_cmpx_t_i32_dpp v0, v2 row_shr:0xe row_mask:0x3 bank_mask:0xa bound_ctrl:0
// GFX11: encoding: [0xfa,0x04,0x8e,0x7d,0x00,0x1e,0x09,0x3a]
v_cmpx_f_f32_dpp v255, v2 quad_perm:[2,3,0,0]
// GFX11: encoding: [0xfa,0x04,0x20,0x7d,0xff,0x0e,0x00,0xff]
// cmpx class
v_cmpx_class_f16_dpp v12, v101 quad_perm:[2,3,0,0] fi:1
// GFX11: encoding: [0xfa,0xca,0xfa,0x7d,0x0c,0x0e,0x04,0xff]
v_cmpx_class_f16_dpp abs(v12), v101 quad_perm:[2,3,0,0]
// GFX11: encoding: [0xfa,0xca,0xfa,0x7d,0x0c,0x0e,0x20,0xff]
// =========================================================
// 64 bit
v_cmp_ge_i16_e64_dpp s[50:51], v1, v22 quad_perm:[2,2,3,1]
// W64: encoding: [0x32,0x00,0x36,0xd4,0xfa,0x2c,0x02,0x00,0x01,0x7a,0x00,0xff]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_cmp_ge_i16_e64_dpp s10, v10, v2 quad_perm:[0,2,3,1] row_mask:0x0
// W32: encoding: [0x0a,0x00,0x36,0xd4,0xfa,0x04,0x02,0x00,0x0a,0x78,0x00,0x0f]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// w32
v_cmp_le_i32_e64_dpp s10, v1, v50 quad_perm:[0,2,3,1] row_mask:0x0
// W32: encoding: [0x0a,0x00,0x43,0xd4,0xfa,0x64,0x02,0x00,0x01,0x78,0x00,0x0f]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_cmp_nle_f32_e64_dpp s10, -v1, v2 quad_perm:[3,2,1,0] row_mask:0x0 bank_mask:0x0
// W32: encoding: [0x0a,0x00,0x1c,0xd4,0xfa,0x04,0x02,0x20,0x01,0x1b,0x00,0x00]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// w64
v_cmp_t_f32_e64_dpp s[10:11], v1, v2 quad_perm:[3,2,1,0] bank_mask:0xe
// W64: encoding: [0x0a,0x00,0x1f,0xd4,0xfa,0x04,0x02,0x00,0x01,0x1b,0x00,0xfe]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_cmp_ngt_f32_e64_dpp s[10:11], v2, v201 row_ror:7 bank_mask:0x1 bound_ctrl:0
// W64: encoding: [0x0a,0x00,0x1b,0xd4,0xfa,0x92,0x03,0x00,0x02,0x27,0x09,0xf1]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
v_cmp_lt_f16_e64_dpp s[10:11], v240, -v2 row_xmask:0x6 row_mask:0x0 fi:1
// W64: encoding: [0x0a,0x00,0x01,0xd4,0xfa,0x04,0x02,0x40,0xf0,0x66,0x05,0x0f]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// cmp class
// w32
v_cmp_class_f32_e64_dpp s10, v10, v2 quad_perm:[2,2,3,1] bound_ctrl:0 fi:1
// W32: encoding: [0x0a,0x00,0x7e,0xd4,0xfa,0x04,0x02,0x00,0x0a,0x7a,0x0c,0xff]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// w64
v_cmp_class_f16_e64_dpp vcc, v1, v2 quad_perm:[1,2,3,1] bank_mask: 0x5 fi:1
// W64: encoding: [0x6a,0x00,0x7d,0xd4,0xfa,0x04,0x02,0x00,0x01,0x79,0x04,0xf5]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error
// cmpx
v_cmpx_eq_i16_e64_dpp v5, v25 quad_perm:[2,2,3,1] bound_ctrl:0 fi:1
// GFX11: encoding: [0x7e,0x00,0xb2,0xd4,0xfa,0x32,0x02,0x00,0x05,0x7a,0x0c,0xff]
v_cmpx_lt_i32_e64_dpp v0, v3 quad_perm:[3,2,1,0] bank_mask:0xe
// GFX11: encoding: [0x7e,0x00,0xc1,0xd4,0xfa,0x06,0x02,0x00,0x00,0x1b,0x00,0xfe]
// cmpx class
v_cmpx_class_f32_e64_dpp v1, v20 row_ror:7 bank_mask:0x1 bound_ctrl:0
// GFX11: encoding: [0x7e,0x00,0xfe,0xd4,0xfa,0x28,0x02,0x00,0x01,0x27,0x09,0xf1]

29
llvm/test/MC/AMDGPU/gfx11_asm_vopc_dpp16.s Normal file
View File

@ -0,0 +1,29 @@
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11,W32 %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11,W64 %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 %s 2>&1 | FileCheck --check-prefix=W32-ERR --implicit-check-not=error: %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 %s 2>&1 | FileCheck --check-prefix=W64-ERR --implicit-check-not=error: %s
v_cmp_gt_u16_dpp v1, v2 row_shl:0x7 row_mask:0x0 bank_mask:0x0 fi:1
// GFX11: encoding: [0xfa,0x04,0x78,0x7c,0x01,0x07,0x05,0x00]
v_cmp_gt_i16_dpp v1, v2 quad_perm:[1,3,1,0] row_mask:0x7
// GFX11: encoding: [0xfa,0x04,0x68,0x7c,0x01,0x1d,0x00,0x7f]
v_cmp_lt_f32 v1, -v2 quad_perm:[0,1,2,2]
// GFX11: encoding: [0xfa,0x04,0x22,0x7c,0x01,0xa4,0x40,0xff]
v_cmp_gt_i32_dpp vcc_lo, v1, v255 row_mirror bank_mask:0x2 fi:1
// W32: encoding: [0xfa,0xfe,0x89,0x7c,0x01,0x40,0x05,0xf2]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error:
v_cmp_f_f32_dpp vcc_lo, v1, v2 row_shl:0x7 row_mask:0x0 bank_mask:0x0
// W32: encoding: [0xfa,0x04,0x20,0x7c,0x01,0x07,0x01,0x00]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error:
v_cmp_ge_u16_dpp vcc, v1, v2 row_share:0xa bound_ctrl:0
// W64: encoding: [0xfa,0x04,0x7c,0x7c,0x01,0x5a,0x09,0xff]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error:
v_cmp_class_f16_dpp vcc, v1, v2 row_half_mirror bound_ctrl:0
// W64: encoding: [0xfa,0x04,0xfa,0x7c,0x01,0x41,0x09,0xff]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error:

31
llvm/test/MC/AMDGPU/gfx11_asm_vopc_dpp8.s Normal file
View File

@ -0,0 +1,31 @@
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11,W32 %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11,W64 %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 %s 2>&1 | FileCheck --check-prefix=W32-ERR --implicit-check-not=error: %s
// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 %s 2>&1 | FileCheck --check-prefix=W64-ERR --implicit-check-not=error: %s
v_cmp_le_u16_dpp v1, v2 dpp8:[7,7,7,3,4,4,6,7] fi:1
// GFX11: encoding: [0xea,0x04,0x76,0x7c,0x01,0xff,0x47,0xfa]
v_cmp_le_i16_dpp v1, v2 dpp8:[7,7,7,3,4,4,6,7]
// GFX11: encoding: [0xe9,0x04,0x66,0x7c,0x01,0xff,0x47,0xfa]
v_cmp_le_i32_dpp vcc_lo, v1, v255 dpp8:[0,2,1,3,4,5,6,7]
// W32: encoding: [0xe9,0xfe,0x87,0x7c,0x01,0x50,0xc6,0xfa]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error:
v_cmp_tru_f32_dpp vcc_lo, v1, v2 dpp8:[0,2,1,3,4,5,6,7]
// W32: encoding: [0xe9,0x04,0x3e,0x7c,0x01,0x50,0xc6,0xfa]
// W64-ERR: :[[@LINE-2]]:{{[0-9]+}}: error:
// check vcc/vcc_lo have been added
v_cmp_lt_f32_dpp v1, v2 dpp8:[2,3,4,1,3,3,3,3]
// W32: v_cmp_lt_f32 vcc_lo, v1, v2 dpp8:[2,3,4,1,3,3,3,3] ; encoding: [0xe9,0x04,0x22,0x7c,0x01,0x1a,0xb3,0x6d]
// W64: v_cmp_lt_f32 vcc, v1, v2 dpp8:[2,3,4,1,3,3,3,3] ; encoding: [0xe9,0x04,0x22,0x7c,0x01,0x1a,0xb3,0x6d]
v_cmp_lt_u16_dpp vcc, v1, v2 dpp8:[7,6,5,3,4,2,1,0] fi:1
// W64: encoding: [0xea,0x04,0x72,0x7c,0x01,0x77,0x47,0x05]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error:
v_cmp_class_f16_dpp vcc, v1, v2 dpp8:[7,6,5,3,4,2,1,0] fi:1
// W64: encoding: [0xea,0x04,0xfa,0x7c,0x01,0x77,0x47,0x05]
// W32-ERR: :[[@LINE-2]]:{{[0-9]+}}: error:

6496
llvm/test/MC/AMDGPU/gfx11_asm_vopcx.s
View File

File diff suppressed because it is too large Load Diff

14
llvm/test/MC/AMDGPU/gfx11_asm_vopcx_dpp16.s Normal file
View File

@ -0,0 +1,14 @@
// RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
// RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
v_cmpx_t_i32_dpp v0, v2 row_shr:0xe row_mask:0x3 bank_mask:0xa bound_ctrl:0
// GFX11: encoding: [0xfa,0x04,0x8e,0x7d,0x00,0x1e,0x09,0x3a]
v_cmpx_f_f32_dpp v255, v2 quad_perm:[2,3,0,0]
// GFX11: encoding: [0xfa,0x04,0x20,0x7d,0xff,0x0e,0x00,0xff]
v_cmpx_class_f16_dpp v12, v101 quad_perm:[2,3,0,0] fi:1
// GFX11: encoding: [0xfa,0xca,0xfa,0x7d,0x0c,0x0e,0x04,0xff]
v_cmpx_class_f16_dpp abs(v12), v101 quad_perm:[2,3,0,0]
// GFX11: encoding: [0xfa,0xca,0xfa,0x7d,0x0c,0x0e,0x20,0xff]

11
llvm/test/MC/AMDGPU/gfx11_asm_vopcx_dpp8.s Normal file
View File

@ -0,0 +1,11 @@
// RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
// RUN: llvm-mc -arch=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -show-encoding %s | FileCheck --check-prefixes=GFX11 %s
v_cmpx_f_i32_dpp v0, v2 dpp8:[7,6,5,3,4,2,1,0] fi:1
// GFX11: encoding: [0xea,0x04,0x80,0x7d,0x00,0x77,0x47,0x05]
v_cmpx_t_f32_dpp v255, v2 dpp8:[7,6,5,3,4,2,1,0]
// GFX11: encoding: [0xe9,0x04,0x3e,0x7d,0xff,0x77,0x47,0x05]
v_cmpx_class_f16_dpp v12, v101 dpp8:[7,6,5,3,4,2,1,0]
// GFX11: encoding: [0xe9,0xca,0xfa,0x7d,0x0c,0x77,0x47,0x05]