llvm/test/MC/AMDGPU/gfx11_asm_vopd_features.s

   1 // RUN: llvm-mc -triple=amdgcn -mcpu=gfx1100 -show-encoding %s | FileCheck -check-prefix=GFX11 %s
   2
   3 //===----------------------------------------------------------------------===//
   4 // A VOPD instruction can use one or more literals,
   5 // provided that they are identical.
   6 //===----------------------------------------------------------------------===//
   7
   8 // LITERAL
   9
  10 v_dual_mul_f32      v11, v1, v2                  ::  v_dual_mul_f32      v10, 0x24681357, v5
  11 // GFX11: encoding: [0x01,0x05,0xc6,0xc8,0xff,0x0a,0x0a,0x0b,0x57,0x13,0x68,0x24]
  12
  13 // LITERAL*2
  14
  15 v_dual_mul_f32      v11, 0x24681357, v2          ::  v_dual_mul_f32      v10, 0x24681357, v5
  16 // GFX11: encoding: [0xff,0x04,0xc6,0xc8,0xff,0x0a,0x0a,0x0b,0x57,0x13,0x68,0x24]
  17
  18 // LITERAL*2 (this is an unclear case because literals have different size, but SP3 accepts this code)
  19
  20 v_dual_add_f32      v6, 0xfe0b, v5               ::  v_dual_dot2acc_f32_f16  v255, 0xfe0b, v4
  21 // GFX11: encoding: [0xff,0x0a,0x18,0xc9,0xff,0x08,0xfe,0x06,0x0b,0xfe,0x00,0x00]
  22
  23 // LITERAL + KIMM
  24
  25 v_dual_add_f32      v5, 0xaf123456, v2           ::  v_dual_fmaak_f32     v6, v3, v1, 0xaf123456 ;
  26 // GFX11: encoding: [0xff,0x04,0x02,0xc9,0x03,0x03,0x06,0x05,0x56,0x34,0x12,0xaf]
  27
  28 // KIMM + LITERAL
  29
  30 v_dual_fmamk_f32    v122, v74, 0xa0172923, v161  ::  v_dual_lshlrev_b32   v247, 0xa0172923, v99
  31 // GFX11: encoding: [0x4a,0x43,0xa3,0xc8,0xff,0xc6,0xf6,0x7a,0x23,0x29,0x17,0xa0]
  32
  33 // KIMM + LITERAL (this is an unclear case because literals have different size, but SP3 accepts this code)
  34
  35 v_dual_fmamk_f32    v122, v74, 0xfe0b, v162      ::  v_dual_dot2acc_f32_f16  v247, 0xfe0b, v99
  36 // GFX11: encoding: [0x4a,0x45,0x99,0xc8,0xff,0xc6,0xf6,0x7a,0x0b,0xfe,0x00,0x00]
  37
  38 // KIMM*2
  39
  40 v_dual_fmamk_f32    v122, 0xdeadbeef, 0xdeadbeef, v161 ::  v_dual_fmamk_f32  v123, 0xdeadbeef, 0xdeadbeef, v162
  41 // GFX11: encoding: [0xff,0x42,0x85,0xc8,0xff,0x44,0x7b,0x7a,0xef,0xbe,0xad,0xde]
  42
  43 //===----------------------------------------------------------------------===//
  44 // A VOPD instruction can use 2 scalar operands,
  45 // but implicit VCC must be counted in.
  46 //===----------------------------------------------------------------------===//
  47
  48 // 2 different SGPRs
  49
  50 v_dual_mul_f32      v0, s1, v2                   ::  v_dual_mul_f32       v3, s4, v5
  51 // GFX11: encoding: [0x01,0x04,0xc6,0xc8,0x04,0x0a,0x02,0x00]
  52
  53 // SGPR + LITERAL
  54
  55 v_dual_fmaak_f32    v122, s74, v161, 2.741       ::  v_dual_and_b32       v247, v160, v98
  56 // GFX11: encoding: [0x4a,0x42,0x65,0xc8,0xa0,0xc5,0xf6,0x7a,0x8b,0x6c,0x2f,0x40]
  57
  58 v_dual_mov_b32      v247, v160                   ::  v_dual_fmaak_f32     v122, s74, v161, 2.741
  59 // GFX11: encoding: [0xa0,0x01,0x02,0xca,0x4a,0x42,0x7b,0xf7,0x8b,0x6c,0x2f,0x40]
  60
  61 // SGPR*2 + LITERAL
  62
  63 v_dual_fmaak_f32    v122, s74, v161, 2.741       ::  v_dual_and_b32       v247, s74, v98
  64 // GFX11: encoding: [0x4a,0x42,0x65,0xc8,0x4a,0xc4,0xf6,0x7a,0x8b,0x6c,0x2f,0x40]
  65
  66 // SGPR + LITERAL*2
  67
  68 v_dual_fmaak_f32    v122, s74, v161, 2.741       ::  v_dual_fmamk_f32     v3, v6, 2.741, v1
  69 // GFX11: encoding: [0x4a,0x42,0x45,0xc8,0x06,0x03,0x02,0x7a,0x8b,0x6c,0x2f,0x40]
  70
  71 // SGPR*2 + LITERAL*2
  72
  73 v_dual_fmaak_f32    v122, s74, v161, 2.741       ::  v_dual_fmamk_f32     v3, s74, 2.741, v1
  74 // GFX11: encoding: [0x4a,0x42,0x45,0xc8,0x4a,0x02,0x02,0x7a,0x8b,0x6c,0x2f,0x40]
  75
  76 // LITERAL + VCC
  77
  78 v_dual_fmaak_f32    v122, v0, v161, 2.741       ::  v_dual_cndmask_b32   v1, v2, v3
  79 // GFX11: encoding: [0x00,0x43,0x53,0xc8,0x02,0x07,0x00,0x7a,0x8b,0x6c,0x2f,0x40]
  80
  81 // LITERAL*2 + VCC
  82
  83 v_dual_fmaak_f32    v122, v0, v161, 2.741       ::  v_dual_cndmask_b32   v1, 2.741, v3
  84 // GFX11: encoding: [0x00,0x43,0x53,0xc8,0xff,0x06,0x00,0x7a,0x8b,0x6c,0x2f,0x40]
  85
  86 // LITERAL*2 + VCC*2
  87
  88 v_dual_cndmask_b32  v255, 0xbabe, v2             ::  v_dual_cndmask_b32   v6, 0xbabe, v3
  89 // GFX11: encoding: [0xff,0x04,0x52,0xca,0xff,0x06,0x06,0xff,0xbe,0xba,0x00,0x00]
  90
  91 // SGPR*2 + VCC
  92
  93 v_dual_add_f32      v255, s105, v2               ::  v_dual_cndmask_b32   v6, s105, v3
  94 // GFX11: encoding: [0x69,0x04,0x12,0xc9,0x69,0x06,0x06,0xff]
  95
  96 // SGPR*2 + VCC*2
  97
  98 v_dual_cndmask_b32  v255, s1, v2                 ::  v_dual_cndmask_b32   v6, s1, v3
  99 // GFX11: encoding: [0x01,0x04,0x52,0xca,0x01,0x06,0x06,0xff]
 100
 101 // VCC*2
 102
 103 v_dual_add_f32      v255, vcc_lo, v2             ::  v_dual_cndmask_b32   v6, v1, v3
 104 // GFX11: encoding: [0x6a,0x04,0x12,0xc9,0x01,0x07,0x06,0xff]
 105
 106 //===----------------------------------------------------------------------===//
 107 // One dst register must be even and the other odd.
 108 //===----------------------------------------------------------------------===//
 109
 110 v_dual_mul_f32      v0, v10, v20                 ::  v_dual_mul_f32      v1, v11, v21
 111 // GFX11: encoding: [0x0a,0x29,0xc6,0xc8,0x0b,0x2b,0x00,0x00]
 112
 113 v_dual_mul_f32      v1, v10, v20                 ::  v_dual_mul_f32      v0, v11, v21
 114 // GFX11: encoding: [0x0a,0x29,0xc6,0xc8,0x0b,0x2b,0x00,0x01]
 115
 116 //===----------------------------------------------------------------------===//
 117 // srcX0 and srcY0 must use different VGPR banks.
 118 //===----------------------------------------------------------------------===//
 119
 120 v_dual_mul_f32      v0, v10, v20                 ::  v_dual_mul_f32      v1, v11, v21
 121 // GFX11: encoding: [0x0a,0x29,0xc6,0xc8,0x0b,0x2b,0x00,0x00]
 122
 123 v_dual_mul_f32      v0, v10, v20                 ::  v_dual_mul_f32      v1, v12, v21
 124 // GFX11: encoding: [0x0a,0x29,0xc6,0xc8,0x0c,0x2b,0x00,0x00]
 125
 126 v_dual_mul_f32      v0, v10, v20                 ::  v_dual_mul_f32      v1, v13, v21
 127 // GFX11: encoding: [0x0a,0x29,0xc6,0xc8,0x0d,0x2b,0x00,0x00]
 128
 129 v_dual_mul_f32      v0, v10, v20                 ::  v_dual_mul_f32      v1, v15, v21
 130 // GFX11: encoding: [0x0a,0x29,0xc6,0xc8,0x0f,0x2b,0x00,0x00]
 131
 132 v_dual_mul_f32      v0, v10, v20                 ::  v_dual_mul_f32      v1, v16, v21
 133 // GFX11: encoding: [0x0a,0x29,0xc6,0xc8,0x10,0x2b,0x00,0x00]
 134
 135 v_dual_mul_f32      v0, v10, v20                 ::  v_dual_mul_f32      v1, v17, v21
 136 // GFX11: encoding: [0x0a,0x29,0xc6,0xc8,0x11,0x2b,0x00,0x00]
 137
 138 //===----------------------------------------------------------------------===//
 139 // srcX1 and srcY1 must use different VGPR banks.
 140 //===----------------------------------------------------------------------===//
 141
 142 v_dual_mul_f32      v0, v10, v20                 ::  v_dual_mul_f32      v1, v11, v21
 143 // GFX11: encoding: [0x0a,0x29,0xc6,0xc8,0x0b,0x2b,0x00,0x00]
 144
 145 v_dual_mul_f32      v0, v10, v20                 ::  v_dual_mul_f32      v1, v11, v22
 146 // GFX11: encoding: [0x0a,0x29,0xc6,0xc8,0x0b,0x2d,0x00,0x00]
 147
 148 v_dual_mul_f32      v0, v10, v20                 ::  v_dual_mul_f32      v1, v11, v23
 149 // GFX11: encoding: [0x0a,0x29,0xc6,0xc8,0x0b,0x2f,0x00,0x00]
 150
 151 v_dual_mul_f32      v0, v10, v20                 ::  v_dual_mul_f32      v1, v11, v25
 152 // GFX11: encoding: [0x0a,0x29,0xc6,0xc8,0x0b,0x33,0x00,0x00]
 153
 154 v_dual_mul_f32      v0, v10, v20                 ::  v_dual_mul_f32      v1, v11, v26
 155 // GFX11: encoding: [0x0a,0x29,0xc6,0xc8,0x0b,0x35,0x00,0x00]
 156
 157 v_dual_mul_f32      v0, v10, v20                 ::  v_dual_mul_f32      v1, v11, v27
 158 // GFX11: encoding: [0x0a,0x29,0xc6,0xc8,0x0b,0x37,0x00,0x00]
 159
 160 //===----------------------------------------------------------------------===//
 161 // srcX2 and srcY2 must use different VGPR banks.
 162 //===----------------------------------------------------------------------===//
 163
 164 v_dual_fmamk_f32    v6, v1, 0xaf123456, v0       :: v_dual_fmamk_f32      v5, v2, 0xaf123456, v1
 165 // GFX11: encoding: [0x01,0x01,0x84,0xc8,0x02,0x03,0x04,0x06,0x56,0x34,0x12,0xaf]
 166
 167 v_dual_fmamk_f32    v6, v1, 0xaf123456, v1       :: v_dual_fmamk_f32      v5, v2, 0xaf123456, v0
 168 // GFX11: encoding: [0x01,0x03,0x84,0xc8,0x02,0x01,0x04,0x06,0x56,0x34,0x12,0xaf]
 169
 170 v_dual_fmac_f32     v6, v1, v2                   :: v_dual_fmamk_f32      v7, v2, 0xaf123456, v7
 171 // GFX11: encoding: [0x01,0x05,0x04,0xc8,0x02,0x0f,0x06,0x06,0x56,0x34,0x12,0xaf]
 172
 173 v_dual_fmac_f32     v7, v1, v2                   :: v_dual_fmamk_f32      v6, v2, 0xaf123456, v6
 174 // GFX11: encoding: [0x01,0x05,0x04,0xc8,0x02,0x0d,0x06,0x07,0x56,0x34,0x12,0xaf]
 175
 176 v_dual_fmamk_f32    v5, v1, 0xaf123456, v5       :: v_dual_fmac_f32       v6, v2, v3
 177 // GFX11: encoding: [0x01,0x0b,0x80,0xc8,0x02,0x07,0x06,0x05,0x56,0x34,0x12,0xaf]
 178
 179 v_dual_fmamk_f32    v6, v1, 0xaf123456, v6       :: v_dual_fmac_f32       v5, v2, v3
 180 // GFX11: encoding: [0x01,0x0d,0x80,0xc8,0x02,0x07,0x04,0x06,0x56,0x34,0x12,0xaf]