llvm/test/MC/AMDGPU/bf16_imm.s

   1 // RUN: llvm-mc -triple=amdgcn -mcpu=gfx1100 -show-encoding %s | FileCheck %s
   2 // RUN: llvm-mc -triple=amdgcn -mcpu=gfx1200 -show-encoding %s | FileCheck %s
   3
   4 v_dot2_bf16_bf16 v5, v1, v2, 100.0
   5 // CHECK: v_dot2_bf16_bf16 v5, v1, v2, 0x42c8 ; encoding: [0x05,0x00,0x67,0xd6,0x01,0x05,0xfe,0x03,0xc8,0x42,0x00,0x00]
   6
   7 v_dot2_bf16_bf16 v2, v0, 1.0, v2
   8 // CHECK: v_dot2_bf16_bf16 v2, v0, 1.0, v2 ; encoding: [0x02,0x00,0x67,0xd6,0x00,0xe5,0x09,0x04]
   9
  10 v_dot2_bf16_bf16 v2, 1.0, v0, v2
  11 // CHECK: v_dot2_bf16_bf16 v2, 1.0, v0, v2 ; encoding: [0x02,0x00,0x67,0xd6,0xf2,0x00,0x0a,0x04]
  12
  13 v_dot2_bf16_bf16 v5, v1, v2, 1.0
  14 // CHECK: v_dot2_bf16_bf16 v5, v1, v2, 1.0 ; encoding: [0x05,0x00,0x67,0xd6,0x01,0x05,0xca,0x03]
  15
  16 v_dot2_bf16_bf16 v2, v0, -1.0, v2
  17 // CHECK: v_dot2_bf16_bf16 v2, v0, -1.0, v2       ; encoding: [0x02,0x00,0x67,0xd6,0x00,0xe7,0x09,0x04]
  18
  19 v_dot2_bf16_bf16 v2, v0, 0.5, v2
  20 // CHECK: v_dot2_bf16_bf16 v2, v0, 0.5, v2        ; encoding: [0x02,0x00,0x67,0xd6,0x00,0xe1,0x09,0x04]
  21
  22 v_dot2_bf16_bf16 v2, v0, -0.5, v2
  23 // CHECK: v_dot2_bf16_bf16 v2, v0, -0.5, v2       ; encoding: [0x02,0x00,0x67,0xd6,0x00,0xe3,0x09,0x04]
  24
  25 v_dot2_bf16_bf16 v2, v0, 2.0, v2
  26 // CHECK: v_dot2_bf16_bf16 v2, v0, 2.0, v2        ; encoding: [0x02,0x00,0x67,0xd6,0x00,0xe9,0x09,0x04]
  27
  28 v_dot2_bf16_bf16 v2, v0, -2.0, v2
  29 // CHECK: v_dot2_bf16_bf16 v2, v0, -2.0, v2       ; encoding: [0x02,0x00,0x67,0xd6,0x00,0xeb,0x09,0x04]
  30
  31 v_dot2_bf16_bf16 v2, v0, 4.0, v2
  32 // CHECK: v_dot2_bf16_bf16 v2, v0, 4.0, v2        ; encoding: [0x02,0x00,0x67,0xd6,0x00,0xed,0x09,0x04]
  33
  34 v_dot2_bf16_bf16 v2, v0, -4.0, v2
  35 // CHECK: v_dot2_bf16_bf16 v2, v0, -4.0, v2       ; encoding: [0x02,0x00,0x67,0xd6,0x00,0xef,0x09,0x04]
  36
  37 // Check 1/(2*pi) rounded value and ideomatic fp32 0.15915494 value
  38 // which cannot be accurately represented in bf16.
  39
  40 v_dot2_bf16_bf16 v2, v0, 0.158203125, v2
  41 // CHECK: v_dot2_bf16_bf16 v2, v0, 0.15915494, v2 ; encoding: [0x02,0x00,0x67,0xd6,0x00,0xf1,0x09,0x04]
  42
  43 v_dot2_bf16_bf16 v2, v0, 0.15915494, v2
  44 // CHECK: v_dot2_bf16_bf16 v2, v0, 0.15915494, v2 ; encoding: [0x02,0x00,0x67,0xd6,0x00,0xf1,0x09,0x04]
  45
  46 v_dot2_bf16_bf16 v2, v0, 0x3e22, v2
  47 // CHECK: v_dot2_bf16_bf16 v2, v0, 0.15915494, v2 ; encoding: [0x02,0x00,0x67,0xd6,0x00,0xf1,0x09,0x04]
  48
  49 v_dot2_bf16_bf16 v2, v0, v2, 0.15915494
  50 // CHECK: v_dot2_bf16_bf16 v2, v0, v2, 0.15915494 ; encoding: [0x02,0x00,0x67,0xd6,0x00,0x05,0xe2,0x03]
  51
  52 v_dot2_f32_bf16 v2, v1, 0, v2
  53 // CHECK: v_dot2_f32_bf16 v2, v1, 0, v2           ; encoding: [0x02,0x40,0x1a,0xcc,0x01,0x01,0x09,0x1c]
  54
  55 v_dot2_f32_bf16 v2, v1, 0.5, v2
  56 // CHECK: v_dot2_f32_bf16 v2, v1, 0.5, v2         ; encoding: [0x02,0x40,0x1a,0xcc,0x01,0xe1,0x09,0x1c]
  57
  58 v_dot2_f32_bf16 v2, v1, -0.5, v2
  59 // CHECK: v_dot2_f32_bf16 v2, v1, -0.5, v2        ; encoding: [0x02,0x40,0x1a,0xcc,0x01,0xe3,0x09,0x1c]
  60
  61 v_dot2_f32_bf16 v2, v1, 1.0, v2
  62 // CHECK: v_dot2_f32_bf16 v2, v1, 1.0, v2         ; encoding: [0x02,0x40,0x1a,0xcc,0x01,0xe5,0x09,0x1c]
  63
  64 v_dot2_f32_bf16 v2, v1, -1.0, v2
  65 // CHECK: v_dot2_f32_bf16 v2, v1, -1.0, v2        ; encoding: [0x02,0x40,0x1a,0xcc,0x01,0xe7,0x09,0x1c]
  66
  67 v_dot2_f32_bf16 v2, v1, 2.0, v2
  68 // CHECK: v_dot2_f32_bf16 v2, v1, 2.0, v2         ; encoding: [0x02,0x40,0x1a,0xcc,0x01,0xe9,0x09,0x1c]
  69
  70 v_dot2_f32_bf16 v2, v1, -2.0, v2
  71 // CHECK: v_dot2_f32_bf16 v2, v1, -2.0, v2        ; encoding: [0x02,0x40,0x1a,0xcc,0x01,0xeb,0x09,0x1c]
  72
  73 v_dot2_f32_bf16 v2, v1, 4.0, v2
  74 // CHECK: v_dot2_f32_bf16 v2, v1, 4.0, v2         ; encoding: [0x02,0x40,0x1a,0xcc,0x01,0xed,0x09,0x1c]
  75
  76 v_dot2_f32_bf16 v2, v1, -4.0, v2
  77 // CHECK: v_dot2_f32_bf16 v2, v1, -4.0, v2        ; encoding: [0x02,0x40,0x1a,0xcc,0x01,0xef,0x09,0x1c]
  78
  79 v_dot2_f32_bf16 v2, v1, 0.15915494, v2
  80 // CHECK: v_dot2_f32_bf16 v2, v1, 0.15915494, v2  ; encoding: [0x02,0x40,0x1a,0xcc,0x01,0xf1,0x09,0x1c]
  81
  82 v_dot2_f32_bf16 v2, v1, 0x3e22, v2
  83 // CHECK: v_dot2_f32_bf16 v2, v1, 0.15915494, v2  ; encoding: [0x02,0x40,0x1a,0xcc,0x01,0xf1,0x09,0x1c]
  84
  85 v_dot2_f32_bf16 v2, 0.5, v1, v2
  86 // CHECK: v_dot2_f32_bf16 v2, 0.5, v1, v2         ; encoding: [0x02,0x40,0x1a,0xcc,0xf0,0x02,0x0a,0x1c]
  87
  88 v_dot2_f32_bf16 v2, -0.5, v1, v2
  89 // CHECK: v_dot2_f32_bf16 v2, -0.5, v1, v2        ; encoding: [0x02,0x40,0x1a,0xcc,0xf1,0x02,0x0a,0x1c]
  90
  91 v_dot2_f32_bf16 v2, 1.0, v1, v2
  92 // CHECK: v_dot2_f32_bf16 v2, 1.0, v1, v2         ; encoding: [0x02,0x40,0x1a,0xcc,0xf2,0x02,0x0a,0x1c]
  93
  94 v_dot2_f32_bf16 v2, -1.0, v1, v2
  95 // CHECK: v_dot2_f32_bf16 v2, -1.0, v1, v2        ; encoding: [0x02,0x40,0x1a,0xcc,0xf3,0x02,0x0a,0x1c]
  96
  97 v_dot2_f32_bf16 v2, 2.0, v1, v2
  98 // CHECK: v_dot2_f32_bf16 v2, 2.0, v1, v2         ; encoding: [0x02,0x40,0x1a,0xcc,0xf4,0x02,0x0a,0x1c]
  99
 100 v_dot2_f32_bf16 v2, -2.0, v1, v2
 101 // CHECK: v_dot2_f32_bf16 v2, -2.0, v1, v2        ; encoding: [0x02,0x40,0x1a,0xcc,0xf5,0x02,0x0a,0x1c]
 102
 103 v_dot2_f32_bf16 v2, 4.0, v1, v2
 104 // CHECK: v_dot2_f32_bf16 v2, 4.0, v1, v2         ; encoding: [0x02,0x40,0x1a,0xcc,0xf6,0x02,0x0a,0x1c]
 105
 106 v_dot2_f32_bf16 v2, -4.0, v1, v2
 107 // CHECK: v_dot2_f32_bf16 v2, -4.0, v1, v2        ; encoding: [0x02,0x40,0x1a,0xcc,0xf7,0x02,0x0a,0x1c]
 108
 109 v_dot2_f32_bf16 v2, 100.0, v1, v2
 110 // CHECK: v_dot2_f32_bf16 v2, 0x42c8, v1, v2      ; encoding: [0x02,0x40,0x1a,0xcc,0xff,0x02,0x0a,0x1c,0xc8,0x42,0x00,0x00]
 111
 112 v_dot2_f32_bf16 v2, v1, 100.0, v2
 113 // CHECK: v_dot2_f32_bf16 v2, v1, 0x42c8, v2      ; encoding: [0x02,0x40,0x1a,0xcc,0x01,0xff,0x09,0x1c,0xc8,0x42,0x00,0x00]