Home | History | Annotate | Download | only in X86 
      1 ; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
      2 ; RUN: opt < %s -enable-no-nans-fp-math  -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=+sse2 | FileCheck %s --check-prefixes=CHECK,SSE,SSE2
      3 ; RUN: opt < %s -enable-no-nans-fp-math  -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=+sse4.2 | FileCheck %s --check-prefixes=CHECK,SSE,SSE42
      4 ; RUN: opt < %s -enable-no-nans-fp-math  -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx | FileCheck %s --check-prefixes=CHECK,AVX,AVX1
      5 ; RUN: opt < %s -enable-no-nans-fp-math  -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX2
      6 ; RUN: opt < %s -enable-no-nans-fp-math  -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512F
      7 ; RUN: opt < %s -enable-no-nans-fp-math  -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=CHECK,AVX512,AVX512BW
      8 ;
      9 ; RUN: opt < %s -enable-no-nans-fp-math  -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mcpu=slm | FileCheck %s --check-prefixes=SLM
     10 ; RUN: opt < %s -enable-no-nans-fp-math  -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mcpu=goldmont | FileCheck %s --check-prefixes=GLM
     11 ; RUN: opt < %s -enable-no-nans-fp-math  -cost-model -analyze -mtriple=x86_64-apple-macosx10.8.0 -mcpu=btver2 | FileCheck %s --check-prefixes=BTVER2
     12 
     13 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
     14 target triple = "x86_64-apple-macosx10.8.0"
     15 
     16 define i32 @fadd(i32 %arg) {
     17 ; SSE2-LABEL: 'fadd'
     18 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = fadd float undef, undef
     19 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fadd <4 x float> undef, undef
     20 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fadd <8 x float> undef, undef
     21 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fadd <16 x float> undef, undef
     22 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = fadd double undef, undef
     23 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fadd <2 x double> undef, undef
     24 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fadd <4 x double> undef, undef
     25 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fadd <8 x double> undef, undef
     26 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
     27 ;
     28 ; SSE42-LABEL: 'fadd'
     29 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fadd float undef, undef
     30 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fadd <4 x float> undef, undef
     31 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fadd <8 x float> undef, undef
     32 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fadd <16 x float> undef, undef
     33 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fadd double undef, undef
     34 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fadd <2 x double> undef, undef
     35 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fadd <4 x double> undef, undef
     36 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fadd <8 x double> undef, undef
     37 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
     38 ;
     39 ; AVX1-LABEL: 'fadd'
     40 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fadd float undef, undef
     41 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fadd <4 x float> undef, undef
     42 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fadd <8 x float> undef, undef
     43 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fadd <16 x float> undef, undef
     44 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fadd double undef, undef
     45 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fadd <2 x double> undef, undef
     46 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fadd <4 x double> undef, undef
     47 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fadd <8 x double> undef, undef
     48 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
     49 ;
     50 ; AVX2-LABEL: 'fadd'
     51 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fadd float undef, undef
     52 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fadd <4 x float> undef, undef
     53 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fadd <8 x float> undef, undef
     54 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = fadd <16 x float> undef, undef
     55 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fadd double undef, undef
     56 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fadd <2 x double> undef, undef
     57 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fadd <4 x double> undef, undef
     58 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = fadd <8 x double> undef, undef
     59 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
     60 ;
     61 ; AVX512-LABEL: 'fadd'
     62 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fadd float undef, undef
     63 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fadd <4 x float> undef, undef
     64 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fadd <8 x float> undef, undef
     65 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = fadd <16 x float> undef, undef
     66 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fadd double undef, undef
     67 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fadd <2 x double> undef, undef
     68 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fadd <4 x double> undef, undef
     69 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F64 = fadd <8 x double> undef, undef
     70 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
     71 ;
     72 ; SLM-LABEL: 'fadd'
     73 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fadd float undef, undef
     74 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fadd <4 x float> undef, undef
     75 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fadd <8 x float> undef, undef
     76 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fadd <16 x float> undef, undef
     77 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fadd double undef, undef
     78 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fadd <2 x double> undef, undef
     79 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fadd <4 x double> undef, undef
     80 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fadd <8 x double> undef, undef
     81 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
     82 ;
     83 ; GLM-LABEL: 'fadd'
     84 ; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fadd float undef, undef
     85 ; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fadd <4 x float> undef, undef
     86 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fadd <8 x float> undef, undef
     87 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fadd <16 x float> undef, undef
     88 ; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fadd double undef, undef
     89 ; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fadd <2 x double> undef, undef
     90 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fadd <4 x double> undef, undef
     91 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fadd <8 x double> undef, undef
     92 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
     93 ;
     94 ; BTVER2-LABEL: 'fadd'
     95 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fadd float undef, undef
     96 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fadd <4 x float> undef, undef
     97 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fadd <8 x float> undef, undef
     98 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fadd <16 x float> undef, undef
     99 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fadd double undef, undef
    100 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fadd <2 x double> undef, undef
    101 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fadd <4 x double> undef, undef
    102 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fadd <8 x double> undef, undef
    103 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    104 ;
    105   %F32 = fadd float undef, undef
    106   %V4F32 = fadd <4 x float> undef, undef
    107   %V8F32 = fadd <8 x float> undef, undef
    108   %V16F32 = fadd <16 x float> undef, undef
    109 
    110   %F64 = fadd double undef, undef
    111   %V2F64 = fadd <2 x double> undef, undef
    112   %V4F64 = fadd <4 x double> undef, undef
    113   %V8F64 = fadd <8 x double> undef, undef
    114 
    115   ret i32 undef
    116 }
    117 
    118 define i32 @fsub(i32 %arg) {
    119 ; SSE2-LABEL: 'fsub'
    120 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = fsub float undef, undef
    121 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fsub <4 x float> undef, undef
    122 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fsub <8 x float> undef, undef
    123 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fsub <16 x float> undef, undef
    124 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = fsub double undef, undef
    125 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fsub <2 x double> undef, undef
    126 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fsub <4 x double> undef, undef
    127 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fsub <8 x double> undef, undef
    128 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    129 ;
    130 ; SSE42-LABEL: 'fsub'
    131 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
    132 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
    133 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> undef, undef
    134 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> undef, undef
    135 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
    136 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
    137 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fsub <4 x double> undef, undef
    138 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fsub <8 x double> undef, undef
    139 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    140 ;
    141 ; AVX1-LABEL: 'fsub'
    142 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
    143 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
    144 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> undef, undef
    145 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> undef, undef
    146 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
    147 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
    148 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fsub <4 x double> undef, undef
    149 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fsub <8 x double> undef, undef
    150 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    151 ;
    152 ; AVX2-LABEL: 'fsub'
    153 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
    154 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
    155 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fsub <8 x float> undef, undef
    156 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = fsub <16 x float> undef, undef
    157 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
    158 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
    159 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fsub <4 x double> undef, undef
    160 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = fsub <8 x double> undef, undef
    161 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    162 ;
    163 ; AVX512-LABEL: 'fsub'
    164 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
    165 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
    166 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fsub <8 x float> undef, undef
    167 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = fsub <16 x float> undef, undef
    168 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
    169 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
    170 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fsub <4 x double> undef, undef
    171 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F64 = fsub <8 x double> undef, undef
    172 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    173 ;
    174 ; SLM-LABEL: 'fsub'
    175 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
    176 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
    177 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> undef, undef
    178 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> undef, undef
    179 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
    180 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fsub <2 x double> undef, undef
    181 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fsub <4 x double> undef, undef
    182 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fsub <8 x double> undef, undef
    183 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    184 ;
    185 ; GLM-LABEL: 'fsub'
    186 ; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
    187 ; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
    188 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> undef, undef
    189 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> undef, undef
    190 ; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
    191 ; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
    192 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fsub <4 x double> undef, undef
    193 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fsub <8 x double> undef, undef
    194 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    195 ;
    196 ; BTVER2-LABEL: 'fsub'
    197 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fsub float undef, undef
    198 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fsub <4 x float> undef, undef
    199 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fsub <8 x float> undef, undef
    200 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fsub <16 x float> undef, undef
    201 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fsub double undef, undef
    202 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fsub <2 x double> undef, undef
    203 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fsub <4 x double> undef, undef
    204 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fsub <8 x double> undef, undef
    205 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    206 ;
    207   %F32 = fsub float undef, undef
    208   %V4F32 = fsub <4 x float> undef, undef
    209   %V8F32 = fsub <8 x float> undef, undef
    210   %V16F32 = fsub <16 x float> undef, undef
    211 
    212   %F64 = fsub double undef, undef
    213   %V2F64 = fsub <2 x double> undef, undef
    214   %V4F64 = fsub <4 x double> undef, undef
    215   %V8F64 = fsub <8 x double> undef, undef
    216 
    217   ret i32 undef
    218 }
    219 
    220 define i32 @fmul(i32 %arg) {
    221 ; SSE2-LABEL: 'fmul'
    222 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = fmul float undef, undef
    223 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fmul <4 x float> undef, undef
    224 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fmul <8 x float> undef, undef
    225 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fmul <16 x float> undef, undef
    226 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = fmul double undef, undef
    227 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fmul <2 x double> undef, undef
    228 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fmul <4 x double> undef, undef
    229 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fmul <8 x double> undef, undef
    230 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    231 ;
    232 ; SSE42-LABEL: 'fmul'
    233 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
    234 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
    235 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fmul <8 x float> undef, undef
    236 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fmul <16 x float> undef, undef
    237 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
    238 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
    239 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fmul <4 x double> undef, undef
    240 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fmul <8 x double> undef, undef
    241 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    242 ;
    243 ; AVX1-LABEL: 'fmul'
    244 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
    245 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
    246 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fmul <8 x float> undef, undef
    247 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fmul <16 x float> undef, undef
    248 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
    249 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
    250 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fmul <4 x double> undef, undef
    251 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fmul <8 x double> undef, undef
    252 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    253 ;
    254 ; AVX2-LABEL: 'fmul'
    255 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
    256 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
    257 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fmul <8 x float> undef, undef
    258 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = fmul <16 x float> undef, undef
    259 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
    260 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
    261 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fmul <4 x double> undef, undef
    262 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = fmul <8 x double> undef, undef
    263 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    264 ;
    265 ; AVX512-LABEL: 'fmul'
    266 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
    267 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
    268 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = fmul <8 x float> undef, undef
    269 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = fmul <16 x float> undef, undef
    270 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
    271 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
    272 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = fmul <4 x double> undef, undef
    273 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F64 = fmul <8 x double> undef, undef
    274 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    275 ;
    276 ; SLM-LABEL: 'fmul'
    277 ; SLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
    278 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fmul <4 x float> undef, undef
    279 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fmul <8 x float> undef, undef
    280 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fmul <16 x float> undef, undef
    281 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = fmul double undef, undef
    282 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F64 = fmul <2 x double> undef, undef
    283 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F64 = fmul <4 x double> undef, undef
    284 ; SLM-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F64 = fmul <8 x double> undef, undef
    285 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    286 ;
    287 ; GLM-LABEL: 'fmul'
    288 ; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
    289 ; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
    290 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fmul <8 x float> undef, undef
    291 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fmul <16 x float> undef, undef
    292 ; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
    293 ; GLM-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
    294 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fmul <4 x double> undef, undef
    295 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fmul <8 x double> undef, undef
    296 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    297 ;
    298 ; BTVER2-LABEL: 'fmul'
    299 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = fmul float undef, undef
    300 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = fmul <4 x float> undef, undef
    301 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = fmul <8 x float> undef, undef
    302 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = fmul <16 x float> undef, undef
    303 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = fmul double undef, undef
    304 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = fmul <2 x double> undef, undef
    305 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = fmul <4 x double> undef, undef
    306 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = fmul <8 x double> undef, undef
    307 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    308 ;
    309   %F32 = fmul float undef, undef
    310   %V4F32 = fmul <4 x float> undef, undef
    311   %V8F32 = fmul <8 x float> undef, undef
    312   %V16F32 = fmul <16 x float> undef, undef
    313 
    314   %F64 = fmul double undef, undef
    315   %V2F64 = fmul <2 x double> undef, undef
    316   %V4F64 = fmul <4 x double> undef, undef
    317   %V8F64 = fmul <8 x double> undef, undef
    318 
    319   ret i32 undef
    320 }
    321 
    322 define i32 @fdiv(i32 %arg) {
    323 ; SSE2-LABEL: 'fdiv'
    324 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 23 for instruction: %F32 = fdiv float undef, undef
    325 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V4F32 = fdiv <4 x float> undef, undef
    326 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 78 for instruction: %V8F32 = fdiv <8 x float> undef, undef
    327 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 156 for instruction: %V16F32 = fdiv <16 x float> undef, undef
    328 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 38 for instruction: %F64 = fdiv double undef, undef
    329 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %V2F64 = fdiv <2 x double> undef, undef
    330 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 138 for instruction: %V4F64 = fdiv <4 x double> undef, undef
    331 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 276 for instruction: %V8F64 = fdiv <8 x double> undef, undef
    332 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    333 ;
    334 ; SSE42-LABEL: 'fdiv'
    335 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %F32 = fdiv float undef, undef
    336 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = fdiv <4 x float> undef, undef
    337 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = fdiv <8 x float> undef, undef
    338 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = fdiv <16 x float> undef, undef
    339 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %F64 = fdiv double undef, undef
    340 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V2F64 = fdiv <2 x double> undef, undef
    341 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V4F64 = fdiv <4 x double> undef, undef
    342 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %V8F64 = fdiv <8 x double> undef, undef
    343 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    344 ;
    345 ; AVX1-LABEL: 'fdiv'
    346 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %F32 = fdiv float undef, undef
    347 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = fdiv <4 x float> undef, undef
    348 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = fdiv <8 x float> undef, undef
    349 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = fdiv <16 x float> undef, undef
    350 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %F64 = fdiv double undef, undef
    351 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V2F64 = fdiv <2 x double> undef, undef
    352 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V4F64 = fdiv <4 x double> undef, undef
    353 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %V8F64 = fdiv <8 x double> undef, undef
    354 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    355 ;
    356 ; AVX2-LABEL: 'fdiv'
    357 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %F32 = fdiv float undef, undef
    358 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4F32 = fdiv <4 x float> undef, undef
    359 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8F32 = fdiv <8 x float> undef, undef
    360 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16F32 = fdiv <16 x float> undef, undef
    361 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %F64 = fdiv double undef, undef
    362 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V2F64 = fdiv <2 x double> undef, undef
    363 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V4F64 = fdiv <4 x double> undef, undef
    364 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V8F64 = fdiv <8 x double> undef, undef
    365 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    366 ;
    367 ; AVX512-LABEL: 'fdiv'
    368 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %F32 = fdiv float undef, undef
    369 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4F32 = fdiv <4 x float> undef, undef
    370 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8F32 = fdiv <8 x float> undef, undef
    371 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = fdiv <16 x float> undef, undef
    372 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %F64 = fdiv double undef, undef
    373 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V2F64 = fdiv <2 x double> undef, undef
    374 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V4F64 = fdiv <4 x double> undef, undef
    375 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = fdiv <8 x double> undef, undef
    376 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    377 ;
    378 ; SLM-LABEL: 'fdiv'
    379 ; SLM-NEXT:  Cost Model: Found an estimated cost of 17 for instruction: %F32 = fdiv float undef, undef
    380 ; SLM-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V4F32 = fdiv <4 x float> undef, undef
    381 ; SLM-NEXT:  Cost Model: Found an estimated cost of 78 for instruction: %V8F32 = fdiv <8 x float> undef, undef
    382 ; SLM-NEXT:  Cost Model: Found an estimated cost of 156 for instruction: %V16F32 = fdiv <16 x float> undef, undef
    383 ; SLM-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %F64 = fdiv double undef, undef
    384 ; SLM-NEXT:  Cost Model: Found an estimated cost of 69 for instruction: %V2F64 = fdiv <2 x double> undef, undef
    385 ; SLM-NEXT:  Cost Model: Found an estimated cost of 138 for instruction: %V4F64 = fdiv <4 x double> undef, undef
    386 ; SLM-NEXT:  Cost Model: Found an estimated cost of 276 for instruction: %V8F64 = fdiv <8 x double> undef, undef
    387 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    388 ;
    389 ; GLM-LABEL: 'fdiv'
    390 ; GLM-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %F32 = fdiv float undef, undef
    391 ; GLM-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %V4F32 = fdiv <4 x float> undef, undef
    392 ; GLM-NEXT:  Cost Model: Found an estimated cost of 70 for instruction: %V8F32 = fdiv <8 x float> undef, undef
    393 ; GLM-NEXT:  Cost Model: Found an estimated cost of 140 for instruction: %V16F32 = fdiv <16 x float> undef, undef
    394 ; GLM-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %F64 = fdiv double undef, undef
    395 ; GLM-NEXT:  Cost Model: Found an estimated cost of 65 for instruction: %V2F64 = fdiv <2 x double> undef, undef
    396 ; GLM-NEXT:  Cost Model: Found an estimated cost of 130 for instruction: %V4F64 = fdiv <4 x double> undef, undef
    397 ; GLM-NEXT:  Cost Model: Found an estimated cost of 260 for instruction: %V8F64 = fdiv <8 x double> undef, undef
    398 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    399 ;
    400 ; BTVER2-LABEL: 'fdiv'
    401 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %F32 = fdiv float undef, undef
    402 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = fdiv <4 x float> undef, undef
    403 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = fdiv <8 x float> undef, undef
    404 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = fdiv <16 x float> undef, undef
    405 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %F64 = fdiv double undef, undef
    406 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V2F64 = fdiv <2 x double> undef, undef
    407 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 44 for instruction: %V4F64 = fdiv <4 x double> undef, undef
    408 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 88 for instruction: %V8F64 = fdiv <8 x double> undef, undef
    409 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    410 ;
    411   %F32 = fdiv float undef, undef
    412   %V4F32 = fdiv <4 x float> undef, undef
    413   %V8F32 = fdiv <8 x float> undef, undef
    414   %V16F32 = fdiv <16 x float> undef, undef
    415 
    416   %F64 = fdiv double undef, undef
    417   %V2F64 = fdiv <2 x double> undef, undef
    418   %V4F64 = fdiv <4 x double> undef, undef
    419   %V8F64 = fdiv <8 x double> undef, undef
    420 
    421   ret i32 undef
    422 }
    423 
    424 define i32 @frem(i32 %arg) {
    425 ; SSE-LABEL: 'frem'
    426 ; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
    427 ; SSE-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
    428 ; SSE-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = frem <8 x float> undef, undef
    429 ; SSE-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = frem <16 x float> undef, undef
    430 ; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
    431 ; SSE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
    432 ; SSE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4F64 = frem <4 x double> undef, undef
    433 ; SSE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8F64 = frem <8 x double> undef, undef
    434 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    435 ;
    436 ; AVX-LABEL: 'frem'
    437 ; AVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
    438 ; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
    439 ; AVX-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V8F32 = frem <8 x float> undef, undef
    440 ; AVX-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V16F32 = frem <16 x float> undef, undef
    441 ; AVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
    442 ; AVX-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
    443 ; AVX-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4F64 = frem <4 x double> undef, undef
    444 ; AVX-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8F64 = frem <8 x double> undef, undef
    445 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    446 ;
    447 ; AVX512-LABEL: 'frem'
    448 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
    449 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
    450 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V8F32 = frem <8 x float> undef, undef
    451 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 62 for instruction: %V16F32 = frem <16 x float> undef, undef
    452 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
    453 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
    454 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4F64 = frem <4 x double> undef, undef
    455 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V8F64 = frem <8 x double> undef, undef
    456 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    457 ;
    458 ; SLM-LABEL: 'frem'
    459 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
    460 ; SLM-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
    461 ; SLM-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = frem <8 x float> undef, undef
    462 ; SLM-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = frem <16 x float> undef, undef
    463 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
    464 ; SLM-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
    465 ; SLM-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4F64 = frem <4 x double> undef, undef
    466 ; SLM-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8F64 = frem <8 x double> undef, undef
    467 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    468 ;
    469 ; GLM-LABEL: 'frem'
    470 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
    471 ; GLM-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
    472 ; GLM-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = frem <8 x float> undef, undef
    473 ; GLM-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = frem <16 x float> undef, undef
    474 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
    475 ; GLM-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
    476 ; GLM-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4F64 = frem <4 x double> undef, undef
    477 ; GLM-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8F64 = frem <8 x double> undef, undef
    478 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    479 ;
    480 ; BTVER2-LABEL: 'frem'
    481 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
    482 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = frem <4 x float> undef, undef
    483 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 30 for instruction: %V8F32 = frem <8 x float> undef, undef
    484 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %V16F32 = frem <16 x float> undef, undef
    485 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
    486 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
    487 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4F64 = frem <4 x double> undef, undef
    488 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8F64 = frem <8 x double> undef, undef
    489 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    490 ;
    491   %F32 = frem float undef, undef
    492   %V4F32 = frem <4 x float> undef, undef
    493   %V8F32 = frem <8 x float> undef, undef
    494   %V16F32 = frem <16 x float> undef, undef
    495 
    496   %F64 = frem double undef, undef
    497   %V2F64 = frem <2 x double> undef, undef
    498   %V4F64 = frem <4 x double> undef, undef
    499   %V8F64 = frem <8 x double> undef, undef
    500 
    501   ret i32 undef
    502 }
    503 
    504 define i32 @fsqrt(i32 %arg) {
    505 ; SSE2-LABEL: 'fsqrt'
    506 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
    507 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
    508 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 112 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
    509 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 224 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
    510 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
    511 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
    512 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
    513 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
    514 ; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    515 ;
    516 ; SSE42-LABEL: 'fsqrt'
    517 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
    518 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
    519 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
    520 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 72 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
    521 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
    522 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
    523 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
    524 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 128 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
    525 ; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    526 ;
    527 ; AVX1-LABEL: 'fsqrt'
    528 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
    529 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
    530 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
    531 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
    532 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
    533 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
    534 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
    535 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 86 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
    536 ; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    537 ;
    538 ; AVX2-LABEL: 'fsqrt'
    539 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
    540 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
    541 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
    542 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
    543 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
    544 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
    545 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
    546 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
    547 ; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    548 ;
    549 ; AVX512-LABEL: 'fsqrt'
    550 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
    551 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
    552 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
    553 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
    554 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
    555 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
    556 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
    557 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
    558 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    559 ;
    560 ; SLM-LABEL: 'fsqrt'
    561 ; SLM-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
    562 ; SLM-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
    563 ; SLM-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
    564 ; SLM-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
    565 ; SLM-NEXT:  Cost Model: Found an estimated cost of 35 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
    566 ; SLM-NEXT:  Cost Model: Found an estimated cost of 70 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
    567 ; SLM-NEXT:  Cost Model: Found an estimated cost of 140 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
    568 ; SLM-NEXT:  Cost Model: Found an estimated cost of 280 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
    569 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    570 ;
    571 ; GLM-LABEL: 'fsqrt'
    572 ; GLM-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
    573 ; GLM-NEXT:  Cost Model: Found an estimated cost of 37 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
    574 ; GLM-NEXT:  Cost Model: Found an estimated cost of 74 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
    575 ; GLM-NEXT:  Cost Model: Found an estimated cost of 148 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
    576 ; GLM-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
    577 ; GLM-NEXT:  Cost Model: Found an estimated cost of 67 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
    578 ; GLM-NEXT:  Cost Model: Found an estimated cost of 134 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
    579 ; GLM-NEXT:  Cost Model: Found an estimated cost of 268 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
    580 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    581 ;
    582 ; BTVER2-LABEL: 'fsqrt'
    583 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %F32 = call float @llvm.sqrt.f32(float undef)
    584 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 14 for instruction: %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
    585 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
    586 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
    587 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %F64 = call double @llvm.sqrt.f64(double undef)
    588 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
    589 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
    590 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 86 for instruction: %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
    591 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    592 ;
    593   %F32 = call float @llvm.sqrt.f32(float undef)
    594   %V4F32 = call <4 x float> @llvm.sqrt.v4f32(<4 x float> undef)
    595   %V8F32 = call <8 x float> @llvm.sqrt.v8f32(<8 x float> undef)
    596   %V16F32 = call <16 x float> @llvm.sqrt.v16f32(<16 x float> undef)
    597 
    598   %F64 = call double @llvm.sqrt.f64(double undef)
    599   %V2F64 = call <2 x double> @llvm.sqrt.v2f64(<2 x double> undef)
    600   %V4F64 = call <4 x double> @llvm.sqrt.v4f64(<4 x double> undef)
    601   %V8F64 = call <8 x double> @llvm.sqrt.v8f64(<8 x double> undef)
    602 
    603   ret i32 undef
    604 }
    605 
    606 define i32 @fabs(i32 %arg) {
    607 ; SSE-LABEL: 'fabs'
    608 ; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
    609 ; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
    610 ; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
    611 ; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
    612 ; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
    613 ; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
    614 ; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
    615 ; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
    616 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    617 ;
    618 ; AVX-LABEL: 'fabs'
    619 ; AVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
    620 ; AVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
    621 ; AVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
    622 ; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
    623 ; AVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
    624 ; AVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
    625 ; AVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
    626 ; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
    627 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    628 ;
    629 ; AVX512-LABEL: 'fabs'
    630 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
    631 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
    632 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
    633 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
    634 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
    635 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
    636 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
    637 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
    638 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    639 ;
    640 ; SLM-LABEL: 'fabs'
    641 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
    642 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
    643 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
    644 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
    645 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
    646 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
    647 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
    648 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
    649 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    650 ;
    651 ; GLM-LABEL: 'fabs'
    652 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
    653 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
    654 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
    655 ; GLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
    656 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
    657 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
    658 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
    659 ; GLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
    660 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    661 ;
    662 ; BTVER2-LABEL: 'fabs'
    663 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.fabs.f32(float undef)
    664 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
    665 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
    666 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
    667 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.fabs.f64(double undef)
    668 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
    669 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
    670 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
    671 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    672 ;
    673   %F32 = call float @llvm.fabs.f32(float undef)
    674   %V4F32 = call <4 x float> @llvm.fabs.v4f32(<4 x float> undef)
    675   %V8F32 = call <8 x float> @llvm.fabs.v8f32(<8 x float> undef)
    676   %V16F32 = call <16 x float> @llvm.fabs.v16f32(<16 x float> undef)
    677 
    678   %F64 = call double @llvm.fabs.f64(double undef)
    679   %V2F64 = call <2 x double> @llvm.fabs.v2f64(<2 x double> undef)
    680   %V4F64 = call <4 x double> @llvm.fabs.v4f64(<4 x double> undef)
    681   %V8F64 = call <8 x double> @llvm.fabs.v8f64(<8 x double> undef)
    682 
    683   ret i32 undef
    684 }
    685 
    686 define i32 @fcopysign(i32 %arg) {
    687 ; SSE-LABEL: 'fcopysign'
    688 ; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
    689 ; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
    690 ; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
    691 ; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
    692 ; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
    693 ; SSE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
    694 ; SSE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
    695 ; SSE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
    696 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    697 ;
    698 ; AVX-LABEL: 'fcopysign'
    699 ; AVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
    700 ; AVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
    701 ; AVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
    702 ; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
    703 ; AVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
    704 ; AVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
    705 ; AVX-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
    706 ; AVX-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
    707 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    708 ;
    709 ; AVX512-LABEL: 'fcopysign'
    710 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
    711 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
    712 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
    713 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
    714 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
    715 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
    716 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
    717 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
    718 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    719 ;
    720 ; SLM-LABEL: 'fcopysign'
    721 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
    722 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
    723 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
    724 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
    725 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
    726 ; SLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
    727 ; SLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
    728 ; SLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
    729 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    730 ;
    731 ; GLM-LABEL: 'fcopysign'
    732 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
    733 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
    734 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
    735 ; GLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
    736 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
    737 ; GLM-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
    738 ; GLM-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
    739 ; GLM-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
    740 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    741 ;
    742 ; BTVER2-LABEL: 'fcopysign'
    743 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
    744 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
    745 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
    746 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
    747 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
    748 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
    749 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
    750 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
    751 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    752 ;
    753   %F32 = call float @llvm.copysign.f32(float undef, float undef)
    754   %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
    755   %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
    756   %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
    757 
    758   %F64 = call double @llvm.copysign.f64(double undef, double undef)
    759   %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
    760   %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
    761   %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
    762 
    763   ret i32 undef
    764 }
    765 
    766 define i32 @fma(i32 %arg) {
    767 ; SSE-LABEL: 'fma'
    768 ; SSE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
    769 ; SSE-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
    770 ; SSE-NEXT:  Cost Model: Found an estimated cost of 86 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
    771 ; SSE-NEXT:  Cost Model: Found an estimated cost of 172 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
    772 ; SSE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
    773 ; SSE-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
    774 ; SSE-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
    775 ; SSE-NEXT:  Cost Model: Found an estimated cost of 84 for instruction: %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
    776 ; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    777 ;
    778 ; AVX-LABEL: 'fma'
    779 ; AVX-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
    780 ; AVX-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
    781 ; AVX-NEXT:  Cost Model: Found an estimated cost of 87 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
    782 ; AVX-NEXT:  Cost Model: Found an estimated cost of 174 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
    783 ; AVX-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
    784 ; AVX-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
    785 ; AVX-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
    786 ; AVX-NEXT:  Cost Model: Found an estimated cost of 86 for instruction: %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
    787 ; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    788 ;
    789 ; AVX512-LABEL: 'fma'
    790 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
    791 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
    792 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
    793 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
    794 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
    795 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
    796 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
    797 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
    798 ; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    799 ;
    800 ; SLM-LABEL: 'fma'
    801 ; SLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
    802 ; SLM-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
    803 ; SLM-NEXT:  Cost Model: Found an estimated cost of 86 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
    804 ; SLM-NEXT:  Cost Model: Found an estimated cost of 172 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
    805 ; SLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
    806 ; SLM-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
    807 ; SLM-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
    808 ; SLM-NEXT:  Cost Model: Found an estimated cost of 84 for instruction: %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
    809 ; SLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    810 ;
    811 ; GLM-LABEL: 'fma'
    812 ; GLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
    813 ; GLM-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
    814 ; GLM-NEXT:  Cost Model: Found an estimated cost of 86 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
    815 ; GLM-NEXT:  Cost Model: Found an estimated cost of 172 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
    816 ; GLM-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
    817 ; GLM-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
    818 ; GLM-NEXT:  Cost Model: Found an estimated cost of 42 for instruction: %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
    819 ; GLM-NEXT:  Cost Model: Found an estimated cost of 84 for instruction: %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
    820 ; GLM-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    821 ;
    822 ; BTVER2-LABEL: 'fma'
    823 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
    824 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
    825 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 87 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
    826 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 174 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
    827 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
    828 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
    829 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
    830 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 86 for instruction: %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
    831 ; BTVER2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
    832 ;
    833   %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
    834   %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
    835   %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
    836   %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
    837 
    838   %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
    839   %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
    840   %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
    841   %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
    842 
    843   ret i32 undef
    844 }
    845 
    846 declare float @llvm.sqrt.f32(float)
    847 declare <4 x float> @llvm.sqrt.v4f32(<4 x float>)
    848 declare <8 x float> @llvm.sqrt.v8f32(<8 x float>)
    849 declare <16 x float> @llvm.sqrt.v16f32(<16 x float>)
    850 
    851 declare double @llvm.sqrt.f64(double)
    852 declare <2 x double> @llvm.sqrt.v2f64(<2 x double>)
    853 declare <4 x double> @llvm.sqrt.v4f64(<4 x double>)
    854 declare <8 x double> @llvm.sqrt.v8f64(<8 x double>)
    855 
    856 declare float @llvm.fabs.f32(float)
    857 declare <4 x float> @llvm.fabs.v4f32(<4 x float>)
    858 declare <8 x float> @llvm.fabs.v8f32(<8 x float>)
    859 declare <16 x float> @llvm.fabs.v16f32(<16 x float>)
    860 
    861 declare double @llvm.fabs.f64(double)
    862 declare <2 x double> @llvm.fabs.v2f64(<2 x double>)
    863 declare <4 x double> @llvm.fabs.v4f64(<4 x double>)
    864 declare <8 x double> @llvm.fabs.v8f64(<8 x double>)
    865 
    866 declare float @llvm.copysign.f32(float, float)
    867 declare <4 x float> @llvm.copysign.v4f32(<4 x float>, <4 x float>)
    868 declare <8 x float> @llvm.copysign.v8f32(<8 x float>, <8 x float>)
    869 declare <16 x float> @llvm.copysign.v16f32(<16 x float>, <16 x float>)
    870 
    871 declare double @llvm.copysign.f64(double, double)
    872 declare <2 x double> @llvm.copysign.v2f64(<2 x double>, <2 x double>)
    873 declare <4 x double> @llvm.copysign.v4f64(<4 x double>, <4 x double>)
    874 declare <8 x double> @llvm.copysign.v8f64(<8 x double>, <8 x double>)
    875 
    876 declare float @llvm.fma.f32(float, float, float)
    877 declare <4 x float> @llvm.fma.v4f32(<4 x float>, <4 x float>, <4 x float>)
    878 declare <8 x float> @llvm.fma.v8f32(<8 x float>, <8 x float>, <8 x float>)
    879 declare <16 x float> @llvm.fma.v16f32(<16 x float>, <16 x float>, <16 x float>)
    880 
    881 declare double @llvm.fma.f64(double, double, double)
    882 declare <2 x double> @llvm.fma.v2f64(<2 x double>, <2 x double>, <2 x double>)
    883 declare <4 x double> @llvm.fma.v4f64(<4 x double>, <4 x double>, <4 x double>)
    884 declare <8 x double> @llvm.fma.v8f64(<8 x double>, <8 x double>, <8 x double>)
    885