Prefetch instructions

Author: Maratyszcza

proposals/simd/BinarySIMD.md

@@ -222,3 +222,5 @@ For example, `ImmLaneIdx16` is a byte with values in the range 0-15 (inclusive).
 | `i32x4.trunc_sat_f32x4_u`  |    `0xf9`| -                  |
 | `f32x4.convert_i32x4_s`    |    `0xfa`| -                  |
 | `f32x4.convert_i32x4_u`    |    `0xfb`| -                  |
+| `prefetch.t`               |    `0xfc`| -                  |
+| `prefetch.nt`              |    `0xfd`| -                  |

proposals/simd/NewOpcodes.md

@@ -128,3 +128,8 @@
 | i32x4.trunc_sat_f32x4_u | 0xf9   |
 | f32x4.convert_i32x4_s   | 0xfa   |
 | f32x4.convert_i32x4_u   | 0xfb   |
+
+| Prefetch Op | opcode |
+| ------------| ------ |
+| prefetch.t  | 0xfc   |
+| prefetch.nt | 0xfd   |

proposals/simd/SIMD.md

@@ -1023,3 +1023,13 @@ def S.widen_low_T_u(a):
 def S.widen_high_T_u(a):
     return S.widen_high_T(Zext, a)
 ```
+
+### Prefetch
+
+* `prefetch.t(memarg)`
+* `prefetch.nt(memarg)`
+
+Prefetch memory location into lowest-level cache. Prefetch with temporal hint
+(`prefetch.t`) should be used for data that will be used only once, and
+prefetch with non-temporal (`prefetch.nt`) hint for data that will be reused
+multiple times.