Implement more SIMD instructions

[dcodeIO]

#508 adds an initial set of SIMD operations as built-in functions that map directly to WebAssembly instructions. Given those as a starting point, it should be relatively straight forward to add the remaining ones.

Instruction	Built-in	Status
v128.const	i8x16 (=v128), i16x8, i32x4, i64x2, f32x4, f64x2	done
i8x16.splat	v128.splat<i8> + alias	done
i16x8.splat	v128.splat<i16> + alias	done
i32x4.splat	v128.splat<i32> + alias	done
i64x2.splat	v128.splat<i64> + alias	done
f32x4.splat	v128.splat<f32> + alias	done
f64x2.splat	v128.splat<f64> + alias	done
v128.load	i8x16 (=v128), i16x8, i32x4, i64x2, f32x4, f64x2	done
v128.store	i8x16 (=v128), i16x8, i32x4, i64x2, f32x4, f64x2	done
i8x16.extract_lane_s	v128.extract_lane<i8> + alias	done
i8x16.extract_lane_u	v128.extract_lane<u8> + alias	done
i16x8.extract_lane_s	v128.extract_lane<i16> + alias	done
i16x8.extract_lane_u	v128.extract_lane<u16> + alias	done
i32x4.extract_lane	v128.extract_lane<i/u32> + alias	done
i64x2.extract_lane	v128.extract_lane<i/u64> + alias	done
f32x4.extract_lane	v128.extract_lane<f32> + alias	done
f64x2.extract_lane	v128.extract_lane<f64> + alias	done
i8x16.replace_lane	v128.replace_lane<i/u8> + alias	done
i16x8.replace_lane	v128.replace_lane<i/u16> + alias	done
i32x4.replace_lane	v128.replace_lane<i/u32> + alias	done
i64x2.replace_lane	v128.replace_lane<i/u64> + alias	done
f32x4.replace_lane	v128.replace_lane<f32> + alias	done
f64x2.replace_lane	v128.replace_lane<f64> + alias	done
i8x16.shuffle	v128.shuffle<i/u8/16/32/f32/64> + aliases	done
i8x16.bitselect	v128.bitselect<i/u8/16/32/f32/64> + aliases	done
i8x16.any_true	v128.any_true<i/u8> + alias	done
i16x8.any_true	v128.any_true<i/u16> + alias	done
i32x4.any_true	v128.any_true<i/u32> + alias	done
i64x2.any_true	v128.any_true<i/u64> + alias	done
i8x16.all_true	v128.all_true<i/u8> + alias	done
i16x8.all_true	v128.all_true<i/u16> + alias	done
i32x4.all_true	v128.all_true<i/u32> + alias	done
i64x2.all_true	v128.all_true<i/u64> + alias	done
i8x16.shl	v128.shl<i/u8> + alias	done
i16x8.shl	v128.shl<i/u16> + alias	done
i32x4.shl	v128.shl<i/u32> + alias	done
i64x2.shl	v128.shl<i/u64> + alias	done
i8x16.shr	v128.shr<i/u8> + alias	done
i16x8.shr	v128.shr<i/u16> + alias	done
i32x4.shr	v128.shr<i/u32> + alias	done
i64x2.shr	v128.shr<i/u64> + alias	done
i8x16.and	v128.and<i/u8> + alias	done
i16x8.and	v128.and<i/u16> + alias	done
i32x4.and	v128.and<i/u32> + alias	done
i64x2.and	v128.and<i/u64> + alias	done
i8x16.or	v128.or<i/u8> + alias	done
i16x8.or	v128.or<i/u16> + alias	done
i32x4.or	v128.or<i/u32> + alias	done
i64x2.or	v128.or<i/u64> + alias	done
i8x16.xor	v128.xor<i/u8> + alias	done
i16x8.xor	v128.xor<i/u16> + alias	done
i32x4.xor	v128.xor<i/u32> + alias	done
i64x2.xor	v128.xor<i/u64> + alias	done
i8x16.not	v128.not<i/u8> + alias	done
i16x8.not	v128.not<i/u16> + alias	done
i32x4.not	v128.not<i/u32> + alias	done
i64x2.not	v128.not<i/u64> + alias	done
i8x16.neg	v128.neg<i/u8> + alias	done
i16x8.neg	v128.neg<i/u16> + alias	done
i32x4.neg	v128.neg<i/u32> + alias	done
i64x2.neg	v128.neg<i/u64> + alias	done
f32x4.neg	v128.neg<f32> + alias	done
f64x2.neg	v128.neg<f64> + alias	done
i8x16.add	v128.add<i/u8> + alias	done
i16x8.add	v128.add<i/u16> + alias	done
i32x4.add	v128.add<i/u32> + alias	done
i64x2.add	v128.add<i/u64> + alias	done
f32x4.add	v128.add<f32> + alias	done
f64x2.add	v128.add<f64> + alias	done
i8x16.sub	v128.sub<i/u8> + alias	done
i16x8.sub	v128.sub<i/u16> + alias	done
i32x4.sub	v128.sub<i/u32> + alias	done
i64x2.sub	v128.sub<i/u64> + alias	done
f32x4.sub	v128.sub<f32> + alias	done
f64x2.sub	v128.sub<f64> + alias	done
i8x16.mul	v128.mul<i/u8> + alias	done
i16x8.mul	v128.mul<i/u16> + alias	done
i32x4.mul	v128.mul<i/u32> + alias	done
i64x2.mul	v128.mul<i/u64> + alias	done
f32x4.mul	v128.mul<f32> + alias	done
f64x2.mul	v128.mul<f64> + alias	done
f32x4.div	v128.div<f32> + alias	done
f64x2.div	v128.div<f64> + alias	done
f32x4.min	v128.min<f32> + alias	done
f64x2.min	v128.min<f64> + alias	done
f32x4.max	v128.max<f32> + alias	done
f64x2.max	v128.max<f64> + alias	done
i8x16.add_saturate	v128.add_saturate<i/u8> + alias	done
i16x8.add_saturate	v128.add_saturate<i/u16> + alias	done
i8x16.sub_saturate	v128.sub_saturate<i/u8> + alias	done
i16x8.sub_saturate	v128.sub_saturate<i/u16> + alias	done
...	...	todo

See also: https://github.com/WebAssembly/simd/blob/master/proposals/simd/SIMD.md

For f32x4, i32x4, etc. I suggest to base class simdop that can be implementable.
For v8x16, v4x32, etc. I suggest another base class, or implement in base class as shuffle4, shuffle8, etc.

[dcodeIO]

Regarding shuffle, and that's what I hinted at here, I think it should be implemented similar to what's done for the other instructions already, even though the underlying instruction operates on 16 lanes exclusively. That is v128.shuffle<T> operating on blocks of lanes of sizeof<T> bytes, with v128.shuffle<i8> naturally operating on all 16 lanes, which is equivalent to the v8x16.shuffle instruction. All possible combinations from i8x16.shuffle, i16x8.shuffle up to f64x2.shuffle could then be defered to the generic built-in as aliases, just as a user might expect from the design of other instructions.

[dcodeIO]

Ok, shuffle is generic as well now, with the following aliases:

• i8x16.shuffle using lanes 0..15, 16..31 of 1 byte each (just as v8x16.shuffle)
• i16x8.shuffle using lanes 0..7, 8..15 of 2 bytes each
• i32x4.shuffle / f32x4.shuffle using lanes 0..3, 4..7 of 4 bytes each
• i64x2.shuffle / f64x2.shuffle using lanes 0..1, 2..3 of 8 bytes each

All of these map to v128.shuffle<T> that takes the respective number of lanes and index ranges depending on T.

[MaxGraey]

@dcodeIO For minimal set I think need implement load/store as well and this pretty enough for experimental and demonstration proposes. Rest of intrinsics could be implement later

[dcodeIO]

Looks like Binaryen simply handles v128 in a normal load/store, so all we have to do there is alias :)

[dcodeIO]

Closing as I'm pretty close to the full set of instructions already. Figured that this work is utterly boring and that nobody would voluntarily jump on this train anyway.