Add i32x4.dot_i16x8_s (#2420)

This experimental instruction is specified in
https://github.com/WebAssembly/simd/pull/127 and is being implemented
to enable further investigation of its performance impact.

1 files changed, 67 insertions, 61 deletions

diff --git a/test/simd.wast.from-wast b/test/simd.wast.from-wast
index 9cbf42d71..b5914a62d 100644
--- a/test/simd.wast.from-wast
+++ b/test/simd.wast.from-wast
@@ -713,336 +713,342 @@
    (local.get $1)
   )
  )
- (func $i64x2.neg (; 123 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $i32x4.dot_i16x8_s (; 123 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+  (i32x4.dot_i16x8_s
+   (local.get $0)
+   (local.get $1)
+  )
+ )
+ (func $i64x2.neg (; 124 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (i64x2.neg
    (local.get $0)
   )
  )
- (func $i64x2.any_true (; 124 ;) (type $FUNCSIG$iV) (param $0 v128) (result i32)
+ (func $i64x2.any_true (; 125 ;) (type $FUNCSIG$iV) (param $0 v128) (result i32)
   (i64x2.any_true
    (local.get $0)
   )
  )
- (func $i64x2.all_true (; 125 ;) (type $FUNCSIG$iV) (param $0 v128) (result i32)
+ (func $i64x2.all_true (; 126 ;) (type $FUNCSIG$iV) (param $0 v128) (result i32)
   (i64x2.all_true
    (local.get $0)
   )
  )
- (func $i64x2.shl (; 126 ;) (type $FUNCSIG$VVi) (param $0 v128) (param $1 i32) (result v128)
+ (func $i64x2.shl (; 127 ;) (type $FUNCSIG$VVi) (param $0 v128) (param $1 i32) (result v128)
   (i64x2.shl
    (local.get $0)
    (local.get $1)
   )
  )
- (func $i64x2.shr_s (; 127 ;) (type $FUNCSIG$VVi) (param $0 v128) (param $1 i32) (result v128)
+ (func $i64x2.shr_s (; 128 ;) (type $FUNCSIG$VVi) (param $0 v128) (param $1 i32) (result v128)
   (i64x2.shr_s
    (local.get $0)
    (local.get $1)
   )
  )
- (func $i64x2.shr_u (; 128 ;) (type $FUNCSIG$VVi) (param $0 v128) (param $1 i32) (result v128)
+ (func $i64x2.shr_u (; 129 ;) (type $FUNCSIG$VVi) (param $0 v128) (param $1 i32) (result v128)
   (i64x2.shr_u
    (local.get $0)
    (local.get $1)
   )
  )
- (func $i64x2.add (; 129 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $i64x2.add (; 130 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (i64x2.add
    (local.get $0)
    (local.get $1)
   )
  )
- (func $i64x2.sub (; 130 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $i64x2.sub (; 131 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (i64x2.sub
    (local.get $0)
    (local.get $1)
   )
  )
- (func $f32x4.add (; 131 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $f32x4.add (; 132 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (f32x4.add
    (local.get $0)
    (local.get $1)
   )
  )
- (func $f32x4.sub (; 132 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $f32x4.sub (; 133 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (f32x4.sub
    (local.get $0)
    (local.get $1)
   )
  )
- (func $f32x4.mul (; 133 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $f32x4.mul (; 134 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (f32x4.mul
    (local.get $0)
    (local.get $1)
   )
  )
- (func $f32x4.div (; 134 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $f32x4.div (; 135 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (f32x4.div
    (local.get $0)
    (local.get $1)
   )
  )
- (func $f32x4.min (; 135 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $f32x4.min (; 136 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (f32x4.min
    (local.get $0)
    (local.get $1)
   )
  )
- (func $f32x4.max (; 136 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $f32x4.max (; 137 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (f32x4.max
    (local.get $0)
    (local.get $1)
   )
  )
- (func $f32x4.abs (; 137 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $f32x4.abs (; 138 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (f32x4.abs
    (local.get $0)
   )
  )
- (func $f32x4.neg (; 138 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $f32x4.neg (; 139 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (f32x4.neg
    (local.get $0)
   )
  )
- (func $f32x4.sqrt (; 139 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $f32x4.sqrt (; 140 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (f32x4.sqrt
    (local.get $0)
   )
  )
- (func $f32x4.qfma (; 140 ;) (type $FUNCSIG$VVVV) (param $0 v128) (param $1 v128) (param $2 v128) (result v128)
+ (func $f32x4.qfma (; 141 ;) (type $FUNCSIG$VVVV) (param $0 v128) (param $1 v128) (param $2 v128) (result v128)
   (f32x4.qfma
    (local.get $0)
    (local.get $1)
    (local.get $2)
   )
  )
- (func $f32x4.qfms (; 141 ;) (type $FUNCSIG$VVVV) (param $0 v128) (param $1 v128) (param $2 v128) (result v128)
+ (func $f32x4.qfms (; 142 ;) (type $FUNCSIG$VVVV) (param $0 v128) (param $1 v128) (param $2 v128) (result v128)
   (f32x4.qfms
    (local.get $0)
    (local.get $1)
    (local.get $2)
   )
  )
- (func $f64x2.add (; 142 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $f64x2.add (; 143 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (f64x2.add
    (local.get $0)
    (local.get $1)
   )
  )
- (func $f64x2.sub (; 143 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $f64x2.sub (; 144 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (f64x2.sub
    (local.get $0)
    (local.get $1)
   )
  )
- (func $f64x2.mul (; 144 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $f64x2.mul (; 145 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (f64x2.mul
    (local.get $0)
    (local.get $1)
   )
  )
- (func $f64x2.div (; 145 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $f64x2.div (; 146 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (f64x2.div
    (local.get $0)
    (local.get $1)
   )
  )
- (func $f64x2.min (; 146 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $f64x2.min (; 147 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (f64x2.min
    (local.get $0)
    (local.get $1)
   )
  )
- (func $f64x2.max (; 147 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $f64x2.max (; 148 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (f64x2.max
    (local.get $0)
    (local.get $1)
   )
  )
- (func $f64x2.abs (; 148 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $f64x2.abs (; 149 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (f64x2.abs
    (local.get $0)
   )
  )
- (func $f64x2.neg (; 149 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $f64x2.neg (; 150 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (f64x2.neg
    (local.get $0)
   )
  )
- (func $f64x2.sqrt (; 150 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $f64x2.sqrt (; 151 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (f64x2.sqrt
    (local.get $0)
   )
  )
- (func $f64x2.qfma (; 151 ;) (type $FUNCSIG$VVVV) (param $0 v128) (param $1 v128) (param $2 v128) (result v128)
+ (func $f64x2.qfma (; 152 ;) (type $FUNCSIG$VVVV) (param $0 v128) (param $1 v128) (param $2 v128) (result v128)
   (f64x2.qfma
    (local.get $0)
    (local.get $1)
    (local.get $2)
   )
  )
- (func $f64x2.qfms (; 152 ;) (type $FUNCSIG$VVVV) (param $0 v128) (param $1 v128) (param $2 v128) (result v128)
+ (func $f64x2.qfms (; 153 ;) (type $FUNCSIG$VVVV) (param $0 v128) (param $1 v128) (param $2 v128) (result v128)
   (f64x2.qfms
    (local.get $0)
    (local.get $1)
    (local.get $2)
   )
  )
- (func $i32x4.trunc_sat_f32x4_s (; 153 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $i32x4.trunc_sat_f32x4_s (; 154 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (i32x4.trunc_sat_f32x4_s
    (local.get $0)
   )
  )
- (func $i32x4.trunc_sat_f32x4_u (; 154 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $i32x4.trunc_sat_f32x4_u (; 155 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (i32x4.trunc_sat_f32x4_u
    (local.get $0)
   )
  )
- (func $i64x2.trunc_sat_f64x2_s (; 155 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $i64x2.trunc_sat_f64x2_s (; 156 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (i64x2.trunc_sat_f64x2_s
    (local.get $0)
   )
  )
- (func $i64x2.trunc_sat_f64x2_u (; 156 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $i64x2.trunc_sat_f64x2_u (; 157 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (i64x2.trunc_sat_f64x2_u
    (local.get $0)
   )
  )
- (func $f32x4.convert_i32x4_s (; 157 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $f32x4.convert_i32x4_s (; 158 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (f32x4.convert_i32x4_s
    (local.get $0)
   )
  )
- (func $f32x4.convert_i32x4_u (; 158 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $f32x4.convert_i32x4_u (; 159 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (f32x4.convert_i32x4_u
    (local.get $0)
   )
  )
- (func $f64x2.convert_i64x2_s (; 159 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $f64x2.convert_i64x2_s (; 160 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (f64x2.convert_i64x2_s
    (local.get $0)
   )
  )
- (func $f64x2.convert_i64x2_u (; 160 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $f64x2.convert_i64x2_u (; 161 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (f64x2.convert_i64x2_u
    (local.get $0)
   )
  )
- (func $v8x16.load_splat (; 161 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
+ (func $v8x16.load_splat (; 162 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
   (v8x16.load_splat
    (local.get $0)
   )
  )
- (func $v16x8.load_splat (; 162 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
+ (func $v16x8.load_splat (; 163 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
   (v16x8.load_splat
    (local.get $0)
   )
  )
- (func $v32x4.load_splat (; 163 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
+ (func $v32x4.load_splat (; 164 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
   (v32x4.load_splat
    (local.get $0)
   )
  )
- (func $v64x2.load_splat (; 164 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
+ (func $v64x2.load_splat (; 165 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
   (v64x2.load_splat
    (local.get $0)
   )
  )
- (func $i8x16.narrow_i16x8_s (; 165 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $i8x16.narrow_i16x8_s (; 166 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (i8x16.narrow_i16x8_s
    (local.get $0)
    (local.get $1)
   )
  )
- (func $i8x16.narrow_i16x8_u (; 166 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $i8x16.narrow_i16x8_u (; 167 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (i8x16.narrow_i16x8_u
    (local.get $0)
    (local.get $1)
   )
  )
- (func $i16x8.narrow_i32x4_s (; 167 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $i16x8.narrow_i32x4_s (; 168 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (i16x8.narrow_i32x4_s
    (local.get $0)
    (local.get $1)
   )
  )
- (func $i16x8.narrow_i32x4_u (; 168 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $i16x8.narrow_i32x4_u (; 169 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (i16x8.narrow_i32x4_u
    (local.get $0)
    (local.get $1)
   )
  )
- (func $i16x8.widen_low_i8x16_s (; 169 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $i16x8.widen_low_i8x16_s (; 170 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (i16x8.widen_low_i8x16_s
    (local.get $0)
   )
  )
- (func $i16x8.widen_high_i8x16_s (; 170 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $i16x8.widen_high_i8x16_s (; 171 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (i16x8.widen_high_i8x16_s
    (local.get $0)
   )
  )
- (func $i16x8.widen_low_i8x16_u (; 171 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $i16x8.widen_low_i8x16_u (; 172 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (i16x8.widen_low_i8x16_u
    (local.get $0)
   )
  )
- (func $i16x8.widen_high_i8x16_u (; 172 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $i16x8.widen_high_i8x16_u (; 173 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (i16x8.widen_high_i8x16_u
    (local.get $0)
   )
  )
- (func $i32x4.widen_low_i16x8_s (; 173 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $i32x4.widen_low_i16x8_s (; 174 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (i32x4.widen_low_i16x8_s
    (local.get $0)
   )
  )
- (func $i32x4.widen_high_i16x8_s (; 174 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $i32x4.widen_high_i16x8_s (; 175 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (i32x4.widen_high_i16x8_s
    (local.get $0)
   )
  )
- (func $i32x4.widen_low_i16x8_u (; 175 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $i32x4.widen_low_i16x8_u (; 176 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (i32x4.widen_low_i16x8_u
    (local.get $0)
   )
  )
- (func $i32x4.widen_high_i16x8_u (; 176 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
+ (func $i32x4.widen_high_i16x8_u (; 177 ;) (type $FUNCSIG$VV) (param $0 v128) (result v128)
   (i32x4.widen_high_i16x8_u
    (local.get $0)
   )
  )
- (func $i16x8.load8x8_u (; 177 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
+ (func $i16x8.load8x8_u (; 178 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
   (i16x8.load8x8_u
    (local.get $0)
   )
  )
- (func $i16x8.load8x8_s (; 178 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
+ (func $i16x8.load8x8_s (; 179 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
   (i16x8.load8x8_s
    (local.get $0)
   )
  )
- (func $i32x4.load16x4_s (; 179 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
+ (func $i32x4.load16x4_s (; 180 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
   (i32x4.load16x4_s
    (local.get $0)
   )
  )
- (func $i32x4.load16x4_u (; 180 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
+ (func $i32x4.load16x4_u (; 181 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
   (i32x4.load16x4_u
    (local.get $0)
   )
  )
- (func $i64x2.load32x2_s (; 181 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
+ (func $i64x2.load32x2_s (; 182 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
   (i64x2.load32x2_s
    (local.get $0)
   )
  )
- (func $i64x2.load32x2_u (; 182 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
+ (func $i64x2.load32x2_u (; 183 ;) (type $FUNCSIG$Vi) (param $0 i32) (result v128)
   (i64x2.load32x2_u
    (local.get $0)
   )
  )
- (func $v8x16.swizzle (; 183 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
+ (func $v8x16.swizzle (; 184 ;) (type $FUNCSIG$VVV) (param $0 v128) (param $1 v128) (result v128)
   (v8x16.swizzle
    (local.get $0)
    (local.get $1)