extmul_high_i8x16_u: Wasm SIMD arithmetic instruction
The extmul_high_i8x16_u SIMD arithmetic instruction takes lanes 8–15 of two unsigned v128 i8x16 value interpretations, multiplies the values in the corresponding lanes, and outputs the result of those operations into an i16x8 value interpretation.
Try it
(module
(import "console" "log" (func $log (param i32)))
(func $main
v128.const i8x16 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3
v128.const i8x16 4 4 4 4 4 4 4 4 5 5 5 5 5 5 5 5
i16x8.extmul_high_i8x16_u
i16x8.extract_lane_s 7
call $log ;; log the result
)
(start $main)
)
WebAssembly.instantiateStreaming(fetch("{%wasm-url%}"), { console });
In the above example, lanes 8–15 of the two i8x16 input values are multiplied together and the products output as an i16x8. Lane 8 of the first input is multiplied by lane 8 of the second input, and the product becomes lane 0 of the output, and so on. As a result, each lane of the output contains the value 15 (3 * 5).
The extmul_high_i8x16_u instruction is a more performant equivalent to passing the results of two extend_high_i8x16_u instructions into a mul instruction.
In other words:
(i16x8.extmul_high_i8x16_u
(input1)
(input2)
)
is equivalent to
(i16x8.mul
(i16x8.extend_high_i8x16_u
(input1)
)
(i16x8.extend_high_i8x16_u
(input2)
)
)
Syntax
i16x8.extmul_high_i8x16_u
i16x8.extmul_high_i8x16_u-
The
i16x8.extmul_high_i8x16_uinstruction.
Type
[input1, input2] -> [output]
Binary encoding
| Instruction | Binary format | Example text => binary |
|---|---|---|
i16x8.extmul_high_i8x16_u |
0xfd 159:u32 |
i16x8.extmul_high_i8x16_u => 0xfd 0x9f 0x01 |