Deno•4y ago

Are pointers supported in Fast API return types?

If no, how do I return pointers?

NeTTOP•10/15/22, 8:34 AM

Nevermind I realized that I can just cast the pointer to an integer and make it work

TTyler Laceby Can I return a pointer to dynamically allocated data for example and store that ...

NeTTOP•10/15/22, 8:36 AM

From what I've seen, it was easier to pass the typed array as a parameter and let the native code do the allocation directly into the typed array

NNeTT If no, how do I return pointers?

NeTTOP•10/15/22, 8:37 AM

now I have a new question. How do I know my FFI call is utilising Fast API?

NeTTOP•10/15/22, 8:43 AM

#[no_mangle]
pub unsafe extern "C" fn no_op() -> isize {
  -97
}

#[no_mangle]
pub unsafe extern "C" fn no_op() -> isize {
  -97
}

With

no_op: {
  parameters: [],
  result: "i64",
},

no_op: {
  parameters: [],
  result: "i64",
},

Takes

26.3ns

26.3ns

on average.

NeTTOP•10/15/22, 8:44 AM

While

no_op: {
  parameters: [],
  result: "pointer",
},

no_op: {
  parameters: [],
  result: "pointer",
},

Takes

16-17ns

16-17ns

on average.

NeTTOP•10/15/22, 8:48 AM

no_op_i64    26.49 ns/iter   (24.97 ns … 70.14 ns)  25.66 ns  43.22 ns  48.23 ns

no_op_ptr    17.58 ns/iter   (16.54 ns … 45.23 ns)  17.03 ns  27.49 ns  28.51 ns

no_op_i64    26.49 ns/iter   (24.97 ns … 70.14 ns)  25.66 ns  43.22 ns  48.23 ns

no_op_ptr    17.58 ns/iter   (16.54 ns … 45.23 ns)  17.03 ns  27.49 ns  28.51 ns

NeTTOP•10/15/22, 8:48 AM

copypasted from

deno bench

deno bench

TTyler Laceby Cause it’s negative I guess it’s just weird for a pointer lol

NeTTOP•10/15/22, 8:49 AM

returning

instead of

-97

-97

is twice as fast btw

NeTTOP•10/15/22, 8:51 AM

no idea

NeTTOP•10/15/22, 8:52 AM

an ACTUAL no-op takes like

NeTTOP•10/15/22, 8:52 AM

4ns

NeTTOP•10/15/22, 8:52 AM

with

#[no_mangle]
pub unsafe extern "C" fn no_op() {
}

#[no_mangle]
pub unsafe extern "C" fn no_op() {
}

NeTTOP•10/15/22, 8:53 AM

and returntype

void

void

when defining symbols

NeTTOP•10/15/22, 9:02 AM

Ig this is actually going through fast api considering my older no-op calls (from a few weeks ago) were taking around

40ns

40ns

AapoAlas•10/15/22, 9:05 AM

Pointers are supported as return types, as are 64 bit integers. Under the hood it works so that an extra buffer argument is passed into the call, our trampoline writes the pointer / integer into the buffer, and then on JS side we check (for positive integers) if the number has no upper bits in which case we return the lower 32 bits as a number. Otherwise, we take a BigInt out of the buffer and check if it can be safely represented as a number. If it can, we return a number. Else, the BigInt.

AAapoAlas Pointers are supported as return types, as are 64 bit integers. Under the hood i...

NeTTOP•10/15/22, 9:11 AM

so uhh

NeTTOP•10/15/22, 9:11 AM

can I make an FFI call explicitly use the slow API?

NeTTOP•10/15/22, 9:11 AM

for comparison

NNeTT can I make an FFI call explicitly use the slow API?

AapoAlas•10/15/22, 9:13 AM

Yes, mark it with

callback: true

callback: true

. Fast API will throw if you try to call into JS so we need a way to mark functions that may call back and use slow calls for those.

NeTTOP•10/15/22, 9:15 AM

yea thanks that works

NeTTOP•10/15/22, 9:15 AM

can see a 10x difference between the fast call and the slow call

AapoAlas•10/15/22, 9:16 AM

You can also force a deopt to slow by using eg. null or Uint32Array for a buffer type, or null for pointer etc.

NeTTOP•10/15/22, 9:16 AM

NeTTOP•10/15/22, 9:18 AM

(that 3.61ns was a fluke)

NeTTOP•10/15/22, 9:18 AM

or wait

NeTTOP•10/15/22, 9:18 AM

NeTTOP•10/15/22, 9:18 AM

becomes slower when called after a dummy bench?

AapoAlas•10/15/22, 9:22 AM

Bench has some issues with the very first bench call.

NeTTOP•10/15/22, 9:23 AM

btw, could something here cause the call to use the ops layer instead of the fast api?

call_fn: {
      parameters: ["buffer", "buffer", "u32", "u32"],
      result: "pointer",
},

call_fn: {
      parameters: ["buffer", "buffer", "u32", "u32"],
      result: "pointer",
},

AapoAlas•10/15/22, 9:26 AM

If either of the first two parameters is not a Uint8Array then it'll deopt to the ops layer.

AapoAlas•10/15/22, 9:27 AM

(Well, not quite ops layer but similar.)

AAapoAlas If either of the first two parameters is not a Uint8Array then it'll deopt to th...

NeTTOP•10/15/22, 9:30 AM

theyre both technically Uint8Arrays

NeTTOP•10/15/22, 9:31 AM

const x_for_ffi = new Uint8Array(x.buffer);
const y_for_ffi = new Uint8Array(y.buffer);

const x_for_ffi = new Uint8Array(x.buffer);
const y_for_ffi = new Uint8Array(y.buffer);

NeTTOP•10/15/22, 9:31 AM

these two variables are the first two params

AapoAlas•10/15/22, 9:32 AM

Okay, that should be fine. Though do note that it's been seen elsewhere that doing

new Uint8Array(buffer)

new Uint8Array(buffer)

can be surprisingly slow. If you can move it up out of the loop / bench it will probably help a lot.

NeTTOP•10/15/22, 9:34 AM

actually now that I see it,

same params but I removed the function body from the rust function

NeTTOP•10/15/22, 9:34 AM

without removing function body from rust function

NeTTOP•10/15/22, 9:38 AM

okay seems some loops were taking too much time