1518289 - ARM64: Implement codegen for visitRoundF().

Status: RESOLVED FIXED

(Core :: JavaScript Engine: JIT, enhancement, P2)

Description

[Sean Stangl [:sstangl]]

Attachment: 0001-Implement-ARM64-CodeGenerator-visitRoundF-.-r-nbp.patch

Just like the visitRound() patch, but for 32-bit floating point.

Comment 1

[Nicolas B. Pierron [:nbp]]

Comment on attachment 9034822 [details] [diff] [review]
0001-Implement-ARM64-CodeGenerator-visitRoundF-.-r-nbp.patch

Review of attachment 9034822 [details] [diff] [review]:
-----------------------------------------------------------------

Looks good to me, nothing more to say expect that this look like a proper port from visitRound to visitRoundF.

::: js/src/jit/arm64/CodeGenerator-arm64.cpp
@@ +1052,5 @@
> +    // In the case of zero, the input may have been NaN or -0, which must bail.
> +    masm.branch32(Assembler::NotEqual, output, Imm32(0), &done);
> +    {
> +      // If input is NaN, comparisons set the C and V bits of the NZCV flags.
> +      masm.Fcmp(ARMFPRegister(input), 0.0f);

note: Nice touch on the 0.0f :)  (hopefully C++ should coerce it to a double at compile time to match Fcmp second parameter type)

Comment 2

[Sean Stangl [:sstangl]]

Attachment: 0001-Bug-1518289-Implement-ARM64-CodeGenerator-visitRound.patch

Same patch, but avoids redundantly creating an ARMFPRegister that already exists.

Comment 3

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/52e46924c0bc
Implement ARM64 CodeGenerator::visitRoundF(). r=nbp

Comment 4

[Razvan Maries]

https://hg.mozilla.org/mozilla-central/rev/52e46924c0bc