Spell out the "literal" operators.

[sunfishcode]

Also, remove a broken link to the binary encoding page. The binary
encoding is not yet determined, but AstSemantics.md doesn't need to
reference it right now anyway.

See also WebAssembly/spec#88

[lukewagner]

Looks good to me, although I wonder if we could get by with the slightly more compact lit.

[sunfishcode]

I'm ok with lit myself.

[rossberg]

A bit more terminology nitpicking from me :) : "literal" usually refers to the textual representation of something. Again this is not really the right level to talk about these opcodes. I would rather call them constants.

[sunfishcode]

"Constant" means unchanging, which can include local variables or other things. "Literals" (or even "immediates") properly express the property that the value is being provided immediately with the instruction.

[rossberg]

That's "constant" as an adjective. Constant as a noun denotes a constant value, I'd say. ;)

[sunfishcode]

I frequently use "constant" as a noun to refer to non-immediate constants. If "literal" sounds too much like a text thing, that's why it's my second choice after "immediate" ;).

[jfbastien]

Should we explicitly point out that this preserve NaN sign, signal/quiet, and payload?

[sunfishcode]

Sounds reasonable. I added a mention of NaN values of all possible bit patterns.

WebAssembly presently has no concept of signaling vs. quiet NaN, so I think it's sufficient to just say all possible bit patterns.

[jfbastien]

lgtm with either term.

I'll suggest cons and walk away slowly.

[rossberg]

On 29 September 2015 at 21:20, JF Bastien notifications@github.com wrote:

I'll suggest cons and walk away slowly.

That would of course have very confusing connotations with the major
S-expression languages. But with that extra t I'd be happy. :)

FWIW, 'constant' also is the terminology that both the JVM (ldc and
{i,l,f,d}const instructions) and the CLR (ldc instruction) use in this
context.

[jfbastien]

I'll suggest cons and walk away slowly.

That would of course have very confusing connotations with the major
S-expression languages. But with that extra t I'd be happy. :)

If that wasn't clear, I was trolling FP programmers, and it looks like I hit my mark dead on :-)

May I instead suggest we use :: ?

[titzer]

"const" also sounds fine to me. v8-native-prototype uses internal names
"I8Const", "I32Const", etc.

On Wed, Sep 30, 2015 at 5:26 PM, rossberg-chromium <notifications@github.com

wrote:

On 29 September 2015 at 21:20, JF Bastien notifications@github.com
wrote:

I'll suggest cons and walk away slowly.

That would of course have very confusing connotations with the major
S-expression languages. But with that extra t I'd be happy. :)

FWIW, 'constant' also is the terminology that both the JVM (ldc and
{i,l,f,d}const instructions) and the CLR (ldc instruction) use in this
context.

—
Reply to this email directly or view it on GitHub
#380 (comment).

[sunfishcode]

Ok. I added a patch to this PR to use const.

[jfbastien]

lgtm, though it's a bit confusion to talk about immediates in the English text, and const as the opcode and header :)

[sunfishcode]

The value is given in an immediate field of the const instruction :).

[rossberg]

lgtm. Thanks!