Add support for hex float literals #1420
Conversation
Fixes gpuweb#709 * Adds literal tokenization to support hex floats
|
This is the hex float part: The exponent suffix is non-optional in this tokenization. The parentheses are matched such that the hex-digit-sequences are a combination that must be preceded by 0x and following by the exponent. Simplifying it: |
|
Discussed in the WGSL meeting of 2021-02-16. Consensus to approve the content of this on the understanding that it reuses C99 syntax. |
* Split FLOAT_LITERAL into DECIMAL_FLOAT_LITERAL and HEX_FLOAT_LITERAL * replace uses of FLOAT_LITERAL with both new tokens
* revert majority of the split of float literal * Add a grammar for FLOAT_LITERAL composed of HEX_FLOAT_LITERAL and DECIMAL_FLOAT_LITERAL
| </thead> | ||
| <tr><td>`FLOAT_LITERAL`<td>`(-?[0-9]*.[0-9]+ | -?[0-9]+.[0-9]*)((e|E)(+|-)?[0-9]+)?` | ||
| <tr><td>`DECIMAL_FLOAT_LITERAL`<td>`(-?[0-9]*.[0-9]+ | -?[0-9]+.[0-9]*)((e|E)(+|-)?[0-9]+)?` | ||
| <tr><td>`HEX_FLOAT_LITERAL`<td>`-?0x([0-9a-fA-F]*.?[0-9a-fA-F]+ | [0-9a-fA-F]+.[0-9a-fA-F]*)(p|P)(+|-)?[0-9]+` |
There was a problem hiding this comment.
Comparing to C99 draft (http://www.open-std.org/jtc1/sc22/WG14/www/docs/n1256.pdf 6.4.42):
-
should allow both 0x and 0X. This worth raising to the group in a new issue. I filed allow both 0x and 0X as hexadecimal literal prefix #1453
-
the significand expression can be simplified to:
([0-9a-fA-F]*.[0-9a-fA-F]+ | [0-9a-fA-F]+.)
In fact, the same simplification can be applied to the decimal float literal.
- C99 doesn't include the leading minus.
The last two points are already tracked in #775
This PR adds unimplemented stubs for the `textureSampleCompareLevel` builtin. Issue gpuweb#1269
Fixes #709