Add float8_e4m3

[apivovarov]

This PR adds f8E4M3 type.

f8E4M3 type follows IEEE 754 convention

f8E4M3 (IEEE 754)
- Exponent bias: 7
- Maximum stored exponent value: 14 (binary 1110)
- Maximum unbiased exponent value: 14 - 7 = 7
- Minimum stored exponent value: 1 (binary 0001)
- Minimum unbiased exponent value: 1 − 7 = −6
- Precision specifies the total number of bits used for the significand (mantisa), 
    including implicit leading integer bit = 3 + 1 = 4
- Follows IEEE 754 conventions for representation of special values
- Has Positive and Negative zero
- Has Positive and Negative infinity
- Has NaNs

Additional details:
- Max exp (unbiased): 7
- Min exp (unbiased): -6
- Infinities (+/-): S.1111.000
- Zeros (+/-): S.0000.000
- NaNs: S.1111.{001, 010, 011, 100, 101, 110, 111}
- Max normal number: S.1110.111 = +/-2^(7) x (1 + 0.875) = +/-240
- Min normal number: S.0001.000 = +/-2^(-6)
- Max subnormal number: S.0000.111 = +/-2^(-6) x 0.875 = +/-2^(-9) x 7
- Min subnormal number: S.0000.001 = +/-2^(-6) x 0.125 = +/-2^(-9)

Related LLVM PRs:

• LLVM PR-97179 [APFloat] Add support for f8E4M3 IEEE 754 type (Merged)
• LLVM PR-97118 [MLIR] Add f8E4M3 IEEE 754 type (Merged)

RFCs:

• StableHLO PR-2486 [RFC] Add f8E4M3 and f8E3M4 types support

Related ml_dtypes PRs:

• PR-57 Add float8_e4m3fnuz and float8_e5m2fnuz

C++ Testing

Tested as described in PR-123 Add CMakeLists for C++ tests

cmake --build build -- all test

100% tests passed, 0 tests failed out of 250

Total Test time (real) =  12.50 sec

[apivovarov]

Peter, Jake, whenever you have time, could you please review this pull request?
f8E4M3 support has already been merged to LLVM APFloat and MLIR.
I also need to add f8E4M3 to XLA. Internally XLA uses tsl which depends on ml_dtypes - tsl/platform/ml_dtypes.h

@hawkinsp @jakevdp

[hawkinsp]

@apivovarov I'm on vacation this week, but I'll take a look next week. Sorry for the delay...

[apivovarov]

@apivovarov I'm on vacation this week, but I'll take a look next week. Sorry for the delay...

Hi Peter, it seems this PR got overlooked. @hawkinsp

BTW, I've also attached a link to the StableHLO [RFC] Add f8E4M3 and f8E3M4 types support openxla/stablehlo#2486

Not sure what change should be merged first StableHLO or ml_dtypes.

[hawkinsp]

This looks good to me. Would you mind also adding an entry to the changelog?

[hawkinsp]

The linter is also sad, please fix.

[hawkinsp]

And it doesn't matter whether the stablehlo change or the ml_dtypes change is merged first, really, they aren't directly coupled.

[hawkinsp]

Linter is still sad.

[apivovarov]

Linter is still sad.

Fixed. pre-commit run --all-files - all green now