Skip to content

Add F4E2M1FN and F8E8M0FNU types #21380

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
wants to merge 1 commit into from
Closed

Add F4E2M1FN and F8E8M0FNU types #21380

wants to merge 1 commit into from

Conversation

sergey-kozub
Copy link
Contributor

Previous PR #19096 was rolled back, re-trying.

This PR adds F4E2M1FN primitive type (4-bit float with 2 bits exponent and 1 bit mantissa), F8E8M0FNU primitive type (8-bit float with 8 bits exponent, no mantissa and no sign) and enables loads/stores in the same way S4/U4 type is implemented.

This will enable using microscaling (MX) formats (RFC), such as MXFP4.

F4E2M1FN
- Exponent bias: 1
- Maximum stored exponent value: 3 (binary 11)
- Maximum unbiased exponent value: 3 - 1 = 2
- Minimum stored exponent value: 1 (binary 01)
- Minimum unbiased exponent value: 11 = 0
- Has Positive and Negative zero
- Doesn't have infinity
- Doesn't have NaNs

Additional details:
- Zeros (+/-): S.00.0
- Max normal number: S.11.1 = ±2^(2) x (1 + 0.5) = ±6.0
- Min normal number: S.01.0 = ±2^(0) = ±1.0
- Min subnormal number: S.00.1 = ±2^(0) x 0.5 = ±0.5

F8E8M0FNU
- Exponent bias: 127
- Maximum stored exponent value: 254 (binary 1111'1110)
- Maximum unbiased exponent value: 254 - 127 = 127
- Minimum stored exponent value: 0 (binary 0000'0000)
- Minimum unbiased exponent value: 0127 = -127
- Doesn't have zero
- Doesn't have infinity
- NaN is encoded as binary 1111'1111

Additional details:
- Zeros cannot be represented
- Negative values cannot be represented
- Mantissa is always 1

Related PRs:

Copy link
Member

@reedwm reedwm left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for updating the PR!

I only skimmed it, but approving since it's a roll-forward of the already-approved PR #19096.

copybara-service bot pushed a commit that referenced this pull request Jan 14, 2025
Imported from GitHub PR #21380

Previous PR #19096 was rolled back, re-trying.

This PR adds F4E2M1FN primitive type (4-bit float with 2 bits exponent and 1 bit mantissa), F8E8M0FNU primitive type (8-bit float with 8 bits exponent, no mantissa and no sign) and enables loads/stores in the same way S4/U4 type is implemented.

This will enable using microscaling (MX) formats ([RFC](#18085)), such as MXFP4.

```c
F4E2M1FN
- Exponent bias: 1
- Maximum stored exponent value: 3 (binary 11)
- Maximum unbiased exponent value: 3 - 1 = 2
- Minimum stored exponent value: 1 (binary 01)
- Minimum unbiased exponent value: 1 − 1 = 0
- Has Positive and Negative zero
- Doesn't have infinity
- Doesn't have NaNs

Additional details:
- Zeros (+/-): S.00.0
- Max normal number: S.11.1 = ±2^(2) x (1 + 0.5) = ±6.0
- Min normal number: S.01.0 = ±2^(0) = ±1.0
- Min subnormal number: S.00.1 = ±2^(0) x 0.5 = ±0.5

F8E8M0FNU
- Exponent bias: 127
- Maximum stored exponent value: 254 (binary 1111'1110)
- Maximum unbiased exponent value: 254 - 127 = 127
- Minimum stored exponent value: 0 (binary 0000'0000)
- Minimum unbiased exponent value: 0 − 127 = -127
- Doesn't have zero
- Doesn't have infinity
- NaN is encoded as binary 1111'1111

Additional details:
- Zeros cannot be represented
- Negative values cannot be represented
- Mantissa is always 1
```

Related PRs:
- openxla/stablehlo#2582
- jax-ml/ml_dtypes#181
- llvm/llvm-project#95392
- llvm/llvm-project#108877
- jax-ml/ml_dtypes#166
- llvm/llvm-project#107127
- llvm/llvm-project#111028
Copybara import of the project:

--
d7e00c4 by Sergey Kozub <[email protected]>:

Add F4E2M1FN and F8E8M0FNU types

Merging this change closes #21380

FUTURE_COPYBARA_INTEGRATE_REVIEW=#21380 from openxla:skozub/e2m1_e8m0 d7e00c4
PiperOrigin-RevId: 715070992
copybara-service bot pushed a commit to tensorflow/tensorflow that referenced this pull request Jan 14, 2025
Imported from GitHub PR openxla/xla#21380

Previous PR openxla/xla#19096 was rolled back, re-trying.

This PR adds F4E2M1FN primitive type (4-bit float with 2 bits exponent and 1 bit mantissa), F8E8M0FNU primitive type (8-bit float with 8 bits exponent, no mantissa and no sign) and enables loads/stores in the same way S4/U4 type is implemented.

This will enable using microscaling (MX) formats ([RFC](openxla/xla#18085)), such as MXFP4.

```c
F4E2M1FN
- Exponent bias: 1
- Maximum stored exponent value: 3 (binary 11)
- Maximum unbiased exponent value: 3 - 1 = 2
- Minimum stored exponent value: 1 (binary 01)
- Minimum unbiased exponent value: 1 − 1 = 0
- Has Positive and Negative zero
- Doesn't have infinity
- Doesn't have NaNs

Additional details:
- Zeros (+/-): S.00.0
- Max normal number: S.11.1 = ±2^(2) x (1 + 0.5) = ±6.0
- Min normal number: S.01.0 = ±2^(0) = ±1.0
- Min subnormal number: S.00.1 = ±2^(0) x 0.5 = ±0.5

F8E8M0FNU
- Exponent bias: 127
- Maximum stored exponent value: 254 (binary 1111'1110)
- Maximum unbiased exponent value: 254 - 127 = 127
- Minimum stored exponent value: 0 (binary 0000'0000)
- Minimum unbiased exponent value: 0 − 127 = -127
- Doesn't have zero
- Doesn't have infinity
- NaN is encoded as binary 1111'1111

Additional details:
- Zeros cannot be represented
- Negative values cannot be represented
- Mantissa is always 1
```

Related PRs:
- openxla/stablehlo#2582
- jax-ml/ml_dtypes#181
- llvm/llvm-project#95392
- llvm/llvm-project#108877
- jax-ml/ml_dtypes#166
- llvm/llvm-project#107127
- llvm/llvm-project#111028
Copybara import of the project:

--
d7e00c49a4b4f26c06266d6bb941275e67464c01 by Sergey Kozub <[email protected]>:

Add F4E2M1FN and F8E8M0FNU types

Merging this change closes #21380

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#21380 from openxla:skozub/e2m1_e8m0 d7e00c49a4b4f26c06266d6bb941275e67464c01
PiperOrigin-RevId: 715070992
copybara-service bot pushed a commit that referenced this pull request Jan 14, 2025
Imported from GitHub PR #21380

Previous PR #19096 was rolled back, re-trying.

This PR adds F4E2M1FN primitive type (4-bit float with 2 bits exponent and 1 bit mantissa), F8E8M0FNU primitive type (8-bit float with 8 bits exponent, no mantissa and no sign) and enables loads/stores in the same way S4/U4 type is implemented.

This will enable using microscaling (MX) formats ([RFC](#18085)), such as MXFP4.

```c
F4E2M1FN
- Exponent bias: 1
- Maximum stored exponent value: 3 (binary 11)
- Maximum unbiased exponent value: 3 - 1 = 2
- Minimum stored exponent value: 1 (binary 01)
- Minimum unbiased exponent value: 1 − 1 = 0
- Has Positive and Negative zero
- Doesn't have infinity
- Doesn't have NaNs

Additional details:
- Zeros (+/-): S.00.0
- Max normal number: S.11.1 = ±2^(2) x (1 + 0.5) = ±6.0
- Min normal number: S.01.0 = ±2^(0) = ±1.0
- Min subnormal number: S.00.1 = ±2^(0) x 0.5 = ±0.5

F8E8M0FNU
- Exponent bias: 127
- Maximum stored exponent value: 254 (binary 1111'1110)
- Maximum unbiased exponent value: 254 - 127 = 127
- Minimum stored exponent value: 0 (binary 0000'0000)
- Minimum unbiased exponent value: 0 − 127 = -127
- Doesn't have zero
- Doesn't have infinity
- NaN is encoded as binary 1111'1111

Additional details:
- Zeros cannot be represented
- Negative values cannot be represented
- Mantissa is always 1
```

Related PRs:
- openxla/stablehlo#2582
- jax-ml/ml_dtypes#181
- llvm/llvm-project#95392
- llvm/llvm-project#108877
- jax-ml/ml_dtypes#166
- llvm/llvm-project#107127
- llvm/llvm-project#111028
Copybara import of the project:

--
d7e00c4 by Sergey Kozub <[email protected]>:

Add F4E2M1FN and F8E8M0FNU types

Merging this change closes #21380

FUTURE_COPYBARA_INTEGRATE_REVIEW=#21380 from openxla:skozub/e2m1_e8m0 d7e00c4
PiperOrigin-RevId: 715070992
copybara-service bot pushed a commit to tensorflow/tensorflow that referenced this pull request Jan 14, 2025
Imported from GitHub PR openxla/xla#21380

Previous PR openxla/xla#19096 was rolled back, re-trying.

This PR adds F4E2M1FN primitive type (4-bit float with 2 bits exponent and 1 bit mantissa), F8E8M0FNU primitive type (8-bit float with 8 bits exponent, no mantissa and no sign) and enables loads/stores in the same way S4/U4 type is implemented.

This will enable using microscaling (MX) formats ([RFC](openxla/xla#18085)), such as MXFP4.

```c
F4E2M1FN
- Exponent bias: 1
- Maximum stored exponent value: 3 (binary 11)
- Maximum unbiased exponent value: 3 - 1 = 2
- Minimum stored exponent value: 1 (binary 01)
- Minimum unbiased exponent value: 1 − 1 = 0
- Has Positive and Negative zero
- Doesn't have infinity
- Doesn't have NaNs

Additional details:
- Zeros (+/-): S.00.0
- Max normal number: S.11.1 = ±2^(2) x (1 + 0.5) = ±6.0
- Min normal number: S.01.0 = ±2^(0) = ±1.0
- Min subnormal number: S.00.1 = ±2^(0) x 0.5 = ±0.5

F8E8M0FNU
- Exponent bias: 127
- Maximum stored exponent value: 254 (binary 1111'1110)
- Maximum unbiased exponent value: 254 - 127 = 127
- Minimum stored exponent value: 0 (binary 0000'0000)
- Minimum unbiased exponent value: 0 − 127 = -127
- Doesn't have zero
- Doesn't have infinity
- NaN is encoded as binary 1111'1111

Additional details:
- Zeros cannot be represented
- Negative values cannot be represented
- Mantissa is always 1
```

Related PRs:
- openxla/stablehlo#2582
- jax-ml/ml_dtypes#181
- llvm/llvm-project#95392
- llvm/llvm-project#108877
- jax-ml/ml_dtypes#166
- llvm/llvm-project#107127
- llvm/llvm-project#111028
Copybara import of the project:

--
d7e00c49a4b4f26c06266d6bb941275e67464c01 by Sergey Kozub <[email protected]>:

Add F4E2M1FN and F8E8M0FNU types

Merging this change closes #21380

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#21380 from openxla:skozub/e2m1_e8m0 d7e00c49a4b4f26c06266d6bb941275e67464c01
PiperOrigin-RevId: 715070992
copybara-service bot pushed a commit that referenced this pull request Jan 14, 2025
Imported from GitHub PR #21380

Previous PR #19096 was rolled back, re-trying.

This PR adds F4E2M1FN primitive type (4-bit float with 2 bits exponent and 1 bit mantissa), F8E8M0FNU primitive type (8-bit float with 8 bits exponent, no mantissa and no sign) and enables loads/stores in the same way S4/U4 type is implemented.

This will enable using microscaling (MX) formats ([RFC](#18085)), such as MXFP4.

```c
F4E2M1FN
- Exponent bias: 1
- Maximum stored exponent value: 3 (binary 11)
- Maximum unbiased exponent value: 3 - 1 = 2
- Minimum stored exponent value: 1 (binary 01)
- Minimum unbiased exponent value: 1 − 1 = 0
- Has Positive and Negative zero
- Doesn't have infinity
- Doesn't have NaNs

Additional details:
- Zeros (+/-): S.00.0
- Max normal number: S.11.1 = ±2^(2) x (1 + 0.5) = ±6.0
- Min normal number: S.01.0 = ±2^(0) = ±1.0
- Min subnormal number: S.00.1 = ±2^(0) x 0.5 = ±0.5

F8E8M0FNU
- Exponent bias: 127
- Maximum stored exponent value: 254 (binary 1111'1110)
- Maximum unbiased exponent value: 254 - 127 = 127
- Minimum stored exponent value: 0 (binary 0000'0000)
- Minimum unbiased exponent value: 0 − 127 = -127
- Doesn't have zero
- Doesn't have infinity
- NaN is encoded as binary 1111'1111

Additional details:
- Zeros cannot be represented
- Negative values cannot be represented
- Mantissa is always 1
```

Related PRs:
- openxla/stablehlo#2582
- jax-ml/ml_dtypes#181
- llvm/llvm-project#95392
- llvm/llvm-project#108877
- jax-ml/ml_dtypes#166
- llvm/llvm-project#107127
- llvm/llvm-project#111028
Copybara import of the project:

--
d7e00c4 by Sergey Kozub <[email protected]>:

Add F4E2M1FN and F8E8M0FNU types

Merging this change closes #21380

FUTURE_COPYBARA_INTEGRATE_REVIEW=#21380 from openxla:skozub/e2m1_e8m0 d7e00c4
PiperOrigin-RevId: 715070992
copybara-service bot pushed a commit to tensorflow/tensorflow that referenced this pull request Jan 14, 2025
Imported from GitHub PR openxla/xla#21380

Previous PR openxla/xla#19096 was rolled back, re-trying.

This PR adds F4E2M1FN primitive type (4-bit float with 2 bits exponent and 1 bit mantissa), F8E8M0FNU primitive type (8-bit float with 8 bits exponent, no mantissa and no sign) and enables loads/stores in the same way S4/U4 type is implemented.

This will enable using microscaling (MX) formats ([RFC](openxla/xla#18085)), such as MXFP4.

```c
F4E2M1FN
- Exponent bias: 1
- Maximum stored exponent value: 3 (binary 11)
- Maximum unbiased exponent value: 3 - 1 = 2
- Minimum stored exponent value: 1 (binary 01)
- Minimum unbiased exponent value: 1 − 1 = 0
- Has Positive and Negative zero
- Doesn't have infinity
- Doesn't have NaNs

Additional details:
- Zeros (+/-): S.00.0
- Max normal number: S.11.1 = ±2^(2) x (1 + 0.5) = ±6.0
- Min normal number: S.01.0 = ±2^(0) = ±1.0
- Min subnormal number: S.00.1 = ±2^(0) x 0.5 = ±0.5

F8E8M0FNU
- Exponent bias: 127
- Maximum stored exponent value: 254 (binary 1111'1110)
- Maximum unbiased exponent value: 254 - 127 = 127
- Minimum stored exponent value: 0 (binary 0000'0000)
- Minimum unbiased exponent value: 0 − 127 = -127
- Doesn't have zero
- Doesn't have infinity
- NaN is encoded as binary 1111'1111

Additional details:
- Zeros cannot be represented
- Negative values cannot be represented
- Mantissa is always 1
```

Related PRs:
- openxla/stablehlo#2582
- jax-ml/ml_dtypes#181
- llvm/llvm-project#95392
- llvm/llvm-project#108877
- jax-ml/ml_dtypes#166
- llvm/llvm-project#107127
- llvm/llvm-project#111028
Copybara import of the project:

--
d7e00c49a4b4f26c06266d6bb941275e67464c01 by Sergey Kozub <[email protected]>:

Add F4E2M1FN and F8E8M0FNU types

Merging this change closes #21380

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#21380 from openxla:skozub/e2m1_e8m0 d7e00c49a4b4f26c06266d6bb941275e67464c01
PiperOrigin-RevId: 715070992
copybara-service bot pushed a commit that referenced this pull request Jan 14, 2025
Imported from GitHub PR #21380

Previous PR #19096 was rolled back, re-trying.

This PR adds F4E2M1FN primitive type (4-bit float with 2 bits exponent and 1 bit mantissa), F8E8M0FNU primitive type (8-bit float with 8 bits exponent, no mantissa and no sign) and enables loads/stores in the same way S4/U4 type is implemented.

This will enable using microscaling (MX) formats ([RFC](#18085)), such as MXFP4.

```c
F4E2M1FN
- Exponent bias: 1
- Maximum stored exponent value: 3 (binary 11)
- Maximum unbiased exponent value: 3 - 1 = 2
- Minimum stored exponent value: 1 (binary 01)
- Minimum unbiased exponent value: 1 − 1 = 0
- Has Positive and Negative zero
- Doesn't have infinity
- Doesn't have NaNs

Additional details:
- Zeros (+/-): S.00.0
- Max normal number: S.11.1 = ±2^(2) x (1 + 0.5) = ±6.0
- Min normal number: S.01.0 = ±2^(0) = ±1.0
- Min subnormal number: S.00.1 = ±2^(0) x 0.5 = ±0.5

F8E8M0FNU
- Exponent bias: 127
- Maximum stored exponent value: 254 (binary 1111'1110)
- Maximum unbiased exponent value: 254 - 127 = 127
- Minimum stored exponent value: 0 (binary 0000'0000)
- Minimum unbiased exponent value: 0 − 127 = -127
- Doesn't have zero
- Doesn't have infinity
- NaN is encoded as binary 1111'1111

Additional details:
- Zeros cannot be represented
- Negative values cannot be represented
- Mantissa is always 1
```

Related PRs:
- openxla/stablehlo#2582
- jax-ml/ml_dtypes#181
- llvm/llvm-project#95392
- llvm/llvm-project#108877
- jax-ml/ml_dtypes#166
- llvm/llvm-project#107127
- llvm/llvm-project#111028
Copybara import of the project:

--
d7e00c4 by Sergey Kozub <[email protected]>:

Add F4E2M1FN and F8E8M0FNU types

Merging this change closes #21380

FUTURE_COPYBARA_INTEGRATE_REVIEW=#21380 from openxla:skozub/e2m1_e8m0 d7e00c4
PiperOrigin-RevId: 715070992
copybara-service bot pushed a commit to tensorflow/tensorflow that referenced this pull request Jan 14, 2025
Imported from GitHub PR openxla/xla#21380

Previous PR openxla/xla#19096 was rolled back, re-trying.

This PR adds F4E2M1FN primitive type (4-bit float with 2 bits exponent and 1 bit mantissa), F8E8M0FNU primitive type (8-bit float with 8 bits exponent, no mantissa and no sign) and enables loads/stores in the same way S4/U4 type is implemented.

This will enable using microscaling (MX) formats ([RFC](openxla/xla#18085)), such as MXFP4.

```c
F4E2M1FN
- Exponent bias: 1
- Maximum stored exponent value: 3 (binary 11)
- Maximum unbiased exponent value: 3 - 1 = 2
- Minimum stored exponent value: 1 (binary 01)
- Minimum unbiased exponent value: 1 − 1 = 0
- Has Positive and Negative zero
- Doesn't have infinity
- Doesn't have NaNs

Additional details:
- Zeros (+/-): S.00.0
- Max normal number: S.11.1 = ±2^(2) x (1 + 0.5) = ±6.0
- Min normal number: S.01.0 = ±2^(0) = ±1.0
- Min subnormal number: S.00.1 = ±2^(0) x 0.5 = ±0.5

F8E8M0FNU
- Exponent bias: 127
- Maximum stored exponent value: 254 (binary 1111'1110)
- Maximum unbiased exponent value: 254 - 127 = 127
- Minimum stored exponent value: 0 (binary 0000'0000)
- Minimum unbiased exponent value: 0 − 127 = -127
- Doesn't have zero
- Doesn't have infinity
- NaN is encoded as binary 1111'1111

Additional details:
- Zeros cannot be represented
- Negative values cannot be represented
- Mantissa is always 1
```

Related PRs:
- openxla/stablehlo#2582
- jax-ml/ml_dtypes#181
- llvm/llvm-project#95392
- llvm/llvm-project#108877
- jax-ml/ml_dtypes#166
- llvm/llvm-project#107127
- llvm/llvm-project#111028
Copybara import of the project:

--
d7e00c49a4b4f26c06266d6bb941275e67464c01 by Sergey Kozub <[email protected]>:

Add F4E2M1FN and F8E8M0FNU types

Merging this change closes #21380

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#21380 from openxla:skozub/e2m1_e8m0 d7e00c49a4b4f26c06266d6bb941275e67464c01
PiperOrigin-RevId: 715070992
copybara-service bot pushed a commit that referenced this pull request Jan 14, 2025
Imported from GitHub PR #21380

Previous PR #19096 was rolled back, re-trying.

This PR adds F4E2M1FN primitive type (4-bit float with 2 bits exponent and 1 bit mantissa), F8E8M0FNU primitive type (8-bit float with 8 bits exponent, no mantissa and no sign) and enables loads/stores in the same way S4/U4 type is implemented.

This will enable using microscaling (MX) formats ([RFC](#18085)), such as MXFP4.

```c
F4E2M1FN
- Exponent bias: 1
- Maximum stored exponent value: 3 (binary 11)
- Maximum unbiased exponent value: 3 - 1 = 2
- Minimum stored exponent value: 1 (binary 01)
- Minimum unbiased exponent value: 1 − 1 = 0
- Has Positive and Negative zero
- Doesn't have infinity
- Doesn't have NaNs

Additional details:
- Zeros (+/-): S.00.0
- Max normal number: S.11.1 = ±2^(2) x (1 + 0.5) = ±6.0
- Min normal number: S.01.0 = ±2^(0) = ±1.0
- Min subnormal number: S.00.1 = ±2^(0) x 0.5 = ±0.5

F8E8M0FNU
- Exponent bias: 127
- Maximum stored exponent value: 254 (binary 1111'1110)
- Maximum unbiased exponent value: 254 - 127 = 127
- Minimum stored exponent value: 0 (binary 0000'0000)
- Minimum unbiased exponent value: 0 − 127 = -127
- Doesn't have zero
- Doesn't have infinity
- NaN is encoded as binary 1111'1111

Additional details:
- Zeros cannot be represented
- Negative values cannot be represented
- Mantissa is always 1
```

Related PRs:
- openxla/stablehlo#2582
- jax-ml/ml_dtypes#181
- llvm/llvm-project#95392
- llvm/llvm-project#108877
- jax-ml/ml_dtypes#166
- llvm/llvm-project#107127
- llvm/llvm-project#111028
Copybara import of the project:

--
d7e00c4 by Sergey Kozub <[email protected]>:

Add F4E2M1FN and F8E8M0FNU types

Merging this change closes #21380

FUTURE_COPYBARA_INTEGRATE_REVIEW=#21380 from openxla:skozub/e2m1_e8m0 d7e00c4
PiperOrigin-RevId: 715070992
copybara-service bot pushed a commit to tensorflow/tensorflow that referenced this pull request Jan 14, 2025
Imported from GitHub PR openxla/xla#21380

Previous PR openxla/xla#19096 was rolled back, re-trying.

This PR adds F4E2M1FN primitive type (4-bit float with 2 bits exponent and 1 bit mantissa), F8E8M0FNU primitive type (8-bit float with 8 bits exponent, no mantissa and no sign) and enables loads/stores in the same way S4/U4 type is implemented.

This will enable using microscaling (MX) formats ([RFC](openxla/xla#18085)), such as MXFP4.

```c
F4E2M1FN
- Exponent bias: 1
- Maximum stored exponent value: 3 (binary 11)
- Maximum unbiased exponent value: 3 - 1 = 2
- Minimum stored exponent value: 1 (binary 01)
- Minimum unbiased exponent value: 1 − 1 = 0
- Has Positive and Negative zero
- Doesn't have infinity
- Doesn't have NaNs

Additional details:
- Zeros (+/-): S.00.0
- Max normal number: S.11.1 = ±2^(2) x (1 + 0.5) = ±6.0
- Min normal number: S.01.0 = ±2^(0) = ±1.0
- Min subnormal number: S.00.1 = ±2^(0) x 0.5 = ±0.5

F8E8M0FNU
- Exponent bias: 127
- Maximum stored exponent value: 254 (binary 1111'1110)
- Maximum unbiased exponent value: 254 - 127 = 127
- Minimum stored exponent value: 0 (binary 0000'0000)
- Minimum unbiased exponent value: 0 − 127 = -127
- Doesn't have zero
- Doesn't have infinity
- NaN is encoded as binary 1111'1111

Additional details:
- Zeros cannot be represented
- Negative values cannot be represented
- Mantissa is always 1
```

Related PRs:
- openxla/stablehlo#2582
- jax-ml/ml_dtypes#181
- llvm/llvm-project#95392
- llvm/llvm-project#108877
- jax-ml/ml_dtypes#166
- llvm/llvm-project#107127
- llvm/llvm-project#111028
Copybara import of the project:

--
d7e00c49a4b4f26c06266d6bb941275e67464c01 by Sergey Kozub <[email protected]>:

Add F4E2M1FN and F8E8M0FNU types

Merging this change closes #21380

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#21380 from openxla:skozub/e2m1_e8m0 d7e00c49a4b4f26c06266d6bb941275e67464c01
PiperOrigin-RevId: 715070992
copybara-service bot pushed a commit to tensorflow/mlir-hlo that referenced this pull request Jan 14, 2025
Imported from GitHub PR openxla/xla#21380

Previous PR openxla/xla#19096 was rolled back, re-trying.

This PR adds F4E2M1FN primitive type (4-bit float with 2 bits exponent and 1 bit mantissa), F8E8M0FNU primitive type (8-bit float with 8 bits exponent, no mantissa and no sign) and enables loads/stores in the same way S4/U4 type is implemented.

This will enable using microscaling (MX) formats ([RFC](openxla/xla#18085)), such as MXFP4.

```c
F4E2M1FN
- Exponent bias: 1
- Maximum stored exponent value: 3 (binary 11)
- Maximum unbiased exponent value: 3 - 1 = 2
- Minimum stored exponent value: 1 (binary 01)
- Minimum unbiased exponent value: 1 − 1 = 0
- Has Positive and Negative zero
- Doesn't have infinity
- Doesn't have NaNs

Additional details:
- Zeros (+/-): S.00.0
- Max normal number: S.11.1 = ±2^(2) x (1 + 0.5) = ±6.0
- Min normal number: S.01.0 = ±2^(0) = ±1.0
- Min subnormal number: S.00.1 = ±2^(0) x 0.5 = ±0.5

F8E8M0FNU
- Exponent bias: 127
- Maximum stored exponent value: 254 (binary 1111'1110)
- Maximum unbiased exponent value: 254 - 127 = 127
- Minimum stored exponent value: 0 (binary 0000'0000)
- Minimum unbiased exponent value: 0 − 127 = -127
- Doesn't have zero
- Doesn't have infinity
- NaN is encoded as binary 1111'1111

Additional details:
- Zeros cannot be represented
- Negative values cannot be represented
- Mantissa is always 1
```

Related PRs:
- openxla/stablehlo#2582
- jax-ml/ml_dtypes#181
- llvm/llvm-project#95392
- llvm/llvm-project#108877
- jax-ml/ml_dtypes#166
- llvm/llvm-project#107127
- llvm/llvm-project#111028
Copybara import of the project:

--
d7e00c49a4b4f26c06266d6bb941275e67464c01 by Sergey Kozub <[email protected]>:

Add F4E2M1FN and F8E8M0FNU types

Merging this change closes #21380

PiperOrigin-RevId: 715434229
copybara-service bot pushed a commit to tensorflow/tensorflow that referenced this pull request Jan 14, 2025
Imported from GitHub PR openxla/xla#21380

Previous PR openxla/xla#19096 was rolled back, re-trying.

This PR adds F4E2M1FN primitive type (4-bit float with 2 bits exponent and 1 bit mantissa), F8E8M0FNU primitive type (8-bit float with 8 bits exponent, no mantissa and no sign) and enables loads/stores in the same way S4/U4 type is implemented.

This will enable using microscaling (MX) formats ([RFC](openxla/xla#18085)), such as MXFP4.

```c
F4E2M1FN
- Exponent bias: 1
- Maximum stored exponent value: 3 (binary 11)
- Maximum unbiased exponent value: 3 - 1 = 2
- Minimum stored exponent value: 1 (binary 01)
- Minimum unbiased exponent value: 1 − 1 = 0
- Has Positive and Negative zero
- Doesn't have infinity
- Doesn't have NaNs

Additional details:
- Zeros (+/-): S.00.0
- Max normal number: S.11.1 = ±2^(2) x (1 + 0.5) = ±6.0
- Min normal number: S.01.0 = ±2^(0) = ±1.0
- Min subnormal number: S.00.1 = ±2^(0) x 0.5 = ±0.5

F8E8M0FNU
- Exponent bias: 127
- Maximum stored exponent value: 254 (binary 1111'1110)
- Maximum unbiased exponent value: 254 - 127 = 127
- Minimum stored exponent value: 0 (binary 0000'0000)
- Minimum unbiased exponent value: 0 − 127 = -127
- Doesn't have zero
- Doesn't have infinity
- NaN is encoded as binary 1111'1111

Additional details:
- Zeros cannot be represented
- Negative values cannot be represented
- Mantissa is always 1
```

Related PRs:
- openxla/stablehlo#2582
- jax-ml/ml_dtypes#181
- llvm/llvm-project#95392
- llvm/llvm-project#108877
- jax-ml/ml_dtypes#166
- llvm/llvm-project#107127
- llvm/llvm-project#111028
Copybara import of the project:

--
d7e00c49a4b4f26c06266d6bb941275e67464c01 by Sergey Kozub <[email protected]>:

Add F4E2M1FN and F8E8M0FNU types

Merging this change closes #21380

PiperOrigin-RevId: 715434229
copybara-service bot pushed a commit to tensorflow/tensorflow that referenced this pull request Jan 14, 2025
…ution error

* When the CPU client has an input buffer's definition event available, do not
ignore it if the event has an error. Instead, take such as definition event as
an input dependency, which will poison output buffers. This behavior is
consistent with how input error buffers poison output buffers.

* If an execution is enqueued before the previous execution has not finished,
and the first execution fails with an error, the second execution does not fail
accidentally. This is done by clearing any error set in `GetLastEnqueueEvent()`
because this last enqueue event is used for force sequential execution, not to
propagate errors across independently enqueued computations.

* An error buffer is created a dummy internal buffer in
`TfrtCpuClient::CreateErrorBuffer`, which meets internal invariants that even
an error `TfrtCpuBuffer` has a buffer and can handle metadata operations
including `TfrtCpuBuffer::BufferSizes()`.

FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#21380 from openxla:skozub/e2m1_e8m0 d7e00c49a4b4f26c06266d6bb941275e67464c01
PiperOrigin-RevId: 715022528
@sergey-kozub sergey-kozub deleted the skozub/e2m1_e8m0 branch January 29, 2025 11:38
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants