[Mesa-dev] [PATCH 01/10] i965: Add functions to convert float <-> VF.

[Francisco Jerez]

Matt Turner <mattst88 at gmail.com> writes:

> ---
>  src/mesa/drivers/dri/i965/Makefile.sources   |  1 +
>  src/mesa/drivers/dri/i965/brw_packed_float.c | 74 ++++++++++++++++++++++++++++
>  src/mesa/drivers/dri/i965/brw_packed_float.h | 25 ++++++++++
>  3 files changed, 100 insertions(+)
>  create mode 100644 src/mesa/drivers/dri/i965/brw_packed_float.c
>  create mode 100644 src/mesa/drivers/dri/i965/brw_packed_float.h
>
> diff --git a/src/mesa/drivers/dri/i965/Makefile.sources b/src/mesa/drivers/dri/i965/Makefile.sources
> index 9c006da..6b0f601 100644
> --- a/src/mesa/drivers/dri/i965/Makefile.sources
> +++ b/src/mesa/drivers/dri/i965/Makefile.sources
> @@ -78,6 +78,7 @@ i965_FILES = \
>  	brw_meta_fast_clear.c \
>  	brw_misc_state.c \
>  	brw_object_purgeable.c \
> +	brw_packed_float.c \
>  	brw_performance_monitor.c \
>  	brw_program.c \
>  	brw_primitive_restart.c \
> diff --git a/src/mesa/drivers/dri/i965/brw_packed_float.c b/src/mesa/drivers/dri/i965/brw_packed_float.c
> new file mode 100644
> index 0000000..d25e7dd
> --- /dev/null
> +++ b/src/mesa/drivers/dri/i965/brw_packed_float.c
> @@ -0,0 +1,74 @@
> +/*
> + * Copyright © 2014 Intel Corporation
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a
> + * copy of this software and associated documentation files (the "Software"),
> + * to deal in the Software without restriction, including without limitation
> + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> + * and/or sell copies of the Software, and to permit persons to whom the
> + * Software is furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice (including the next
> + * paragraph) shall be included in all copies or substantial portions of the
> + * Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
> + * IN THE SOFTWARE.
> + */
> +#include "brw_packed_float.h"
> +
> +union fu {
> +   float f;
> +   unsigned u;
> +   struct {
> +      unsigned mantissa:23;
> +      unsigned exponent:8;
> +      unsigned sign:1;
> +   };
> +};

Whew, you managed to rely on three different kinds of unspecified
behavior with this ;).  Writing the value of a union member and reading
back a different one, assuming that 'f' is IEEE 754, and that bitfields
are laid out in memory LSB to MSB packed in the same dword.  I'm not
saying that any of these assumptions won't hold on any of the platforms
we care about, but wouldn't it be better style and easier to just use
frexp() and ldexp()?

> +
> +int
> +brw_float_to_vf(float f)
> +{

Are you aware of the int_to_float8() function in brw_reg.h?  It's less
powerful than yours because it doesn't handle floating point arguments
so you could probably get rid of it -- though I think its name makes
more sense because you only get a single 8-bit float as a result from
this function rather than a full VF.

> +   union fu fu = { .f = f };
> +
> +   /* ±0.0f is special cased. */
> +   if (f == 0.0f)
> +      return fu.sign << 7;
> +
> +   unsigned mantissa = fu.mantissa >> (23 - 4);
> +   unsigned exponent = fu.exponent - (127 - 3);
> +   unsigned vf = (fu.sign << 7) | (exponent << 4) | mantissa;
> +
> +   /* 0.125 would have had the same representation as 0.0, so reject it. */
> +   if (vf & 0x7f)

As Ville mentioned already this is most likely not what you want.

> +      return -1;
> +
> +   if (fu.u & 0x7ffff || exponent >= 8)
> +      return -1;
> +
> +   return vf;
> +}
> +
> +float
> +brw_vf_to_float(unsigned char vf)
> +{
> +   union fu fu;
> +
> +   /* ±0.0f is special cased. */
> +   if (vf == 0x00 || vf == 0x80) {
> +      fu.u = vf << 24;
> +      return fu.f;
> +   }

No need to special-case 0, AFAICT this just repeats the same you're
doing below.

> +
> +   fu.sign = vf >> 7;
> +   fu.exponent = ((vf & 0xf0) >> 4) + (127 - 3);

The sign bit is not getting masked off as Ken said.

> +   fu.mantissa = (vf & 0xf) << (23 - 4);
> +
> +   return fu.f;
> +}
> diff --git a/src/mesa/drivers/dri/i965/brw_packed_float.h b/src/mesa/drivers/dri/i965/brw_packed_float.h
> new file mode 100644
> index 0000000..8b6825f
> --- /dev/null
> +++ b/src/mesa/drivers/dri/i965/brw_packed_float.h
> @@ -0,0 +1,25 @@
> +/*
> + * Copyright © 2014 Intel Corporation
> + *
> + * Permission is hereby granted, free of charge, to any person obtaining a
> + * copy of this software and associated documentation files (the "Software"),
> + * to deal in the Software without restriction, including without limitation
> + * the rights to use, copy, modify, merge, publish, distribute, sublicense,
> + * and/or sell copies of the Software, and to permit persons to whom the
> + * Software is furnished to do so, subject to the following conditions:
> + *
> + * The above copyright notice and this permission notice (including the next
> + * paragraph) shall be included in all copies or substantial portions of the
> + * Software.
> + *
> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
> + * IN THE SOFTWARE.
> + */
> +
> +int brw_float_to_vf(float f);
> +float brw_vf_to_float(unsigned char vf);
> -- 
> 2.0.4
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 212 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/mesa-dev/attachments/20141024/e9c28490/attachment.sig>