[Mesa-dev] [PATCH] nir: fix divide by zero crash during constant folding

[Connor Abbott]

Floating point division shouldn't signal on division by zero, it
should just return an appropriately-signed infinity, which seems like
a sane thing to do, and way better than just returning 0. So we
shouldn't do this with fdiv. I guess 0 is as good a result as any for
the integer division, though -- there aren't really any great choices.

On Tue, Feb 27, 2018 at 10:07 PM, Timothy Arceri <tarceri at itsqueeze.com> wrote:
> From the GLSL 4.60 spec Section 5.9 (Expressions):
>
>    "Dividing by zero does not cause an exception but does result in
>     an unspecified value."
>
> Fixes: 89285e4d47a6 "nir: add new constant folding infrastructure"
>
> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=105271
> ---
>  src/compiler/nir/nir_opcodes.py | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/src/compiler/nir/nir_opcodes.py b/src/compiler/nir/nir_opcodes.py
> index 278562b2bd..dcc5b07d05 100644
> --- a/src/compiler/nir/nir_opcodes.py
> +++ b/src/compiler/nir/nir_opcodes.py
> @@ -403,9 +403,9 @@ binop("imul_high", tint32, commutative,
>  binop("umul_high", tuint32, commutative,
>        "(uint32_t)(((uint64_t) src0 * (uint64_t) src1) >> 32)")
>
> -binop("fdiv", tfloat, "", "src0 / src1")
> -binop("idiv", tint, "", "src0 / src1")
> -binop("udiv", tuint, "", "src0 / src1")
> +binop("fdiv", tfloat, "", "src1 == 0 ? 0 : (src0 / src1)")
> +binop("idiv", tint, "", "src1 == 0 ? 0 : (src0 / src1)")
> +binop("udiv", tuint, "", "src1 == 0 ? 0 : (src0 / src1)")
>
>  # returns a boolean representing the carry resulting from the addition of
>  # the two unsigned arguments.
> --
> 2.14.3
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev