https://bugs.freedesktop.org/show_bug.cgi?id=83570
Roland Scheidegger srol...@vmware.com changed:
What|Removed |Added
Status|NEW |RESOLVED
https://bugs.freedesktop.org/show_bug.cgi?id=83570
--- Comment #13 from Roland Scheidegger srol...@vmware.com ---
(In reply to comment #12)
Roland,
Perhaps you can help me here. I'm looking at udiv_emit_cpu and I'm just
not getting it. The comment says:
/* udiv by zero is guaranteed to
https://bugs.freedesktop.org/show_bug.cgi?id=83570
--- Comment #14 from rcond...@hotmail.com ---
(In reply to comment #13)
(comparison masks are either all ones or all zeros).
AHH!!! That was the piece I was missing. My hair follicles thank you.
--
You are receiving this mail because:
You are
https://bugs.freedesktop.org/show_bug.cgi?id=83570
--- Comment #12 from rcond...@hotmail.com ---
Roland,
Perhaps you can help me here. I'm looking at udiv_emit_cpu and I'm just not
getting it. The comment says:
/* udiv by zero is guaranteed to return 0x */
but the way I read it, udiv
https://bugs.freedesktop.org/show_bug.cgi?id=83570
--- Comment #4 from Roland Scheidegger srol...@vmware.com ---
Ok, I just was thinking it is probably something along these lines, and could
reproduce this with some hacked up piglit test doing integer division by zero.
Though I got a SIGFPE
https://bugs.freedesktop.org/show_bug.cgi?id=83570
--- Comment #5 from Roland Scheidegger srol...@vmware.com ---
Actually, I think it would be a good idea if you could return 0x
instead of 0. This is typically what GPUs will do, since the d3d10 docs claim
this is required, both for div
https://bugs.freedesktop.org/show_bug.cgi?id=83570
--- Comment #6 from Roland Scheidegger srol...@vmware.com ---
(In reply to comment #5)
Actually, I think it would be a good idea if you could return 0x
instead of 0. This is typically what GPUs will do, since the d3d10 docs
claim this
https://bugs.freedesktop.org/show_bug.cgi?id=83570
--- Comment #7 from Roland Scheidegger srol...@vmware.com ---
(In reply to comment #6)
(In reply to comment #5)
Actually, I think it would be a good idea if you could return 0x
instead of 0. This is typically what GPUs will do, since
https://bugs.freedesktop.org/show_bug.cgi?id=83570
--- Comment #8 from rcond...@hotmail.com ---
for the llvm version I was going to return bld-undef. I figure I might as well
change tgsi_exec.c version to 0x for consistency...but I don't have
strong feelings either way.
--
You are
https://bugs.freedesktop.org/show_bug.cgi?id=83570
--- Comment #10 from rcond...@hotmail.com ---
Could be totally borked, but this is what I have at the moment:
if (type.floating) {
return LLVMBuildFDiv(builder, a, b, );
}
else {
//The spec says divide by zero is undefined,
https://bugs.freedesktop.org/show_bug.cgi?id=83570
--- Comment #11 from Roland Scheidegger srol...@vmware.com ---
(In reply to comment #10)
Could be totally borked, but this is what I have at the moment:
if (type.floating) {
return LLVMBuildFDiv(builder, a, b, );
}
else {
https://bugs.freedesktop.org/show_bug.cgi?id=83570
--- Comment #2 from rcond...@hotmail.com ---
Actually, that only fixes the non-llvm build. I'm not sure how to track down
the equivalent piece of code in the llvm variant.
--
You are receiving this mail because:
You are the assignee for the
https://bugs.freedesktop.org/show_bug.cgi?id=83570
--- Comment #3 from rcond...@hotmail.com ---
I figured out the LLVM side - I'll have a patch in the next few days.
--
You are receiving this mail because:
You are the assignee for the bug.
___
https://bugs.freedesktop.org/show_bug.cgi?id=83570
--- Comment #1 from rcond...@hotmail.com ---
I haven't really gotten a grasp on the code base, but my assessment is that the
issue is:
tgsi_exec.c
static void
micro_idiv(union tgsi_exec_channel *dst,
const union tgsi_exec_channel
14 matches
Mail list logo