Re: [PING] target/ppc: Fix bugs in VSX_CVT_FP_TO_INT and VSX_CVT_FP_TO_INT2 macros

[Cédric Le Goater]

Hello John,

On 11/9/23 23:18, John Platts wrote:
> What is the status of integrating the fixes to the VSX_CVT_FP_TO_INT and 
> VSX_CVT_FP_TO_INT2 macros in target/ppc/fpu_helper.c?

You sent a couple of times the same patch :
 
 https://patchwork.ozlabs.org/project/qemu-ppc/list/?submitter=87426

I suppose v3 is the one to consider.

> The bug that is currently there with the VSX_CVT_FP_TO_INT and 
> VSX_CVT_FP_TO_INT2 macros is that float_invalid_cvt is incorrectly called if 
> the current source value is a non-NaN value and a previous NaN source value 
> from the same source vector was encountered.
>
> The link to a patch that fixes the above bug in the VSX_CVT_FP_TO_INT and 
> VSX_CVT_FP_TO_INT2 macros can be found at 
> SN6PR05MB5837B5EF0D11E0A787EC30179DA5A@SN6PR05MB5837.namprd05.prod.outlook.com/">https://patchew.org/QEMU/SN6PR05MB5837B5EF0D11E0A787EC30179DA5A@SN6PR05MB5837.namprd05.prod.outlook.com/.
>
> A description of this bug can be found at 
> https://gitlab.com/qemu-project/qemu/-/issues/1941.
>
> Here is a link to a test program that will detect the bugs that are currently 
> there with the emulation of the xvcvspsxws, xvcvspuxws, xvcvspsxds, xvcvspuxds, 
> xvcvdpsxws, xvcvdpuxws, xvcvdpsxds, and xvcvdpuxds instructions if the source 
> vector contains at least one NaN value:
> https://gitlab.com/qemu-project/qemu/uploads/fcbb97896ff2f4ab435affae94467f4d/vsx_f2i_nan_test_102523.c

If you have time, could you please respin a v4 and include the
test case under tests/tcg/ppc64 ? Please add a Fixes: tag with
the commit id also.

The plan is to merge fixes, such as this one, in the coming weeks,
and I will send one PPC PR before the end of the 8.2 cycle.

Thanks,

C.