[PATCH v2] x86: Ignore rounding for vcvt[,u]si2sd under r32 and vcvt[,u]dq2pd instead of reporting bad

[Jan Beulich]

On 16.01.2025 03:44, Haochen Jiang wrote:
> As said in previous thread, HW would accept those rounding encodings so
> it would be safe to do that in toolchain.
> 
> Changes and patch descrption are embedded following.
> 
> Ok for trunk?

I'm sorry, but no. I don't see justification for the assembler side change.
For SAE-only insns we don't permit rounding control other than through the
-mrcig= command line option. As a consequence I don't think we should
permit rounding control via insn operand modifiers here either. Whether we
want to make -mrcig= work for these insns I'm uncertain about; if so, that
still shouldn't be by permitting {sae} operand modifiers on the insn
operands, imo.

Therefore, with the assembler change (and resulting testsuite adjustments)
dropped, this is okay. If you want to pursue the assembler change, then
please re-submit as an incremental change.

> Note: It seems to me that omitting 512 when using .insn with vcvt[,u]dq2pd
> will lead to 256 size. That is why I am not omitting them. It might be a
> bug for .insn but I did not look into that.

I'll look into this.

Jan