Add sse_fp_cost into i386_rtx_costs

2025-04-18 Thread Jan Hubicka
Hi, Znver5 has addss cost of 2 while other common floating point SSE operations costs 3 cycles. We currently have only one entry in the costs tables which makes it impossible to model this. This patch adds sse_fp_op which is used for other common FP operations (basically conversions) and updates

Re: Add sse_fp_cost into i386_rtx_costs

2025-04-17 Thread Jan Hubicka
> On Thu, 17 Apr 2025, Jan Hubicka wrote: > > > Hi, > > Znver5 has addss cost of 2 while other common floating point SSE operations > > costs 3 cycles. We currently have only one entry in the costs tables which > > makes it impossible to model this. This patch adds sse_fp_op which is used > > f

Re: Add sse_fp_cost into i386_rtx_costs

2025-04-17 Thread Richard Biener
On Thu, 17 Apr 2025, Jan Hubicka wrote: > Hi, > Znver5 has addss cost of 2 while other common floating point SSE operations > costs 3 cycles. We currently have only one entry in the costs tables which > makes it impossible to model this. This patch adds sse_fp_op which is used > for > other com