|
From: | Bastian Koppelmann |
Subject: | Re: [Qemu-devel] [PATCH 17/28] target/riscv: Convert quadrant 2 of RVXC insns to decodetree |
Date: | Fri, 19 Oct 2018 17:10:36 +0200 |
User-agent: | Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101 Thunderbird/60.2.1 |
On 10/13/18 9:34 PM, Richard Henderson wrote:
On 10/12/18 10:30 AM, Bastian Koppelmann wrote:address@hidden ... . ..... ..... .. &c_flwsp_ldsp uimm_flwsp=%uimm_6bit_lw \ + uimm_ldsp=%uimm_6bit_ld %rd address@hidden ... . ..... ..... .. &c_fswsp_sdsp uimm_fswsp=%uimm_6bit_sw \ + uimm_sdsp=%uimm_6bit_sd rs2=%rs2_5I'm not especially keen about decoding the same bits twice. I wonder if it's better to call the proper extract_foo function within the trans_ function like you do elsewhere. Written as c_flwsp_ldsp 011 - ----- ----- 10
Hmm, Do you mean split the formats up into: @c_fswsp ... . ..... ..... .. &fswsp_sdsp uimm=%uimm_6bit_sw and @c_sdsp ... . ..... ..... .. &fswsp_sdsp uimm=%uimm_6bit_sd rs2=%rs2_5And then call extract_c_sdsp and extract_c_fswsp in each of the TARGET_RVX cases. However then we end up with unused functions in either TARGET_RV32 and TARGET_RV64.
I don't think its worth it to go through all that trouble for a small optimization.
Cheers, Bastian
[Prev in Thread] | Current Thread | [Next in Thread] |