From mboxrd@z Thu Jan 1 00:00:00 1970 From: Albert ARIBAUD Date: Thu, 13 Feb 2014 03:56:56 +0100 Subject: [U-Boot] [RFC] Add 64-bit data support for memory commands In-Reply-To: References: <1392240023-25205-1-git-send-email-yorksun@freescale.com> <20140212221101.A5F0338060E@gemini.denx.de> Message-ID: <20140213035656.71992b6f@lilith> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de Hi FengHua, On Thu, 13 Feb 2014 10:41:07 +0800 (GMT+08:00), FengHua wrote: > > > > > -----Original Messages----- > > From: "Wolfgang Denk" > > Sent Time: 2014-02-13 06:11:01 (Thursday) > > To: "York Sun" > > Cc: scottwood at freescale.com, u-boot at lists.denx.de > > Subject: Re: [U-Boot] [RFC] Add 64-bit data support for memory commands > > > > Dear York Sun, > > > > In message <1392240023-25205-1-git-send-email-yorksun@freescale.com> you wrote: > > > For aarch64, unsigned long is 64-bit data. Memory commands should be fixed > > > with u32 for 32-bit address access. A double word data size is added to > > > support 64-bit data. > > > > > > Signed-off-by: York Sun > > > > Can you please make all this code conditional for 64 bit architectures > > only, so that 32 bit systems do not suffer from the increased code > > size? > > > > Thanks. > > > > Best regards, > > > > Wolfgang Denk > How about use some macro from compiler of sizeof(long)? > Maybe this could avoid conditional switch, and cmd_mem work with 32 and 64 bit platform. That would make "md.l" be 32 or 64 bits depending on the target's natural int size, wouldn't it? But then, "md.w" and "md.b" would remain fixed-size, respectively 16 and 8 bits. That would not be consistent IMO. I think "md.l" should clearly remain 32 bits, and if 64-bit is introduced in "md" and others it should be clearly recognizable e.g. as ".d" whatever the target's natural int size. > Best Wishes Amicalement, -- Albert.