Date: Mon, 16 Apr 2012 14:03:09 -0600 From: Ian Lepore <freebsd@damnhippie.dyndns.org> To: Juli Mallett <jmallett@freebsd.org> Cc: Andrew Turner <andrew@freebsd.org>, svn-src-head@freebsd.org, svn-src-all@freebsd.org, src-committers@freebsd.org Subject: Re: svn commit: r234337 - in head: lib/libc/arm/gen sys/arm/include Message-ID: <1334606589.1082.201.camel@revolution.hippie.lan> In-Reply-To: <CACVs6=8BYYN0ZW7zzHZduns-ZH9hUF=M1EmXfoOkCZVMh1ubpg@mail.gmail.com> References: <201204160938.q3G9cLmw026093@svn.freebsd.org> <CACVs6=8BYYN0ZW7zzHZduns-ZH9hUF=M1EmXfoOkCZVMh1ubpg@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 2012-04-16 at 12:41 -0700, Juli Mallett wrote: > On Mon, Apr 16, 2012 at 02:38, Andrew Turner <andrew@freebsd.org> wrote: > > Log: > > Replace the C implementation of __aeabi_read_tp with an assembly version. > > This ensures we follow the ABI by preserving registers r1-r3. > > > +ENTRY(__aeabi_read_tp) > > + ldr r0, .Larm_tp_address > > + ldr r0, [r0] > > + RET > > + > > +.Larm_tp_address: > > + .word ARM_TP_ADDRESS > > + > > Why is this indirection required? Can't you just use ARM_TP_ADDRESS > instead of loading it from data? Also, is our convention for ARM to > use END() with ENTRY() or not? It used to be possible to directly load ARM_TP_ADDRESS into a register when it had a cleverly-crafted value that made use of ARM's ability to encode a constant value into an instruction if it can be expressed as an 8-bit value shifted by anything up to 31 bits. (iirc, it used to be 0xe0000000). At some point that address became a problem for someone and it got changed to a value which can no longer be encoded within the instruction. The alternative is to load it with a single instruction by using a pc-relative address, or use a series of load-and-shift instructions to form the new more complex constant. Ick. -- Ian
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1334606589.1082.201.camel>