From: slash.tmp@free.fr (Mason)
To: linux-arm-kernel@lists.infradead.org
Subject: read_cpuid_id() in arch/arm/kernel/setup.c
Date: Fri, 13 Mar 2015 17:03:52 +0100 [thread overview]
Message-ID: <55030A68.6070002@free.fr> (raw)
Hello everyone,
As far as I can tell, read_cpuid_id() resolves to read_cpuid(CPUID_ID)
which resolves to mrc 15, 0, rN, cr0, cr0, {0}
Consider this:
/*
* The CPU ID never changes at run time, so we might as well tell the
* compiler that it's constant. Use this function to read the CPU ID
* rather than directly reading processor_id or read_cpuid() directly.
*/
static inline unsigned int __attribute_const__ read_cpuid_id(void)
{
return read_cpuid(CPUID_ID);
}
Despite the comment and attribute, my compiler(*) still reloads the
value every time.
(*) gcc version 4.9.3 20141031 (prerelease) (Linaro GCC 2014.11)
e.g.
static int __get_cpu_architecture(void)
{
int cpu_arch;
unsigned int id = read_cpuid_id();
if ((id & 0x0008f000) == 0) {
cpu_arch = CPU_ARCH_UNKNOWN;
} else if ((id & 0x0008f000) == 0x00007000) {
cpu_arch = (id & (1 << 23)) ? CPU_ARCH_ARMv4T : CPU_ARCH_ARMv3;
} else if ((id & 0x00080000) == 0x00000000) {
cpu_arch = (id >> 16) & 7;
if (cpu_arch)
cpu_arch += CPU_ARCH_ARMv3;
} else if ((id & 0x000f0000) == 0x000f0000) {
resolves to
c01fec74: ee10cf10 mrc 15, 0, ip, cr0, cr0, {0}
c01fec78: e21cca8f ands ip, ip, #585728 ; 0x8f000
c01fec7c: e34c3023 movt r3, #49187 ; 0xc023
c01fec80: e5837008 str r7, [r3, #8]
c01fec84: e50b304c str r3, [fp, #-76] ; 0x4c
c01fec88: 0a000022 beq c01fed18 <setup_arch+0xe4>
c01fec8c: ee103f10 mrc 15, 0, r3, cr0, cr0, {0}
c01fec90: e2033a8f and r3, r3, #585728 ; 0x8f000
c01fec94: e3530a07 cmp r3, #28672 ; 0x7000
c01fec98: 1a000004 bne c01fecb0 <setup_arch+0x7c>
c01fec9c: ee103f10 mrc 15, 0, r3, cr0, cr0, {0}
c01feca0: e3130502 tst r3, #8388608 ; 0x800000
c01feca4: 13a0c003 movne ip, #3
c01feca8: 03a0c001 moveq ip, #1
c01fecac: ea000019 b c01fed18 <setup_arch+0xe4>
c01fecb0: ee103f10 mrc 15, 0, r3, cr0, cr0, {0}
c01fecb4: e3130702 tst r3, #524288 ; 0x80000
So I thought it would be nice to give the poor compiler a break,
and just stuff the result in a local variable:
--- setup.c 2015-03-03 18:04:59.000000000 +0100
+++ setup.foo.c 2015-03-13 16:26:56.413380663 +0100
@@ -237,15 +237,16 @@
{
int cpu_arch;
- if ((read_cpuid_id() & 0x0008f000) == 0) {
+ unsigned int id = read_cpuid_id();
+ if ((id & 0x0008f000) == 0) {
cpu_arch = CPU_ARCH_UNKNOWN;
- } else if ((read_cpuid_id() & 0x0008f000) == 0x00007000) {
- cpu_arch = (read_cpuid_id() & (1 << 23)) ? CPU_ARCH_ARMv4T : CPU_ARCH_ARMv3;
- } else if ((read_cpuid_id() & 0x00080000) == 0x00000000) {
- cpu_arch = (read_cpuid_id() >> 16) & 7;
+ } else if ((id & 0x0008f000) == 0x00007000) {
+ cpu_arch = (id & (1 << 23)) ? CPU_ARCH_ARMv4T : CPU_ARCH_ARMv3;
+ } else if ((id & 0x00080000) == 0x00000000) {
+ cpu_arch = (id >> 16) & 7;
if (cpu_arch)
cpu_arch += CPU_ARCH_ARMv3;
- } else if ((read_cpuid_id() & 0x000f0000) == 0x000f0000) {
+ } else if ((id & 0x000f0000) == 0x000f0000) {
unsigned int mmfr0;
/* Revised CPUID format. Read the Memory Model Feature
which compiles to
c01fec74: ee102f10 mrc 15, 0, r2, cr0, cr0, {0}
c01fec78: e212ca8f ands ip, r2, #585728 ; 0x8f000
c01fec7c: e34c3023 movt r3, #49187 ; 0xc023
c01fec80: e5837008 str r7, [r3, #8]
c01fec84: e50b304c str r3, [fp, #-76] ; 0x4c
c01fec88: 0a00001c beq c01fed00 <setup_arch+0xcc>
c01fec8c: e35c0a07 cmp ip, #28672 ; 0x7000
c01fec90: 1a000003 bne c01feca4 <setup_arch+0x70>
c01fec94: e3120502 tst r2, #8388608 ; 0x800000
c01fec98: 13a0c003 movne ip, #3
c01fec9c: 03a0c001 moveq ip, #1
c01feca0: ea000016 b c01fed00 <setup_arch+0xcc>
c01feca4: e3120702 tst r2, #524288 ; 0x80000
Is this nano-optimization worth considering?
Regards.
next reply other threads:[~2015-03-13 16:03 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-03-13 16:03 Mason [this message]
2015-03-13 16:19 ` read_cpuid_id() in arch/arm/kernel/setup.c Russell King - ARM Linux
2015-03-13 16:39 ` Mason
2015-03-13 16:45 ` Russell King - ARM Linux
2015-03-13 17:06 ` Mason
2015-03-15 17:40 ` Mason
2015-03-16 8:44 ` Mason
2015-03-16 16:54 ` Paul Walmsley
2015-03-16 22:17 ` Mason
2015-03-16 23:30 ` Mason
2015-03-13 16:56 ` Mark Rutland
2015-03-13 17:02 ` Russell King - ARM Linux
2015-03-13 18:26 ` Mark Rutland
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=55030A68.6070002@free.fr \
--to=slash.tmp@free.fr \
--cc=linux-arm-kernel@lists.infradead.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.