| Message ID | 1446036929-19163-3-git-send-email-yszhou4tech@gmail.com |
|---|---|
| State | Superseded |
| Headers | show |
On 28 October 2015 at 12:55, Yousong Zhou <yszhou4tech@gmail.com> wrote: > --- a/include/target.mk > +++ b/include/target.mk > @@ -212,6 +212,7 @@ ifeq ($(DUMP),1) > CPU_CFLAGS_mips32 = -mips32 -mtune=mips32 > CPU_CFLAGS_mips32r2 = -mips32r2 -mtune=mips32r2 > CPU_CFLAGS_mips64 = -mips64 -mtune=mips64 -mabi=64 > + CPU_CFLAGS_24kc = -mips32r2 -mtune=24kc > CPU_CFLAGS_24kec = -mips32r2 -mtune=24kec > CPU_CFLAGS_34kc = -mips32r2 -mtune=34kc > CPU_CFLAGS_74kc = -mips32r2 -mtune=74kc Nothing against the patch itself. But was wondering, since the builds are so device specific, what's the point of building for a generic mips32(r2)? Isn't just CPU_CFLAGS_74kc = -march=74kc (which the gcc manual says implies -mtune=74kc) going to generate faster code than CPU_CFLAGS_74kc = -mips32r2 -mtune=74kc without any drawback?
Fwd to openwrt-devel list. ---------- Forwarded message ---------- From: Yousong Zhou <yszhou4tech@gmail.com> Date: 29 October 2015 at 09:59 Subject: Re: [OpenWrt-Devel] [PATCH 2/5] target.mk: add optimization flags for MIPS 24Kc. To: Cristian Morales Vega <cristian@samknows.com> On 28 October 2015 at 22:18, Cristian Morales Vega <cristian@samknows.com> wrote: > On 28 October 2015 at 12:55, Yousong Zhou <yszhou4tech@gmail.com> wrote: >> --- a/include/target.mk >> +++ b/include/target.mk >> @@ -212,6 +212,7 @@ ifeq ($(DUMP),1) >> CPU_CFLAGS_mips32 = -mips32 -mtune=mips32 >> CPU_CFLAGS_mips32r2 = -mips32r2 -mtune=mips32r2 >> CPU_CFLAGS_mips64 = -mips64 -mtune=mips64 -mabi=64 >> + CPU_CFLAGS_24kc = -mips32r2 -mtune=24kc >> CPU_CFLAGS_24kec = -mips32r2 -mtune=24kec >> CPU_CFLAGS_34kc = -mips32r2 -mtune=34kc >> CPU_CFLAGS_74kc = -mips32r2 -mtune=74kc > > Nothing against the patch itself. But was wondering, since the builds > are so device specific, what's the point of building for a generic > mips32(r2)? > > Isn't just > > CPU_CFLAGS_74kc = -march=74kc > (which the gcc manual says implies -mtune=74kc) > > going to generate faster code than > > CPU_CFLAGS_74kc = -mips32r2 -mtune=74kc > > without any drawback? Previously I also wondered why ar71xx set the CPU_TYPE to 34kc while there are 24kc SoCs like AR9331. According to the GCC manual -mips32r2 is equivalent to -march=mips32r2 ... By using ‘-march’ and ‘-mtune’ together, it is possible to generate code that will run on a family of processors, but optimize the code for one particular member of that family. So I guess it's just that `-mips32r2 -mtune=74kc` is safer than `-march=74kc` yousong > _______________________________________________ > openwrt-devel mailing list > openwrt-devel@lists.openwrt.org > https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
diff --git a/include/target.mk b/include/target.mk index 3e7f17d..119f88c 100644 --- a/include/target.mk +++ b/include/target.mk @@ -212,6 +212,7 @@ ifeq ($(DUMP),1) CPU_CFLAGS_mips32 = -mips32 -mtune=mips32 CPU_CFLAGS_mips32r2 = -mips32r2 -mtune=mips32r2 CPU_CFLAGS_mips64 = -mips64 -mtune=mips64 -mabi=64 + CPU_CFLAGS_24kc = -mips32r2 -mtune=24kc CPU_CFLAGS_24kec = -mips32r2 -mtune=24kec CPU_CFLAGS_34kc = -mips32r2 -mtune=34kc CPU_CFLAGS_74kc = -mips32r2 -mtune=74kc
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com> --- include/target.mk | 1 + 1 file changed, 1 insertion(+)