[PATCH] Additional compiler optimization options

[PATCH] Additional compiler optimization options

[John]

Came across this and wanted to show it to others more knowledgeable for feedback.  Is this value-added or not?  Upstream URL is: https://gitorious.org/rummage/random_patches

Please cc me on replies as I am not a regular subscriber to lkml.  Thank you.

--- linux-3.6/arch/x86/include/asm/module.h2012-04-29 18:43:58.239336240 -0700
+++ linux-3.6.mod/arch/x86/include/asm/module.h2012-04-29 18:43:41.609545221 -0700
@@ -17,6 +17,14 @@
 #define MODULE_PROC_FAMILY "586MMX "
 #elif defined CONFIG_MCORE2
 #define MODULE_PROC_FAMILY "CORE2 "
+#elif defined CONFIG_MCOREI7
+#define MODULE_PROC_FAMILY "COREI7 "
+#elif defined CONFIG_MCOREI7AVX
+#define MODULE_PROC_FAMILY "COREI7AVX "
+#elif defined CONFIG_MCOREAVXI
+#define MODULE_PROC_FAMILY "COREAVXI "
+#elif defined CONFIG_MCOREAVX2
+#define MODULE_PROC_FAMILY "COREAVX2 "
 #elif defined CONFIG_MATOM
 #define MODULE_PROC_FAMILY "ATOM "
 #elif defined CONFIG_M686
@@ -35,6 +43,14 @@
 #define MODULE_PROC_FAMILY "K7 "
 #elif defined CONFIG_MK8
 #define MODULE_PROC_FAMILY "K8 "
+#elif defined CONFIG_MBARCELONA
+#define MODULE_PROC_FAMILY "BARCELONA "
+#elif defined CONFIG_MBOBCAT
+#define MODULE_PROC_FAMILY "BOBCAT "
+#elif defined CONFIG_MBULLDOZER
+#define MODULE_PROC_FAMILY "BULLDOZER "
+#elif defined CONFIG_MPILEDRIVER
+#define MODULE_PROC_FAMILY "PILEDRIVER "
 #elif defined CONFIG_MELAN
 #define MODULE_PROC_FAMILY "ELAN "
 #elif defined CONFIG_MCRUSOE
--- linux-3.6/arch/x86/Kconfig.cpu2012-04-29 18:43:58.249336198 -0700
+++ linux-3.6.mod/arch/x86/Kconfig.cpu2012-04-29 18:40:46.091751798 -0700
@@ -147,7 +147,7 @@
 
 
 config MK6
-bool "K6/K6-II/K6-III"
+bool "AMD K6/K6-II/K6-III"
 depends on X86_32
 ---help---
   Select this for an AMD K6-family processor.  Enables use of
@@ -155,7 +155,7 @@
   flags to GCC.
 
 config MK7
-bool "Athlon/Duron/K7"
+bool "AMD Athlon/Duron/K7"
 depends on X86_32
 ---help---
   Select this for an AMD Athlon K7-family processor.  Enables use of
@@ -163,12 +163,40 @@
   flags to GCC.
 
 config MK8
-bool "Opteron/Athlon64/Hammer/K8"
+bool "AMD Opteron/Athlon64/Hammer/K8"
 ---help---
   Select this for an AMD Opteron or Athlon64 Hammer-family processor.
   Enables use of some extended instructions, and passes appropriate
   optimization flags to GCC.
 
+config MBARCELONA
+bool "AMD Barcelona"
+---help---
+  Select this for AMD Barcelona and newer processors.
+
+  Enables -march=barcelona
+
+config MBOBCAT
+bool "AMD Bobcat"
+---help---
+  Select this for AMD Bobcat processors.
+
+  Enables -march=btver1
+
+config MBULLDOZER
+bool "AMD Bulldozer"
+---help---
+  Select this for AMD Bulldozer processors.
+
+  Enables -march=bdver1
+
+config MPILEDRIVER
+bool "AMD Piledriver"
+---help---
+  Select this for AMD Piledriver processors.
+
+  Enables -march=bdver2
+
 config MCRUSOE
 bool "Crusoe"
 depends on X86_32
@@ -260,7 +288,7 @@
   in /proc/cpuinfo. Family 15 is an older Xeon, Family 6 a newer one.
 
 config MCORE2
-bool "Core 2/newer Xeon"
+bool "Intel Core 2"
 ---help---
 
   Select this for Intel Core 2 and newer Core 2 Xeons (Xeon 51xx and
@@ -268,6 +296,41 @@
   family in /proc/cpuinfo. Newer ones have 6 and older ones 15
   (not a typo)
 
+  Enables -march=core2
+
+config MCOREI7
+bool "Intel Core i7"
+---help---
+
+  Select this for the Intel Nehalem platform. Intel Nehalem proecessors
+  include Core i3, i5, i7, Xeon: 34xx, 35xx, 55xx, 56xx, 75xx processors.
+
+  Enables -march=corei7
+
+config MCOREI7AVX
+bool "Intel Core 2nd Gen AVX"
+---help---
+
+  Select this for 2nd Gen Core processors including Sandy Bridge.
+
+  Enables -march=corei7-avx
+
+config MCOREAVXI
+bool "Intel Core 3rd Gen AVX"
+---help---
+
+  Select this for 3rd Gen Core processors including Ivy Bridge.
+
+  Enables -march=corei7-avx
+
+config MCOREAVX2
+bool "Intel Core AVX-2"
+---help---
+
+  Select this for AVX-2 enabled processors including Haswell.
+
+  Enables -march=corei7-avx-2
+
 config MATOM
 bool "Intel Atom"
 ---help---
@@ -312,7 +375,7 @@
 config X86_L1_CACHE_SHIFT
 int
 default "7" if MPENTIUM4 || MPSC
-default "6" if MK7 || MK8 || MPENTIUMM || MCORE2 || MATOM || MVIAC7 || X86_GENERIC || GENERIC_CPU
+default "6" if MK7 || MK8 || MBARCELONA || MBOBCAT || MBULLDOZER || MPILEDRIVER || MPENTIUMM || MCORE2 || MCOREI7 || MCOREI7AVX || MCOREAVXI || MCOREAVX2 || MATOM || MVIAC7 || X86_GENERIC || GENERIC_CPU
 default "4" if MELAN || M486 || M386 || MGEODEGX1
 default "5" if MWINCHIP3D || MWINCHIPC6 || MCRUSOE || MEFFICEON || MCYRIXIII || MK6 || MPENTIUMIII || MPENTIUMII || M686 || M586MMX || M586TSC || M586 || MVIAC3_2 || MGEODE_LX
 
@@ -363,11 +426,11 @@
 
 config X86_INTEL_USERCOPY
 def_bool y
-depends on MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M586MMX || X86_GENERIC || MK8 || MK7 || MEFFICEON || MCORE2
+depends on MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M586MMX || X86_GENERIC || MK8 || MK7 || MBARCELONA || MEFFICEON || MCORE2 || MCOREI7 || MCOREI7AVX || MCOREAVXI || MCOREAVX2
 
 config X86_USE_PPRO_CHECKSUM
 def_bool y
-depends on MWINCHIP3D || MWINCHIPC6 || MCYRIXIII || MK7 || MK6 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || MK8 || MVIAC3_2 || MVIAC7 || MEFFICEON || MGEODE_LX || MCORE2 || MATOM
+depends on MWINCHIP3D || MWINCHIPC6 || MCYRIXIII || MK7 || MK6 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || MK8 || MVIAC3_2 || MVIAC7 || MEFFICEON || MGEODE_LX || MCORE2 || MCOREI7 || MCOREI7AVX || MCOREAVXI || MCOREAVX2 || MATOM
 
 config X86_USE_3DNOW
 def_bool y
@@ -395,17 +458,17 @@
 
 config X86_TSC
 def_bool y
-depends on ((MWINCHIP3D || MCRUSOE || MEFFICEON || MCYRIXIII || MK7 || MK6 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || M586MMX || M586TSC || MK8 || MVIAC3_2 || MVIAC7 || MGEODEGX1 || MGEODE_LX || MCORE2 || MATOM) && !X86_NUMAQ) || X86_64
+depends on ((MWINCHIP3D || MCRUSOE || MEFFICEON || MCYRIXIII || MK7 || MK6 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || M586MMX || M586TSC || MK8 || MBARCELONA || MBOBCAT || MBULLDOZER || MPILEDRIVER || MVIAC3_2 || MVIAC7 || MGEODEGX1 || MGEODE_LX || MCORE2 || MCOREI7 || MCOREI7-AVX || MATOM) && !X86_NUMAQ) || X86_64
 
 config X86_CMPXCHG64
 def_bool y
-depends on X86_PAE || X86_64 || MCORE2 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || MATOM
+depends on X86_PAE || X86_64 || MCORE2 || MCOREI7 || MCOREI7AVX || MCOREAVXI || MCOREAVX2 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || MATOM
 
 # this should be set for all -march=.. options where the compiler
 # generates cmov.
 config X86_CMOV
 def_bool y
-depends on (MK8 || MK7 || MCORE2 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || MVIAC3_2 || MVIAC7 || MCRUSOE || MEFFICEON || X86_64 || MATOM || MGEODE_LX)
+depends on (MK8 || MBARCELONA || MBOBCAT || MBULLDOZER || MPILEDRIVER || MK7 || MCORE2 || MCOREI7 || MCOREI7AVX || MCOREAVXI || MCOREAVX2 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || MVIAC3_2 || MVIAC7 || MCRUSOE || MEFFICEON || X86_64 || MATOM || MGEODE_LX)
 
 config X86_MINIMUM_CPU_FAMILY
 int
--- linux-3.6/arch/x86/Makefile2012-04-29 18:43:58.249336198 -0700
+++ linux-3.6.mod/arch/x86/Makefile2012-04-29 18:41:51.260932642 -0700
@@ -51,10 +51,22 @@
 
         # FIXME - should be integrated in Makefile.cpu (Makefile_32.cpu)
         cflags-$(CONFIG_MK8) += $(call cc-option,-march=k8)
+        cflags-$(CONFIG_MBARCELONA) += $(call cc-option,-march=barcelona)
+        cflags-$(CONFIG_MBOBCAT) += $(call cc-option,-march=btver1)
+        cflags-$(CONFIG_MBULLDOZER) += $(call cc-option,-march=bdver1)
+        cflags-$(CONFIG_MPILEDRIVER) += $(call cc-option,-march=bdver2)
         cflags-$(CONFIG_MPSC) += $(call cc-option,-march=nocona)
 
         cflags-$(CONFIG_MCORE2) += \
-                $(call cc-option,-march=core2,$(call cc-option,-mtune=generic))
+                $(call cc-option,-march=core2,$(call cc-option,-mtune=core2))
+        cflags-$(CONFIG_MCOREI7) += \
+                $(call cc-option,-march=corei7,$(call cc-option,-mtune=corei7))
+        cflags-$(CONFIG_MCOREI7AVX) += \
+                $(call cc-option,-march=corei7-avx,$(call cc-option,-mtune=corei7-avx))
+        cflags-$(CONFIG_MCOREAVXI) += \
+                $(call cc-option,-march=core-avx-i,$(call cc-option,-mtune=core-avx-i))
+        cflags-$(CONFIG_MCOREAVX2) += \
+                $(call cc-option,-march=core-avx2,$(call cc-option,-mtune=core-avx2))
 cflags-$(CONFIG_MATOM) += $(call cc-option,-march=atom) \
 $(call cc-option,-mtune=atom,$(call cc-option,-mtune=generic))
         cflags-$(CONFIG_GENERIC_CPU) += $(call cc-option,-mtune=generic)
--- linux-3.6/arch/x86/Makefile_32.cpu2012-04-29 18:43:58.269335977 -0700
+++ linux-3.6.mod/arch/x86/Makefile_32.cpu2012-04-29 18:42:39.120330975 -0700
@@ -25,6 +25,10 @@
 # They make zero difference whatsosever to performance at this time.
 cflags-$(CONFIG_MK7)+= -march=athlon
 cflags-$(CONFIG_MK8)+= $(call cc-option,-march=k8,-march=athlon)
+cflags-$(CONFIG_MBARCELONA)+= $(call cc-option,-march=barcelona,-march=athlon)
+cflags-$(CONFIG_MBOBCAT)+= $(call cc-option,-march=btver1,-march=athlon)
+cflags-$(CONFIG_MBULLDOZER)+= $(call cc-option,-march=bdver1,-march=athlon)
+cflags-$(CONFIG_MPILEDRIVER)+= $(call cc-option,-march=bdver2,-march=athlon)
 cflags-$(CONFIG_MCRUSOE)+= -march=i686 $(align)-functions=0 $(align)-jumps=0 $(align)-loops=0
 cflags-$(CONFIG_MEFFICEON)+= -march=i686 $(call tune,pentium3) $(align)-functions=0 $(align)-jumps=0 $(align)-loops=0
 cflags-$(CONFIG_MWINCHIPC6)+= $(call cc-option,-march=winchip-c6,-march=i586)
@@ -33,6 +37,10 @@
 cflags-$(CONFIG_MVIAC3_2)+= $(call cc-option,-march=c3-2,-march=i686)
 cflags-$(CONFIG_MVIAC7)+= -march=i686
 cflags-$(CONFIG_MCORE2)+= -march=i686 $(call tune,core2)
+cflags-$(CONFIG_MCOREI7)+= -march=i686 $(call tune,corei7)
+cflags-$(CONFIG_MCOREI7AVX)+= -march=i686 $(call tune,corei7-avx)
+cflags-$(CONFIG_MCOREAVXI)+= -march=i686 $(call tune,core-avx-i)
+cflags-$(CONFIG_MCOREAVX2)+= -march=i686 $(call tune,core-avx-2)
 cflags-$(CONFIG_MATOM)+= $(call cc-option,-march=atom,$(call cc-option,-march=core2,-march=i686)) \
 $(call cc-option,-mtune=atom,$(call cc-option,-mtune=generic))

Re: [PATCH] Additional compiler optimization options

[Borislav Petkov]

On Wed, Dec 05, 2012 at 02:09:15PM -0800, John wrote:
> Came across this and wanted to show it to others
> more knowledgeable for feedback.  Is this value-added or not?
>  Upstream URL is: https://gitorious.org/rummage/random_patches

This only makes sense if you can show a noticeable performance
improvement using one of the those -march options with at least one
benchmark from those which we have now in the kernel.

And even that use case is limited because the majority of the kernel
builds out there are distro kernels built with -mtune=generic so it
would only make difference for people building their own kernels, who
are a small minority, relatively.

Thanks.

-- 
Regards/Gruss,
    Boris.

Re: [PATCH] Additional compiler optimization options

[John]

> This only makes sense if you can show a noticeable performance

> improvement using one of the those -march options with at least one
> benchmark from those which we have now in the kernel.

Thank you for the feedback, Boris.  Can you point me to the built in benchmarks to which you referred?  I am interested in probing this further.

Re: [PATCH] Additional compiler optimization options

[Borislav Petkov]

On Wed, Dec 05, 2012 at 03:00:12PM -0800, John wrote:
> > This only makes sense if you can show a noticeable performance
> 
> > improvement using one of the those -march options with at least one
> > benchmark from those which we have now in the kernel.
>
> Thank you for the feedback, Boris.  Can you point me to the built in
> benchmarks to which you referred?  I am interested in probing this
> further.

Maybe the easiest thing to do is build the kernel:

make -j<number_of_cores+1>

in your kernel directory and time it. You could get more precise output
if you trace that workload with perf:

$ perf stat --repeat 10 make -j<number_of_cores+1>

And look at how the perf output looks like.

For all those runs, you need hardware supporting any of these options
and the running kernel should each time be built with a different
optimization option.

This will show us whether differently optimized kernels show any
performance difference.

And then you can go and try other benchmarks like kernbench, hackbench,
etc, whatever your favourite search engine gives you for "linux
benchmarks"

HTH.

-- 
Regards/Gruss,
    Boris.

[PATCH] Expand CPU compiler options

[John]

Please cc me on replies as I am not a regular subscriber to lkml.  Thank you.



I tested the attached patch written by André Ramnitz using three different machines running a generic x86-64 kernel and an otherwise identical kernel running with the optimized gcc options. 

Conclusion: There are small but real speed increases using a make endpoint to running with this patch.

Details:
1) Three test machines: Intel Xeon X3360, Intel i7-2620M, Intel Core i7-3660K.

2) All ran the make benchmark (linked below) 35 times while booted into a 'generic' kernel. Then all ran the same make benchmark 35 times after booting into an optimized kernel. Below are the optimizations chosen for each machine.
2a) X3360 = core2
2b) i7-2620M = corei7-avx
2c) i7-3660K = core-avx-i
3) Analyzed resulting distributions for statistical significance via ANOVA plots that clearly show statistically significant albeit small differences.

Links to ANOVA plots:
http://s19.postimage.org/68urcofzn/corei7_avx.png
http://s19.postimage.org/ozwomuak3/core_avx_i.png

http://s19.postimage.org/d0l6fj4z7/core2.png


References:

Bash script that controls the benchmark: https://github.com/graysky2/bin/blob/master/bench
Log file generated by script: http://repo-ck.com/bench/compile_time_optimization.txt.gz

---
--- linux-3.6/arch/x86/include/asm/module.h2012-04-29 18:43:58.239336240 -0700
+++ linux-3.6.mod/arch/x86/include/asm/module.h2012-04-29 18:43:41.609545221 -0700
@@ -17,6 +17,14 @@
 #define MODULE_PROC_FAMILY "586MMX "
 #elif defined CONFIG_MCORE2
 #define MODULE_PROC_FAMILY "CORE2 "
+#elif defined CONFIG_MCOREI7
+#define MODULE_PROC_FAMILY "COREI7 "
+#elif defined CONFIG_MCOREI7AVX
+#define MODULE_PROC_FAMILY "COREI7AVX "
+#elif defined CONFIG_MCOREAVXI
+#define MODULE_PROC_FAMILY "COREAVXI "
+#elif defined CONFIG_MCOREAVX2
+#define MODULE_PROC_FAMILY "COREAVX2 "
 #elif defined CONFIG_MATOM
 #define MODULE_PROC_FAMILY "ATOM "
 #elif defined CONFIG_M686
@@ -35,6 +43,14 @@
 #define MODULE_PROC_FAMILY "K7 "
 #elif defined CONFIG_MK8
 #define MODULE_PROC_FAMILY "K8 "
+#elif defined CONFIG_MBARCELONA
+#define MODULE_PROC_FAMILY "BARCELONA "
+#elif defined CONFIG_MBOBCAT
+#define MODULE_PROC_FAMILY "BOBCAT "
+#elif defined CONFIG_MBULLDOZER
+#define MODULE_PROC_FAMILY "BULLDOZER "
+#elif defined CONFIG_MPILEDRIVER
+#define MODULE_PROC_FAMILY "PILEDRIVER "
 #elif defined CONFIG_MELAN
 #define MODULE_PROC_FAMILY "ELAN "
 #elif defined CONFIG_MCRUSOE
--- linux-3.6/arch/x86/Kconfig.cpu2012-04-29 18:43:58.249336198 -0700
+++ linux-3.6.mod/arch/x86/Kconfig.cpu2012-04-29 18:40:46.091751798 -0700
@@ -147,7 +147,7 @@
 
 
 config MK6
-bool "K6/K6-II/K6-III"
+bool "AMD K6/K6-II/K6-III"
 depends on X86_32
 ---help---
   Select this for an AMD K6-family processor.  Enables use of
@@ -155,7 +155,7 @@
   flags to GCC.
 
 config MK7
-bool "Athlon/Duron/K7"
+bool "AMD Athlon/Duron/K7"
 depends on X86_32
 ---help---
   Select this for an AMD Athlon K7-family processor.  Enables use of
@@ -163,12 +163,40 @@
   flags to GCC.
 
 config MK8
-bool "Opteron/Athlon64/Hammer/K8"
+bool "AMD Opteron/Athlon64/Hammer/K8"
 ---help---
   Select this for an AMD Opteron or Athlon64 Hammer-family processor.
   Enables use of some extended instructions, and passes appropriate
   optimization flags to GCC.
 
+config MBARCELONA
+bool "AMD Barcelona"
+---help---
+  Select this for AMD Barcelona and newer processors.
+
+  Enables -march=barcelona
+
+config MBOBCAT
+bool "AMD Bobcat"
+---help---
+  Select this for AMD Bobcat processors.
+
+  Enables -march=btver1
+
+config MBULLDOZER
+bool "AMD Bulldozer"
+---help---
+  Select this for AMD Bulldozer processors.
+
+  Enables -march=bdver1
+
+config MPILEDRIVER
+bool "AMD Piledriver"
+---help---
+  Select this for AMD Piledriver processors.
+
+  Enables -march=bdver2
+
 config MCRUSOE
 bool "Crusoe"
 depends on X86_32
@@ -260,7 +288,7 @@
   in /proc/cpuinfo. Family 15 is an older Xeon, Family 6 a newer one.
 
 config MCORE2
-bool "Core 2/newer Xeon"
+bool "Intel Core 2"
 ---help---
 
   Select this for Intel Core 2 and newer Core 2 Xeons (Xeon 51xx and
@@ -268,6 +296,41 @@
   family in /proc/cpuinfo. Newer ones have 6 and older ones 15
   (not a typo)
 
+  Enables -march=core2
+
+config MCOREI7
+bool "Intel Core i7"
+---help---
+
+  Select this for the Intel Nehalem platform. Intel Nehalem proecessors
+  include Core i3, i5, i7, Xeon: 34xx, 35xx, 55xx, 56xx, 75xx processors.
+
+  Enables -march=corei7
+
+config MCOREI7AVX
+bool "Intel Core 2nd Gen AVX"
+---help---
+
+  Select this for 2nd Gen Core processors including Sandy Bridge.
+
+  Enables -march=corei7-avx
+
+config MCOREAVXI
+bool "Intel Core 3rd Gen AVX"
+---help---
+
+  Select this for 3rd Gen Core processors including Ivy Bridge.
+
+  Enables -march=core-avx-i
+
+config MCOREAVX2
+bool "Intel Core AVX-2"
+---help---
+
+  Select this for AVX-2 enabled processors including Haswell.
+
+  Enables -march=corei7-avx-2
+
 config MATOM
 bool "Intel Atom"
 ---help---
@@ -312,7 +375,7 @@
 config X86_L1_CACHE_SHIFT
 int
 default "7" if MPENTIUM4 || MPSC
-default "6" if MK7 || MK8 || MPENTIUMM || MCORE2 || MATOM || MVIAC7 || X86_GENERIC || GENERIC_CPU
+default "6" if MK7 || MK8 || MBARCELONA || MBOBCAT || MBULLDOZER || MPILEDRIVER || MPENTIUMM || MCORE2 || MCOREI7 || MCOREI7AVX || MCOREAVXI || MCOREAVX2 || MATOM || MVIAC7 || X86_GENERIC || GENERIC_CPU
 default "4" if MELAN || M486 || M386 || MGEODEGX1
 default "5" if MWINCHIP3D || MWINCHIPC6 || MCRUSOE || MEFFICEON || MCYRIXIII || MK6 || MPENTIUMIII || MPENTIUMII || M686 || M586MMX || M586TSC || M586 || MVIAC3_2 || MGEODE_LX
 
@@ -363,11 +426,11 @@
 
 config X86_INTEL_USERCOPY
 def_bool y
-depends on MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M586MMX || X86_GENERIC || MK8 || MK7 || MEFFICEON || MCORE2
+depends on MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M586MMX || X86_GENERIC || MK8 || MK7 || MBARCELONA || MEFFICEON || MCORE2 || MCOREI7 || MCOREI7AVX || MCOREAVXI || MCOREAVX2
 
 config X86_USE_PPRO_CHECKSUM
 def_bool y
-depends on MWINCHIP3D || MWINCHIPC6 || MCYRIXIII || MK7 || MK6 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || MK8 || MVIAC3_2 || MVIAC7 || MEFFICEON || MGEODE_LX || MCORE2 || MATOM
+depends on MWINCHIP3D || MWINCHIPC6 || MCYRIXIII || MK7 || MK6 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || MK8 || MVIAC3_2 || MVIAC7 || MEFFICEON || MGEODE_LX || MCORE2 || MCOREI7 || MCOREI7AVX || MCOREAVXI || MCOREAVX2 || MATOM
 
 config X86_USE_3DNOW
 def_bool y
@@ -395,17 +458,17 @@
 
 config X86_TSC
 def_bool y
-depends on ((MWINCHIP3D || MCRUSOE || MEFFICEON || MCYRIXIII || MK7 || MK6 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || M586MMX || M586TSC || MK8 || MVIAC3_2 || MVIAC7 || MGEODEGX1 || MGEODE_LX || MCORE2 || MATOM) && !X86_NUMAQ) || X86_64
+depends on ((MWINCHIP3D || MCRUSOE || MEFFICEON || MCYRIXIII || MK7 || MK6 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || M586MMX || M586TSC || MK8 || MBARCELONA || MBOBCAT || MBULLDOZER || MPILEDRIVER || MVIAC3_2 || MVIAC7 || MGEODEGX1 || MGEODE_LX || MCORE2 || MCOREI7 || MCOREI7-AVX || MATOM) && !X86_NUMAQ) || X86_64
 
 config X86_CMPXCHG64
 def_bool y
-depends on X86_PAE || X86_64 || MCORE2 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || MATOM
+depends on X86_PAE || X86_64 || MCORE2 || MCOREI7 || MCOREI7AVX || MCOREAVXI || MCOREAVX2 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || MATOM
 
 # this should be set for all -march=.. options where the compiler
 # generates cmov.
 config X86_CMOV
 def_bool y
-depends on (MK8 || MK7 || MCORE2 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || MVIAC3_2 || MVIAC7 || MCRUSOE || MEFFICEON || X86_64 || MATOM || MGEODE_LX)
+depends on (MK8 || MBARCELONA || MBOBCAT || MBULLDOZER || MPILEDRIVER || MK7 || MCORE2 || MCOREI7 || MCOREI7AVX || MCOREAVXI || MCOREAVX2 || MPENTIUM4 || MPENTIUMM || MPENTIUMIII || MPENTIUMII || M686 || MVIAC3_2 || MVIAC7 || MCRUSOE || MEFFICEON || X86_64 || MATOM || MGEODE_LX)
 
 config X86_MINIMUM_CPU_FAMILY
 int
--- linux-3.6/arch/x86/Makefile2012-04-29 18:43:58.249336198 -0700
+++ linux-3.6.mod/arch/x86/Makefile2012-04-29 18:41:51.260932642 -0700
@@ -51,10 +51,22 @@
 
         # FIXME - should be integrated in Makefile.cpu (Makefile_32.cpu)
         cflags-$(CONFIG_MK8) += $(call cc-option,-march=k8)
+        cflags-$(CONFIG_MBARCELONA) += $(call cc-option,-march=barcelona)
+        cflags-$(CONFIG_MBOBCAT) += $(call cc-option,-march=btver1)
+        cflags-$(CONFIG_MBULLDOZER) += $(call cc-option,-march=bdver1)
+        cflags-$(CONFIG_MPILEDRIVER) += $(call cc-option,-march=bdver2)
         cflags-$(CONFIG_MPSC) += $(call cc-option,-march=nocona)
 
         cflags-$(CONFIG_MCORE2) += \
-                $(call cc-option,-march=core2,$(call cc-option,-mtune=generic))
+                $(call cc-option,-march=core2,$(call cc-option,-mtune=core2))
+        cflags-$(CONFIG_MCOREI7) += \
+                $(call cc-option,-march=corei7,$(call cc-option,-mtune=corei7))
+        cflags-$(CONFIG_MCOREI7AVX) += \
+                $(call cc-option,-march=corei7-avx,$(call cc-option,-mtune=corei7-avx))
+        cflags-$(CONFIG_MCOREAVXI) += \
+                $(call cc-option,-march=core-avx-i,$(call cc-option,-mtune=core-avx-i))
+        cflags-$(CONFIG_MCOREAVX2) += \
+                $(call cc-option,-march=core-avx2,$(call cc-option,-mtune=core-avx2))
 cflags-$(CONFIG_MATOM) += $(call cc-option,-march=atom) \
 $(call cc-option,-mtune=atom,$(call cc-option,-mtune=generic))
         cflags-$(CONFIG_GENERIC_CPU) += $(call cc-option,-mtune=generic)
--- linux-3.6/arch/x86/Makefile_32.cpu2012-04-29 18:43:58.269335977 -0700
+++ linux-3.6.mod/arch/x86/Makefile_32.cpu2012-04-29 18:42:39.120330975 -0700
@@ -25,6 +25,10 @@
 # They make zero difference whatsosever to performance at this time.
 cflags-$(CONFIG_MK7)+= -march=athlon
 cflags-$(CONFIG_MK8)+= $(call cc-option,-march=k8,-march=athlon)
+cflags-$(CONFIG_MBARCELONA)+= $(call cc-option,-march=barcelona,-march=athlon)
+cflags-$(CONFIG_MBOBCAT)+= $(call cc-option,-march=btver1,-march=athlon)
+cflags-$(CONFIG_MBULLDOZER)+= $(call cc-option,-march=bdver1,-march=athlon)
+cflags-$(CONFIG_MPILEDRIVER)+= $(call cc-option,-march=bdver2,-march=athlon)
 cflags-$(CONFIG_MCRUSOE)+= -march=i686 $(align)-functions=0 $(align)-jumps=0 $(align)-loops=0
 cflags-$(CONFIG_MEFFICEON)+= -march=i686 $(call tune,pentium3) $(align)-functions=0 $(align)-jumps=0 $(align)-loops=0
 cflags-$(CONFIG_MWINCHIPC6)+= $(call cc-option,-march=winchip-c6,-march=i586)
@@ -33,6 +37,10 @@
 cflags-$(CONFIG_MVIAC3_2)+= $(call cc-option,-march=c3-2,-march=i686)
 cflags-$(CONFIG_MVIAC7)+= -march=i686
 cflags-$(CONFIG_MCORE2)+= -march=i686 $(call tune,core2)
+cflags-$(CONFIG_MCOREI7)+= -march=i686 $(call tune,corei7)
+cflags-$(CONFIG_MCOREI7AVX)+= -march=i686 $(call tune,corei7-avx)
+cflags-$(CONFIG_MCOREAVXI)+= -march=i686 $(call tune,core-avx-i)
+cflags-$(CONFIG_MCOREAVX2)+= -march=i686 $(call tune,core-avx-2)
 cflags-$(CONFIG_MATOM)+= $(call cc-option,-march=atom,$(call cc-option,-march=core2,-march=i686)) \
 $(call cc-option,-mtune=atom,$(call cc-option,-mtune=generic))

Re: [PATCH] Expand CPU compiler options

[Borislav Petkov]

On Sat, Dec 08, 2012 at 04:13:59AM -0800, John wrote:
> I tested the attached patch written by André Ramnitz using three
> different machines running a generic x86-64 kernel and an otherwise
> identical kernel running with the optimized gcc options. 
>
> Conclusion: There are small but real speed increases using a make
> endpoint to running with this patch.
>
> Details: 1) Three test machines: Intel Xeon X3360, Intel i7-2620M,
> Intel Core i7-3660K.
>
> 2) All ran the make benchmark (linked below) 35 times while booted
> into a 'generic' kernel. Then all ran the same make benchmark
> 35 times after booting into an optimized kernel. Below are the
> optimizations chosen for each machine. 2a) X3360 = core2 2b)
> i7-2620M = corei7-avx 2c) i7-3660K = core-avx-i 3) Analyzed
> resulting distributions for statistical significance via ANOVA
> plots that clearly show statistically significant albeit small
> differences.
>
> Links to ANOVA plots:
> http://s19.postimage.org/68urcofzn/corei7_avx.png
> http://s19.postimage.org/ozwomuak3/core_avx_i.png
>
> http://s19.postimage.org/d0l6fj4z7/core2.png
>
>
> References:
>
> Bash script that controls the
> benchmark: https://github.com/graysky2/bin/blob/master/bench
> Log file generated by script:
> http://repo-ck.com/bench/compile_time_optimization.txt.gz

Let's see, if I'm reading the log file correctly, the average values of
each test run differ by ~ 0.1 seconds tops.

For example, i7-3770K generic build gives on average 69.41404 while
the more optimized version 69.33554. The diff between the two is even
less than 0.1 second. The other two machines' diff is a bit higher. And
from looking at your graphs, this is all eaten up by stddev so I'd say
there aren't any improvements from using a different uarch target - just
noise. AFAICT, at least.

You could trace this same workload with perf as I told you originally to
see whether there are some other uarch benefits, and have a more precise
time measurement than using 'date' but I'm very sceptical.

In any case, these results are too marginal to warrant any code change
since they're basically disappearing in noise.

Btw, you might look into what optimizations exactly went into those
different compiler options - they might not be improving a lot, if at
all, performance-wise but be adding support for new instructions, etc,
etc, i.e. features which are not related to performance. And if that
is the case, there's no need for those different uarch build targets
in the kernel. Remember, the majority of linux kernels out there are
generic-x86_64 builds.

Thanks.

-- 
Regards/Gruss,
    Boris.

Re: [PATCH] Expand CPU compiler options

[John]

> Let's see, if I'm reading the log file correctly, the average values of

> each test run differ by ~ 0.1 seconds tops.
> 
> For example, i7-3770K generic build gives on average 69.41404 while
> the more optimized version 69.33554. The diff between the two is even
> less than 0.1 second. The other two machines' diff is a bit higher. And
> from looking at your graphs, this is all eaten up by stddev so I'd say
> there aren't any improvements from using a different uarch target - just
> noise. AFAICT, at least.


While I agree that the differences as small - on the order of ms - they are not insignificant nor are they in the noise of the measurements.  
1) All the assumptions for ANOVA are met:
*Data are normally distributed as show in the normal quantile plots.

*The population variances are fairly equal (Levene and Barlett tests).

2) The ANOVA plots clearly show significance.
*Pair-wise analysis by Tukey-Kramer shows significance at the 0.05 level for all CPUs compared.  Below are the differences in median values:
core2 +87.5 ms
corei7-avx +79.7 ms
core-avx-i+257.2 ms

> In any case, these results are too marginal to warrant any code change
> since they're basically disappearing in noise.


Currently, the 'core2' optimization is included in as an option for a CPU Family.  In the case of core-avx-i, I am showing an improvement of ~3x over that.  And in the corei7-avx case, the improvement is approximately equal to it.  I don't understand why these wouldn't be considered value-added since they preform at least as good as the option already included.

Thank you for the consideration and suggestions.

Re: [PATCH] Expand CPU compiler options

[Borislav Petkov]

On Sun, Dec 09, 2012 at 04:08:55AM -0800, John wrote:
> While I agree that the differences as small - on the order of
> ms - they are not insignificant nor are they in the noise of the
> measurements.

Oh, maybe I wasn't clear - I wasn't talking about statistical
significance but rather about practical significance.

A minuscule speedup (a lot less than 1%) showing only in a *single*
workload so far and relevant only for a *very* *small* number of linux
users - only those who build their own kernels - is simply not worth
including in the kernel.

-- 
Regards/Gruss,
    Boris.

Re: [PATCH] Expand CPU compiler options

[John]

> Oh, maybe I wasn't clear - I wasn't talking about statistical

> significance but rather about practical significance.

OK.  I will not argue with that :)

> A minuscule speedup (a lot less than 1%) showing only in a *single*
> workload so far and relevant only for a *very* *small* number of linux
> users - only those who build their own kernels - is simply not worth
> including in the kernel.


My only confusion is why some of these are already included in the linux kernel source today, for example CORE2.  As I stated in my previous email, the two 'new' ones I tested preform as-good-as or better-than the CORE2 which is already included.  Does my confusion make sense?

Thanks again for the consideration!

Re: [PATCH] Expand CPU compiler options

[Borislav Petkov]

On Sun, Dec 09, 2012 at 04:59:41AM -0800, John wrote:
> My only confusion is why some of these are already included in the
> linux kernel source today, for example CORE2.  As I stated in my
> previous email, the two 'new' ones I tested preform as-good-as or
> better-than the CORE2 which is already included.  Does my confusion
> make sense?

That's a good question. Hmm, next question :-).

In any case, I've yet to see significant perf improvements to the kernel
due to it being built with specific uarch -mtune settings.

-- 
Regards/Gruss,
    Boris.