Profiling of a page fault scalability microbenchmark shows flush_hash_range
is not calling the batch hpte invalidate hcall (H_BULK_REMOVE).

It turns out we have a duplicate firmware feature for hcall-bulk and the
current setup code stops after finding the first match. This meant we never
batch and always do individual invalidates.

The patch below removes the duplicate and shifts FW_FEATURE_CMO to close
the gap. With the patch applied the single threaded page fault rate improves
from 217169 to 238755 per second on a POWER5 test box, a 10% improvement.

Signed-off-by: Anton Blanchard <an...@samba.org>
---

Index: linux.trees.git/arch/powerpc/include/asm/firmware.h
===================================================================
--- linux.trees.git.orig/arch/powerpc/include/asm/firmware.h    2009-10-12 
18:11:56.000000000 +1100
+++ linux.trees.git/arch/powerpc/include/asm/firmware.h 2009-10-12 
18:21:20.000000000 +1100
@@ -37,7 +37,7 @@
 #define FW_FEATURE_VIO         ASM_CONST(0x0000000000004000)
 #define FW_FEATURE_RDMA                ASM_CONST(0x0000000000008000)
 #define FW_FEATURE_LLAN                ASM_CONST(0x0000000000010000)
-#define FW_FEATURE_BULK                ASM_CONST(0x0000000000020000)
+#define FW_FEATURE_BULK_REMOVE ASM_CONST(0x0000000000020000)
 #define FW_FEATURE_XDABR       ASM_CONST(0x0000000000040000)
 #define FW_FEATURE_MULTITCE    ASM_CONST(0x0000000000080000)
 #define FW_FEATURE_SPLPAR      ASM_CONST(0x0000000000100000)
@@ -45,8 +45,7 @@
 #define FW_FEATURE_LPAR                ASM_CONST(0x0000000000400000)
 #define FW_FEATURE_PS3_LV1     ASM_CONST(0x0000000000800000)
 #define FW_FEATURE_BEAT                ASM_CONST(0x0000000001000000)
-#define FW_FEATURE_BULK_REMOVE ASM_CONST(0x0000000002000000)
-#define FW_FEATURE_CMO         ASM_CONST(0x0000000004000000)
+#define FW_FEATURE_CMO         ASM_CONST(0x0000000002000000)
 
 #ifndef __ASSEMBLY__
 
@@ -58,8 +57,9 @@ enum {
                FW_FEATURE_PERF | FW_FEATURE_DUMP | FW_FEATURE_INTERRUPT |
                FW_FEATURE_MIGRATE | FW_FEATURE_PERFMON | FW_FEATURE_CRQ |
                FW_FEATURE_VIO | FW_FEATURE_RDMA | FW_FEATURE_LLAN |
-               FW_FEATURE_BULK | FW_FEATURE_XDABR | FW_FEATURE_MULTITCE |
-               FW_FEATURE_SPLPAR | FW_FEATURE_LPAR | FW_FEATURE_CMO,
+               FW_FEATURE_BULK_REMOVE | FW_FEATURE_XDABR |
+               FW_FEATURE_MULTITCE | FW_FEATURE_SPLPAR | FW_FEATURE_LPAR |
+               FW_FEATURE_CMO,
        FW_FEATURE_PSERIES_ALWAYS = 0,
        FW_FEATURE_ISERIES_POSSIBLE = FW_FEATURE_ISERIES | FW_FEATURE_LPAR,
        FW_FEATURE_ISERIES_ALWAYS = FW_FEATURE_ISERIES | FW_FEATURE_LPAR,
Index: linux.trees.git/arch/powerpc/platforms/pseries/firmware.c
===================================================================
--- linux.trees.git.orig/arch/powerpc/platforms/pseries/firmware.c      
2009-10-12 18:08:37.000000000 +1100
+++ linux.trees.git/arch/powerpc/platforms/pseries/firmware.c   2009-10-12 
18:12:54.000000000 +1100
@@ -51,11 +51,10 @@ firmware_features_table[FIRMWARE_MAX_FEA
        {FW_FEATURE_VIO,                "hcall-vio"},
        {FW_FEATURE_RDMA,               "hcall-rdma"},
        {FW_FEATURE_LLAN,               "hcall-lLAN"},
-       {FW_FEATURE_BULK,               "hcall-bulk"},
+       {FW_FEATURE_BULK_REMOVE,        "hcall-bulk"},
        {FW_FEATURE_XDABR,              "hcall-xdabr"},
        {FW_FEATURE_MULTITCE,           "hcall-multi-tce"},
        {FW_FEATURE_SPLPAR,             "hcall-splpar"},
-       {FW_FEATURE_BULK_REMOVE,        "hcall-bulk"},
 };
 
 /* Build up the firmware features bitmask using the contents of
_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev@lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev

Reply via email to