diff options
| author |   Damjan Marion <damarion@cisco.com> | 2016-05-11 23:07:18 +0200 |
|---|---|---|
| committer | Damjan Marion <damarion@cisco.com> | 2016-05-19 18:14:38 +0200 |
| commit | 1c80e831b728ab378949714d5059a0b5b1822a0a (patch) | |
| tree | e4f3ecfee143f7dee0e9905570d41ca3ee345c83 /build-data/platforms | |
| parent | 82e29c455833b5b12e04c89d2dec1106b499e6b0 (diff) | |
Add support for multiple microarchitectures in single binary
* compiler -march= parameter is changed from native to corei7
so code is always genereted with instructions which are available
on the Nehalem microarchitecture (up to SSE4.2)
* compiler -mtune= parameter is added so code is optimized for
corei7-avx which equals to Sandy Bridge microarchitecture
* set of macros is added which allows run-time detection of available
cpu instructions (e.g. clib_cpu_supports_avx())
* set of macros is added which allows us to clone graph node funcitons
where cloned function is optmized for different microarchitecture
Those macros are using following attributes:
__attribute__((flatten))
__attribute__((target("arch=core-avx2)))
I.e. If applied to foo_node_fn() macro will generate cloned
functions foo_node_fn_avx2() and foo_node_fn_avx512() (future)
It will also generate function void * foo_node_fn_multiarch_select()
which detects available instruction set and returns pointer to the
best matching function clone.
Change-Id: I2dce0ac92a5ede95fcb56f47f3d1f3c4c040bac0
Signed-off-by: Damjan Marion <damarion@cisco.com>
Diffstat (limited to 'build-data/platforms')
| -rw-r--r-- | build-data/platforms/vpp.mk | 7 | ||||
| -rw-r--r-- | build-data/platforms/vpp_lite.mk | 6 |
2 files changed, 9 insertions, 4 deletions
diff --git a/build-data/platforms/vpp.mk b/build-data/platforms/vpp.mk index c381be19b42..ec0d874db4b 100644 --- a/build-data/platforms/vpp.mk +++ b/build-data/platforms/vpp.mk @@ -13,6 +13,9 @@ # vector packet processor vpp_arch = native +vpp_march = corei7 # Nehalem Instruction set +vpp_mtune = corei7-avx # Optimize for Sandy Bridge +vpp_dpdk_arch = corei7 vpp_native_tools = vppapigen vpp_uses_dpdk = yes @@ -40,9 +43,9 @@ vpp_debug_TAG_CFLAGS = -g -O0 -DCLIB_DEBUG -DFORTIFY_SOURCE=2 -march=$(MARCH) \ vpp_debug_TAG_LDFLAGS = -g -O0 -DCLIB_DEBUG -DFORTIFY_SOURCE=2 -march=$(MARCH) \ -fstack-protector-all -fPIC -Werror -vpp_TAG_CFLAGS = -g -O2 -DFORTIFY_SOURCE=2 -march=$(MARCH) \ +vpp_TAG_CFLAGS = -g -O2 -DFORTIFY_SOURCE=2 -march=$(MARCH) -mtune=$(MTUNE) \ -fstack-protector -fPIC -Werror -vpp_TAG_LDFLAGS = -g -O2 -DFORTIFY_SOURCE=2 -march=$(MARCH) \ +vpp_TAG_LDFLAGS = -g -O2 -DFORTIFY_SOURCE=2 -march=$(MARCH) -mtune=$(MTUNE) \ -fstack-protector -fPIC -Werror vpp_gcov_TAG_CFLAGS = -g -O0 -DCLIB_DEBUG -march=$(MARCH) \ diff --git a/build-data/platforms/vpp_lite.mk b/build-data/platforms/vpp_lite.mk index 4183700e681..d35d2347e69 100644 --- a/build-data/platforms/vpp_lite.mk +++ b/build-data/platforms/vpp_lite.mk @@ -13,6 +13,8 @@ # vector packet processor vpp_lite_arch = native +vpp_lite_march = corei7 # Nehalem Instruction set +vpp_lite_mtune = corei7-avx # Optimize for Sandy Bridge vpp_lite_native_tools = vppapigen vpp_lite_uses_dpdk = no @@ -30,7 +32,7 @@ vpp_lite_debug_TAG_CFLAGS = -g -O0 -DCLIB_DEBUG -DFORTIFY_SOURCE=2 -march=$(MARC vpp_lite_debug_TAG_LDFLAGS = -g -O0 -DCLIB_DEBUG -DFORTIFY_SOURCE=2 -march=$(MARCH) \ -fstack-protector-all -fPIC -Werror -vpp_lite_TAG_CFLAGS = -g -O2 -DFORTIFY_SOURCE=2 -march=$(MARCH) \ +vpp_lite_TAG_CFLAGS = -g -O2 -DFORTIFY_SOURCE=2 -march=$(MARCH) -mtune=$(MTUNE) \ -fstack-protector -fPIC -Werror -vpp_lite_TAG_LDFLAGS = -g -O2 -DFORTIFY_SOURCE=2 -march=$(MARCH) \ +vpp_lite_TAG_LDFLAGS = -g -O2 -DFORTIFY_SOURCE=2 -march=$(MARCH) -mtune=$(MTUNE) \ -fstack-protector -fPIC -Werror |