From: Tony Lu <zlu@ezchip.com>
To: 'Neil Horman' <nhorman@tuxdriver.com>
Cc: dev@dpdk.org
Subject: Re: [dpdk-dev] [PATCH 14/15] app/test: turn off cpu flag checks for tile architecture
Date: Thu, 11 Dec 2014 12:43:36 +0800 [thread overview]
Message-ID: <000901d014fd$106f85d0$314e9170$@com> (raw)
In-Reply-To: <20141209150321.GC28871@hmsreliant.think-freely.org>
>-----Original Message-----
>From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Neil Horman
>Sent: Tuesday, December 09, 2014 11:03 PM
>To: Zhigang Lu
>Cc: dev@dpdk.org
>Subject: Re: [dpdk-dev] [PATCH 14/15] app/test: turn off cpu flag checks
for tile
>architecture
>
>On Mon, Dec 08, 2014 at 04:59:37PM +0800, Zhigang Lu wrote:
>> Tile processor doesn't have CPU flag hardware registers, so this patch
>> turns off cpu flag checks for tile.
>>
>> Signed-off-by: Zhigang Lu <zlu@ezchip.com>
>> Signed-off-by: Cyril Chemparathy <cchemparathy@ezchip.com>
>> ---
>> app/test/test_cpuflags.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/app/test/test_cpuflags.c b/app/test/test_cpuflags.c index
>> 5aeba5d..da93af5 100644
>> --- a/app/test/test_cpuflags.c
>> +++ b/app/test/test_cpuflags.c
>> @@ -113,7 +113,7 @@ test_cpuflags(void)
>>
>> printf("Check for ICACHE_SNOOP:\t\t");
>> CHECK_FOR_FLAG(RTE_CPUFLAG_ICACHE_SNOOP);
>> -#else
>> +#elif !defined(RTE_ARCH_TILE)
>> printf("Check for SSE:\t\t");
>> CHECK_FOR_FLAG(RTE_CPUFLAG_SSE);
>>
>Please stop this. It doesn't make sense for a library that supports
multiple
>arches, we need some way to generically test for flags that doesn't involve
>forcing applications to do ton's of ifdeffing. Perhaps
rte_cpu_get_flag_enabled
>needs to do a flag table lookup based on the detected arch at run time, and
>return the appropriate response. In the case of tile, it can just be an
empty
>table, so 0 is always returned. But making an application responsible for
doing
>arch checks is a guarantee to write non-portable applications
>
>Neil
>
Thanks for taking a look at this.
This change just follows what PPC did in commit 9ae15538. The root cause is
that
the test_cpuflags.c explicitly tests X86-specific CPU flags, so we might
need to revise
this test case to make it architecture-independent.
A alternative change to this test case is as follows.
--- a/app/test/test_cpuflags.c
+++ b/app/test/test_cpuflags.c
@@ -77,81 +77,13 @@ cpu_flag_result(int result)
static int
test_cpuflags(void)
{
- int result;
+ int i, result;
printf("\nChecking for flags from different registers...\n");
-#ifdef RTE_ARCH_PPC_64
- printf("Check for PPC64:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_PPC64);
-
- printf("Check for PPC32:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_PPC32);
-
- printf("Check for VSX:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_VSX);
-
- printf("Check for DFP:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_DFP);
-
- printf("Check for FPU:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_FPU);
-
- printf("Check for SMT:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_SMT);
-
- printf("Check for MMU:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_MMU);
-
- printf("Check for ALTIVEC:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_ALTIVEC);
-
- printf("Check for ARCH_2_06:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_ARCH_2_06);
-
- printf("Check for ARCH_2_07:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_ARCH_2_07);
-
- printf("Check for ICACHE_SNOOP:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_ICACHE_SNOOP);
-#else
- printf("Check for SSE:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_SSE);
-
- printf("Check for SSE2:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_SSE2);
-
- printf("Check for SSE3:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_SSE3);
-
- printf("Check for SSE4.1:\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_SSE4_1);
-
- printf("Check for SSE4.2:\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_SSE4_2);
-
- printf("Check for AVX:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_AVX);
-
- printf("Check for AVX2:\t\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_AVX2);
-
- printf("Check for TRBOBST:\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_TRBOBST);
-
- printf("Check for ENERGY_EFF:\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_ENERGY_EFF);
-
- printf("Check for LAHF_SAHF:\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_LAHF_SAHF);
-
- printf("Check for 1GB_PG:\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_1GB_PG);
-
- printf("Check for INVTSC:\t");
- CHECK_FOR_FLAG(RTE_CPUFLAG_INVTSC);
-
-
-#endif
+ for (i = 0; i < RTE_CPUFLAG_NUMFLAGS; i++) {
+ printf("Check for %s:\t\t", cpu_feature_table[i].name);
+ CHECK_FOR_FLAG(i);
+ }
Thanks
-Zhigang
next prev parent reply other threads:[~2014-12-11 4:44 UTC|newest]
Thread overview: 32+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-12-08 8:59 [dpdk-dev] [PATCH 00/15] Patches for DPDK to support " Zhigang Lu
2014-12-08 8:59 ` [dpdk-dev] [PATCH 01/15] mk: introduce Tilera Tile architecture Zhigang Lu
2014-12-08 11:09 ` Bruce Richardson
2014-12-08 14:25 ` Neil Horman
2014-12-08 21:34 ` Cyril Chemparathy
2014-12-08 8:59 ` [dpdk-dev] [PATCH 02/15] eal/tile: add atomic operations for TileGx Zhigang Lu
2014-12-08 14:28 ` Neil Horman
2014-12-08 21:29 ` Cyril Chemparathy
2014-12-08 8:59 ` [dpdk-dev] [PATCH 03/15] eal/tile: add byte order " Zhigang Lu
2014-12-08 8:59 ` [dpdk-dev] [PATCH 04/15] eal/tile: add spinlock " Zhigang Lu
2014-12-08 8:59 ` [dpdk-dev] [PATCH 05/15] eal/tile: add prefetch " Zhigang Lu
2014-12-08 8:59 ` [dpdk-dev] [PATCH 06/15] eal/tile: add memcpy " Zhigang Lu
2014-12-08 8:59 ` [dpdk-dev] [PATCH 07/15] eal/tile: add CPU flags " Zhigang Lu
2014-12-08 8:59 ` [dpdk-dev] [PATCH 08/15] eal/tile: add cycle " Zhigang Lu
2014-12-08 8:59 ` [dpdk-dev] [PATCH 09/15] eal: split vector operations to architecture specific Zhigang Lu
2014-12-08 8:59 ` [dpdk-dev] [PATCH 10/15] eal/tile: add vector operations for TileGx Zhigang Lu
2014-12-08 8:59 ` [dpdk-dev] [PATCH 11/15] eal/tile: add EAL support for global mPIPE initialization Zhigang Lu
2014-12-08 20:03 ` Neil Horman
2014-12-08 21:32 ` Cyril Chemparathy
2014-12-09 11:36 ` Neil Horman
2014-12-08 8:59 ` [dpdk-dev] [PATCH 12/15] eal/tile: add mPIPE buffer stack mempool provider Zhigang Lu
2014-12-09 14:07 ` Neil Horman
2014-12-12 8:30 ` Tony Lu
2014-12-12 13:03 ` Neil Horman
2014-12-08 8:59 ` [dpdk-dev] [PATCH 13/15] pmd/tile: add mPIPE poll mode driver for TileGx Zhigang Lu
2014-12-08 8:59 ` [dpdk-dev] [PATCH 14/15] app/test: turn off cpu flag checks for tile architecture Zhigang Lu
2014-12-09 15:03 ` Neil Horman
2014-12-11 4:43 ` Tony Lu [this message]
2014-12-11 13:39 ` Neil Horman
2014-12-12 8:10 ` Tony Lu
2014-12-12 13:07 ` Neil Horman
2014-12-08 8:59 ` [dpdk-dev] [PATCH 15/15] eal: allow empty set of compile time cpuflags Zhigang Lu
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='000901d014fd$106f85d0$314e9170$@com' \
--to=zlu@ezchip.com \
--cc=dev@dpdk.org \
--cc=nhorman@tuxdriver.com \
/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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).