acl: fix build with GCC 6.3
Checks
Commit Message
--buildtype=debug with gcc 6.3 produces the following error:
../lib/librte_acl/acl_run_avx512_common.h: In function
‘resolve_match_idx_avx512x16’:
../lib/librte_acl/acl_run_avx512x16.h:33:18: error:
the last argument must be an 8-bit immediate
^
../lib/librte_acl/acl_run_avx512_common.h:373:9: note:
in expansion of macro ‘_M_I_’
return _M_I_(slli_epi32)(mi, match_log);
^~~~~
Seems like gcc-6.3 complains about the following construct:
static const uint32_t match_log = 5;
...
_mm512_slli_epi32(mi, match_log);
It can't substitute constant variable 'match_log' with its actual value.
The fix replaces constant variable with its immediate value.
Bugzilla ID: 717
Fixes: b64c2295f7fc ("acl: add 256-bit AVX512 classify method")
Fixes: 45da22e42ec3 ("acl: add 512-bit AVX512 classify method")
Cc: stable@dpdk.org
Reported-by: Liang Ma <liangma@liangbit.com>
Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
---
lib/acl/acl_run_avx512.c | 8 ++++----
lib/acl/acl_run_avx512_common.h | 4 ++--
lib/acl/acl_run_avx512x16.h | 6 ++----
3 files changed, 8 insertions(+), 10 deletions(-)
Comments
On Fri, May 21, 2021 at 03:42:07PM +0100, Konstantin Ananyev wrote:
<snip>
I apply this patch but it caused i40e pmd avx512 build failure.
not sure about the root cause. I build debug version with latest repo.
21/05/2021 16:42, Konstantin Ananyev:
> --buildtype=debug with gcc 6.3 produces the following error:
>
> ../lib/librte_acl/acl_run_avx512_common.h: In function
> ‘resolve_match_idx_avx512x16’:
> ../lib/librte_acl/acl_run_avx512x16.h:33:18: error:
> the last argument must be an 8-bit immediate
> ^
> ../lib/librte_acl/acl_run_avx512_common.h:373:9: note:
> in expansion of macro ‘_M_I_’
> return _M_I_(slli_epi32)(mi, match_log);
> ^~~~~
>
> Seems like gcc-6.3 complains about the following construct:
>
> static const uint32_t match_log = 5;
> ...
> _mm512_slli_epi32(mi, match_log);
>
> It can't substitute constant variable 'match_log' with its actual value.
> The fix replaces constant variable with its immediate value.
>
> Bugzilla ID: 717
> Fixes: b64c2295f7fc ("acl: add 256-bit AVX512 classify method")
> Fixes: 45da22e42ec3 ("acl: add 512-bit AVX512 classify method")
> Cc: stable@dpdk.org
>
> Reported-by: Liang Ma <liangma@liangbit.com>
> Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
How much critical is it?
It looks safer to wait for 21.08 cycle?
> -----Original Message-----
> From: Thomas Monjalon <thomas@monjalon.net>
> Sent: Friday, May 21, 2021 4:11 PM
> To: liangma@liangbit.com; Ananyev, Konstantin <konstantin.ananyev@intel.com>
> Cc: dev@dpdk.org; stable@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH] acl: fix build with GCC 6.3
>
> 21/05/2021 16:42, Konstantin Ananyev:
> > --buildtype=debug with gcc 6.3 produces the following error:
> >
> > ../lib/librte_acl/acl_run_avx512_common.h: In function
> > ‘resolve_match_idx_avx512x16’:
> > ../lib/librte_acl/acl_run_avx512x16.h:33:18: error:
> > the last argument must be an 8-bit immediate
> > ^
> > ../lib/librte_acl/acl_run_avx512_common.h:373:9: note:
> > in expansion of macro ‘_M_I_’
> > return _M_I_(slli_epi32)(mi, match_log);
> > ^~~~~
> >
> > Seems like gcc-6.3 complains about the following construct:
> >
> > static const uint32_t match_log = 5;
> > ...
> > _mm512_slli_epi32(mi, match_log);
> >
> > It can't substitute constant variable 'match_log' with its actual value.
> > The fix replaces constant variable with its immediate value.
> >
> > Bugzilla ID: 717
> > Fixes: b64c2295f7fc ("acl: add 256-bit AVX512 classify method")
> > Fixes: 45da22e42ec3 ("acl: add 512-bit AVX512 classify method")
> > Cc: stable@dpdk.org
> >
> > Reported-by: Liang Ma <liangma@liangbit.com>
> > Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
>
> How much critical is it?
> It looks safer to wait for 21.08 cycle?
I think we can wait till 21.08 cycle.
On Fri, May 21, 2021 at 05:10:41PM +0200, Thomas Monjalon wrote:
> 21/05/2021 16:42, Konstantin Ananyev:
> > --buildtype=debug with gcc 6.3 produces the following error:
> >
> > ../lib/librte_acl/acl_run_avx512_common.h: In function
> > ‘resolve_match_idx_avx512x16’:
> > ../lib/librte_acl/acl_run_avx512x16.h:33:18: error:
> > the last argument must be an 8-bit immediate
> > ^
> > ../lib/librte_acl/acl_run_avx512_common.h:373:9: note:
> > in expansion of macro ‘_M_I_’
> > return _M_I_(slli_epi32)(mi, match_log);
> > ^~~~~
> >
> > Seems like gcc-6.3 complains about the following construct:
> >
> > static const uint32_t match_log = 5;
> > ...
> > _mm512_slli_epi32(mi, match_log);
> >
> > It can't substitute constant variable 'match_log' with its actual value.
> > The fix replaces constant variable with its immediate value.
> >
> > Bugzilla ID: 717
> > Fixes: b64c2295f7fc ("acl: add 256-bit AVX512 classify method")
> > Fixes: 45da22e42ec3 ("acl: add 512-bit AVX512 classify method")
> > Cc: stable@dpdk.org
> >
> > Reported-by: Liang Ma <liangma@liangbit.com>
> > Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
>
> How much critical is it?
> It looks safer to wait for 21.08 cycle?
+1. gcc8.30 is OK. It's a bit late now.
>
>
21/05/2021 16:42, Konstantin Ananyev:
> --buildtype=debug with gcc 6.3 produces the following error:
>
> ../lib/librte_acl/acl_run_avx512_common.h: In function
> ‘resolve_match_idx_avx512x16’:
> ../lib/librte_acl/acl_run_avx512x16.h:33:18: error:
> the last argument must be an 8-bit immediate
> ^
> ../lib/librte_acl/acl_run_avx512_common.h:373:9: note:
> in expansion of macro ‘_M_I_’
> return _M_I_(slli_epi32)(mi, match_log);
> ^~~~~
>
> Seems like gcc-6.3 complains about the following construct:
>
> static const uint32_t match_log = 5;
> ...
> _mm512_slli_epi32(mi, match_log);
>
> It can't substitute constant variable 'match_log' with its actual value.
> The fix replaces constant variable with its immediate value.
>
> Bugzilla ID: 717
> Fixes: b64c2295f7fc ("acl: add 256-bit AVX512 classify method")
> Fixes: 45da22e42ec3 ("acl: add 512-bit AVX512 classify method")
> Cc: stable@dpdk.org
>
> Reported-by: Liang Ma <liangma@liangbit.com>
> Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Applied, thanks
@@ -4,8 +4,8 @@
#include "acl_run_sse.h"
-/*sizeof(uint32_t) << match_log == sizeof(struct rte_acl_match_results)*/
-static const uint32_t match_log = 5;
+/*sizeof(uint32_t) << ACL_MATCH_LOG == sizeof(struct rte_acl_match_results)*/
+#define ACL_MATCH_LOG 5
struct acl_flow_avx512 {
uint32_t num_packets; /* number of packets processed */
@@ -82,7 +82,7 @@ resolve_mcle8_avx512x1(uint32_t result[],
for (k = 0; k != nb_pkt; k++, result += nb_cat) {
- mi = match[k] << match_log;
+ mi = match[k] << ACL_MATCH_LOG;
for (j = 0; j != nb_cat; j += RTE_ACL_RESULTS_MULTIPLIER) {
@@ -92,7 +92,7 @@ resolve_mcle8_avx512x1(uint32_t result[],
for (i = 1, pm = match + nb_pkt; i != nb_trie;
i++, pm += nb_pkt) {
- mn = j + (pm[k] << match_log);
+ mn = j + (pm[k] << ACL_MATCH_LOG);
nr = _mm_loadu_si128((const xmm_t *)(res + mn));
np = _mm_loadu_si128((const xmm_t *)(pri + mn));
@@ -393,8 +393,8 @@ static inline _T_simd
_F_(resolve_match_idx)(_T_simd mi)
{
RTE_BUILD_BUG_ON(sizeof(struct rte_acl_match_results) !=
- 1 << (match_log + 2));
- return _M_I_(slli_epi32)(mi, match_log);
+ 1 << (ACL_MATCH_LOG + 2));
+ return _M_I_(slli_epi32)(mi, ACL_MATCH_LOG);
}
/*
@@ -252,8 +252,6 @@ resolve_mcgt8_avx512x1(uint32_t result[],
__mmask16 cm, sm;
__m512i cp, cr, np, nr;
- const uint32_t match_log = 5;
-
res = pr->results;
pri = pr->priority;
@@ -261,7 +259,7 @@ resolve_mcgt8_avx512x1(uint32_t result[],
for (k = 0; k != nb_pkt; k++, result += nb_cat) {
- mi = match[k] << match_log;
+ mi = match[k] << ACL_MATCH_LOG;
cr = _mm512_maskz_loadu_epi32(cm, res + mi);
cp = _mm512_maskz_loadu_epi32(cm, pri + mi);
@@ -269,7 +267,7 @@ resolve_mcgt8_avx512x1(uint32_t result[],
for (i = 1, pm = match + nb_pkt; i != nb_trie;
i++, pm += nb_pkt) {
- mi = pm[k] << match_log;
+ mi = pm[k] << ACL_MATCH_LOG;
nr = _mm512_maskz_loadu_epi32(cm, res + mi);
np = _mm512_maskz_loadu_epi32(cm, pri + mi);