* [PATCH 0/2] Instruction set detection for clang.
@ 2020-07-23 0:27 Shu-Chun Weng
2020-07-23 0:27 ` [PATCH 1/2] configure: avx2 and avx512f " Shu-Chun Weng
` (2 more replies)
0 siblings, 3 replies; 8+ messages in thread
From: Shu-Chun Weng @ 2020-07-23 0:27 UTC (permalink / raw)
To: rth; +Cc: Shu-Chun Weng, qemu-devel
Currently when configuring QEMU with clang, AVX2, AVX512F, ATOMIC64, and
ATOMIC128 are all disabled because the detection code is GCC-only. With these
two patches, I am able to configure, build, and run tests with clang with all of
the above enabled.
Shu-Chun Weng (2):
configure: avx2 and avx512f detection for clang
configure: atomic64/128 detection for clang
configure | 34 +++++++++++++++++++++++-----------
util/bufferiszero.c | 33 +++++++++++++++++++++++----------
2 files changed, 46 insertions(+), 21 deletions(-)
--
2.28.0.rc0.105.gf9edc3c819-goog
^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH 1/2] configure: avx2 and avx512f detection for clang
2020-07-23 0:27 [PATCH 0/2] Instruction set detection for clang Shu-Chun Weng
@ 2020-07-23 0:27 ` Shu-Chun Weng
2020-07-23 4:54 ` Thomas Huth
2020-07-23 0:27 ` [PATCH 2/2] configure: atomic64/128 " Shu-Chun Weng
2020-08-05 23:23 ` [PATCH 0/2] Instruction set " Shu-Chun Weng
2 siblings, 1 reply; 8+ messages in thread
From: Shu-Chun Weng @ 2020-07-23 0:27 UTC (permalink / raw)
To: rth; +Cc: Shu-Chun Weng, qemu-devel
Since clang does not support "#pragma GCC", the instruction sets are
always disabled. In this change, we
1. wrap "#pragma GCC" inside "#ifndef __clang__",
2. only retain them around "#include <{e,i,s}mmintrin.h>" to work
around gcc bug,
3. and annotate each function with `__attribute__((target(*)))` which
is recognized by both gcc and clang.
Signed-off-by: Shu-Chun Weng <scw@google.com>
---
configure | 16 ++++++++++++++--
util/bufferiszero.c | 33 +++++++++++++++++++++++----------
2 files changed, 37 insertions(+), 12 deletions(-)
diff --git a/configure b/configure
index 4bd80ed507..d9ce3aa5db 100755
--- a/configure
+++ b/configure
@@ -5808,10 +5808,16 @@ fi
if test "$cpuid_h" = "yes" && test "$avx2_opt" != "no"; then
cat > $TMPC << EOF
+#include <cpuid.h>
+#ifndef __clang__
#pragma GCC push_options
#pragma GCC target("avx2")
-#include <cpuid.h>
+#endif
#include <immintrin.h>
+#ifndef __clang__
+#pragma GCC pop_options
+#endif
+__attribute__((target("avx2")))
static int bar(void *a) {
__m256i x = *(__m256i *)a;
return _mm256_testz_si256(x, x);
@@ -5835,10 +5841,16 @@ fi
if test "$cpuid_h" = "yes" && test "$avx512f_opt" = "yes"; then
cat > $TMPC << EOF
+#include <cpuid.h>
+#ifndef __clang__
#pragma GCC push_options
#pragma GCC target("avx512f")
-#include <cpuid.h>
+#endif
#include <immintrin.h>
+#ifndef __clang__
+#pragma GCC pop_options
+#endif
+__attribute__((target("avx512f")))
static int bar(void *a) {
__m512i x = *(__m512i *)a;
return _mm512_test_epi64_mask(x, x);
diff --git a/util/bufferiszero.c b/util/bufferiszero.c
index 695bb4ce28..ca836b6e8c 100644
--- a/util/bufferiszero.c
+++ b/util/bufferiszero.c
@@ -64,17 +64,18 @@ buffer_zero_int(const void *buf, size_t len)
}
#if defined(CONFIG_AVX512F_OPT) || defined(CONFIG_AVX2_OPT) || defined(__SSE2__)
-/* Do not use push_options pragmas unnecessarily, because clang
- * does not support them.
- */
-#if defined(CONFIG_AVX512F_OPT) || defined(CONFIG_AVX2_OPT)
+#ifndef __clang__
#pragma GCC push_options
#pragma GCC target("sse2")
#endif
#include <emmintrin.h>
+#ifndef __clang__
+#pragma GCC pop_options
+#endif
/* Note that each of these vectorized functions require len >= 64. */
+__attribute__((target("sse2")))
static bool
buffer_zero_sse2(const void *buf, size_t len)
{
@@ -104,19 +105,22 @@ buffer_zero_sse2(const void *buf, size_t len)
return _mm_movemask_epi8(_mm_cmpeq_epi8(t, zero)) == 0xFFFF;
}
-#if defined(CONFIG_AVX512F_OPT) || defined(CONFIG_AVX2_OPT)
-#pragma GCC pop_options
-#endif
#ifdef CONFIG_AVX2_OPT
/* Note that due to restrictions/bugs wrt __builtin functions in gcc <= 4.8,
* the includes have to be within the corresponding push_options region, and
* therefore the regions themselves have to be ordered with increasing ISA.
*/
+#ifndef __clang__
#pragma GCC push_options
#pragma GCC target("sse4")
+#endif
#include <smmintrin.h>
+#ifndef __clang__
+#pragma GCC pop_options
+#endif
+__attribute__((target("sse4")))
static bool
buffer_zero_sse4(const void *buf, size_t len)
{
@@ -145,11 +149,16 @@ buffer_zero_sse4(const void *buf, size_t len)
return _mm_testz_si128(t, t);
}
-#pragma GCC pop_options
+#ifndef __clang__
#pragma GCC push_options
#pragma GCC target("avx2")
+#endif
#include <immintrin.h>
+#ifndef __clang__
+#pragma GCC pop_options
+#endif
+__attribute__((target("avx2")))
static bool
buffer_zero_avx2(const void *buf, size_t len)
{
@@ -176,14 +185,19 @@ buffer_zero_avx2(const void *buf, size_t len)
return _mm256_testz_si256(t, t);
}
-#pragma GCC pop_options
#endif /* CONFIG_AVX2_OPT */
#ifdef CONFIG_AVX512F_OPT
+#ifndef __clang__
#pragma GCC push_options
#pragma GCC target("avx512f")
+#endif
#include <immintrin.h>
+#ifndef __clang__
+#pragma GCC pop_options
+#endif
+__attribute__((target("avx512f")))
static bool
buffer_zero_avx512(const void *buf, size_t len)
{
@@ -210,7 +224,6 @@ buffer_zero_avx512(const void *buf, size_t len)
return !_mm512_test_epi64_mask(t, t);
}
-#pragma GCC pop_options
#endif
--
2.28.0.rc0.105.gf9edc3c819-goog
^ permalink raw reply related [flat|nested] 8+ messages in thread
* [PATCH 2/2] configure: atomic64/128 detection for clang
2020-07-23 0:27 [PATCH 0/2] Instruction set detection for clang Shu-Chun Weng
2020-07-23 0:27 ` [PATCH 1/2] configure: avx2 and avx512f " Shu-Chun Weng
@ 2020-07-23 0:27 ` Shu-Chun Weng
2020-08-05 23:23 ` [PATCH 0/2] Instruction set " Shu-Chun Weng
2 siblings, 0 replies; 8+ messages in thread
From: Shu-Chun Weng @ 2020-07-23 0:27 UTC (permalink / raw)
To: rth; +Cc: Shu-Chun Weng, qemu-devel
The public interface for __atomic_* and __sync_* do not contain the
explicit *_{number} versions:
https://gcc.gnu.org/onlinedocs/gcc/_005f_005fatomic-Builtins.html
https://gcc.gnu.org/onlinedocs/gcc/_005f_005fsync-Builtins.html
They appear to be GCC's internal symbols which happen to work. However,
clang does not recognize them. Replace the existing usages with the `_n`
versions (or no suffix) which are the documented API.
Signed-off-by: Shu-Chun Weng <scw@google.com>
---
configure | 18 +++++++++---------
1 file changed, 9 insertions(+), 9 deletions(-)
diff --git a/configure b/configure
index d9ce3aa5db..0613a049e9 100755
--- a/configure
+++ b/configure
@@ -5894,9 +5894,9 @@ if test "$int128" = "yes"; then
int main(void)
{
unsigned __int128 x = 0, y = 0;
- y = __atomic_load_16(&x, 0);
- __atomic_store_16(&x, y, 0);
- __atomic_compare_exchange_16(&x, &y, x, 0, 0, 0);
+ y = __atomic_load_n(&x, 0);
+ __atomic_store_n(&x, y, 0);
+ __atomic_compare_exchange_n(&x, &y, x, 0, 0, 0);
return 0;
}
EOF
@@ -5911,7 +5911,7 @@ if test "$int128" = yes && test "$atomic128" = no; then
int main(void)
{
unsigned __int128 x = 0, y = 0;
- __sync_val_compare_and_swap_16(&x, y, x);
+ __sync_val_compare_and_swap(&x, y, x);
return 0;
}
EOF
@@ -5931,11 +5931,11 @@ int main(void)
{
uint64_t x = 0, y = 0;
#ifdef __ATOMIC_RELAXED
- y = __atomic_load_8(&x, 0);
- __atomic_store_8(&x, y, 0);
- __atomic_compare_exchange_8(&x, &y, x, 0, 0, 0);
- __atomic_exchange_8(&x, y, 0);
- __atomic_fetch_add_8(&x, y, 0);
+ y = __atomic_load_n(&x, 0);
+ __atomic_store_n(&x, y, 0);
+ __atomic_compare_exchange_n(&x, &y, x, 0, 0, 0);
+ __atomic_exchange_n(&x, y, 0);
+ __atomic_fetch_add(&x, y, 0);
#else
typedef char is_host64[sizeof(void *) >= sizeof(uint64_t) ? 1 : -1];
__sync_lock_test_and_set(&x, y);
--
2.28.0.rc0.105.gf9edc3c819-goog
^ permalink raw reply related [flat|nested] 8+ messages in thread
* Re: [PATCH 1/2] configure: avx2 and avx512f detection for clang
2020-07-23 0:27 ` [PATCH 1/2] configure: avx2 and avx512f " Shu-Chun Weng
@ 2020-07-23 4:54 ` Thomas Huth
2020-07-23 6:04 ` Shu-Chun Weng
0 siblings, 1 reply; 8+ messages in thread
From: Thomas Huth @ 2020-07-23 4:54 UTC (permalink / raw)
To: Shu-Chun Weng, rth; +Cc: qemu-devel
On 23/07/2020 02.27, Shu-Chun Weng wrote:
> Since clang does not support "#pragma GCC", the instruction sets are
> always disabled. In this change, we
>
> 1. wrap "#pragma GCC" inside "#ifndef __clang__",
> 2. only retain them around "#include <{e,i,s}mmintrin.h>" to work
> around gcc bug,
> 3. and annotate each function with `__attribute__((target(*)))` which
> is recognized by both gcc and clang.
>
> Signed-off-by: Shu-Chun Weng <scw@google.com>
> ---
> configure | 16 ++++++++++++++--
> util/bufferiszero.c | 33 +++++++++++++++++++++++----------
> 2 files changed, 37 insertions(+), 12 deletions(-)
>
> diff --git a/configure b/configure
> index 4bd80ed507..d9ce3aa5db 100755
> --- a/configure
> +++ b/configure
> @@ -5808,10 +5808,16 @@ fi
>
> if test "$cpuid_h" = "yes" && test "$avx2_opt" != "no"; then
> cat > $TMPC << EOF
> +#include <cpuid.h>
> +#ifndef __clang__
> #pragma GCC push_options
> #pragma GCC target("avx2")
> -#include <cpuid.h>
> +#endif
> #include <immintrin.h>
> +#ifndef __clang__
> +#pragma GCC pop_options
> +#endif
> +__attribute__((target("avx2")))
> static int bar(void *a) {
> __m256i x = *(__m256i *)a;
> return _mm256_testz_si256(x, x);
I wonder whether it would make more sense to pass "-mavx2" to the
compile_object call afterwards and simply remove the #pragmas here?
Did you try that already?
Thomas
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH 1/2] configure: avx2 and avx512f detection for clang
2020-07-23 4:54 ` Thomas Huth
@ 2020-07-23 6:04 ` Shu-Chun Weng
2020-07-23 6:24 ` Thomas Huth
0 siblings, 1 reply; 8+ messages in thread
From: Shu-Chun Weng @ 2020-07-23 6:04 UTC (permalink / raw)
To: Thomas Huth; +Cc: rth, qemu-devel
[-- Attachment #1.1: Type: text/plain, Size: 1801 bytes --]
Do we have the flexibility to do that for util/bufferiszero.c as well?
Otherwise, we are using different mechanisms to detect (compile test.c with
-mavx2) and actually use (GCC pragma & __attribute__((target(*))))) the
feature in production.
Shu-Chun
On Wed, Jul 22, 2020 at 9:55 PM Thomas Huth <thuth@redhat.com> wrote:
> On 23/07/2020 02.27, Shu-Chun Weng wrote:
> > Since clang does not support "#pragma GCC", the instruction sets are
> > always disabled. In this change, we
> >
> > 1. wrap "#pragma GCC" inside "#ifndef __clang__",
> > 2. only retain them around "#include <{e,i,s}mmintrin.h>" to work
> > around gcc bug,
> > 3. and annotate each function with `__attribute__((target(*)))` which
> > is recognized by both gcc and clang.
> >
> > Signed-off-by: Shu-Chun Weng <scw@google.com>
> > ---
> > configure | 16 ++++++++++++++--
> > util/bufferiszero.c | 33 +++++++++++++++++++++++----------
> > 2 files changed, 37 insertions(+), 12 deletions(-)
> >
> > diff --git a/configure b/configure
> > index 4bd80ed507..d9ce3aa5db 100755
> > --- a/configure
> > +++ b/configure
> > @@ -5808,10 +5808,16 @@ fi
> >
> > if test "$cpuid_h" = "yes" && test "$avx2_opt" != "no"; then
> > cat > $TMPC << EOF
> > +#include <cpuid.h>
> > +#ifndef __clang__
> > #pragma GCC push_options
> > #pragma GCC target("avx2")
> > -#include <cpuid.h>
> > +#endif
> > #include <immintrin.h>
> > +#ifndef __clang__
> > +#pragma GCC pop_options
> > +#endif
> > +__attribute__((target("avx2")))
> > static int bar(void *a) {
> > __m256i x = *(__m256i *)a;
> > return _mm256_testz_si256(x, x);
>
> I wonder whether it would make more sense to pass "-mavx2" to the
> compile_object call afterwards and simply remove the #pragmas here?
> Did you try that already?
>
> Thomas
>
>
[-- Attachment #1.2: Type: text/html, Size: 2645 bytes --]
[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 3844 bytes --]
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH 1/2] configure: avx2 and avx512f detection for clang
2020-07-23 6:04 ` Shu-Chun Weng
@ 2020-07-23 6:24 ` Thomas Huth
0 siblings, 0 replies; 8+ messages in thread
From: Thomas Huth @ 2020-07-23 6:24 UTC (permalink / raw)
To: Shu-Chun Weng; +Cc: qemu-devel, rth
On 23/07/2020 08.04, Shu-Chun Weng wrote:
> Do we have the flexibility to do that for util/bufferiszero.c as well?
> Otherwise, we are using different mechanisms to detect (compile test.c
> with -mavx2) and actually use (GCC pragma & __attribute__((target(*)))))
> the feature in production.
That's true ... so it's likely better to keep the pragmas in the
configure script, indeed!
Thanks,
Thomas
> Shu-Chun
>
> On Wed, Jul 22, 2020 at 9:55 PM Thomas Huth <thuth@redhat.com
> <mailto:thuth@redhat.com>> wrote:
>
> On 23/07/2020 02.27, Shu-Chun Weng wrote:
> > Since clang does not support "#pragma GCC", the instruction sets are
> > always disabled. In this change, we
> >
> > 1. wrap "#pragma GCC" inside "#ifndef __clang__",
> > 2. only retain them around "#include <{e,i,s}mmintrin.h>" to work
> > around gcc bug,
> > 3. and annotate each function with `__attribute__((target(*)))` which
> > is recognized by both gcc and clang.
> >
> > Signed-off-by: Shu-Chun Weng <scw@google.com <mailto:scw@google.com>>
> > ---
> > configure | 16 ++++++++++++++--
> > util/bufferiszero.c | 33 +++++++++++++++++++++++----------
> > 2 files changed, 37 insertions(+), 12 deletions(-)
> >
> > diff --git a/configure b/configure
> > index 4bd80ed507..d9ce3aa5db 100755
> > --- a/configure
> > +++ b/configure
> > @@ -5808,10 +5808,16 @@ fi
> >
> > if test "$cpuid_h" = "yes" && test "$avx2_opt" != "no"; then
> > cat > $TMPC << EOF
> > +#include <cpuid.h>
> > +#ifndef __clang__
> > #pragma GCC push_options
> > #pragma GCC target("avx2")
> > -#include <cpuid.h>
> > +#endif
> > #include <immintrin.h>
> > +#ifndef __clang__
> > +#pragma GCC pop_options
> > +#endif
> > +__attribute__((target("avx2")))
> > static int bar(void *a) {
> > __m256i x = *(__m256i *)a;
> > return _mm256_testz_si256(x, x);
>
> I wonder whether it would make more sense to pass "-mavx2" to the
> compile_object call afterwards and simply remove the #pragmas here?
> Did you try that already?
>
> Thomas
>
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH 0/2] Instruction set detection for clang.
2020-07-23 0:27 [PATCH 0/2] Instruction set detection for clang Shu-Chun Weng
2020-07-23 0:27 ` [PATCH 1/2] configure: avx2 and avx512f " Shu-Chun Weng
2020-07-23 0:27 ` [PATCH 2/2] configure: atomic64/128 " Shu-Chun Weng
@ 2020-08-05 23:23 ` Shu-Chun Weng
2020-09-17 7:25 ` Shu-Chun Weng
2 siblings, 1 reply; 8+ messages in thread
From: Shu-Chun Weng @ 2020-08-05 23:23 UTC (permalink / raw)
To: rth; +Cc: qemu-devel
[-- Attachment #1.1: Type: text/plain, Size: 752 bytes --]
Ping: https://patchew.org/QEMU/cover.1595463707.git.scw@google.com/
On Wed, Jul 22, 2020 at 5:27 PM Shu-Chun Weng <scw@google.com> wrote:
> Currently when configuring QEMU with clang, AVX2, AVX512F, ATOMIC64, and
> ATOMIC128 are all disabled because the detection code is GCC-only. With
> these
> two patches, I am able to configure, build, and run tests with clang with
> all of
> the above enabled.
>
> Shu-Chun Weng (2):
> configure: avx2 and avx512f detection for clang
> configure: atomic64/128 detection for clang
>
> configure | 34 +++++++++++++++++++++++-----------
> util/bufferiszero.c | 33 +++++++++++++++++++++++----------
> 2 files changed, 46 insertions(+), 21 deletions(-)
>
> --
> 2.28.0.rc0.105.gf9edc3c819-goog
>
>
[-- Attachment #1.2: Type: text/html, Size: 1176 bytes --]
[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 3844 bytes --]
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH 0/2] Instruction set detection for clang.
2020-08-05 23:23 ` [PATCH 0/2] Instruction set " Shu-Chun Weng
@ 2020-09-17 7:25 ` Shu-Chun Weng
0 siblings, 0 replies; 8+ messages in thread
From: Shu-Chun Weng @ 2020-09-17 7:25 UTC (permalink / raw)
To: rth; +Cc: qemu-devel
[-- Attachment #1.1: Type: text/plain, Size: 854 bytes --]
Ping.
On Wed, Aug 5, 2020 at 4:23 PM Shu-Chun Weng <scw@google.com> wrote:
> Ping: https://patchew.org/QEMU/cover.1595463707.git.scw@google.com/
>
> On Wed, Jul 22, 2020 at 5:27 PM Shu-Chun Weng <scw@google.com> wrote:
>
>> Currently when configuring QEMU with clang, AVX2, AVX512F, ATOMIC64, and
>> ATOMIC128 are all disabled because the detection code is GCC-only. With
>> these
>> two patches, I am able to configure, build, and run tests with clang with
>> all of
>> the above enabled.
>>
>> Shu-Chun Weng (2):
>> configure: avx2 and avx512f detection for clang
>> configure: atomic64/128 detection for clang
>>
>> configure | 34 +++++++++++++++++++++++-----------
>> util/bufferiszero.c | 33 +++++++++++++++++++++++----------
>> 2 files changed, 46 insertions(+), 21 deletions(-)
>>
>> --
>> 2.28.0.rc0.105.gf9edc3c819-goog
>>
>>
[-- Attachment #1.2: Type: text/html, Size: 1544 bytes --]
[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 3990 bytes --]
^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2020-09-17 7:28 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-07-23 0:27 [PATCH 0/2] Instruction set detection for clang Shu-Chun Weng
2020-07-23 0:27 ` [PATCH 1/2] configure: avx2 and avx512f " Shu-Chun Weng
2020-07-23 4:54 ` Thomas Huth
2020-07-23 6:04 ` Shu-Chun Weng
2020-07-23 6:24 ` Thomas Huth
2020-07-23 0:27 ` [PATCH 2/2] configure: atomic64/128 " Shu-Chun Weng
2020-08-05 23:23 ` [PATCH 0/2] Instruction set " Shu-Chun Weng
2020-09-17 7:25 ` Shu-Chun Weng
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).