[PATCH][AArch64 Testsuite] Add test of vld[234]q? intrinsic

[PATCH][AArch64 Testsuite] Add test of vld[234]q? intrinsic

[Alan Lawrence]

[-- Attachment #1: Type: text/plain, Size: 411 bytes --]

This adds a test of all the variants of vld2, vld2q, vld3, vld3q, vld4, and 
vld4q. These all use typexNxM structs and the OI/CI/XImode mechanism, so the 
test cross-checks this against plain ol' vst1(q?).

Cross-tested on aarch64-none-elf (passing), also on aarch64_be-none-elf 
(https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59810).

gcc/testsuite/ChangeLog:

	* gcc.target/aarch64/vldN_1.c: New test.

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: test_vldN.patch --]
[-- Type: text/x-patch; name=test_vldN.patch, Size: 2520 bytes --]

diff --git a/gcc/testsuite/gcc.target/aarch64/vldN_1.c b/gcc/testsuite/gcc.target/aarch64/vldN_1.c
new file mode 100644
index 0000000000000000000000000000000000000000..b64de16a1658da166175288bc1378a57d220801f
--- /dev/null
+++ b/gcc/testsuite/gcc.target/aarch64/vldN_1.c
@@ -0,0 +1,79 @@
+/* { dg-do run } */
+/* { dg-options "-O3" } */
+
+#include <arm_neon.h>
+
+extern void abort (void);
+
+#define TESTMETH(BASE, ELTS, STRUCT, SUFFIX)	\
+int __attribute__ ((noinline))			\
+test_vld##STRUCT##SUFFIX ()			\
+{						\
+  BASE##_t data[ELTS * STRUCT];			\
+  BASE##_t temp[ELTS];				\
+  BASE##x##ELTS##x##STRUCT##_t vectors;		\
+  int i,j;					\
+  for (i = 0; i < STRUCT * ELTS; i++)		\
+    data [i] = (BASE##_t) 2*i + 1;		\
+  asm volatile ("" : : : "memory");		\
+  vectors = vld##STRUCT##SUFFIX (data);		\
+  for (i = 0; i < STRUCT; i++)			\
+    {						\
+      vst1##SUFFIX (temp, vectors.val[i]);	\
+      asm volatile ("" : : : "memory");		\
+      for (j = 0; j < ELTS; j++)		\
+        if (temp[j] != data[i + STRUCT*j])	\
+          return 1;				\
+    }						\
+  return 0;					\
+}
+
+#define VARIANTS(VARIANT, STRUCT)	\
+VARIANT (uint8, 8, STRUCT, _u8)		\
+VARIANT (uint16, 4, STRUCT, _u16)	\
+VARIANT (uint32, 2, STRUCT, _u32)	\
+VARIANT (uint64, 1, STRUCT, _u64)	\
+VARIANT (int8, 8, STRUCT, _s8)		\
+VARIANT (int16, 4, STRUCT, _s16)	\
+VARIANT (int32, 2, STRUCT, _s32)	\
+VARIANT (int64, 1, STRUCT, _s64)	\
+VARIANT (poly8, 8, STRUCT, _p8)		\
+VARIANT (poly16, 4, STRUCT, _p16)	\
+VARIANT (float32, 2, STRUCT, _f32)	\
+VARIANT (float64, 1, STRUCT, _f64)	\
+VARIANT (uint8, 16, STRUCT, q_u8)	\
+VARIANT (uint16, 8, STRUCT, q_u16)	\
+VARIANT (uint32, 4, STRUCT, q_u32)	\
+VARIANT (uint64, 2, STRUCT, q_u64)	\
+VARIANT (int8, 16, STRUCT, q_s8)	\
+VARIANT (int16, 8, STRUCT, q_s16)	\
+VARIANT (int32, 4, STRUCT, q_s32)	\
+VARIANT (int64, 2, STRUCT, q_s64)	\
+VARIANT (poly8, 16, STRUCT, q_p8)	\
+VARIANT (poly16, 8, STRUCT, q_p16)	\
+VARIANT (float32, 4, STRUCT, q_f32)	\
+VARIANT (float64, 2, STRUCT, q_f64)
+
+/* Tests of vld2 and vld2q.  */
+VARIANTS (TESTMETH, 2)
+
+/* Tests of vld3 and vld3q.  */
+VARIANTS (TESTMETH, 3)
+
+/* Tests of vld4 and vld4q.  */
+VARIANTS (TESTMETH, 4)
+
+#define CHECK(BASE, ELTS, STRUCT, SUFFIX)	\
+  if (test_vld##STRUCT##SUFFIX () != 0)		\
+    abort ();
+
+int
+main (int argc, char **argv)
+{
+  VARIANTS (CHECK, 2)
+  VARIANTS (CHECK, 3)
+  VARIANTS (CHECK, 4)
+
+  return 0;
+}
+

Re: [PATCH][AArch64 Testsuite] Add test of vld[234]q? intrinsic

[Marcus Shawcroft]

On 8 September 2014 11:35, Alan Lawrence <alan.lawrence@arm.com> wrote:
> This adds a test of all the variants of vld2, vld2q, vld3, vld3q, vld4, and
> vld4q. These all use typexNxM structs and the OI/CI/XImode mechanism, so the
> test cross-checks this against plain ol' vst1(q?).
>
> Cross-tested on aarch64-none-elf (passing), also on aarch64_be-none-elf
> (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59810).
>
> gcc/testsuite/ChangeLog:
>
>         * gcc.target/aarch64/vldN_1.c: New test.

OK /Marcus

Re: [PATCH][AArch64 Testsuite] Add test of vld[234]q? intrinsic

[Christophe Lyon]

On 9 September 2014 12:19, Marcus Shawcroft <marcus.shawcroft@gmail.com> wrote:
> On 8 September 2014 11:35, Alan Lawrence <alan.lawrence@arm.com> wrote:
>> This adds a test of all the variants of vld2, vld2q, vld3, vld3q, vld4, and
>> vld4q. These all use typexNxM structs and the OI/CI/XImode mechanism, so the
>> test cross-checks this against plain ol' vst1(q?).
>>
>> Cross-tested on aarch64-none-elf (passing), also on aarch64_be-none-elf
>> (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59810).
>>

Hi Alan,

On my side, your new test fails at execution on aarch64_be-none-elf:
http://cbuild.validation.linaro.org/build/cross-validation/gcc/trunk/215072/report-build-info.html

This seems strange since you tested it too.
I am using the Foundation Model.

Christophe.

>> gcc/testsuite/ChangeLog:
>>
>>         * gcc.target/aarch64/vldN_1.c: New test.
>
> OK /Marcus

Re: [PATCH][AArch64 Testsuite] Add test of vld[234]q? intrinsic

[Alan Lawrence]

Yes, I had seen that, and the failure is expected. AFAICT the test is correct; 
but the implementation of vld[234] is incorrect on bigendian, because of 
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59810 .

HTH, Alan.



Christophe Lyon wrote:
> On 9 September 2014 12:19, Marcus Shawcroft <marcus.shawcroft@gmail.com> wrote:
>> On 8 September 2014 11:35, Alan Lawrence <alan.lawrence@arm.com> wrote:
>>> This adds a test of all the variants of vld2, vld2q, vld3, vld3q, vld4, and
>>> vld4q. These all use typexNxM structs and the OI/CI/XImode mechanism, so the
>>> test cross-checks this against plain ol' vst1(q?).
>>>
>>> Cross-tested on aarch64-none-elf (passing), also on aarch64_be-none-elf
>>> (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59810).
>>>
> 
> Hi Alan,
> 
> On my side, your new test fails at execution on aarch64_be-none-elf:
> http://cbuild.validation.linaro.org/build/cross-validation/gcc/trunk/215072/report-build-info.html
> 
> This seems strange since you tested it too.
> I am using the Foundation Model.
> 
> Christophe.
> 
>>> gcc/testsuite/ChangeLog:
>>>
>>>         * gcc.target/aarch64/vldN_1.c: New test.
>> OK /Marcus
> 

Re: [PATCH][AArch64 Testsuite] Add test of vld[234]q? intrinsic

[Christophe Lyon]

Ha OK, I had misunderstood your first email, and thought you had the
test also pass in big endian.

Thanks for the clarification.


On 11 September 2014 15:56, Alan Lawrence <alan.lawrence@arm.com> wrote:
> Yes, I had seen that, and the failure is expected. AFAICT the test is
> correct; but the implementation of vld[234] is incorrect on bigendian,
> because of https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59810 .
>
> HTH, Alan.
>
>
>
>
> Christophe Lyon wrote:
>>
>> On 9 September 2014 12:19, Marcus Shawcroft <marcus.shawcroft@gmail.com>
>> wrote:
>>>
>>> On 8 September 2014 11:35, Alan Lawrence <alan.lawrence@arm.com> wrote:
>>>>
>>>> This adds a test of all the variants of vld2, vld2q, vld3, vld3q, vld4,
>>>> and
>>>> vld4q. These all use typexNxM structs and the OI/CI/XImode mechanism, so
>>>> the
>>>> test cross-checks this against plain ol' vst1(q?).
>>>>
>>>> Cross-tested on aarch64-none-elf (passing), also on aarch64_be-none-elf
>>>> (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=59810).
>>>>
>>
>> Hi Alan,
>>
>> On my side, your new test fails at execution on aarch64_be-none-elf:
>>
>> http://cbuild.validation.linaro.org/build/cross-validation/gcc/trunk/215072/report-build-info.html
>>
>> This seems strange since you tested it too.
>> I am using the Foundation Model.
>>
>> Christophe.
>>
>>>> gcc/testsuite/ChangeLog:
>>>>
>>>>         * gcc.target/aarch64/vldN_1.c: New test.
>>>
>>> OK /Marcus
>>
>>
>
>