[Patch, AArch64, AArch64-4.7] Map scalar types in arm_neon.h to standard types.

[Patch, AArch64, AArch64-4.7] Map scalar types in arm_neon.h to standard types.

[Tejas Belagod]

[-- Attachment #1: Type: text/plain, Size: 912 bytes --]


Hi,

The attached patch fixes one of arm_neon.h's issues with C++. All scalar types 
<type>x1_t are mapped to internal __builtin_aarch64_simd_<type> types. This 
creates duplicate function name symbols when one attempts overloading a standard 
type and one of these builtin types that are of the same size. For example, 
int8x1_t when overloaded with signed char creates a duplicate function name 
symbol. This patch increases the strictness and maps scalar types <type>x1_t to 
standard types. This means overloading is not possible, but that will not make a 
difference as we can now define it this way and needless to say it was broken 
till now anyway.

Tested for aarch64-none-elf. OK for trunk and aarch64-4.7-branch?

Thanks,
Tejas Belagod
ARM.

Changelog:

2013-01-18  Tejas Belagod  <tejas.belagod@arm.com>

gcc/
	* config/aarch64/arm_neon.h: Map scalar types to standard types.

[-- Attachment #2: builtin-type.txt --]
[-- Type: text/plain, Size: 1550 bytes --]

diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
index 5a72029..50e0a7b 100644
--- a/gcc/config/aarch64/arm_neon.h
+++ b/gcc/config/aarch64/arm_neon.h
@@ -35,11 +35,11 @@ typedef __builtin_aarch64_simd_hi int16x4_t
   __attribute__ ((__vector_size__ (8)));
 typedef __builtin_aarch64_simd_si int32x2_t
   __attribute__ ((__vector_size__ (8)));
-typedef __builtin_aarch64_simd_di int64x1_t;
-typedef __builtin_aarch64_simd_si int32x1_t;
-typedef __builtin_aarch64_simd_hi int16x1_t;
-typedef __builtin_aarch64_simd_qi int8x1_t;
-typedef __builtin_aarch64_simd_df float64x1_t;
+typedef int64_t int64x1_t;
+typedef int32_t int32x1_t;
+typedef int16_t int16x1_t;
+typedef int8_t int8x1_t;
+typedef double float64x1_t;
 typedef __builtin_aarch64_simd_sf float32x2_t
   __attribute__ ((__vector_size__ (8)));
 typedef __builtin_aarch64_simd_poly8 poly8x8_t
@@ -52,10 +52,10 @@ typedef __builtin_aarch64_simd_uhi uint16x4_t
   __attribute__ ((__vector_size__ (8)));
 typedef __builtin_aarch64_simd_usi uint32x2_t
   __attribute__ ((__vector_size__ (8)));
-typedef __builtin_aarch64_simd_udi uint64x1_t;
-typedef __builtin_aarch64_simd_usi uint32x1_t;
-typedef __builtin_aarch64_simd_uhi uint16x1_t;
-typedef __builtin_aarch64_simd_uqi uint8x1_t;
+typedef uint64_t uint64x1_t;
+typedef uint32_t uint32x1_t;
+typedef uint16_t uint16x1_t;
+typedef uint8_t uint8x1_t;
 typedef __builtin_aarch64_simd_qi int8x16_t
   __attribute__ ((__vector_size__ (16)));
 typedef __builtin_aarch64_simd_hi int16x8_t

Re: [Patch, AArch64, AArch64-4.7] Map scalar types in arm_neon.h to standard types.

[Marcus Shawcroft]

On 18/01/13 10:25, Tejas Belagod wrote:
>
> Hi,
>
> The attached patch fixes one of arm_neon.h's issues with C++. All scalar types
> <type>x1_t are mapped to internal __builtin_aarch64_simd_<type> types. This
> creates duplicate function name symbols when one attempts overloading a standard
> type and one of these builtin types that are of the same size. For example,
> int8x1_t when overloaded with signed char creates a duplicate function name
> symbol. This patch increases the strictness and maps scalar types <type>x1_t to
> standard types. This means overloading is not possible, but that will not make a
> difference as we can now define it this way and needless to say it was broken
> till now anyway.
>
> Tested for aarch64-none-elf. OK for trunk and aarch64-4.7-branch?
>
> Thanks,
> Tejas Belagod
> ARM.
>
> Changelog:
>
> 2013-01-18  Tejas Belagod  <tejas.belagod@arm.com>
>
> gcc/
> 	* config/aarch64/arm_neon.h: Map scalar types to standard types.
>

OK