Re: [PATCH][arm][committed] Fix use of CRC32 intrinsics with Armv8-a and hard-float

[Kyrill Tkachov]

On 8/22/19 4:53 PM, Kyrill Tkachov wrote:

Hi all,

We currently have a nasty error when trying to use the __crc* 
intrinsics with an -mfloat-abi=hard.
That is because the target pragma guarding them uses armv8-a+crc that 
does not include fp by default.

So we get errors like:
error: '-mfloat-abi=hard': selected processor lacks an FPU

This patch fixes that by using an FP-enabled arch target pragma to 
guard these intrinsics when floating-point is available.

That way both the softfloat and hardfloat variants work.

Bootstrapped and tested on arm-none-linux-gnueabihf.

Committing to trunk. Will backport to branches later.

Now backported to GCC 9 and 8 branches after bootstrapping and testing 
there.


Will do GCC 7 once I get more testing cycles.

Thanks,

Kyrill



Thanks,
Kyrill

2019-08-22  Kyrylo Tkachov 

    * config/arm/arm_acle.h: Use arch=armv8-a+crc+simd pragma for CRC32
    intrinsics if __ARM_FP.
    Use __ARM_FEATURE_CRC32 ifdef guard.

2019-08-22  Kyrylo Tkachov 

    * gcc.target/arm/acle/crc_hf_1.c: New test.

[PATCH][arm][committed] Fix use of CRC32 intrinsics with Armv8-a and hard-float

[Kyrill Tkachov]

Hi all,

We currently have a nasty error when trying to use the __crc* intrinsics 
with an -mfloat-abi=hard.
That is because the target pragma guarding them uses armv8-a+crc that 
does not include fp by default.

So we get errors like:
error: '-mfloat-abi=hard': selected processor lacks an FPU

This patch fixes that by using an FP-enabled arch target pragma to guard 
these intrinsics when floating-point is available.

That way both the softfloat and hardfloat variants work.

Bootstrapped and tested on arm-none-linux-gnueabihf.

Committing to trunk. Will backport to branches later.

Thanks,
Kyrill

2019-08-22  Kyrylo Tkachov 

    * config/arm/arm_acle.h: Use arch=armv8-a+crc+simd pragma for CRC32
    intrinsics if __ARM_FP.
    Use __ARM_FEATURE_CRC32 ifdef guard.

2019-08-22  Kyrylo Tkachov 

    * gcc.target/arm/acle/crc_hf_1.c: New test.

diff --git a/gcc/config/arm/arm_acle.h b/gcc/config/arm/arm_acle.h
index 2c7acc698ea50e0cd539a7d9647498746cd1536f..6857ab1787df0ffa672e5078e5a0b9c9cc52e695 100644
--- a/gcc/config/arm/arm_acle.h
+++ b/gcc/config/arm/arm_acle.h
@@ -174,8 +174,12 @@ __arm_mrrc2 (const unsigned int __coproc, const unsigned int __opc1,
 #endif /* (!__thumb__ || __thumb2__) &&  __ARM_ARCH >= 4.  */
 
 #pragma GCC push_options
-#if __ARM_ARCH >= 8
+#ifdef __ARM_FEATURE_CRC32
+#ifdef __ARM_FP
+#pragma GCC target ("arch=armv8-a+crc+simd")
+#else
 #pragma GCC target ("arch=armv8-a+crc")
+#endif
 
 __extension__ static __inline uint32_t __attribute__ ((__always_inline__))
 __crc32b (uint32_t __a, uint8_t __b)
@@ -235,7 +239,7 @@ __crc32cd (uint32_t __a, uint64_t __b)
 }
 #endif
 
-#endif /* __ARM_ARCH >= 8.  */
+#endif /* __ARM_FEATURE_CRC32  */
 #pragma GCC pop_options
 
 #ifdef __cplusplus
diff --git a/gcc/testsuite/gcc.target/arm/acle/crc_hf_1.c b/gcc/testsuite/gcc.target/arm/acle/crc_hf_1.c
new file mode 100644
index ..e6cbfc0b33e56e4275b96978ca1823d7682792fb
--- /dev/null
+++ b/gcc/testsuite/gcc.target/arm/acle/crc_hf_1.c
@@ -0,0 +1,14 @@
+/* Test that using an Armv8-a hard-float target doesn't
+   break CRC intrinsics.  */
+
+/* { dg-do compile } */
+/* { dg-require-effective-target arm_hard_vfp_ok }  */
+/* { dg-options "-mfloat-abi=hard -march=armv8-a+simd+crc" } */
+
+#include 
+
+uint32_t
+foo (uint32_t a, uint32_t b)
+{
+  return __crc32cw (a, b);
+}