[PATCH] testsuites/unit: Add tests for compiler builtins

Joel Sherrill joel at rtems.org
Thu Nov 2 14:57:14 UTC 2023


I noted a few cases where I don't think the value used is interesting
enough or an edge case. Hopefully I caught them all.

On Thu, Nov 2, 2023 at 7:12 AM Sebastian Huber <
sebastian.huber at embedded-brains.de> wrote:

> Test more compiler builtins which may use integer library routins:
>

routines.

>
> https://gcc.gnu.org/onlinedocs/gccint/Integer-library-routines.html
>
> Update #3716.
> ---
>  testsuites/unit/tc-compiler-builtins.c | 377 ++++++++++++++++++++++++-
>  1 file changed, 369 insertions(+), 8 deletions(-)
>
> diff --git a/testsuites/unit/tc-compiler-builtins.c
> b/testsuites/unit/tc-compiler-builtins.c
> index 13a9c4a248..4bf8570eb0 100644
> --- a/testsuites/unit/tc-compiler-builtins.c
> +++ b/testsuites/unit/tc-compiler-builtins.c
> @@ -78,6 +78,39 @@
>   *
>   * - Check the return value of __builtin_ctz() for a sample set of inputs.
>   *
> + * - Check the return value of __builtin_ctzll() for a sample set of
> inputs.
> + *
> + * - Check the return value of __builtin_ffs() for a sample set of inputs.
> + *
> + * - Check the return value of __builtin_ffsll() for a sample set of
> inputs.
> + *
> + * - Check the return value of __builtin_parity() for a sample set of
> inputs.
> + *
> + * - Check the return value of __builtin_parityll() for a sample set of
> inputs.
> + *
> + * - Check the return value of __builtin_popcount() for a sample set of
> inputs.
> + *
> + * - Check the return value of __builtin_popcountll() for a sample set of
> + *   inputs.
> + *
> + * - Check the return value of __builtin_bswap32() for a sample set of
> inputs.
> + *
> + * - Check the return value of __builtin_bswap64() for a sample set of
> inputs.
> + *
> + * - Check signed 64-bit comparisons for a sample set of values.
> + *
> + * - Check unsigned 64-bit comparisons for a sample set of values.
> + *
> + * - Check signed 64-bit arithmetic left shift for a sample set of values.
> + *
> + * - Check signed 64-bit arithmetic right shift for a sample set of
> values.
> + *
> + * - Check unsigned 64-bit logical right shift for a sample set of values.
> + *
> + * - Check signed 64-bit multiplication for a sample set of values.
> + *
> + * - Check signed 64-bit negation for a sample set of values.
> + *
>   * - Check signed 64-bit divisions for a sample set of values.
>   *
>   * - Check unsigned 64-bit divisions for a sample set of values.
> @@ -180,22 +213,334 @@ static void CompilerUnitBuiltins_Action_1( void )
>   */
>  static void CompilerUnitBuiltins_Action_2( void )
>  {
> -  volatile int n;
> +  volatile unsigned int n;
>
> -  n = 1;
> +  n = 1U;
>    T_eq_int( __builtin_ctz( n ), 0 );
>
> -  n = 1 << 31;
> +  n = 1U << 31;
>    T_eq_int( __builtin_ctz( n ), 31 );
>
> -  n = ~0;
> +  n = ~0U;
>    T_eq_int( __builtin_ctz( n ), 0 );
>  }
>
>  /**
> - * @brief Check signed 64-bit divisions for a sample set of values.
> + * @brief Check the return value of __builtin_ctzll() for a sample set of
> + *   inputs.
>   */
>  static void CompilerUnitBuiltins_Action_3( void )
> +{
> +  volatile unsigned long long n;
> +
> +  n = 1ULL;
> +  T_eq_int( __builtin_ctzll( n ), 0 );
> +
> +  n = 1ULL << 31;
> +  T_eq_int( __builtin_ctzll( n ), 31 );
> +
> +  n = 1ULL << 32;
> +  T_eq_int( __builtin_ctzll( n ), 32 );
> +
> +  n = 1ULL << 63;
> +  T_eq_int( __builtin_ctzll( n ), 63 );
> +
> +  n = ~0ULL;
> +  T_eq_int( __builtin_ctzll( n ), 0 );
> +}
> +
> +/**
> + * @brief Check the return value of __builtin_ffs() for a sample set of
> inputs.
> + */
> +static void CompilerUnitBuiltins_Action_4( void )
> +{
> +  volatile unsigned int n;
> +
> +  n = 1U;
> +  T_eq_int( __builtin_ffs( n ), 1 );
> +
> +  n = 1U << 31;
> +  T_eq_int( __builtin_ffs( n ), 32 );
> +
> +  n = 0U;
> +  T_eq_int( __builtin_ffs( n ), 0 );
> +}
> +
> +/**
> + * @brief Check the return value of __builtin_ffsll() for a sample set of
> + *   inputs.
> + */
> +static void CompilerUnitBuiltins_Action_5( void )
> +{
> +  volatile unsigned long long n;
> +
> +  n = 1ULL;
> +  T_eq_int( __builtin_ffsll( n ), 1 );
> +
> +  n = 1ULL << 31;
> +  T_eq_int( __builtin_ffsll( n ), 32 );
> +
> +  n = 1ULL << 32;
> +  T_eq_int( __builtin_ffsll( n ), 33 );
> +
> +  n = 1ULL << 63;
> +  T_eq_int( __builtin_ffsll( n ), 64 );
> +
> +  n = 0ULL;
> +  T_eq_int( __builtin_ffsll( n ), 0 );
> +}
> +
> +/**
> + * @brief Check the return value of __builtin_parity() for a sample set of
> + *   inputs.
> + */
> +static void CompilerUnitBuiltins_Action_6( void )
> +{
> +  volatile unsigned int n;
> +
> +  n = 1U;
> +  T_eq_int( __builtin_parity( n ), 1 );
> +
> +  n = ~0U;
> +  T_eq_int( __builtin_parity( n ), 0 );
> +}
> +
> +/**
> + * @brief Check the return value of __builtin_parityll() for a sample set
> of
> + *   inputs.
> + */
> +static void CompilerUnitBuiltins_Action_7( void )
> +{
> +  volatile unsigned long long n;
> +
> +  n = 1ULL;
> +  T_eq_int( __builtin_parityll( n ), 1 );
> +
> +  n = ~0ULL;
> +  T_eq_int( __builtin_parityll( n ), 0 );
> +}
> +
> +/**
> + * @brief Check the return value of __builtin_popcount() for a sample set
> of
> + *   inputs.
> + */
> +static void CompilerUnitBuiltins_Action_8( void )
> +{
> +  volatile unsigned int n;
> +
> +  n = 0U;
> +  T_eq_int( __builtin_popcount( n ), 0 );
> +
> +  n = 1U;
> +  T_eq_int( __builtin_popcount( n ), 1 );
> +
> +  n = ~0U;
> +  T_eq_int( __builtin_popcount( n ), 32 );
> +}
> +
> +/**
> + * @brief Check the return value of __builtin_popcountll() for a sample
> set of
> + *   inputs.
> + */
> +static void CompilerUnitBuiltins_Action_9( void )
> +{
> +  volatile unsigned long long n;
> +
> +  n = 0ULL;
> +  T_eq_int( __builtin_popcountll( n ), 0 );
> +
> +  n = 1ULL;
> +  T_eq_int( __builtin_popcountll( n ), 1 );
> +
> +  n = ~0ULL;
> +  T_eq_int( __builtin_popcountll( n ), 64 );
> +}
> +
> +/**
> + * @brief Check the return value of __builtin_bswap32() for a sample set
> of
> + *   inputs.
> + */
> +static void CompilerUnitBuiltins_Action_10( void )
> +{
> +  volatile int32_t n;
> +
> +  n = INT32_C( 0 );
> +  T_eq_i32( __builtin_bswap32( n ), n );
> +
> +  n = INT32_C( 1 );
> +  T_eq_i32( __builtin_bswap32( n ), n << 24 );
> +
> +  n = ~INT32_C( 0 );
> +  T_eq_i32( __builtin_bswap32( n ), n );
>

More interesting value needed to see if the other bytes changed.


> +}
> +
> +/**
> + * @brief Check the return value of __builtin_bswap64() for a sample set
> of
> + *   inputs.
> + */
> +static void CompilerUnitBuiltins_Action_11( void )
> +{
> +  volatile int64_t n;
> +
> +  n = INT64_C( 0 );
> +  T_eq_i64( __builtin_bswap64( n ), n );
> +
> +  n = INT64_C( 1 );
> +  T_eq_i64( __builtin_bswap64( n ), n << 56 );
> +
> +  n = ~INT64_C( 0 );
> +  T_eq_i64( __builtin_bswap64( n ), n );
>

You probably should check some value that uses more than one of the 8
bytes in the data type.


> +}
> +
> +/**
> + * @brief Check signed 64-bit comparisons for a sample set of values.
> + */
> +static void CompilerUnitBuiltins_Action_12( void )
> +{
> +  volatile int64_t a;
> +  volatile int64_t b;
> +
> +  a = INT64_C( 0 );
> +  b = INT64_C( 0 );
> +  T_false( a < b );
> +
> +  a = INT64_C( 0 );
> +  b = INT64_C( 1 );
> +  T_true( a < b );
>

Same.

I tried to catch cases below here when the values tested might not
be interesting enough or an edge case.


> +}
> +
> +/**
> + * @brief Check unsigned 64-bit comparisons for a sample set of values.
> + */
> +static void CompilerUnitBuiltins_Action_13( void )
> +{
> +  volatile uint64_t a;
> +  volatile uint64_t b;
> +
> +  a = UINT64_C( 0 );
> +  b = UINT64_C( 0 );
> +  T_false( a < b );
> +
> +  a = UINT64_C( 0 );
> +  b = UINT64_C( 1 );
> +  T_true( a < b );
> +}
> +
> +/**
> + * @brief Check signed 64-bit arithmetic left shift for a sample set of
> values.
> + */
> +static void CompilerUnitBuiltins_Action_14( void )
> +{
> +  volatile int64_t i;
> +  volatile int s;
> +
> +  i = INT64_C( 1 );
> +  s = 0;
> +  T_eq_i64( i << s, INT64_C( 1 ) );
> +
> +  i = -INT64_C( 1 );
> +  s = 0;
> +  T_eq_i64( i << s, -INT64_C( 1 ) );
> +
> +  i = INT64_C( 1 );
> +  s = 1;
> +  T_eq_i64( i << s, INT64_C( 2 ) );
> +
> +  i = -INT64_C( 1 );
> +  s = 1;
> +  T_eq_i64( i << s, -INT64_C( 2 ) );
> +}
> +
> +/**
> + * @brief Check signed 64-bit arithmetic right shift for a sample set of
> + *   values.
> + */
> +static void CompilerUnitBuiltins_Action_15( void )
> +{
> +  volatile int64_t i;
> +  volatile int s;
> +
> +  i = INT64_C( 1 );
> +  s = 0;
> +  T_eq_i64( i >> s, INT64_C( 1 ) );
> +
> +  i = -INT64_C( 1 );
> +  s = 0;
> +  T_eq_i64( i >> s, -INT64_C( 1 ) );
> +
> +  i = INT64_C( 2 );
> +  s = 1;
> +  T_eq_i64( i >> s, INT64_C( 1 ) );
> +
> +  i = -INT64_C( 2 );
> +  s = 1;
> +  T_eq_i64( i >> s, -INT64_C( 1 ) );
> +}
> +
> +/**
> + * @brief Check unsigned 64-bit logical right shift for a sample set of
> values.
> + */
> +static void CompilerUnitBuiltins_Action_16( void )
> +{
> +  volatile uint64_t i;
> +  volatile int s;
> +
> +  i = UINT64_C( 1 );
> +  s = 0;
> +  T_eq_u64( i >> s, UINT64_C( 1 ) );
> +
> +  i = -UINT64_C( 1 );
> +  s = 0;
> +  T_eq_u64( i >> s, UINT64_C( 0xffffffffffffffff ) );
> +
> +  i = UINT64_C( 2 );
> +  s = 1;
> +  T_eq_u64( i >> s, UINT64_C( 1 ) );
> +
> +  i = -UINT64_C( 2 );
> +  s = 1;
> +  T_eq_u64( i >> s, UINT64_C( 0x7fffffffffffffff ) );
> +}
> +
> +/**
> + * @brief Check signed 64-bit multiplication for a sample set of values.
> + */
> +static void CompilerUnitBuiltins_Action_17( void )
> +{
> +  volatile int64_t a;
> +  volatile int64_t b;
> +
> +  a = INT64_C( 1 );
> +  b = INT64_C( 1 );
> +  T_eq_i64( a * b, INT64_C( 1 ) );
> +
> +  a = INT64_C( 1 );
> +  b = INT64_C( 0 );
> +  T_eq_i64( a * b, INT64_C( 0 ) );
> +
> +  a = INT64_C( 0 );
> +  b = INT64_C( 1 );
> +  T_eq_i64( a * b, INT64_C( 0 ) );
> +}
> +
> +/**
> + * @brief Check signed 64-bit negation for a sample set of values.
> + */
> +static void CompilerUnitBuiltins_Action_18( void )
> +{
> +  volatile int64_t i;
> +
> +  i = INT64_C( 1 );
> +  T_eq_i64( -i, -INT64_C( 1 ) );
> +
> +  i = -INT64_C( 1 );
> +  T_eq_i64( -i, INT64_C( 1 ) );
> +}
> +
> +/**
> + * @brief Check signed 64-bit divisions for a sample set of values.
> + */
> +static void CompilerUnitBuiltins_Action_19( void )
>  {
>    volatile int64_t n;
>    volatile int64_t d;
> @@ -304,7 +649,7 @@ static void CompilerUnitBuiltins_Action_3( void )
>  /**
>   * @brief Check unsigned 64-bit divisions for a sample set of values.
>   */
> -static void CompilerUnitBuiltins_Action_4( void )
> +static void CompilerUnitBuiltins_Action_20( void )
>  {
>    volatile uint64_t n;
>    volatile uint64_t d;
> @@ -381,7 +726,7 @@ static void CompilerUnitBuiltins_Action_4( void )
>  /**
>   * @brief Check signed 64-bit modulo operations for a sample set of
> values.
>   */
> -static void CompilerUnitBuiltins_Action_5( void )
> +static void CompilerUnitBuiltins_Action_21( void )
>  {
>    volatile int64_t n;
>    volatile int64_t d;
> @@ -502,7 +847,7 @@ static void CompilerUnitBuiltins_Action_5( void )
>  /**
>   * @brief Check unsigned 64-bit modulo operations for a sample set of
> values.
>   */
> -static void CompilerUnitBuiltins_Action_6( void )
> +static void CompilerUnitBuiltins_Action_22( void )
>  {
>    volatile uint64_t n;
>    volatile uint64_t d;
> @@ -560,6 +905,22 @@ T_TEST_CASE_FIXTURE( CompilerUnitBuiltins,
> &CompilerUnitBuiltins_Fixture )
>    CompilerUnitBuiltins_Action_4();
>    CompilerUnitBuiltins_Action_5();
>    CompilerUnitBuiltins_Action_6();
> +  CompilerUnitBuiltins_Action_7();
> +  CompilerUnitBuiltins_Action_8();
> +  CompilerUnitBuiltins_Action_9();
> +  CompilerUnitBuiltins_Action_10();
> +  CompilerUnitBuiltins_Action_11();
> +  CompilerUnitBuiltins_Action_12();
> +  CompilerUnitBuiltins_Action_13();
> +  CompilerUnitBuiltins_Action_14();
> +  CompilerUnitBuiltins_Action_15();
> +  CompilerUnitBuiltins_Action_16();
> +  CompilerUnitBuiltins_Action_17();
> +  CompilerUnitBuiltins_Action_18();
> +  CompilerUnitBuiltins_Action_19();
> +  CompilerUnitBuiltins_Action_20();
> +  CompilerUnitBuiltins_Action_21();
> +  CompilerUnitBuiltins_Action_22();
>  }
>
>  /** @} */
> --
> 2.35.3
>
> _______________________________________________
> devel mailing list
> devel at rtems.org
> http://lists.rtems.org/mailman/listinfo/devel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/devel/attachments/20231102/b2f2f1d0/attachment-0001.htm>


More information about the devel mailing list