[PATCH 02/11] rs6000, fix arguments, add documentation for vector, element conversions
Kewen.Lin
linkw@linux.ibm.com
Wed Feb 28 09:23:28 GMT 2024
Hi,
on 2024/2/21 01:56, Carl Love wrote:
>
> GCC maintainers:
>
> This patch fixes the return type for the __builtin_vsx_xvcvdpuxws and __builtin_vsx_xvcvspuxds built-ins. They were defined as signed but should have been defined as unsigned.
>
> The patch has been tested on Power 10 with no regressions.
>
> Please let me know if this patch is acceptable for mainline. Thanks.
>
> Carl
>
> -----------------------------------------------------
> rs6000, fix arguments, add documentation for vector element conversions
>
> The return type for the __builtin_vsx_xvcvdpuxws, __builtin_vsx_xvcvspuxds,
> __builtin_vsx_xvcvspuxws built-ins should be unsigned. This patch changes
> the return values from signed to unsigned.
>
> The documentation for the vector element conversion built-ins:
>
> __builtin_vsx_xvcvspsxws
> __builtin_vsx_xvcvspsxds
> __builtin_vsx_xvcvspuxds
> __builtin_vsx_xvcvdpsxws
> __builtin_vsx_xvcvdpuxws
> __builtin_vsx_xvcvdpuxds_uns
> __builtin_vsx_xvcvspdp
> __builtin_vsx_xvcvdpsp
> __builtin_vsx_xvcvspuxws
> __builtin_vsx_xvcvsxwdp
> __builtin_vsx_xvcvuxddp_uns
> __builtin_vsx_xvcvuxwdp
>
> is missing from extend.texi. This patch adds the missing documentation.
I think we should recommend users to adopt the recommended built-ins in
PVIPR, by checking the corresponding mnemonic in PVIPR, I got:
__builtin_vsx_xvcvspsxws -> vec_signed
__builtin_vsx_xvcvspsxds -> N/A
__builtin_vsx_xvcvspuxds -> N/A
__builtin_vsx_xvcvdpsxws -> vec_signed{e,o}
__builtin_vsx_xvcvdpuxws -> vec_unsigned{e,o}
__builtin_vsx_xvcvdpuxds_uns -> vec_unsigned
__builtin_vsx_xvcvspdp -> vec_double{e,o}
__builtin_vsx_xvcvdpsp -> vec_float{e,o}
__builtin_vsx_xvcvspuxws -> vec_unsigned
__builtin_vsx_xvcvsxwdp -> vec_double{e,o}
__builtin_vsx_xvcvuxddp_uns> vec_double
For __builtin_vsx_xvcvspsxds and __builtin_vsx_xvcvspuxds which don't have
the according PVIPR built-ins, we can extend the current vec_{un,}signed{e,o}
to cover them and document them following the section mentioning PVIPR.
BR,
Kewen
>
> This patch also adds runnable test cases for each of the built-ins.
>
> gcc/ChangeLog:
> * config/rs6000/rs6000-builtins.def (__builtin_vsx_xvcvdpuxws,
> __builtin_vsx_xvcvspuxds, __builtin_vsx_xvcvspuxws): Change
> return type from signed to unsigned.
> * doc/extend.texi (__builtin_vsx_xvcvspsxws,
> __builtin_vsx_xvcvspsxds, __builtin_vsx_xvcvspuxds,
> __builtin_vsx_xvcvdpsxws, __builtin_vsx_xvcvdpuxws,
> __builtin_vsx_xvcvdpuxds_uns, __builtin_vsx_xvcvspdp,
> __builtin_vsx_xvcvdpsp, __builtin_vsx_xvcvspuxws,
> __builtin_vsx_xvcvsxwdp, __builtin_vsx_xvcvuxddp_uns,
> __builtin_vsx_xvcvuxwdp): Add documentation for builtins.
>
> gcc/testsuite/ChangeLog:
> * gcc.target/powerpc/vsx-builtin-runnable-1.c: New test file.
> ---
> gcc/config/rs6000/rs6000-builtins.def | 6 +-
> gcc/doc/extend.texi | 135 ++++++++++
> .../powerpc/vsx-builtin-runnable-1.c | 233 ++++++++++++++++++
> 3 files changed, 371 insertions(+), 3 deletions(-)
> create mode 100644 gcc/testsuite/gcc.target/powerpc/vsx-builtin-runnable-1.c
>
> diff --git a/gcc/config/rs6000/rs6000-builtins.def b/gcc/config/rs6000/rs6000-builtins.def
> index d66a53a0fab..fd316f629e5 100644
> --- a/gcc/config/rs6000/rs6000-builtins.def
> +++ b/gcc/config/rs6000/rs6000-builtins.def
> @@ -1724,7 +1724,7 @@
> const vull __builtin_vsx_xvcvdpuxds_uns (vd);
> XVCVDPUXDS_UNS vsx_fixuns_truncv2dfv2di2 {}
>
> - const vsi __builtin_vsx_xvcvdpuxws (vd);
> + const vui __builtin_vsx_xvcvdpuxws (vd);
> XVCVDPUXWS vsx_xvcvdpuxws {}
>
> const vd __builtin_vsx_xvcvspdp (vf);
> @@ -1736,10 +1736,10 @@
> const vsi __builtin_vsx_xvcvspsxws (vf);
> XVCVSPSXWS vsx_fix_truncv4sfv4si2 {}
>
> - const vsll __builtin_vsx_xvcvspuxds (vf);
> + const vull __builtin_vsx_xvcvspuxds (vf);
> XVCVSPUXDS vsx_xvcvspuxds {}
>
> - const vsi __builtin_vsx_xvcvspuxws (vf);
> + const vui __builtin_vsx_xvcvspuxws (vf);
> XVCVSPUXWS vsx_fixuns_truncv4sfv4si2 {}
>
> const vd __builtin_vsx_xvcvsxddp (vsll);
> diff --git a/gcc/doc/extend.texi b/gcc/doc/extend.texi
> index 4d8610f6aa8..583b1d890bf 100644
> --- a/gcc/doc/extend.texi
> +++ b/gcc/doc/extend.texi
> @@ -21360,6 +21360,141 @@ __float128 __builtin_sqrtf128 (__float128);
> __float128 __builtin_fmaf128 (__float128, __float128, __float128);
> @end smallexample
>
> +@smallexample
> +vector int __builtin_vsx_xvcvspsxws (vector float);
> +@end smallexample
> +
> +The @code{__builtin_vsx_xvcvspsxws} converts the single precision floating
> +point vector element i to a signed single-precision integer value using
> +round to zero storing the result in element i. If the source element is NaN
> +the result is set to 0x80000000 and VXCI is set to 1. If the source
> +element is SNaN then VXSNAN is also set to 1. If the rounded value is greater
> +than 2^31 - 1 the result is 0x7FFFFFFF and VXCVI is set to 1. If the
> +rounded value is less than -2^31, the result is set to 0x80000000 and
> +VXCVI is set to 1. If the rounded result is inexact then XX is set to 1.
>> +@smallexample
> +vector signed long long int __builtin_vsx_xvcvspsxds (vector float);
> +@end smallexample
> +
> +The @code{__builtin_vsx_xvcvspsxds} converts the single precision floating
> +point vector element to a double precision signed integer value using the
> +round to zero rounding mode. If the source element is NaN the result
> +is set to 0x8000000000000000 and VXCI is set to 1. If the source element is
> +SNaN then VXSNAN is also set to 1. If the rounded value is greater than
> +2^63 - 1 the result is 0x7FFFFFFFFFFFFFFF and VXCVI is set to 1. If the
> +rounded value is less than zero, the result is set to 0x8000000000000000 and
> +VXCVI is set to 1. If the rounded result is inexact then XX is set to 1.
> +
> +@smallexample
> +vector unsigned long long __builtin_vsx_xvcvspuxds (vector float);
> +@end smallexample
> +
> +The @code{__builtin_vsx_xvcvspuxds} converts the single precision floating
> +point vector element 2*i to an unsigned double-precision integer value using
> +round to zero storing the result in element i. If the source element is NaN
> +the result is set to 0x0000000000000000 and VXCI is set to 1. If the source
> +element is SNaN then VXSNAN is also set to 1. If the rounded value is greater
> +than 2^63 - 1 the result is 0xFFFFFFFFFFFFFFFF and VXCVI is set to 1. If the
> +rounded value is less than -2^63, the result is set to 0x0000000000000000 and
> +VXCVI is set to 1. If the rounded result is inexact then XX is set to 1.
> +
> +@smallexample
> +vector signed int __builtin_vsx_xvcvdpsxws (vector double);
> +@end smallexample
> +
> +The @code{__builtin_vsx_xvcvdpsxws} converts the ith double precision floating
> +point vector element to a single-precision integer value using the round to
> +zero rounding mode. The single precision integer value is placed into vector
> +elements j and j+1 where j = i*2. If the source element is NaN the result
> +is set to 0x80000000 and VXCI is set to 1. If the source element is SNaN then
> +VXSNAN is also set to 1. If the rounded value is greater than 2^31 - 1 the
> +result is 0x7FFFFFFF and VXCVI is set to 1. If the rounded value is less than
> +-2^31, the result is set to 0x80000000 and VXCVI is set to 1. If the rounded
> +result is inexact then XX is set to 1.
> +
> +@smallexample
> +vector unsigned int __builtin_vsx_xvcvdpuxws (vector double);
> +@end smallexample
> +
> +The @code{__builtin_vsx_xvcvdpuxws} converts the ith double precision floating
> +point vector element to a unsigned single-precision integer value using the
> +round to zero rounding mode. The single precision integer value is placed into
> +vector elements j and j+1 where j = i*2. If the source element is NaN the
> +result is set to 0x00000000 and VXCI is set to 1. If the source element is
> +SNaN then VXSNAN is also set to 1. If the rounded value is greater than
> +2^31 - 1 the result is 0xFFFFFFFF and VXCVI is set to 1. If the rounded value
> +is less than zero, the result is set to 0x00000000 and VXCVI is set to 1. If
> +the rounded result is inexact then XX is set to 1.
> +
> +@smallexample
> +vector unsigned long long int __builtin_vsx_xvcvdpuxds_uns (vector double);
> +@end smallexample
> +
> +The @code{__builtin_vsx_xvcvdpuxds_uns} converts the double precision floating
> +point vector element to a double precision unsigned integer value using the
> +round to zero rounding mode. If the source element is NaN the result is set
> +to 0x0000000000000000 and VXCI is set to 1. If the source element is SNaN
> +then VXSNAN is also set to 1. If the rounded value is greater than 2^63 - 1
> +the result is 0xFFFFFFFFFFFFFFFF and VXCVI is set to 1. If the rounded value
> +is less than zero, the result is set to 0x0000000000000000 and VXCVI is set to
> +1. If the rounded result is inexact then XX is set to 1.
> +
> +@smallexample
> +vector double __builtin_vsx_xvcvspdp (vector float);
> +@end smallexample
> +
> +The @code{__builtin_vsx_xvcvspdp} converts single precision floating
> +point vector element to a double precision floating point value. Input element
> +at index 2*i is stored in the destination element i.
> +
> +@smallexample
> +vector float __builtin_vsx_xvcvdpsp (vector double);
> +@end smallexample
> +
> +The @code{__builtin_vsx_xvcvdpsp} converts the ith double precision floating
> +point vector element to a single-precision floating point value using the
> +rounding mode specified by RN. The single precision value is placed into
> +vector elements j and j+1 where j = i*2. The rounding mode, RN, is specified
> +by bits [62:63] of the FPSCR.
> +
> +@smallexample
> +vector unsigned int __builtin_vsx_xvcvspuxws (vector float);
> +@end smallexample
> +
> +The @code{__builtin_vsx_xvcvspuxws} converts the single precision floating
> +point vector element i to an unsigned single-precision integer value using
> +round to zero storing the result in element i. If the source element is NaN
> +the result is set to 0x00000000 and VXCI is set to 1. If the source element
> +is SNaN then VXSNAN is also set to 1. If the rounded value is greater than
> +2^31 - 1 the result is 0xFFFFFFFF and VXCVI is set to 1. If the rounded
> +value is less than -2^31, the result is set to 0x00000000 and VXCVI is set
> +to 1. If the rounded result is inexact then XX is set to 1.
> +
> +@smallexample
> +vector double __builtin_vsx_xvcvsxwdp (vector signed int);
> +@end smallexample
> +
> +The @code{__builtin_vsx_xvcvsxwdp} converts single precision integer value
> +to a double precision floating point value. Input element at index 2*i is
> +stored in the destination element i.
> +
> +@smallexample
> +vector double __builtin_vsx_xvcvuxddp_uns (vector unsigned long long);
> +@end smallexample
> +
> +The @code{__builtin_vsx_xvcvuxddp_uns} converts unsigned double-precision
> +integer value to a double precision floating point value. Input element
> +at index i is stored in the destination element i.
> +
> +@smallexample
> +vector double __builtin_vsx_xvcvuxwdp (vector unsigned int);
> +@end smallexample
> +
> +The @code{__builtin_vsx_xvcvuxwdp} converts single precision unsigned integer
> +value to a double precision floating point value. Input element at index 2*i
> +is stored in the destination element i.
> +
> @node Basic PowerPC Built-in Functions Available on ISA 2.07
> @subsubsection Basic PowerPC Built-in Functions Available on ISA 2.07
>
> diff --git a/gcc/testsuite/gcc.target/powerpc/vsx-builtin-runnable-1.c b/gcc/testsuite/gcc.target/powerpc/vsx-builtin-runnable-1.c
> new file mode 100644
> index 00000000000..91d16c3ba72
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/powerpc/vsx-builtin-runnable-1.c
> @@ -0,0 +1,233 @@
> +/* { dg-do run { target { lp64 } } } */
> +/* { dg-require-effective-target powerpc_vsx_ok } */
> +/* { dg-options "-O2 -mdejagnu-cpu=power7" } */
> +
> +#define DEBUG 0
> +
> +#if DEBUG
> +#include <stdio.h>
> +#include <stdlib.h>
> +#endif
> +
> +void abort (void);
> +
> +int main ()
> +{
> + int i;
> + vector double vd_arg1, vd_result, vd_expected_result;
> + vector float vf_arg1, vf_result, vf_expected_result;
> + vector int vsi_arg1;
> + vector unsigned int vui_arg1;
> + vector int vsi_result, vsi_expected_result;
> + vector unsigned int vui_result, vui_expected_result;
> + vector signed long long int vsll_result, vsll_expected_result;
> + vector unsigned long long int vull_arg1;
> + vector unsigned long long int vull_result, vull_expected_result;
> +
> + /* VSX Vector Convert with round to zero Single-Precision floating point to
> + Single-Precision signed integer format using the round to zero mode. */
> +
> + vf_arg1 = (vector float) {12345.98, 7654.321, -2.1234,
> + 9999999999956789012345678.9};
> + vsi_result = __builtin_vsx_xvcvspsxws (vf_arg1);
> + vsi_expected_result = (vector signed int) {12345, 7654, -2, 0x7fffffff};
> +
> + for (i = 0; i < 4; i++)
> + if (vsi_result[i] != vsi_expected_result[i])
> +#if DEBUG
> + printf("ERROR, __builtin_vsx_xvcvspsxws: vsi_result[%d] = 0x%x, vsi_expected_result[%d] = 0x%x\n",
> + i, vsi_result[i], i, vsi_expected_result[i]);
> +#else
> + abort();
> +#endif
> +
> + /* VSX Vector Convert with round Single-Precision floating point to
> + Double-Precision signed integer format using the round to zero mode. */
> +
> + vf_arg1 = (vector float) {12345.98, 7654.321, -2.1234,
> + 9999999999956789012345678.9};
> + vsll_result = __builtin_vsx_xvcvspsxds (vf_arg1);
> + vsll_expected_result = (vector signed long long) {7654, 0x7fffffffffffffff};
> +
> + for (i = 0; i < 2; i++)
> + if (vsll_result[i] != vsll_expected_result[i])
> +#if DEBUG
> + printf("ERROR, __builtin_vsx_xvcvspsxds: vsll_result[%d] = 0x%llx, vsll_expected_result[%d] = 0x%llx\n",
> + i, vsll_result[i], i, vsll_expected_result[i]);
> +#else
> + abort();
> +#endif
> +
> + /* VSX Vector Convert with round Single-Precision floating point to
> + Double-Precision unsigned integer format using the round to zero mode. */
> +
> + vf_arg1 = (vector float) {12345.98, 764.321, -2.1234,
> + 9999999999956789012345678.9};
> + vull_result = __builtin_vsx_xvcvspuxds (vf_arg1);
> + vull_expected_result = (vector unsigned long long) {764, 0xffffffffffffffff};
> +
> + for (i = 0; i < 2; i++)
> + if (vull_result[i] != vull_expected_result[i])
> +#if DEBUG
> + printf("ERROR, __builtin_vsx_xvcvspuxds: vull_result[%d] = 0x%llx, vull_expected_result[%d] = 0x%llx\n",
> + i, vull_result[i], i, vull_expected_result[i]);
> +#else
> + abort();
> +#endif
> +
> + /* VSX Vector Convert with round Double-Precision floating point to
> + signed Single-Precision integer format using the round to zero mode. */
> +
> + vd_arg1 = (vector double) {12345.987654321, -2.1234567890123456789};
> + /* Each double-precision value, i, is converted to single precision integer
> + and placed in vector elements j and j+1 where j = i*2. The round to
> + zero rounding mode is used. */
> + vsi_result = __builtin_vsx_xvcvdpsxws (vd_arg1);
> + vsi_expected_result = (vector int) {12345, 12345, -2, -2};
> +
> + for (i = 0; i < 4; i++)
> + if (vsi_result[i] != vsi_expected_result[i])
> +#if DEBUG
> + printf("ERROR, __builtin_vsx_xvcvdpsxws: vsi_result[%d] = %d, vsi_expected_result[%d] = %d\n",
> + i, vsi_result[i], i, vsi_expected_result[i]);
> +#else
> + abort();
> +#endif
> +
> + /* VSX Vector Convert with round Double-Precision floating point to
> + unsigned Single-Precision integer format using the round to zero mode. */
> +
> + vd_arg1 = (vector double) {12345.987654321, -2.1234567890123456789};
> + /* Each double-precision value, i, is converted to single precision integer
> + and placed in vector elements j and j+1 where j = i*2. The round to
> + zero rounding mode is used. */
> + vui_result = __builtin_vsx_xvcvdpuxws (vd_arg1);
> + vui_expected_result = (vector unsigned int) {12345, 12345, 0, 0};
> +
> + for (i = 0; i < 4; i++)
> + if (vui_result[i] != vui_expected_result[i])
> +#if DEBUG
> + printf("ERROR, __builtin_vsx_xvcvdpuxws: vui_result[%d] = %d, vui_expected_result[%d] = %d\n",
> + i, vui_result[i], i, vui_expected_result[i]);
> +#else
> + abort();
> +#endif
> +
> + /* VSX Vector Convert with round Double-Precision floating point to
> + Double-Precision unsigned integer format using the round to zero mode. */
> +
> + vd_arg1 = (vector double) {12345.987654321, -2.1234567890123456789};
> + vull_result = __builtin_vsx_xvcvdpuxds_uns (vd_arg1);
> + vull_expected_result = (vector unsigned long long) {12345, 0};
> +
> + for (i = 0; i < 2; i++)
> + if (vull_result[i] != vull_expected_result[i])
> +#if DEBUG
> + printf("ERROR, __builtin_vsx_xvcvdpuxds_uns: vull_result[%d] = %lld, vull_expected_result[%d] = %lld\n",
> + i, vull_result[i], i, vull_expected_result[i]);
> +#else
> + abort();
> +#endif
> +
> + /* VSX Vector Convert Single-Precision floating point to Double-Precision
> + floating point */
> +
> + vf_arg1 = (vector float) {12345.98, -2.0, 31.11, -55.5};
> + vd_result = __builtin_vsx_xvcvspdp (vf_arg1);
> + vd_expected_result = (vector double) {-2.0, -55.5};
> +
> + for (i = 0; i < 2; i++)
> + if (vd_result[i] != vd_expected_result[i])
> +#if DEBUG
> + printf("ERROR, __builtin_vsx_xvcvspdp: vd_result[%d] = %f, vf_expected_result[%d] = %f\n",
> + i, vd_result[i], i, vd_expected_result[i]);
> +#else
> + abort();
> +#endif
> +
> + /* VSX Vector Convert with round Double-Precision float point format to
> + Single-Precision floating point format using the rounding mode specified
> + by the RN field of the FPSCR. */
> +
> + vd_arg1 = (vector double) {12345.12345, -0.1234567890123456789};
> + /* Each double-precision value, i, is converted to single precision and
> + placed in vector elements j and j+1 where j = i*2. */
> + vf_result = __builtin_vsx_xvcvdpsp (vd_arg1);
> + vf_expected_result = (vector float) {12345.12345, 12345.12345,
> + -0.1234567890, -0.1234567890};
> +
> + for (i = 0; i < 4; i++)
> + if (vf_result[i] != vf_expected_result[i])
> +#if DEBUG
> + printf("ERROR, __builtin_vsx_xvcvdpsp: vf_result[%d] = %f, vf_expected_result[%d] = %f\n",
> + i, vf_result[i], i, vf_expected_result[i]);
> +#else
> + abort();
> +#endif
> +
> + /* VSX Vector Convert with round Single-Precision floating point to
> + Single-Precision unsigned integer format using the round to zero mode. */
> +
> + vf_arg1 = (vector float) {12345.98, 7654.321, -2.1234,
> + 9999999999956789012345678.9};
> + vui_result = __builtin_vsx_xvcvspuxws (vf_arg1);
> + vui_expected_result = (vector unsigned int) {12345, 7654, 0x0, 0xffffffff};
> +
> + for (i = 0; i < 4; i++)
> + if (vui_result[i] != vui_expected_result[i])
> +#if DEBUG
> + printf("ERROR, __builtin_vsx_xvcvspuxws: vui_result[%d] = 0x%x, vui_expected_result[%d] = 0x%x\n",
> + i, vui_result[i], i, vui_expected_result[i]);
> +#else
> + abort();
> +#endif
> +
> + /* VSX Vector Convert Signed integer word to Double-Precision floating point
> + format. */
> +
> + vsi_arg1 = (vector int) {2345, 98, -2, -55};
> + vd_result = __builtin_vsx_xvcvsxwdp (vsi_arg1);
> + vd_expected_result = (vector double) {98.0, -55.0};
> +
> + for (i = 0; i < 2; i++)
> + if (vd_result[i] != vd_expected_result[i])
> +#if DEBUG
> + printf("ERROR, __builtin_vsx_xvcvsxwdp: vd_result[%d] = %f, vd_expected_result[%d] = %f\n",
> + i, vd_result[i], i, vd_expected_result[i]);
> +#else
> + abort();
> +#endif
> +
> + /* VSX Vector Convert with round unsigned Double-Word integer to
> + Double-Precision floating point format. */
> +
> + vull_arg1 = (vector unsigned long long) {12398, 22255};
> + vd_result = __builtin_vsx_xvcvuxddp_uns (vull_arg1);
> + vd_expected_result = (vector double) {12398.0, 22255.0};
> +
> + for (i = 0; i < 2; i++)
> + if (vd_result[i] != vd_expected_result[i])
> +#if DEBUG
> + printf("ERROR, __builtin_vsx_xvcvuxddp_uns: vd_result[%d] = %f, vd_expected_result[%d] = %f\n",
> + i, vd_result[i], i, vd_expected_result[i]);
> +#else
> + abort();
> +#endif
> +
> + /* VSX Vector Convert unsigned Single-Precision integer to Double-Precision
> + floating point format. */
> +
> + vui_arg1 = (vector unsigned int) {12398, 22255, 345, 87};
> + vd_result = __builtin_vsx_xvcvuxwdp (vui_arg1);
> + vd_expected_result = (vector double) {22255.0, 87.0};
> +
> + for (i = 0; i < 2; i++)
> + if (vd_result[i] != vd_expected_result[i])
> +#if DEBUG
> + printf("ERROR, __builtin_vsx_xvcvuxwdp: vd_result[%d] = %f, vd_expected_result[%d] = %f\n",
> + i, vd_result[i], i, vd_expected_result[i]);
> +#else
> + abort();
> +#endif
> + return 0;
> +}
More information about the Gcc-patches
mailing list