Pat Haugen <pthaugen@us.ibm.com>
Peter Bergner <bergner@vnet.ibm.com>
+ * gcc.target/powerpc/p8vector-builtin-1.c: New test to test
+ power8 builtin functions.
+ * gcc/testsuite/gcc.target/powerpc/p8vector-builtin-2.c: Likewise.
+ * gcc/testsuite/gcc.target/powerpc/p8vector-builtin-3.c: Likewise.
+ * gcc/testsuite/gcc.target/powerpc/p8vector-builtin-4.c: Likewise.
+ * gcc/testsuite/gcc.target/powerpc/p8vector-builtin-5.c: Likewise.
+ * gcc/testsuite/gcc.target/powerpc/p8vector-builtin-6.c: Likewise.
+ * gcc/testsuite/gcc.target/powerpc/p8vector-builtin-7.c: Likewise.
+ * gcc/testsuite/gcc.target/powerpc/p8vector-vectorize-1.c: New
+ tests to test power8 auto-vectorization.
+ * gcc/testsuite/gcc.target/powerpc/p8vector-vectorize-2.c: Likewise.
+ * gcc/testsuite/gcc.target/powerpc/p8vector-vectorize-3.c: Likewise.
+ * gcc/testsuite/gcc.target/powerpc/p8vector-vectorize-4.c: Likewise.
+ * gcc/testsuite/gcc.target/powerpc/p8vector-vectorize-5.c: Likewise.
+
* gcc.target/powerpc/crypto-builtin-1.c: Use effective target
powerpc_p8vector_ok instead of powerpc_vsx_ok.
--- /dev/null
+/* { dg-do compile { target { powerpc*-*-* } } } */
+/* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */
+/* { dg-require-effective-target powerpc_p8vector_ok } */
+/* { dg-options "-mcpu=power8 -O2 -ftree-vectorize -fvect-cost-model -fno-unroll-loops -fno-unroll-all-loops" } */
+
+#ifndef TYPE
+#define TYPE long long
+#endif
+
+#ifndef SIGN_TYPE
+#define SIGN_TYPE signed TYPE
+#endif
+
+#ifndef UNS_TYPE
+#define UNS_TYPE unsigned TYPE
+#endif
+
+typedef vector SIGN_TYPE v_sign;
+typedef vector UNS_TYPE v_uns;
+
+v_sign sign_add (v_sign a, v_sign b)
+{
+ return a + b;
+}
+
+v_sign sign_sub (v_sign a, v_sign b)
+{
+ return a - b;
+}
+
+v_sign sign_shift_left (v_sign a, v_sign b)
+{
+ return a << b;
+}
+
+v_sign sign_shift_right (v_sign a, v_sign b)
+{
+ return a >> b;
+}
+
+v_uns uns_add (v_uns a, v_uns b)
+{
+ return a + b;
+}
+
+v_uns uns_sub (v_uns a, v_uns b)
+{
+ return a - b;
+}
+
+v_uns uns_shift_left (v_uns a, v_uns b)
+{
+ return a << b;
+}
+
+v_uns uns_shift_right (v_uns a, v_uns b)
+{
+ return a >> b;
+}
+
+/* { dg-final { scan-assembler-times "vaddudm" 2 } } */
+/* { dg-final { scan-assembler-times "vsubudm" 2 } } */
+/* { dg-final { scan-assembler-times "vsld" 2 } } */
+/* { dg-final { scan-assembler-times "vsrad" 1 } } */
+/* { dg-final { scan-assembler-times "vsrd" 1 } } */
--- /dev/null
+/* { dg-do compile { target { powerpc*-*-* } } } */
+/* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */
+/* { dg-require-effective-target powerpc_p8vector_ok } */
+/* { dg-options "-mcpu=power8 -O2 -ftree-vectorize -fvect-cost-model -fno-unroll-loops -fno-unroll-all-loops" } */
+
+#include <altivec.h>
+
+typedef vector long long v_sign;
+typedef vector unsigned long long v_uns;
+typedef vector bool long long v_bool;
+
+v_sign sign_add_1 (v_sign a, v_sign b)
+{
+ return __builtin_altivec_vaddudm (a, b);
+}
+
+v_sign sign_add_2 (v_sign a, v_sign b)
+{
+ return vec_add (a, b);
+}
+
+v_sign sign_add_3 (v_sign a, v_sign b)
+{
+ return vec_vaddudm (a, b);
+}
+
+v_sign sign_sub_1 (v_sign a, v_sign b)
+{
+ return __builtin_altivec_vsubudm (a, b);
+}
+
+v_sign sign_sub_2 (v_sign a, v_sign b)
+{
+ return vec_sub (a, b);
+}
+
+
+v_sign sign_sub_3 (v_sign a, v_sign b)
+{
+ return vec_vsubudm (a, b);
+}
+
+v_sign sign_min_1 (v_sign a, v_sign b)
+{
+ return __builtin_altivec_vminsd (a, b);
+}
+
+v_sign sign_min_2 (v_sign a, v_sign b)
+{
+ return vec_min (a, b);
+}
+
+v_sign sign_min_3 (v_sign a, v_sign b)
+{
+ return vec_vminsd (a, b);
+}
+
+v_sign sign_max_1 (v_sign a, v_sign b)
+{
+ return __builtin_altivec_vmaxsd (a, b);
+}
+
+v_sign sign_max_2 (v_sign a, v_sign b)
+{
+ return vec_max (a, b);
+}
+
+v_sign sign_max_3 (v_sign a, v_sign b)
+{
+ return vec_vmaxsd (a, b);
+}
+
+v_sign sign_abs (v_sign a)
+{
+ return vec_abs (a); /* xor, vsubudm, vmaxsd. */
+}
+
+v_bool sign_eq (v_sign a, v_sign b)
+{
+ return vec_cmpeq (a, b);
+}
+
+v_bool sign_lt (v_sign a, v_sign b)
+{
+ return vec_cmplt (a, b);
+}
+
+v_uns uns_add_2 (v_uns a, v_uns b)
+{
+ return vec_add (a, b);
+}
+
+v_uns uns_add_3 (v_uns a, v_uns b)
+{
+ return vec_vaddudm (a, b);
+}
+
+v_uns uns_sub_2 (v_uns a, v_uns b)
+{
+ return vec_sub (a, b);
+}
+
+v_uns uns_sub_3 (v_uns a, v_uns b)
+{
+ return vec_vsubudm (a, b);
+}
+
+v_uns uns_min_2 (v_uns a, v_uns b)
+{
+ return vec_min (a, b);
+}
+
+v_uns uns_min_3 (v_uns a, v_uns b)
+{
+ return vec_vminud (a, b);
+}
+
+v_uns uns_max_2 (v_uns a, v_uns b)
+{
+ return vec_max (a, b);
+}
+
+v_uns uns_max_3 (v_uns a, v_uns b)
+{
+ return vec_vmaxud (a, b);
+}
+
+v_bool uns_eq (v_uns a, v_uns b)
+{
+ return vec_cmpeq (a, b);
+}
+
+v_bool uns_lt (v_uns a, v_uns b)
+{
+ return vec_cmplt (a, b);
+}
+
+v_sign sign_rl_1 (v_sign a, v_sign b)
+{
+ return __builtin_altivec_vrld (a, b);
+}
+
+v_sign sign_rl_2 (v_sign a, v_uns b)
+{
+ return vec_rl (a, b);
+}
+
+v_uns uns_rl_2 (v_uns a, v_uns b)
+{
+ return vec_rl (a, b);
+}
+
+v_sign sign_sl_1 (v_sign a, v_sign b)
+{
+ return __builtin_altivec_vsld (a, b);
+}
+
+v_sign sign_sl_2 (v_sign a, v_uns b)
+{
+ return vec_sl (a, b);
+}
+
+v_sign sign_sl_3 (v_sign a, v_uns b)
+{
+ return vec_vsld (a, b);
+}
+
+v_uns uns_sl_2 (v_uns a, v_uns b)
+{
+ return vec_sl (a, b);
+}
+
+v_uns uns_sl_3 (v_uns a, v_uns b)
+{
+ return vec_vsld (a, b);
+}
+
+v_sign sign_sra_1 (v_sign a, v_sign b)
+{
+ return __builtin_altivec_vsrad (a, b);
+}
+
+v_sign sign_sra_2 (v_sign a, v_uns b)
+{
+ return vec_sra (a, b);
+}
+
+v_sign sign_sra_3 (v_sign a, v_uns b)
+{
+ return vec_vsrad (a, b);
+}
+
+/* { dg-final { scan-assembler-times "vaddudm" 5 } } */
+/* { dg-final { scan-assembler-times "vsubudm" 6 } } */
+/* { dg-final { scan-assembler-times "vmaxsd" 4 } } */
+/* { dg-final { scan-assembler-times "vminsd" 3 } } */
+/* { dg-final { scan-assembler-times "vmaxud" 2 } } */
+/* { dg-final { scan-assembler-times "vminud" 2 } } */
+/* { dg-final { scan-assembler-times "vcmpequd" 2 } } */
+/* { dg-final { scan-assembler-times "vcmpgtsd" 1 } } */
+/* { dg-final { scan-assembler-times "vcmpgtud" 1 } } */
+/* { dg-final { scan-assembler-times "vrld" 3 } } */
+/* { dg-final { scan-assembler-times "vsld" 5 } } */
+/* { dg-final { scan-assembler-times "vsrad" 3 } } */
--- /dev/null
+/* { dg-do compile { target { powerpc*-*-* } } } */
+/* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */
+/* { dg-require-effective-target powerpc_p8vector_ok } */
+/* { dg-options "-mcpu=power8 -O3 -ftree-vectorize -fvect-cost-model" } */
+
+#include <altivec.h>
+
+typedef vector long long vll_sign;
+typedef vector unsigned long long vll_uns;
+typedef vector bool long long vll_bool;
+
+typedef vector int vi_sign;
+typedef vector unsigned int vi_uns;
+typedef vector bool int vi_bool;
+
+typedef vector short vs_sign;
+typedef vector unsigned short vs_uns;
+typedef vector bool short vs_bool;
+
+typedef vector signed char vc_sign;
+typedef vector unsigned char vc_uns;
+typedef vector bool char vc_bool;
+
+
+vi_sign vi_pack_1 (vll_sign a, vll_sign b)
+{
+ return __builtin_altivec_vpkudum (a, b);
+}
+
+vi_sign vi_pack_2 (vll_sign a, vll_sign b)
+{
+ return vec_pack (a, b);
+}
+
+vi_sign vi_pack_3 (vll_sign a, vll_sign b)
+{
+ return vec_vpkudum (a, b);
+}
+
+vs_sign vs_pack_1 (vi_sign a, vi_sign b)
+{
+ return __builtin_altivec_vpkuwum (a, b);
+}
+
+vs_sign vs_pack_2 (vi_sign a, vi_sign b)
+{
+ return vec_pack (a, b);
+}
+
+vs_sign vs_pack_3 (vi_sign a, vi_sign b)
+{
+ return vec_vpkuwum (a, b);
+}
+
+vc_sign vc_pack_1 (vs_sign a, vs_sign b)
+{
+ return __builtin_altivec_vpkuhum (a, b);
+}
+
+vc_sign vc_pack_2 (vs_sign a, vs_sign b)
+{
+ return vec_pack (a, b);
+}
+
+vc_sign vc_pack_3 (vs_sign a, vs_sign b)
+{
+ return vec_vpkuhum (a, b);
+}
+
+vll_sign vll_unpack_hi_1 (vi_sign a)
+{
+ return __builtin_altivec_vupkhsw (a);
+}
+
+vll_sign vll_unpack_hi_2 (vi_sign a)
+{
+ return vec_unpackh (a);
+}
+
+vll_sign vll_unpack_hi_3 (vi_sign a)
+{
+ return __builtin_vec_vupkhsw (a);
+}
+
+vll_sign vll_unpack_lo_1 (vi_sign a)
+{
+ return vec_vupklsw (a);
+}
+
+vll_sign vll_unpack_lo_2 (vi_sign a)
+{
+ return vec_unpackl (a);
+}
+
+vll_sign vll_unpack_lo_3 (vi_sign a)
+{
+ return vec_vupklsw (a);
+}
+
+/* { dg-final { scan-assembler-times "vpkudum" 3 } } */
+/* { dg-final { scan-assembler-times "vpkuwum" 3 } } */
+/* { dg-final { scan-assembler-times "vpkuhum" 3 } } */
+/* { dg-final { scan-assembler-times "vupklsw" 3 } } */
+/* { dg-final { scan-assembler-times "vupkhsw" 3 } } */
--- /dev/null
+/* { dg-do compile { target { powerpc*-*-* } } } */
+/* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */
+/* { dg-require-effective-target powerpc_p8vector_ok } */
+/* { dg-options "-mcpu=power8 -O3 -ftree-vectorize -fvect-cost-model" } */
+
+#include <altivec.h>
+
+typedef vector long long vll_sign;
+typedef vector unsigned long long vll_uns;
+typedef vector bool long long vll_bool;
+
+typedef vector int vi_sign;
+typedef vector unsigned int vi_uns;
+typedef vector bool int vi_bool;
+
+typedef vector short vs_sign;
+typedef vector unsigned short vs_uns;
+typedef vector bool short vs_bool;
+
+typedef vector signed char vc_sign;
+typedef vector unsigned char vc_uns;
+typedef vector bool char vc_bool;
+
+vll_sign vll_clz_1 (vll_sign a)
+{
+ return __builtin_altivec_vclzd (a);
+}
+
+vll_sign vll_clz_2 (vll_sign a)
+{
+ return vec_vclz (a);
+}
+
+vll_sign vll_clz_3 (vll_sign a)
+{
+ return vec_vclzd (a);
+}
+
+vll_uns vll_clz_4 (vll_uns a)
+{
+ return vec_vclz (a);
+}
+
+vll_uns vll_clz_5 (vll_uns a)
+{
+ return vec_vclzd (a);
+}
+
+vi_sign vi_clz_1 (vi_sign a)
+{
+ return __builtin_altivec_vclzw (a);
+}
+
+vi_sign vi_clz_2 (vi_sign a)
+{
+ return vec_vclz (a);
+}
+
+vi_sign vi_clz_3 (vi_sign a)
+{
+ return vec_vclzw (a);
+}
+
+vi_uns vi_clz_4 (vi_uns a)
+{
+ return vec_vclz (a);
+}
+
+vi_uns vi_clz_5 (vi_uns a)
+{
+ return vec_vclzw (a);
+}
+
+vs_sign vs_clz_1 (vs_sign a)
+{
+ return __builtin_altivec_vclzh (a);
+}
+
+vs_sign vs_clz_2 (vs_sign a)
+{
+ return vec_vclz (a);
+}
+
+vs_sign vs_clz_3 (vs_sign a)
+{
+ return vec_vclzh (a);
+}
+
+vs_uns vs_clz_4 (vs_uns a)
+{
+ return vec_vclz (a);
+}
+
+vs_uns vs_clz_5 (vs_uns a)
+{
+ return vec_vclzh (a);
+}
+
+vc_sign vc_clz_1 (vc_sign a)
+{
+ return __builtin_altivec_vclzb (a);
+}
+
+vc_sign vc_clz_2 (vc_sign a)
+{
+ return vec_vclz (a);
+}
+
+vc_sign vc_clz_3 (vc_sign a)
+{
+ return vec_vclzb (a);
+}
+
+vc_uns vc_clz_4 (vc_uns a)
+{
+ return vec_vclz (a);
+}
+
+vc_uns vc_clz_5 (vc_uns a)
+{
+ return vec_vclzb (a);
+}
+
+vll_sign vll_popcnt_1 (vll_sign a)
+{
+ return __builtin_altivec_vpopcntd (a);
+}
+
+vll_sign vll_popcnt_2 (vll_sign a)
+{
+ return vec_vpopcnt (a);
+}
+
+vll_sign vll_popcnt_3 (vll_sign a)
+{
+ return vec_vpopcntd (a);
+}
+
+vll_uns vll_popcnt_4 (vll_uns a)
+{
+ return vec_vpopcnt (a);
+}
+
+vll_uns vll_popcnt_5 (vll_uns a)
+{
+ return vec_vpopcntd (a);
+}
+
+vi_sign vi_popcnt_1 (vi_sign a)
+{
+ return __builtin_altivec_vpopcntw (a);
+}
+
+vi_sign vi_popcnt_2 (vi_sign a)
+{
+ return vec_vpopcnt (a);
+}
+
+vi_sign vi_popcnt_3 (vi_sign a)
+{
+ return vec_vpopcntw (a);
+}
+
+vi_uns vi_popcnt_4 (vi_uns a)
+{
+ return vec_vpopcnt (a);
+}
+
+vi_uns vi_popcnt_5 (vi_uns a)
+{
+ return vec_vpopcntw (a);
+}
+
+vs_sign vs_popcnt_1 (vs_sign a)
+{
+ return __builtin_altivec_vpopcnth (a);
+}
+
+vs_sign vs_popcnt_2 (vs_sign a)
+{
+ return vec_vpopcnt (a);
+}
+
+vs_sign vs_popcnt_3 (vs_sign a)
+{
+ return vec_vpopcnth (a);
+}
+
+vs_uns vs_popcnt_4 (vs_uns a)
+{
+ return vec_vpopcnt (a);
+}
+
+vs_uns vs_popcnt_5 (vs_uns a)
+{
+ return vec_vpopcnth (a);
+}
+
+vc_sign vc_popcnt_1 (vc_sign a)
+{
+ return __builtin_altivec_vpopcntb (a);
+}
+
+vc_sign vc_popcnt_2 (vc_sign a)
+{
+ return vec_vpopcnt (a);
+}
+
+vc_sign vc_popcnt_3 (vc_sign a)
+{
+ return vec_vpopcntb (a);
+}
+
+vc_uns vc_popcnt_4 (vc_uns a)
+{
+ return vec_vpopcnt (a);
+}
+
+vc_uns vc_popcnt_5 (vc_uns a)
+{
+ return vec_vpopcntb (a);
+}
+
+vc_uns vc_gbb_1 (vc_uns a)
+{
+ return __builtin_altivec_vgbbd (a);
+}
+
+vc_sign vc_gbb_2 (vc_sign a)
+{
+ return vec_vgbbd (a);
+}
+
+vc_uns vc_gbb_3 (vc_uns a)
+{
+ return vec_vgbbd (a);
+}
+
+/* { dg-final { scan-assembler-times "vclzd" 5 } } */
+/* { dg-final { scan-assembler-times "vclzw" 5 } } */
+/* { dg-final { scan-assembler-times "vclzh" 5 } } */
+/* { dg-final { scan-assembler-times "vclzb" 5 } } */
+
+/* { dg-final { scan-assembler-times "vpopcntd" 5 } } */
+/* { dg-final { scan-assembler-times "vpopcntw" 5 } } */
+/* { dg-final { scan-assembler-times "vpopcnth" 5 } } */
+/* { dg-final { scan-assembler-times "vpopcntb" 5 } } */
+
+/* { dg-final { scan-assembler-times "vgbbd" 3 } } */
--- /dev/null
+/* { dg-do compile { target { powerpc*-*-* } } } */
+/* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */
+/* { dg-require-effective-target powerpc_p8vector_ok } */
+/* { dg-options "-mcpu=power8 -O2 -ftree-vectorize -fvect-cost-model -fno-unroll-loops -fno-unroll-all-loops" } */
+
+#include <altivec.h>
+
+#ifndef SIZE
+#define SIZE 1024
+#endif
+
+#ifndef ALIGN
+#define ALIGN 32
+#endif
+
+#ifndef ATTR_ALIGN
+#define ATTR_ALIGN __attribute__((__aligned__(ALIGN)))
+#endif
+
+#define DOIT(TYPE, PREFIX) \
+TYPE PREFIX ## _eqv_builtin (TYPE a, TYPE b) \
+{ \
+ return vec_eqv (a, b); \
+} \
+ \
+TYPE PREFIX ## _eqv_arith (TYPE a, TYPE b) \
+{ \
+ return ~(a ^ b); \
+} \
+ \
+TYPE PREFIX ## _nand_builtin (TYPE a, TYPE b) \
+{ \
+ return vec_nand (a, b); \
+} \
+ \
+TYPE PREFIX ## _nand_arith1 (TYPE a, TYPE b) \
+{ \
+ return ~(a & b); \
+} \
+ \
+TYPE PREFIX ## _nand_arith2 (TYPE a, TYPE b) \
+{ \
+ return (~a) | (~b); \
+} \
+ \
+TYPE PREFIX ## _orc_builtin (TYPE a, TYPE b) \
+{ \
+ return vec_orc (a, b); \
+} \
+ \
+TYPE PREFIX ## _orc_arith1 (TYPE a, TYPE b) \
+{ \
+ return (~ a) | b; \
+} \
+ \
+TYPE PREFIX ## _orc_arith2 (TYPE a, TYPE b) \
+{ \
+ return a | (~ b); \
+}
+
+#define DOIT_FLOAT(TYPE, PREFIX) \
+TYPE PREFIX ## _eqv_builtin (TYPE a, TYPE b) \
+{ \
+ return vec_eqv (a, b); \
+} \
+ \
+TYPE PREFIX ## _nand_builtin (TYPE a, TYPE b) \
+{ \
+ return vec_nand (a, b); \
+} \
+ \
+TYPE PREFIX ## _orc_builtin (TYPE a, TYPE b) \
+{ \
+ return vec_orc (a, b); \
+}
+
+typedef vector signed char sign_char_vec;
+typedef vector short sign_short_vec;
+typedef vector int sign_int_vec;
+typedef vector long long sign_llong_vec;
+
+typedef vector unsigned char uns_char_vec;
+typedef vector unsigned short uns_short_vec;
+typedef vector unsigned int uns_int_vec;
+typedef vector unsigned long long uns_llong_vec;
+
+typedef vector float float_vec;
+typedef vector double double_vec;
+
+DOIT(sign_char_vec, sign_char)
+DOIT(sign_short_vec, sign_short)
+DOIT(sign_int_vec, sign_int)
+DOIT(sign_llong_vec, sign_llong)
+
+DOIT(uns_char_vec, uns_char)
+DOIT(uns_short_vec, uns_short)
+DOIT(uns_int_vec, uns_int)
+DOIT(uns_llong_vec, uns_llong)
+
+DOIT_FLOAT(float_vec, float)
+DOIT_FLOAT(double_vec, double)
+
+/* { dg-final { scan-assembler-times "xxleqv" 18 } } */
+/* { dg-final { scan-assembler-times "xxlnand" 26 } } */
+/* { dg-final { scan-assembler-times "xxlorc" 26 } } */
--- /dev/null
+/* { dg-do compile { target { powerpc*-*-* } } } */
+/* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */
+/* { dg-require-effective-target powerpc_p8vector_ok } */
+/* { dg-options "-mcpu=power8 -O2" } */
+
+vector float dbl_to_float_p8 (double x) { return __builtin_vsx_xscvdpspn (x); }
+double float_to_dbl_p8 (vector float x) { return __builtin_vsx_xscvspdpn (x); }
+
+/* { dg-final { scan-assembler "xscvdpspn" } } */
+/* { dg-final { scan-assembler "xscvspdpn" } } */
--- /dev/null
+/* { dg-do compile { target { powerpc*-*-* } } } */
+/* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */
+/* { dg-require-effective-target powerpc_p8vector_ok } */
+/* { dg-options "-mcpu=power8 -O2" } */
+
+#include <altivec.h>
+
+typedef vector int v_sign;
+typedef vector unsigned int v_uns;
+
+v_sign even_sign (v_sign a, v_sign b)
+{
+ return vec_vmrgew (a, b);
+}
+
+v_uns even_uns (v_uns a, v_uns b)
+{
+ return vec_vmrgew (a, b);
+}
+
+v_sign odd_sign (v_sign a, v_sign b)
+{
+ return vec_vmrgow (a, b);
+}
+
+v_uns odd_uns (v_uns a, v_uns b)
+{
+ return vec_vmrgow (a, b);
+}
+
+/* { dg-final { scan-assembler-times "vmrgew" 2 } } */
+/* { dg-final { scan-assembler-times "vmrgow" 2 } } */
--- /dev/null
+/* { dg-do compile { target { powerpc*-*-* } } } */
+/* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */
+/* { dg-require-effective-target powerpc_p8vector_ok } */
+/* { dg-options "-mcpu=power8 -O2 -ftree-vectorize -fvect-cost-model -fno-unroll-loops -fno-unroll-all-loops" } */
+
+#ifndef SIZE
+#define SIZE 1024
+#endif
+
+#ifndef ALIGN
+#define ALIGN 32
+#endif
+
+#ifndef TYPE
+#define TYPE long long
+#endif
+
+#ifndef SIGN_TYPE
+#define SIGN_TYPE signed TYPE
+#endif
+
+#ifndef UNS_TYPE
+#define UNS_TYPE unsigned TYPE
+#endif
+
+#define ALIGN_ATTR __attribute__((__aligned__(ALIGN)))
+
+SIGN_TYPE sa[SIZE] ALIGN_ATTR;
+SIGN_TYPE sb[SIZE] ALIGN_ATTR;
+SIGN_TYPE sc[SIZE] ALIGN_ATTR;
+
+UNS_TYPE ua[SIZE] ALIGN_ATTR;
+UNS_TYPE ub[SIZE] ALIGN_ATTR;
+UNS_TYPE uc[SIZE] ALIGN_ATTR;
+
+void
+sign_add (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ sa[i] = sb[i] + sc[i];
+}
+
+void
+sign_sub (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ sa[i] = sb[i] - sc[i];
+}
+
+void
+sign_shift_left (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ sa[i] = sb[i] << sc[i];
+}
+
+void
+sign_shift_right (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ sa[i] = sb[i] >> sc[i];
+}
+
+void
+sign_max (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ sa[i] = (sb[i] > sc[i]) ? sb[i] : sc[i];
+}
+
+void
+sign_min (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ sa[i] = (sb[i] < sc[i]) ? sb[i] : sc[i];
+}
+
+void
+sign_abs (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ sa[i] = (sb[i] < 0) ? -sb[i] : sb[i]; /* xor, vsubudm, vmaxsd. */
+}
+
+void
+sign_eq (SIGN_TYPE val1, SIGN_TYPE val2)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ sa[i] = (sb[i] == sc[i]) ? val1 : val2;
+}
+
+void
+sign_lt (SIGN_TYPE val1, SIGN_TYPE val2)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ sa[i] = (sb[i] < sc[i]) ? val1 : val2;
+}
+
+void
+uns_add (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ ua[i] = ub[i] + uc[i];
+}
+
+void
+uns_sub (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ ua[i] = ub[i] - uc[i];
+}
+
+void
+uns_shift_left (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ ua[i] = ub[i] << uc[i];
+}
+
+void
+uns_shift_right (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ ua[i] = ub[i] >> uc[i];
+}
+
+void
+uns_max (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ ua[i] = (ub[i] > uc[i]) ? ub[i] : uc[i];
+}
+
+void
+uns_min (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ ua[i] = (ub[i] < uc[i]) ? ub[i] : uc[i];
+}
+
+void
+uns_eq (UNS_TYPE val1, UNS_TYPE val2)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ ua[i] = (ub[i] == uc[i]) ? val1 : val2;
+}
+
+void
+uns_lt (UNS_TYPE val1, UNS_TYPE val2)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ ua[i] = (ub[i] < uc[i]) ? val1 : val2;
+}
+
+/* { dg-final { scan-assembler-times "\[\t \]vaddudm\[\t \]" 2 } } */
+/* { dg-final { scan-assembler-times "\[\t \]vsubudm\[\t \]" 3 } } */
+/* { dg-final { scan-assembler-times "\[\t \]vmaxsd\[\t \]" 2 } } */
+/* { dg-final { scan-assembler-times "\[\t \]vmaxud\[\t \]" 1 } } */
+/* { dg-final { scan-assembler-times "\[\t \]vminsd\[\t \]" 1 } } */
+/* { dg-final { scan-assembler-times "\[\t \]vminud\[\t \]" 1 } } */
+/* { dg-final { scan-assembler-times "\[\t \]vsld\[\t \]" 2 } } */
+/* { dg-final { scan-assembler-times "\[\t \]vsrad\[\t \]" 1 } } */
+/* { dg-final { scan-assembler-times "\[\t \]vsrd\[\t \]" 1 } } */
+/* { dg-final { scan-assembler-times "\[\t \]vcmpequd\[\t \]" 2 } } */
+/* { dg-final { scan-assembler-times "\[\t \]vcmpgtsd\[\t \]" 1 } } */
+/* { dg-final { scan-assembler-times "\[\t \]vcmpgtud\[\t \]" 1 } } */
--- /dev/null
+/* { dg-do compile { target { powerpc*-*-* } } } */
+/* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */
+/* { dg-require-effective-target powerpc_p8vector_ok } */
+/* { dg-options "-mcpu=power8 -O2 -ftree-vectorize -fvect-cost-model" } */
+
+#include <stddef.h>
+
+#ifndef SIZE
+#define SIZE 1024
+#endif
+
+#ifndef ALIGN
+#define ALIGN 32
+#endif
+
+#define ALIGN_ATTR __attribute__((__aligned__(ALIGN)))
+
+long long sign_ll[SIZE] ALIGN_ATTR;
+int sign_i [SIZE] ALIGN_ATTR;
+
+void copy_int_to_long_long (void)
+{
+ size_t i;
+
+ for (i = 0; i < SIZE; i++)
+ sign_ll[i] = sign_i[i];
+}
+
+/* { dg-final { scan-assembler "vupkhsw" } } */
+/* { dg-final { scan-assembler "vupklsw" } } */
--- /dev/null
+/* { dg-do compile { target { powerpc*-*-* } } } */
+/* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */
+/* { dg-require-effective-target powerpc_p8vector_ok } */
+/* { dg-options "-mcpu=power8 -O2 -ftree-vectorize -fvect-cost-model" } */
+
+#include <stddef.h>
+
+#ifndef SIZE
+#define SIZE 1024
+#endif
+
+#ifndef ALIGN
+#define ALIGN 32
+#endif
+
+#define ALIGN_ATTR __attribute__((__aligned__(ALIGN)))
+
+long long sign_ll[SIZE] ALIGN_ATTR;
+int sign_i [SIZE] ALIGN_ATTR;
+
+void copy_long_long_to_int (void)
+{
+ size_t i;
+
+ for (i = 0; i < SIZE; i++)
+ sign_i[i] = sign_ll[i];
+}
+
+/* { dg-final { scan-assembler "vpkudum" } } */
--- /dev/null
+/* { dg-do compile { target { powerpc*-*-* } } } */
+/* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */
+/* { dg-require-effective-target powerpc_p8vector_ok } */
+/* { dg-options "-mcpu=power8 -O2 -ftree-vectorize -fvect-cost-model -fno-unroll-loops -fno-unroll-all-loops" } */
+
+#ifndef SIZE
+#define SIZE 1024
+#endif
+
+#ifndef ALIGN
+#define ALIGN 32
+#endif
+
+#define ALIGN_ATTR __attribute__((__aligned__(ALIGN)))
+
+#define DO_BUILTIN(PREFIX, TYPE, CLZ, POPCNT) \
+TYPE PREFIX ## _a[SIZE] ALIGN_ATTR; \
+TYPE PREFIX ## _b[SIZE] ALIGN_ATTR; \
+ \
+void \
+PREFIX ## _clz (void) \
+{ \
+ unsigned long i; \
+ \
+ for (i = 0; i < SIZE; i++) \
+ PREFIX ## _a[i] = CLZ (PREFIX ## _b[i]); \
+} \
+ \
+void \
+PREFIX ## _popcnt (void) \
+{ \
+ unsigned long i; \
+ \
+ for (i = 0; i < SIZE; i++) \
+ PREFIX ## _a[i] = POPCNT (PREFIX ## _b[i]); \
+}
+
+#if !defined(DO_LONG_LONG) && !defined(DO_LONG) && !defined(DO_INT) && !defined(DO_SHORT) && !defined(DO_CHAR)
+#define DO_INT 1
+#endif
+
+#if DO_LONG_LONG
+/* At the moment, only int is auto vectorized. */
+DO_BUILTIN (sll, long long, __builtin_clzll, __builtin_popcountll)
+DO_BUILTIN (ull, unsigned long long, __builtin_clzll, __builtin_popcountll)
+#endif
+
+#if defined(_ARCH_PPC64) && DO_LONG
+DO_BUILTIN (sl, long, __builtin_clzl, __builtin_popcountl)
+DO_BUILTIN (ul, unsigned long, __builtin_clzl, __builtin_popcountl)
+#endif
+
+#if DO_INT
+DO_BUILTIN (si, int, __builtin_clz, __builtin_popcount)
+DO_BUILTIN (ui, unsigned int, __builtin_clz, __builtin_popcount)
+#endif
+
+#if DO_SHORT
+DO_BUILTIN (ss, short, __builtin_clz, __builtin_popcount)
+DO_BUILTIN (us, unsigned short, __builtin_clz, __builtin_popcount)
+#endif
+
+#if DO_CHAR
+DO_BUILTIN (sc, signed char, __builtin_clz, __builtin_popcount)
+DO_BUILTIN (uc, unsigned char, __builtin_clz, __builtin_popcount)
+#endif
+
+/* { dg-final { scan-assembler-times "vclzw" 2 } } */
+/* { dg-final { scan-assembler-times "vpopcntw" 2 } } */
--- /dev/null
+/* { dg-do compile { target { powerpc*-*-* } } } */
+/* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */
+/* { dg-require-effective-target powerpc_p8vector_ok } */
+/* { dg-options "-mcpu=power8 -O2 -ftree-vectorize -fvect-cost-model -fno-unroll-loops -fno-unroll-all-loops" } */
+
+#ifndef SIZE
+#define SIZE 1024
+#endif
+
+#ifndef ALIGN
+#define ALIGN 32
+#endif
+
+#ifndef ATTR_ALIGN
+#define ATTR_ALIGN __attribute__((__aligned__(ALIGN)))
+#endif
+
+#ifndef TYPE
+#define TYPE unsigned int
+#endif
+
+TYPE in1 [SIZE] ATTR_ALIGN;
+TYPE in2 [SIZE] ATTR_ALIGN;
+TYPE eqv [SIZE] ATTR_ALIGN;
+TYPE nand1[SIZE] ATTR_ALIGN;
+TYPE nand2[SIZE] ATTR_ALIGN;
+TYPE orc1 [SIZE] ATTR_ALIGN;
+TYPE orc2 [SIZE] ATTR_ALIGN;
+
+void
+do_eqv (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ {
+ eqv[i] = ~(in1[i] ^ in2[i]);
+ }
+}
+
+void
+do_nand1 (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ {
+ nand1[i] = ~(in1[i] & in2[i]);
+ }
+}
+
+void
+do_nand2 (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ {
+ nand2[i] = (~in1[i]) | (~in2[i]);
+ }
+}
+
+void
+do_orc1 (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ {
+ orc1[i] = (~in1[i]) | in2[i];
+ }
+}
+
+void
+do_orc2 (void)
+{
+ unsigned long i;
+
+ for (i = 0; i < SIZE; i++)
+ {
+ orc1[i] = in1[i] | (~in2[i]);
+ }
+}
+
+/* { dg-final { scan-assembler-times "xxleqv" 1 } } */
+/* { dg-final { scan-assembler-times "xxlnand" 2 } } */
+/* { dg-final { scan-assembler-times "xxlorc" 2 } } */
projects / thirdparty / gcc.git / commitdiff
