https://gcc.gnu.org/g:c0e167041b054400959aa8b3923df76934197fec
commit r17-5791-gc0e167041b054400959aa8b3923df76934197fec Author: Pan Li <[email protected]> Date: Fri Oct 30 16:20:05 2026 +0800 RISC-V: Add tests for no reg overlap of fp widen ternary [PR target/128162] Add test cases for the register group overlap of the floating-point widening ternary insns, to ensure the vs1 and vs2 will not fall inside the vd register group, as the vd is also read as the accumulator with a wider EEW. The test cases ask for the overlap by taking a piece of the vd register group as the vs1 or vs2, and cover below insns. vfwmacc.vv, vfwmsac.vv, vfwnmacc.vv, vfwnmsac.vv vfwmacc.vf, vfwmsac.vf, vfwnmacc.vf, vfwnmsac.vf vfwmaccbf16.vv, vfwmaccbf16.vf PR target/128162 gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vf-bf16-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vf-bf16-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vf-bf16-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vv-bf16-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vv-bf16-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vv-bf16-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f16-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f16-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f16-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f32-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f32-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f32-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f16-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f16-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f16-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f32-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f32-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f32-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f16-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f16-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f16-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f32-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f32-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f32-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f16-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f16-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f16-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f32-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f32-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f32-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f16-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f16-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f16-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f32-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f32-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f32-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f16-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f16-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f16-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f32-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f32-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f32-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f16-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f16-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f16-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f32-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f32-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f32-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f16-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f16-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f16-m4.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f32-m1.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f32-m2.c: New test. * gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f32-m4.c: New test. Signed-off-by: Pan Li <[email protected]> Diff: --- .../rvv/autovec/group_overlap/vfwmacc_vf-f16-m1.c | 29 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmacc_vf-f16-m2.c | 29 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmacc_vf-f16-m4.c | 29 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmacc_vf-f32-m1.c | 29 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmacc_vf-f32-m2.c | 29 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmacc_vf-f32-m4.c | 29 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmacc_vv-f16-m1.c | 32 ++++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmacc_vv-f16-m2.c | 32 ++++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmacc_vv-f16-m4.c | 32 ++++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmacc_vv-f32-m1.c | 32 ++++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmacc_vv-f32-m2.c | 32 ++++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmacc_vv-f32-m4.c | 32 ++++++++++++++++++++++ .../autovec/group_overlap/vfwmaccbf16_vf-bf16-m1.c | 29 ++++++++++++++++++++ .../autovec/group_overlap/vfwmaccbf16_vf-bf16-m2.c | 29 ++++++++++++++++++++ .../autovec/group_overlap/vfwmaccbf16_vf-bf16-m4.c | 29 ++++++++++++++++++++ .../autovec/group_overlap/vfwmaccbf16_vv-bf16-m1.c | 32 ++++++++++++++++++++++ .../autovec/group_overlap/vfwmaccbf16_vv-bf16-m2.c | 32 ++++++++++++++++++++++ .../autovec/group_overlap/vfwmaccbf16_vv-bf16-m4.c | 32 ++++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmsac_vf-f16-m1.c | 29 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmsac_vf-f16-m2.c | 29 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmsac_vf-f16-m4.c | 29 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmsac_vf-f32-m1.c | 29 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmsac_vf-f32-m2.c | 29 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmsac_vf-f32-m4.c | 29 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmsac_vv-f16-m1.c | 32 ++++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmsac_vv-f16-m2.c | 32 ++++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmsac_vv-f16-m4.c | 32 ++++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmsac_vv-f32-m1.c | 32 ++++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmsac_vv-f32-m2.c | 32 ++++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwmsac_vv-f32-m4.c | 32 ++++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmacc_vf-f16-m1.c | 28 +++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmacc_vf-f16-m2.c | 28 +++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmacc_vf-f16-m4.c | 28 +++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmacc_vf-f32-m1.c | 28 +++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmacc_vf-f32-m2.c | 28 +++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmacc_vf-f32-m4.c | 28 +++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmacc_vv-f16-m1.c | 30 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmacc_vv-f16-m2.c | 30 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmacc_vv-f16-m4.c | 30 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmacc_vv-f32-m1.c | 30 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmacc_vv-f32-m2.c | 30 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmacc_vv-f32-m4.c | 30 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmsac_vf-f16-m1.c | 28 +++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmsac_vf-f16-m2.c | 28 +++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmsac_vf-f16-m4.c | 28 +++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmsac_vf-f32-m1.c | 28 +++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmsac_vf-f32-m2.c | 28 +++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmsac_vf-f32-m4.c | 28 +++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmsac_vv-f16-m1.c | 30 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmsac_vv-f16-m2.c | 30 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmsac_vv-f16-m4.c | 30 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmsac_vv-f32-m1.c | 30 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmsac_vv-f32-m2.c | 30 ++++++++++++++++++++ .../rvv/autovec/group_overlap/vfwnmsac_vv-f32-m4.c | 30 ++++++++++++++++++++ 54 files changed, 1611 insertions(+) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f16-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f16-m1.c new file mode 100644 index 000000000000..6bf00acdfd5a --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f16-m1.c @@ -0,0 +1,29 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m2_f16m2 ( \ + __riscv_vreinterpret_v_u32m2_u16m2 (__riscv_vreinterpret_v_f32m2_u32m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e16m1, + vfloat16m1_t, + vfloat32m2_t, + vfloat16m2_t, + _Float16, + __riscv_vle32_v_f32m2, + RI_F, + __riscv_vget_v_f16m2_f16m1, + 1, + __riscv_vfwmacc_vf_f32m2, + __riscv_vse32_v_f32m2, + vfwmacc_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmacc\.vf\s+v6,fa0,v5([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmacc\.vf\s+v4,fa0,v3([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v6,[a-z0-9]+,v(6|7)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f16-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f16-m2.c new file mode 100644 index 000000000000..e4c553e08ce3 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f16-m2.c @@ -0,0 +1,29 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m4_f16m4 ( \ + __riscv_vreinterpret_v_u32m4_u16m4 (__riscv_vreinterpret_v_f32m4_u32m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e16m2, + vfloat16m2_t, + vfloat32m4_t, + vfloat16m4_t, + _Float16, + __riscv_vle32_v_f32m4, + RI_F, + __riscv_vget_v_f16m4_f16m2, + 1, + __riscv_vfwmacc_vf_f32m4, + __riscv_vse32_v_f32m4, + vfwmacc_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmacc\.vf\s+v12,fa0,v10([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmacc\.vf\s+v8,fa0,v6([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v12,[a-z0-9]+,v(12|13|14|15)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f16-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f16-m4.c new file mode 100644 index 000000000000..38e109ed5b78 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f16-m4.c @@ -0,0 +1,29 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m8_f16m8 ( \ + __riscv_vreinterpret_v_u32m8_u16m8 (__riscv_vreinterpret_v_f32m8_u32m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e16m4, + vfloat16m4_t, + vfloat32m8_t, + vfloat16m8_t, + _Float16, + __riscv_vle32_v_f32m8, + RI_F, + __riscv_vget_v_f16m8_f16m4, + 1, + __riscv_vfwmacc_vf_f32m8, + __riscv_vse32_v_f32m8, + vfwmacc_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmacc\.vf\s+v24,fa0,v20([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmacc\.vf\s+v16,fa0,v12([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v24,[a-z0-9]+,v(24|25|26|27|28|29|30|31)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f32-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f32-m1.c new file mode 100644 index 000000000000..b0a48643f776 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f32-m1.c @@ -0,0 +1,29 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m2_f32m2 ( \ + __riscv_vreinterpret_v_u64m2_u32m2 (__riscv_vreinterpret_v_f64m2_u64m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e32m1, + vfloat32m1_t, + vfloat64m2_t, + vfloat32m2_t, + float, + __riscv_vle64_v_f64m2, + RI_F, + __riscv_vget_v_f32m2_f32m1, + 1, + __riscv_vfwmacc_vf_f64m2, + __riscv_vse64_v_f64m2, + vfwmacc_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmacc\.vf\s+v6,fa0,v5([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmacc\.vf\s+v4,fa0,v3([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v6,[a-z0-9]+,v(6|7)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f32-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f32-m2.c new file mode 100644 index 000000000000..e990238cbbbe --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f32-m2.c @@ -0,0 +1,29 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m4_f32m4 ( \ + __riscv_vreinterpret_v_u64m4_u32m4 (__riscv_vreinterpret_v_f64m4_u64m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e32m2, + vfloat32m2_t, + vfloat64m4_t, + vfloat32m4_t, + float, + __riscv_vle64_v_f64m4, + RI_F, + __riscv_vget_v_f32m4_f32m2, + 1, + __riscv_vfwmacc_vf_f64m4, + __riscv_vse64_v_f64m4, + vfwmacc_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmacc\.vf\s+v12,fa0,v10([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmacc\.vf\s+v8,fa0,v6([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v12,[a-z0-9]+,v(12|13|14|15)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f32-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f32-m4.c new file mode 100644 index 000000000000..35a3f8ba1e01 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vf-f32-m4.c @@ -0,0 +1,29 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m8_f32m8 ( \ + __riscv_vreinterpret_v_u64m8_u32m8 (__riscv_vreinterpret_v_f64m8_u64m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e32m4, + vfloat32m4_t, + vfloat64m8_t, + vfloat32m8_t, + float, + __riscv_vle64_v_f64m8, + RI_F, + __riscv_vget_v_f32m8_f32m4, + 1, + __riscv_vfwmacc_vf_f64m8, + __riscv_vse64_v_f64m8, + vfwmacc_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmacc\.vf\s+v24,fa0,v20([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmacc\.vf\s+v16,fa0,v12([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v24,[a-z0-9]+,v(24|25|26|27|28|29|30|31)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vf\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f16-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f16-m1.c new file mode 100644 index 000000000000..4d38faa8fdc2 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f16-m1.c @@ -0,0 +1,32 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m2_f16m2 ( \ + __riscv_vreinterpret_v_u32m2_u16m2 (__riscv_vreinterpret_v_f32m2_u32m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e16m1, + vfloat16m1_t, + vfloat32m2_t, + vfloat16m2_t, + __riscv_vle16_v_f16m1, + __riscv_vle32_v_f32m2, + RI_F, + __riscv_vget_v_f16m2_f16m1, + 1, + __riscv_vfwmacc_vv_f32m2, + __riscv_vse32_v_f32m2, + vfwmacc_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmacc\.vv\s+v6,v5,v8([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmacc\.vv\s+v4,v1,v3([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v6,v(6|7),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v6,[a-z0-9]+,v(6|7)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v4,v(4|5),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v2,v(2|3),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f16-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f16-m2.c new file mode 100644 index 000000000000..e01b862b4414 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f16-m2.c @@ -0,0 +1,32 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m4_f16m4 ( \ + __riscv_vreinterpret_v_u32m4_u16m4 (__riscv_vreinterpret_v_f32m4_u32m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e16m2, + vfloat16m2_t, + vfloat32m4_t, + vfloat16m4_t, + __riscv_vle16_v_f16m2, + __riscv_vle32_v_f32m4, + RI_F, + __riscv_vget_v_f16m4_f16m2, + 1, + __riscv_vfwmacc_vv_f32m4, + __riscv_vse32_v_f32m4, + vfwmacc_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmacc\.vv\s+v12,v10,v16([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmacc\.vv\s+v8,v2,v6([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v12,v(12|13|14|15),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v12,[a-z0-9]+,v(12|13|14|15)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v8,v(8|9|10|11),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v4,v(4|5|6|7),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f16-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f16-m4.c new file mode 100644 index 000000000000..e81e0a08f14e --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f16-m4.c @@ -0,0 +1,32 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m8_f16m8 ( \ + __riscv_vreinterpret_v_u32m8_u16m8 (__riscv_vreinterpret_v_f32m8_u32m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e16m4, + vfloat16m4_t, + vfloat32m8_t, + vfloat16m8_t, + __riscv_vle16_v_f16m4, + __riscv_vle32_v_f32m8, + RI_F, + __riscv_vget_v_f16m8_f16m4, + 1, + __riscv_vfwmacc_vv_f32m8, + __riscv_vse32_v_f32m8, + vfwmacc_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmacc\.vv\s+v24,v20,v0([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmacc\.vv\s+v16,v4,v12([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v24,v(24|25|26|27|28|29|30|31),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v24,[a-z0-9]+,v(24|25|26|27|28|29|30|31)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v16,v(16|17|18|19|20|21|22|23),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v8,v(8|9|10|11|12|13|14|15),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f32-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f32-m1.c new file mode 100644 index 000000000000..b68285e0ca4b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f32-m1.c @@ -0,0 +1,32 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m2_f32m2 ( \ + __riscv_vreinterpret_v_u64m2_u32m2 (__riscv_vreinterpret_v_f64m2_u64m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e32m1, + vfloat32m1_t, + vfloat64m2_t, + vfloat32m2_t, + __riscv_vle32_v_f32m1, + __riscv_vle64_v_f64m2, + RI_F, + __riscv_vget_v_f32m2_f32m1, + 1, + __riscv_vfwmacc_vv_f64m2, + __riscv_vse64_v_f64m2, + vfwmacc_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmacc\.vv\s+v6,v5,v8([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmacc\.vv\s+v4,v1,v3([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v6,v(6|7),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v6,[a-z0-9]+,v(6|7)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v4,v(4|5),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v2,v(2|3),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f32-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f32-m2.c new file mode 100644 index 000000000000..e8a628979a67 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f32-m2.c @@ -0,0 +1,32 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m4_f32m4 ( \ + __riscv_vreinterpret_v_u64m4_u32m4 (__riscv_vreinterpret_v_f64m4_u64m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e32m2, + vfloat32m2_t, + vfloat64m4_t, + vfloat32m4_t, + __riscv_vle32_v_f32m2, + __riscv_vle64_v_f64m4, + RI_F, + __riscv_vget_v_f32m4_f32m2, + 1, + __riscv_vfwmacc_vv_f64m4, + __riscv_vse64_v_f64m4, + vfwmacc_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmacc\.vv\s+v12,v10,v16([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmacc\.vv\s+v8,v2,v6([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v12,v(12|13|14|15),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v12,[a-z0-9]+,v(12|13|14|15)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v8,v(8|9|10|11),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v4,v(4|5|6|7),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f32-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f32-m4.c new file mode 100644 index 000000000000..c5063417d019 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmacc_vv-f32-m4.c @@ -0,0 +1,32 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m8_f32m8 ( \ + __riscv_vreinterpret_v_u64m8_u32m8 (__riscv_vreinterpret_v_f64m8_u64m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e32m4, + vfloat32m4_t, + vfloat64m8_t, + vfloat32m8_t, + __riscv_vle32_v_f32m4, + __riscv_vle64_v_f64m8, + RI_F, + __riscv_vget_v_f32m8_f32m4, + 1, + __riscv_vfwmacc_vv_f64m8, + __riscv_vse64_v_f64m8, + vfwmacc_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmacc\.vv\s+v24,v20,v0([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmacc\.vv\s+v16,v4,v12([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v24,v(24|25|26|27|28|29|30|31),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v24,[a-z0-9]+,v(24|25|26|27|28|29|30|31)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v16,v(16|17|18|19|20|21|22|23),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v8,v(8|9|10|11|12|13|14|15),} } } */ +/* { dg-final { scan-assembler-not {vfwmacc\.vv\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vf-bf16-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vf-bf16-m1.c new file mode 100644 index 000000000000..16bf30fea642 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vf-bf16-m1.c @@ -0,0 +1,29 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfbfwma -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m2_bf16m2 ( \ + __riscv_vreinterpret_v_u32m2_u16m2 (__riscv_vreinterpret_v_f32m2_u32m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e16m1, + vbfloat16m1_t, + vfloat32m2_t, + vbfloat16m2_t, + __bf16, + __riscv_vle32_v_f32m2, + RI_F, + __riscv_vget_v_bf16m2_bf16m1, + 1, + __riscv_vfwmaccbf16_vf_f32m2, + __riscv_vse32_v_f32m2, + vfwmaccbf16_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmaccbf16\.vf\s+v6,fa0,v5([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmaccbf16\.vf\s+v4,fa0,v3([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vf\s+v6,[a-z0-9]+,v(6|7)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vf\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vf\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vf-bf16-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vf-bf16-m2.c new file mode 100644 index 000000000000..755cf0da7514 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vf-bf16-m2.c @@ -0,0 +1,29 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfbfwma -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m4_bf16m4 ( \ + __riscv_vreinterpret_v_u32m4_u16m4 (__riscv_vreinterpret_v_f32m4_u32m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e16m2, + vbfloat16m2_t, + vfloat32m4_t, + vbfloat16m4_t, + __bf16, + __riscv_vle32_v_f32m4, + RI_F, + __riscv_vget_v_bf16m4_bf16m2, + 1, + __riscv_vfwmaccbf16_vf_f32m4, + __riscv_vse32_v_f32m4, + vfwmaccbf16_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmaccbf16\.vf\s+v12,fa0,v10([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmaccbf16\.vf\s+v8,fa0,v6([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vf\s+v12,[a-z0-9]+,v(12|13|14|15)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vf\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vf\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vf-bf16-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vf-bf16-m4.c new file mode 100644 index 000000000000..23584141c4fa --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vf-bf16-m4.c @@ -0,0 +1,29 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfbfwma -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m8_bf16m8 ( \ + __riscv_vreinterpret_v_u32m8_u16m8 (__riscv_vreinterpret_v_f32m8_u32m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e16m4, + vbfloat16m4_t, + vfloat32m8_t, + vbfloat16m8_t, + __bf16, + __riscv_vle32_v_f32m8, + RI_F, + __riscv_vget_v_bf16m8_bf16m4, + 1, + __riscv_vfwmaccbf16_vf_f32m8, + __riscv_vse32_v_f32m8, + vfwmaccbf16_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmaccbf16\.vf\s+v24,fa0,v20([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmaccbf16\.vf\s+v16,fa0,v12([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vf\s+v24,[a-z0-9]+,v(24|25|26|27|28|29|30|31)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vf\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vf\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vv-bf16-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vv-bf16-m1.c new file mode 100644 index 000000000000..5877a969f972 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vv-bf16-m1.c @@ -0,0 +1,32 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfbfwma -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m2_bf16m2 ( \ + __riscv_vreinterpret_v_u32m2_u16m2 (__riscv_vreinterpret_v_f32m2_u32m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e16m1, + vbfloat16m1_t, + vfloat32m2_t, + vbfloat16m2_t, + __riscv_vle16_v_bf16m1, + __riscv_vle32_v_f32m2, + RI_F, + __riscv_vget_v_bf16m2_bf16m1, + 1, + __riscv_vfwmaccbf16_vv_f32m2, + __riscv_vse32_v_f32m2, + vfwmaccbf16_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmaccbf16\.vv\s+v6,v5,v8([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmaccbf16\.vv\s+v4,v1,v3([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v6,v(6|7),} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v6,[a-z0-9]+,v(6|7)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v4,v(4|5),} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v2,v(2|3),} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vv-bf16-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vv-bf16-m2.c new file mode 100644 index 000000000000..673cc34dfb8d --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vv-bf16-m2.c @@ -0,0 +1,32 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfbfwma -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m4_bf16m4 ( \ + __riscv_vreinterpret_v_u32m4_u16m4 (__riscv_vreinterpret_v_f32m4_u32m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e16m2, + vbfloat16m2_t, + vfloat32m4_t, + vbfloat16m4_t, + __riscv_vle16_v_bf16m2, + __riscv_vle32_v_f32m4, + RI_F, + __riscv_vget_v_bf16m4_bf16m2, + 1, + __riscv_vfwmaccbf16_vv_f32m4, + __riscv_vse32_v_f32m4, + vfwmaccbf16_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmaccbf16\.vv\s+v12,v10,v16([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmaccbf16\.vv\s+v8,v2,v6([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v12,v(12|13|14|15),} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v12,[a-z0-9]+,v(12|13|14|15)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v8,v(8|9|10|11),} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v4,v(4|5|6|7),} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vv-bf16-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vv-bf16-m4.c new file mode 100644 index 000000000000..22bb2c504d07 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmaccbf16_vv-bf16-m4.c @@ -0,0 +1,32 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfbfwma -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m8_bf16m8 ( \ + __riscv_vreinterpret_v_u32m8_u16m8 (__riscv_vreinterpret_v_f32m8_u32m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e16m4, + vbfloat16m4_t, + vfloat32m8_t, + vbfloat16m8_t, + __riscv_vle16_v_bf16m4, + __riscv_vle32_v_f32m8, + RI_F, + __riscv_vget_v_bf16m8_bf16m4, + 1, + __riscv_vfwmaccbf16_vv_f32m8, + __riscv_vse32_v_f32m8, + vfwmaccbf16_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmaccbf16\.vv\s+v24,v20,v0([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmaccbf16\.vv\s+v16,v4,v12([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v24,v(24|25|26|27|28|29|30|31),} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v24,[a-z0-9]+,v(24|25|26|27|28|29|30|31)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v16,v(16|17|18|19|20|21|22|23),} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v8,v(8|9|10|11|12|13|14|15),} } } */ +/* { dg-final { scan-assembler-not {vfwmaccbf16\.vv\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f16-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f16-m1.c new file mode 100644 index 000000000000..d426cfddd649 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f16-m1.c @@ -0,0 +1,29 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m2_f16m2 ( \ + __riscv_vreinterpret_v_u32m2_u16m2 (__riscv_vreinterpret_v_f32m2_u32m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e16m1, + vfloat16m1_t, + vfloat32m2_t, + vfloat16m2_t, + _Float16, + __riscv_vle32_v_f32m2, + RI_F, + __riscv_vget_v_f16m2_f16m1, + 1, + __riscv_vfwmsac_vf_f32m2, + __riscv_vse32_v_f32m2, + vfwmsac_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmsac\.vf\s+v6,fa0,v5([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmsac\.vf\s+v4,fa0,v3([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v6,[a-z0-9]+,v(6|7)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f16-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f16-m2.c new file mode 100644 index 000000000000..f58011b2ea2e --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f16-m2.c @@ -0,0 +1,29 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m4_f16m4 ( \ + __riscv_vreinterpret_v_u32m4_u16m4 (__riscv_vreinterpret_v_f32m4_u32m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e16m2, + vfloat16m2_t, + vfloat32m4_t, + vfloat16m4_t, + _Float16, + __riscv_vle32_v_f32m4, + RI_F, + __riscv_vget_v_f16m4_f16m2, + 1, + __riscv_vfwmsac_vf_f32m4, + __riscv_vse32_v_f32m4, + vfwmsac_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmsac\.vf\s+v12,fa0,v10([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmsac\.vf\s+v8,fa0,v6([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v12,[a-z0-9]+,v(12|13|14|15)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f16-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f16-m4.c new file mode 100644 index 000000000000..8c03ae857204 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f16-m4.c @@ -0,0 +1,29 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m8_f16m8 ( \ + __riscv_vreinterpret_v_u32m8_u16m8 (__riscv_vreinterpret_v_f32m8_u32m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e16m4, + vfloat16m4_t, + vfloat32m8_t, + vfloat16m8_t, + _Float16, + __riscv_vle32_v_f32m8, + RI_F, + __riscv_vget_v_f16m8_f16m4, + 1, + __riscv_vfwmsac_vf_f32m8, + __riscv_vse32_v_f32m8, + vfwmsac_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmsac\.vf\s+v24,fa0,v20([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmsac\.vf\s+v16,fa0,v12([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v24,[a-z0-9]+,v(24|25|26|27|28|29|30|31)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f32-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f32-m1.c new file mode 100644 index 000000000000..931f13b03aa0 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f32-m1.c @@ -0,0 +1,29 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m2_f32m2 ( \ + __riscv_vreinterpret_v_u64m2_u32m2 (__riscv_vreinterpret_v_f64m2_u64m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e32m1, + vfloat32m1_t, + vfloat64m2_t, + vfloat32m2_t, + float, + __riscv_vle64_v_f64m2, + RI_F, + __riscv_vget_v_f32m2_f32m1, + 1, + __riscv_vfwmsac_vf_f64m2, + __riscv_vse64_v_f64m2, + vfwmsac_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmsac\.vf\s+v6,fa0,v5([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmsac\.vf\s+v4,fa0,v3([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v6,[a-z0-9]+,v(6|7)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f32-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f32-m2.c new file mode 100644 index 000000000000..435171867008 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f32-m2.c @@ -0,0 +1,29 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m4_f32m4 ( \ + __riscv_vreinterpret_v_u64m4_u32m4 (__riscv_vreinterpret_v_f64m4_u64m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e32m2, + vfloat32m2_t, + vfloat64m4_t, + vfloat32m4_t, + float, + __riscv_vle64_v_f64m4, + RI_F, + __riscv_vget_v_f32m4_f32m2, + 1, + __riscv_vfwmsac_vf_f64m4, + __riscv_vse64_v_f64m4, + vfwmsac_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmsac\.vf\s+v12,fa0,v10([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmsac\.vf\s+v8,fa0,v6([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v12,[a-z0-9]+,v(12|13|14|15)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f32-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f32-m4.c new file mode 100644 index 000000000000..d11513e5de53 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vf-f32-m4.c @@ -0,0 +1,29 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m8_f32m8 ( \ + __riscv_vreinterpret_v_u64m8_u32m8 (__riscv_vreinterpret_v_f64m8_u64m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e32m4, + vfloat32m4_t, + vfloat64m8_t, + vfloat32m8_t, + float, + __riscv_vle64_v_f64m8, + RI_F, + __riscv_vget_v_f32m8_f32m4, + 1, + __riscv_vfwmsac_vf_f64m8, + __riscv_vse64_v_f64m8, + vfwmsac_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmsac\.vf\s+v24,fa0,v20([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmsac\.vf\s+v16,fa0,v12([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v24,[a-z0-9]+,v(24|25|26|27|28|29|30|31)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vf\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f16-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f16-m1.c new file mode 100644 index 000000000000..e984236ef721 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f16-m1.c @@ -0,0 +1,32 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m2_f16m2 ( \ + __riscv_vreinterpret_v_u32m2_u16m2 (__riscv_vreinterpret_v_f32m2_u32m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e16m1, + vfloat16m1_t, + vfloat32m2_t, + vfloat16m2_t, + __riscv_vle16_v_f16m1, + __riscv_vle32_v_f32m2, + RI_F, + __riscv_vget_v_f16m2_f16m1, + 1, + __riscv_vfwmsac_vv_f32m2, + __riscv_vse32_v_f32m2, + vfwmsac_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmsac\.vv\s+v6,v5,v8([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmsac\.vv\s+v4,v1,v3([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v6,v(6|7),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v6,[a-z0-9]+,v(6|7)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v4,v(4|5),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v2,v(2|3),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f16-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f16-m2.c new file mode 100644 index 000000000000..5ae77869b199 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f16-m2.c @@ -0,0 +1,32 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m4_f16m4 ( \ + __riscv_vreinterpret_v_u32m4_u16m4 (__riscv_vreinterpret_v_f32m4_u32m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e16m2, + vfloat16m2_t, + vfloat32m4_t, + vfloat16m4_t, + __riscv_vle16_v_f16m2, + __riscv_vle32_v_f32m4, + RI_F, + __riscv_vget_v_f16m4_f16m2, + 1, + __riscv_vfwmsac_vv_f32m4, + __riscv_vse32_v_f32m4, + vfwmsac_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmsac\.vv\s+v12,v10,v16([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmsac\.vv\s+v8,v2,v6([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v12,v(12|13|14|15),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v12,[a-z0-9]+,v(12|13|14|15)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v8,v(8|9|10|11),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v4,v(4|5|6|7),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f16-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f16-m4.c new file mode 100644 index 000000000000..834498302276 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f16-m4.c @@ -0,0 +1,32 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m8_f16m8 ( \ + __riscv_vreinterpret_v_u32m8_u16m8 (__riscv_vreinterpret_v_f32m8_u32m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e16m4, + vfloat16m4_t, + vfloat32m8_t, + vfloat16m8_t, + __riscv_vle16_v_f16m4, + __riscv_vle32_v_f32m8, + RI_F, + __riscv_vget_v_f16m8_f16m4, + 1, + __riscv_vfwmsac_vv_f32m8, + __riscv_vse32_v_f32m8, + vfwmsac_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmsac\.vv\s+v24,v20,v0([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmsac\.vv\s+v16,v4,v12([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v24,v(24|25|26|27|28|29|30|31),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v24,[a-z0-9]+,v(24|25|26|27|28|29|30|31)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v16,v(16|17|18|19|20|21|22|23),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v8,v(8|9|10|11|12|13|14|15),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f32-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f32-m1.c new file mode 100644 index 000000000000..e667042a6629 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f32-m1.c @@ -0,0 +1,32 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m2_f32m2 ( \ + __riscv_vreinterpret_v_u64m2_u32m2 (__riscv_vreinterpret_v_f64m2_u64m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e32m1, + vfloat32m1_t, + vfloat64m2_t, + vfloat32m2_t, + __riscv_vle32_v_f32m1, + __riscv_vle64_v_f64m2, + RI_F, + __riscv_vget_v_f32m2_f32m1, + 1, + __riscv_vfwmsac_vv_f64m2, + __riscv_vse64_v_f64m2, + vfwmsac_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmsac\.vv\s+v6,v5,v8([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmsac\.vv\s+v4,v1,v3([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v6,v(6|7),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v6,[a-z0-9]+,v(6|7)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v4,v(4|5),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v2,v(2|3),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f32-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f32-m2.c new file mode 100644 index 000000000000..08bb2b74a9f7 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f32-m2.c @@ -0,0 +1,32 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m4_f32m4 ( \ + __riscv_vreinterpret_v_u64m4_u32m4 (__riscv_vreinterpret_v_f64m4_u64m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e32m2, + vfloat32m2_t, + vfloat64m4_t, + vfloat32m4_t, + __riscv_vle32_v_f32m2, + __riscv_vle64_v_f64m4, + RI_F, + __riscv_vget_v_f32m4_f32m2, + 1, + __riscv_vfwmsac_vv_f64m4, + __riscv_vse64_v_f64m4, + vfwmsac_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmsac\.vv\s+v12,v10,v16([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmsac\.vv\s+v8,v2,v6([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v12,v(12|13|14|15),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v12,[a-z0-9]+,v(12|13|14|15)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v8,v(8|9|10|11),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v4,v(4|5|6|7),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f32-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f32-m4.c new file mode 100644 index 000000000000..ab50b7a1376a --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwmsac_vv-f32-m4.c @@ -0,0 +1,32 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m8_f32m8 ( \ + __riscv_vreinterpret_v_u64m8_u32m8 (__riscv_vreinterpret_v_f64m8_u64m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e32m4, + vfloat32m4_t, + vfloat64m8_t, + vfloat32m8_t, + __riscv_vle32_v_f32m4, + __riscv_vle64_v_f64m8, + RI_F, + __riscv_vget_v_f32m8_f32m4, + 1, + __riscv_vfwmsac_vv_f64m8, + __riscv_vse64_v_f64m8, + vfwmsac_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwmsac\.vv\s+v24,v20,v0([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwmsac\.vv\s+v16,v4,v12([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v24,v(24|25|26|27|28|29|30|31),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v24,[a-z0-9]+,v(24|25|26|27|28|29|30|31)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v16,v(16|17|18|19|20|21|22|23),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v8,v(8|9|10|11|12|13|14|15),} } } */ +/* { dg-final { scan-assembler-not {vfwmsac\.vv\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f16-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f16-m1.c new file mode 100644 index 000000000000..a3a0ee162321 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f16-m1.c @@ -0,0 +1,28 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m2_f16m2 ( \ + __riscv_vreinterpret_v_u32m2_u16m2 (__riscv_vreinterpret_v_f32m2_u32m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e16m1, + vfloat16m1_t, + vfloat32m2_t, + vfloat16m2_t, + _Float16, + __riscv_vle32_v_f32m2, + RI_F, + __riscv_vget_v_f16m2_f16m1, + 1, + __riscv_vfwnmacc_vf_f32m2, + __riscv_vse32_v_f32m2, + vfwnmacc_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmacc\.vf\s+v4,fa0,v6([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmacc\.vf\s+v2,fa0,v1([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vf\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vf\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f16-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f16-m2.c new file mode 100644 index 000000000000..a94d138871ce --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f16-m2.c @@ -0,0 +1,28 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m4_f16m4 ( \ + __riscv_vreinterpret_v_u32m4_u16m4 (__riscv_vreinterpret_v_f32m4_u32m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e16m2, + vfloat16m2_t, + vfloat32m4_t, + vfloat16m4_t, + _Float16, + __riscv_vle32_v_f32m4, + RI_F, + __riscv_vget_v_f16m4_f16m2, + 1, + __riscv_vfwnmacc_vf_f32m4, + __riscv_vse32_v_f32m4, + vfwnmacc_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmacc\.vf\s+v8,fa0,v12([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmacc\.vf\s+v4,fa0,v2([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vf\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vf\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f16-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f16-m4.c new file mode 100644 index 000000000000..f3c29ec9272c --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f16-m4.c @@ -0,0 +1,28 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m8_f16m8 ( \ + __riscv_vreinterpret_v_u32m8_u16m8 (__riscv_vreinterpret_v_f32m8_u32m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e16m4, + vfloat16m4_t, + vfloat32m8_t, + vfloat16m8_t, + _Float16, + __riscv_vle32_v_f32m8, + RI_F, + __riscv_vget_v_f16m8_f16m4, + 1, + __riscv_vfwnmacc_vf_f32m8, + __riscv_vse32_v_f32m8, + vfwnmacc_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmacc\.vf\s+v16,fa0,v24([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmacc\.vf\s+v8,fa0,v4([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vf\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vf\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f32-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f32-m1.c new file mode 100644 index 000000000000..da0e0ee1cb41 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f32-m1.c @@ -0,0 +1,28 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m2_f32m2 ( \ + __riscv_vreinterpret_v_u64m2_u32m2 (__riscv_vreinterpret_v_f64m2_u64m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e32m1, + vfloat32m1_t, + vfloat64m2_t, + vfloat32m2_t, + float, + __riscv_vle64_v_f64m2, + RI_F, + __riscv_vget_v_f32m2_f32m1, + 1, + __riscv_vfwnmacc_vf_f64m2, + __riscv_vse64_v_f64m2, + vfwnmacc_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmacc\.vf\s+v4,fa0,v6([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmacc\.vf\s+v2,fa0,v1([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vf\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vf\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f32-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f32-m2.c new file mode 100644 index 000000000000..ef15aacaf1bd --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f32-m2.c @@ -0,0 +1,28 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m4_f32m4 ( \ + __riscv_vreinterpret_v_u64m4_u32m4 (__riscv_vreinterpret_v_f64m4_u64m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e32m2, + vfloat32m2_t, + vfloat64m4_t, + vfloat32m4_t, + float, + __riscv_vle64_v_f64m4, + RI_F, + __riscv_vget_v_f32m4_f32m2, + 1, + __riscv_vfwnmacc_vf_f64m4, + __riscv_vse64_v_f64m4, + vfwnmacc_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmacc\.vf\s+v8,fa0,v12([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmacc\.vf\s+v4,fa0,v2([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vf\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vf\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f32-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f32-m4.c new file mode 100644 index 000000000000..981cb49caef9 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vf-f32-m4.c @@ -0,0 +1,28 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m8_f32m8 ( \ + __riscv_vreinterpret_v_u64m8_u32m8 (__riscv_vreinterpret_v_f64m8_u64m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e32m4, + vfloat32m4_t, + vfloat64m8_t, + vfloat32m8_t, + float, + __riscv_vle64_v_f64m8, + RI_F, + __riscv_vget_v_f32m8_f32m4, + 1, + __riscv_vfwnmacc_vf_f64m8, + __riscv_vse64_v_f64m8, + vfwnmacc_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmacc\.vf\s+v16,fa0,v24([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmacc\.vf\s+v8,fa0,v4([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vf\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vf\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f16-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f16-m1.c new file mode 100644 index 000000000000..e85b8f82a687 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f16-m1.c @@ -0,0 +1,30 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m2_f16m2 ( \ + __riscv_vreinterpret_v_u32m2_u16m2 (__riscv_vreinterpret_v_f32m2_u32m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e16m1, + vfloat16m1_t, + vfloat32m2_t, + vfloat16m2_t, + __riscv_vle16_v_f16m1, + __riscv_vle32_v_f32m2, + RI_F, + __riscv_vget_v_f16m2_f16m1, + 1, + __riscv_vfwnmacc_vv_f32m2, + __riscv_vse32_v_f32m2, + vfwnmacc_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmacc\.vv\s+v4,v8,v7([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmacc\.vv\s+v2,v6,v1([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v4,v(4|5),} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v2,v(2|3),} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f16-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f16-m2.c new file mode 100644 index 000000000000..8e741ec232b6 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f16-m2.c @@ -0,0 +1,30 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m4_f16m4 ( \ + __riscv_vreinterpret_v_u32m4_u16m4 (__riscv_vreinterpret_v_f32m4_u32m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e16m2, + vfloat16m2_t, + vfloat32m4_t, + vfloat16m4_t, + __riscv_vle16_v_f16m2, + __riscv_vle32_v_f32m4, + RI_F, + __riscv_vget_v_f16m4_f16m2, + 1, + __riscv_vfwnmacc_vv_f32m4, + __riscv_vse32_v_f32m4, + vfwnmacc_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmacc\.vv\s+v8,v16,v14([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmacc\.vv\s+v4,v12,v2([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v8,v(8|9|10|11),} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v4,v(4|5|6|7),} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f16-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f16-m4.c new file mode 100644 index 000000000000..46450bc4c1a5 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f16-m4.c @@ -0,0 +1,30 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m8_f16m8 ( \ + __riscv_vreinterpret_v_u32m8_u16m8 (__riscv_vreinterpret_v_f32m8_u32m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e16m4, + vfloat16m4_t, + vfloat32m8_t, + vfloat16m8_t, + __riscv_vle16_v_f16m4, + __riscv_vle32_v_f32m8, + RI_F, + __riscv_vget_v_f16m8_f16m4, + 1, + __riscv_vfwnmacc_vv_f32m8, + __riscv_vse32_v_f32m8, + vfwnmacc_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmacc\.vv\s+v16,v0,v28([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmacc\.vv\s+v8,v24,v4([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v16,v(16|17|18|19|20|21|22|23),} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v8,v(8|9|10|11|12|13|14|15),} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f32-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f32-m1.c new file mode 100644 index 000000000000..408bc2e189c8 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f32-m1.c @@ -0,0 +1,30 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m2_f32m2 ( \ + __riscv_vreinterpret_v_u64m2_u32m2 (__riscv_vreinterpret_v_f64m2_u64m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e32m1, + vfloat32m1_t, + vfloat64m2_t, + vfloat32m2_t, + __riscv_vle32_v_f32m1, + __riscv_vle64_v_f64m2, + RI_F, + __riscv_vget_v_f32m2_f32m1, + 1, + __riscv_vfwnmacc_vv_f64m2, + __riscv_vse64_v_f64m2, + vfwnmacc_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmacc\.vv\s+v4,v8,v7([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmacc\.vv\s+v2,v6,v1([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v4,v(4|5),} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v2,v(2|3),} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f32-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f32-m2.c new file mode 100644 index 000000000000..af100f0283f9 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f32-m2.c @@ -0,0 +1,30 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m4_f32m4 ( \ + __riscv_vreinterpret_v_u64m4_u32m4 (__riscv_vreinterpret_v_f64m4_u64m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e32m2, + vfloat32m2_t, + vfloat64m4_t, + vfloat32m4_t, + __riscv_vle32_v_f32m2, + __riscv_vle64_v_f64m4, + RI_F, + __riscv_vget_v_f32m4_f32m2, + 1, + __riscv_vfwnmacc_vv_f64m4, + __riscv_vse64_v_f64m4, + vfwnmacc_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmacc\.vv\s+v8,v16,v14([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmacc\.vv\s+v4,v12,v2([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v8,v(8|9|10|11),} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v4,v(4|5|6|7),} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f32-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f32-m4.c new file mode 100644 index 000000000000..af8ebffd6589 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmacc_vv-f32-m4.c @@ -0,0 +1,30 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m8_f32m8 ( \ + __riscv_vreinterpret_v_u64m8_u32m8 (__riscv_vreinterpret_v_f64m8_u64m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e32m4, + vfloat32m4_t, + vfloat64m8_t, + vfloat32m8_t, + __riscv_vle32_v_f32m4, + __riscv_vle64_v_f64m8, + RI_F, + __riscv_vget_v_f32m8_f32m4, + 1, + __riscv_vfwnmacc_vv_f64m8, + __riscv_vse64_v_f64m8, + vfwnmacc_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmacc\.vv\s+v16,v0,v28([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmacc\.vv\s+v8,v24,v4([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v16,v(16|17|18|19|20|21|22|23),} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v8,v(8|9|10|11|12|13|14|15),} } } */ +/* { dg-final { scan-assembler-not {vfwnmacc\.vv\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f16-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f16-m1.c new file mode 100644 index 000000000000..14cbb83ba1cb --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f16-m1.c @@ -0,0 +1,28 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m2_f16m2 ( \ + __riscv_vreinterpret_v_u32m2_u16m2 (__riscv_vreinterpret_v_f32m2_u32m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e16m1, + vfloat16m1_t, + vfloat32m2_t, + vfloat16m2_t, + _Float16, + __riscv_vle32_v_f32m2, + RI_F, + __riscv_vget_v_f16m2_f16m1, + 1, + __riscv_vfwnmsac_vf_f32m2, + __riscv_vse32_v_f32m2, + vfwnmsac_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmsac\.vf\s+v4,fa0,v6([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmsac\.vf\s+v2,fa0,v1([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vf\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vf\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f16-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f16-m2.c new file mode 100644 index 000000000000..127bb123d8d8 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f16-m2.c @@ -0,0 +1,28 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m4_f16m4 ( \ + __riscv_vreinterpret_v_u32m4_u16m4 (__riscv_vreinterpret_v_f32m4_u32m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e16m2, + vfloat16m2_t, + vfloat32m4_t, + vfloat16m4_t, + _Float16, + __riscv_vle32_v_f32m4, + RI_F, + __riscv_vget_v_f16m4_f16m2, + 1, + __riscv_vfwnmsac_vf_f32m4, + __riscv_vse32_v_f32m4, + vfwnmsac_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmsac\.vf\s+v8,fa0,v12([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmsac\.vf\s+v4,fa0,v2([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vf\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vf\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f16-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f16-m4.c new file mode 100644 index 000000000000..dcbf3ddc6f32 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f16-m4.c @@ -0,0 +1,28 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m8_f16m8 ( \ + __riscv_vreinterpret_v_u32m8_u16m8 (__riscv_vreinterpret_v_f32m8_u32m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e16m4, + vfloat16m4_t, + vfloat32m8_t, + vfloat16m8_t, + _Float16, + __riscv_vle32_v_f32m8, + RI_F, + __riscv_vget_v_f16m8_f16m4, + 1, + __riscv_vfwnmsac_vf_f32m8, + __riscv_vse32_v_f32m8, + vfwnmsac_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmsac\.vf\s+v16,fa0,v24([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmsac\.vf\s+v8,fa0,v4([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vf\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vf\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f32-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f32-m1.c new file mode 100644 index 000000000000..2e8df3dd3618 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f32-m1.c @@ -0,0 +1,28 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m2_f32m2 ( \ + __riscv_vreinterpret_v_u64m2_u32m2 (__riscv_vreinterpret_v_f64m2_u64m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e32m1, + vfloat32m1_t, + vfloat64m2_t, + vfloat32m2_t, + float, + __riscv_vle64_v_f64m2, + RI_F, + __riscv_vget_v_f32m2_f32m1, + 1, + __riscv_vfwnmsac_vf_f64m2, + __riscv_vse64_v_f64m2, + vfwnmsac_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmsac\.vf\s+v4,fa0,v6([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmsac\.vf\s+v2,fa0,v1([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vf\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vf\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f32-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f32-m2.c new file mode 100644 index 000000000000..119997d72938 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f32-m2.c @@ -0,0 +1,28 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m4_f32m4 ( \ + __riscv_vreinterpret_v_u64m4_u32m4 (__riscv_vreinterpret_v_f64m4_u64m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e32m2, + vfloat32m2_t, + vfloat64m4_t, + vfloat32m4_t, + float, + __riscv_vle64_v_f64m4, + RI_F, + __riscv_vget_v_f32m4_f32m2, + 1, + __riscv_vfwnmsac_vf_f64m4, + __riscv_vse64_v_f64m4, + vfwnmsac_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmsac\.vf\s+v8,fa0,v12([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmsac\.vf\s+v4,fa0,v2([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vf\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vf\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f32-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f32-m4.c new file mode 100644 index 000000000000..12beee094f52 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vf-f32-m4.c @@ -0,0 +1,28 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m8_f32m8 ( \ + __riscv_vreinterpret_v_u64m8_u32m8 (__riscv_vreinterpret_v_f64m8_u64m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_2( + __riscv_vsetvlmax_e32m4, + vfloat32m4_t, + vfloat64m8_t, + vfloat32m8_t, + float, + __riscv_vle64_v_f64m8, + RI_F, + __riscv_vget_v_f32m8_f32m4, + 1, + __riscv_vfwnmsac_vf_f64m8, + __riscv_vse64_v_f64m8, + vfwnmsac_vf, + LOOP_WIDEN_TERNARY_VX_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmsac\.vf\s+v16,fa0,v24([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmsac\.vf\s+v8,fa0,v4([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vf\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vf\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f16-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f16-m1.c new file mode 100644 index 000000000000..294873d1e6eb --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f16-m1.c @@ -0,0 +1,30 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m2_f16m2 ( \ + __riscv_vreinterpret_v_u32m2_u16m2 (__riscv_vreinterpret_v_f32m2_u32m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e16m1, + vfloat16m1_t, + vfloat32m2_t, + vfloat16m2_t, + __riscv_vle16_v_f16m1, + __riscv_vle32_v_f32m2, + RI_F, + __riscv_vget_v_f16m2_f16m1, + 1, + __riscv_vfwnmsac_vv_f32m2, + __riscv_vse32_v_f32m2, + vfwnmsac_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmsac\.vv\s+v4,v8,v7([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmsac\.vv\s+v2,v6,v1([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v4,v(4|5),} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v2,v(2|3),} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f16-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f16-m2.c new file mode 100644 index 000000000000..640b9eb14f98 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f16-m2.c @@ -0,0 +1,30 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m4_f16m4 ( \ + __riscv_vreinterpret_v_u32m4_u16m4 (__riscv_vreinterpret_v_f32m4_u32m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e16m2, + vfloat16m2_t, + vfloat32m4_t, + vfloat16m4_t, + __riscv_vle16_v_f16m2, + __riscv_vle32_v_f32m4, + RI_F, + __riscv_vget_v_f16m4_f16m2, + 1, + __riscv_vfwnmsac_vv_f32m4, + __riscv_vse32_v_f32m4, + vfwnmsac_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmsac\.vv\s+v8,v16,v14([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmsac\.vv\s+v4,v12,v2([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v8,v(8|9|10|11),} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v4,v(4|5|6|7),} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f16-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f16-m4.c new file mode 100644 index 000000000000..391ae6de1286 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f16-m4.c @@ -0,0 +1,30 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u16m8_f16m8 ( \ + __riscv_vreinterpret_v_u32m8_u16m8 (__riscv_vreinterpret_v_f32m8_u32m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e16m4, + vfloat16m4_t, + vfloat32m8_t, + vfloat16m8_t, + __riscv_vle16_v_f16m4, + __riscv_vle32_v_f32m8, + RI_F, + __riscv_vget_v_f16m8_f16m4, + 1, + __riscv_vfwnmsac_vv_f32m8, + __riscv_vse32_v_f32m8, + vfwnmsac_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmsac\.vv\s+v16,v0,v28([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmsac\.vv\s+v8,v24,v4([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v16,v(16|17|18|19|20|21|22|23),} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v8,v(8|9|10|11|12|13|14|15),} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f32-m1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f32-m1.c new file mode 100644 index 000000000000..47a11457eb8b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f32-m1.c @@ -0,0 +1,30 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m2_f32m2 ( \ + __riscv_vreinterpret_v_u64m2_u32m2 (__riscv_vreinterpret_v_f64m2_u64m2 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e32m1, + vfloat32m1_t, + vfloat64m2_t, + vfloat32m2_t, + __riscv_vle32_v_f32m1, + __riscv_vle64_v_f64m2, + RI_F, + __riscv_vget_v_f32m2_f32m1, + 1, + __riscv_vfwnmsac_vv_f64m2, + __riscv_vse64_v_f64m2, + vfwnmsac_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmsac\.vv\s+v4,v8,v7([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmsac\.vv\s+v2,v6,v1([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v4,v(4|5),} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v4,[a-z0-9]+,v(4|5)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v2,v(2|3),} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v2,[a-z0-9]+,v(2|3)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f32-m2.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f32-m2.c new file mode 100644 index 000000000000..d360b7f0e2ea --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f32-m2.c @@ -0,0 +1,30 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m4_f32m4 ( \ + __riscv_vreinterpret_v_u64m4_u32m4 (__riscv_vreinterpret_v_f64m4_u64m4 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e32m2, + vfloat32m2_t, + vfloat64m4_t, + vfloat32m4_t, + __riscv_vle32_v_f32m2, + __riscv_vle64_v_f64m4, + RI_F, + __riscv_vget_v_f32m4_f32m2, + 1, + __riscv_vfwnmsac_vv_f64m4, + __riscv_vse64_v_f64m4, + vfwnmsac_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmsac\.vv\s+v8,v16,v14([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmsac\.vv\s+v4,v12,v2([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v8,v(8|9|10|11),} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v8,[a-z0-9]+,v(8|9|10|11)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v4,v(4|5|6|7),} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v4,[a-z0-9]+,v(4|5|6|7)([^0-9]|$)} } } */ diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f32-m4.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f32-m4.c new file mode 100644 index 000000000000..203226593f21 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/group_overlap/vfwnmsac_vv-f32-m4.c @@ -0,0 +1,30 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64gcv_zvfh -mabi=lp64d" } */ + +#include "group_overlap.h" + +#define RI_F(v) \ + __riscv_vreinterpret_v_u32m8_f32m8 ( \ + __riscv_vreinterpret_v_u64m8_u32m8 (__riscv_vreinterpret_v_f64m8_u64m8 (v))) + +DEF_GROUP_OVERLAP_TERNARY_0( + __riscv_vsetvlmax_e32m4, + vfloat32m4_t, + vfloat64m8_t, + vfloat32m8_t, + __riscv_vle32_v_f32m4, + __riscv_vle64_v_f64m8, + RI_F, + __riscv_vget_v_f32m8_f32m4, + 1, + __riscv_vfwnmsac_vv_f64m8, + __riscv_vse64_v_f64m8, + vfwnmsac_vv, + LOOP_WIDEN_TERNARY_BODY_OVERLAP_X2) + +/* { dg-final { scan-assembler-times {vfwnmsac\.vv\s+v16,v0,v28([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-times {vfwnmsac\.vv\s+v8,v24,v4([^0-9]|$)} 1 } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v16,v(16|17|18|19|20|21|22|23),} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v16,[a-z0-9]+,v(16|17|18|19|20|21|22|23)([^0-9]|$)} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v8,v(8|9|10|11|12|13|14|15),} } } */ +/* { dg-final { scan-assembler-not {vfwnmsac\.vv\s+v8,[a-z0-9]+,v(8|9|10|11|12|13|14|15)([^0-9]|$)} } } */
