/* { dg-do assemble { target aarch64_asm_sve2p1_ok } } */ /* { dg-do compile { target { ! aarch64_asm_sve2p1_ok } } } */ /* { dg-skip-if "" { *-*-* } { "-DSTREAMING_COMPATIBLE" } { "" } } */ /* { dg-final { check-function-bodies "**" "" "-DCHECK_ASM" } } */ #include "test_sve_acle.h" #pragma GCC target "+sve2p1" /* ** st4q_f32_base: ** st4q {z0\.q - z3\.q}, p0, \[x0\] ** ret */ TEST_STORE (st4q_f32_base, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0, z0), svst4q (p0, x0, z0)) /* ** st4q_f32_index: ** add (x[0-9]), x0, x1, lsl #?2 ** st4q {z0\.q - z3\.q}, p0, \[\1\] ** ret */ TEST_STORE (st4q_f32_index, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0 + x1, z0), svst4q (p0, x0 + x1, z0)) /* ** st4q_f32_index2: ** add (x[0-9]), x0, x1, lsl #?3 ** st4q {z0\.q - z3\.q}, p0, \[\1\] ** ret */ TEST_STORE (st4q_f32_index2, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0 + x1 * 2, z0), svst4q (p0, x0 + x1 * 2, z0)) /* ** st4q_f32_index4: ** st4q {z0\.q - z3\.q}, p0, \[x0, x1, lsl #?4\] ** ret */ TEST_STORE (st4q_f32_index4, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0 + x1 * 4, z0), svst4q (p0, x0 + x1 * 4, z0)) /* Moving the constant into a register would also be OK. */ /* ** st4q_f32_1: ** incb x0 ** st4q {z0\.q - z3\.q}, p0, \[x0\] ** ret */ TEST_STORE (st4q_f32_1, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0 + svcntw (), z0), svst4q (p0, x0 + svcntw (), z0)) /* Moving the constant into a register would also be OK. */ /* ** st4q_f32_2: ** incb x0, all, mul #2 ** st4q {z0\.q - z3\.q}, p0, \[x0\] ** ret */ TEST_STORE (st4q_f32_2, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0 + svcntw () * 2, z0), svst4q (p0, x0 + svcntw () * 2, z0)) /* Moving the constant into a register would also be OK. */ /* ** st4q_f32_3: ** incb x0, all, mul #3 ** st4q {z0\.q - z3\.q}, p0, \[x0\] ** ret */ TEST_STORE (st4q_f32_3, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0 + svcntw () * 3, z0), svst4q (p0, x0 + svcntw () * 3, z0)) /* ** st4q_f32_4: ** st4q {z0\.q - z3\.q}, p0, \[x0, #4, mul vl\] ** ret */ TEST_STORE (st4q_f32_4, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0 + svcntw () * 4, z0), svst4q (p0, x0 + svcntw () * 4, z0)) /* ** st4q_f32_28: ** st4q {z0\.q - z3\.q}, p0, \[x0, #28, mul vl\] ** ret */ TEST_STORE (st4q_f32_28, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0 + svcntw () * 28, z0), svst4q (p0, x0 + svcntw () * 28, z0)) /* ** st4q_f32_32: ** [^{]* ** st4q {z0\.q - z3\.q}, p0, \[x[0-9]+\] ** ret */ TEST_STORE (st4q_f32_32, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0 + svcntw () * 32, z0), svst4q (p0, x0 + svcntw () * 32, z0)) /* Moving the constant into a register would also be OK. */ /* ** st4q_f32_m1: ** decb x0 ** st4q {z0\.q - z3\.q}, p0, \[x0\] ** ret */ TEST_STORE (st4q_f32_m1, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0 - svcntw (), z0), svst4q (p0, x0 - svcntw (), z0)) /* Moving the constant into a register would also be OK. */ /* ** st4q_f32_m2: ** decb x0, all, mul #2 ** st4q {z0\.q - z3\.q}, p0, \[x0\] ** ret */ TEST_STORE (st4q_f32_m2, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0 - svcntw () * 2, z0), svst4q (p0, x0 - svcntw () * 2, z0)) /* Moving the constant into a register would also be OK. */ /* ** st4q_f32_m3: ** decb x0, all, mul #3 ** st4q {z0\.q - z3\.q}, p0, \[x0\] ** ret */ TEST_STORE (st4q_f32_m3, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0 - svcntw () * 3, z0), svst4q (p0, x0 - svcntw () * 3, z0)) /* ** st4q_f32_m4: ** st4q {z0\.q - z3\.q}, p0, \[x0, #-4, mul vl\] ** ret */ TEST_STORE (st4q_f32_m4, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0 - svcntw () * 4, z0), svst4q (p0, x0 - svcntw () * 4, z0)) /* ** st4q_f32_m32: ** st4q {z0\.q - z3\.q}, p0, \[x0, #-32, mul vl\] ** ret */ TEST_STORE (st4q_f32_m32, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0 - svcntw () * 32, z0), svst4q (p0, x0 - svcntw () * 32, z0)) /* ** st4q_f32_m36: ** [^{]* ** st4q {z0\.q - z3\.q}, p0, \[x[0-9]+\] ** ret */ TEST_STORE (st4q_f32_m36, svfloat32x4_t, float32_t, svst4q_f32 (p0, x0 - svcntw () * 36, z0), svst4q (p0, x0 - svcntw () * 36, z0)) /* ** st4q_vnum_f32_0: ** st4q {z0\.q - z3\.q}, p0, \[x0\] ** ret */ TEST_STORE (st4q_vnum_f32_0, svfloat32x4_t, float32_t, svst4q_vnum_f32 (p0, x0, 0, z0), svst4q_vnum (p0, x0, 0, z0)) /* Moving the constant into a register would also be OK. */ /* ** st4q_vnum_f32_1: ** incb x0 ** st4q {z0\.q - z3\.q}, p0, \[x0\] ** ret */ TEST_STORE (st4q_vnum_f32_1, svfloat32x4_t, float32_t, svst4q_vnum_f32 (p0, x0, 1, z0), svst4q_vnum (p0, x0, 1, z0)) /* Moving the constant into a register would also be OK. */ /* ** st4q_vnum_f32_2: ** incb x0, all, mul #2 ** st4q {z0\.q - z3\.q}, p0, \[x0\] ** ret */ TEST_STORE (st4q_vnum_f32_2, svfloat32x4_t, float32_t, svst4q_vnum_f32 (p0, x0, 2, z0), svst4q_vnum (p0, x0, 2, z0)) /* Moving the constant into a register would also be OK. */ /* ** st4q_vnum_f32_3: ** incb x0, all, mul #3 ** st4q {z0\.q - z3\.q}, p0, \[x0\] ** ret */ TEST_STORE (st4q_vnum_f32_3, svfloat32x4_t, float32_t, svst4q_vnum_f32 (p0, x0, 3, z0), svst4q_vnum (p0, x0, 3, z0)) /* ** st4q_vnum_f32_4: ** st4q {z0\.q - z3\.q}, p0, \[x0, #4, mul vl\] ** ret */ TEST_STORE (st4q_vnum_f32_4, svfloat32x4_t, float32_t, svst4q_vnum_f32 (p0, x0, 4, z0), svst4q_vnum (p0, x0, 4, z0)) /* ** st4q_vnum_f32_28: ** st4q {z0\.q - z3\.q}, p0, \[x0, #28, mul vl\] ** ret */ TEST_STORE (st4q_vnum_f32_28, svfloat32x4_t, float32_t, svst4q_vnum_f32 (p0, x0, 28, z0), svst4q_vnum (p0, x0, 28, z0)) /* ** st4q_vnum_f32_32: ** [^{]* ** st4q {z0\.q - z3\.q}, p0, \[x[0-9]+\] ** ret */ TEST_STORE (st4q_vnum_f32_32, svfloat32x4_t, float32_t, svst4q_vnum_f32 (p0, x0, 32, z0), svst4q_vnum (p0, x0, 32, z0)) /* Moving the constant into a register would also be OK. */ /* ** st4q_vnum_f32_m1: ** decb x0 ** st4q {z0\.q - z3\.q}, p0, \[x0\] ** ret */ TEST_STORE (st4q_vnum_f32_m1, svfloat32x4_t, float32_t, svst4q_vnum_f32 (p0, x0, -1, z0), svst4q_vnum (p0, x0, -1, z0)) /* Moving the constant into a register would also be OK. */ /* ** st4q_vnum_f32_m2: ** decb x0, all, mul #2 ** st4q {z0\.q - z3\.q}, p0, \[x0\] ** ret */ TEST_STORE (st4q_vnum_f32_m2, svfloat32x4_t, float32_t, svst4q_vnum_f32 (p0, x0, -2, z0), svst4q_vnum (p0, x0, -2, z0)) /* Moving the constant into a register would also be OK. */ /* ** st4q_vnum_f32_m3: ** decb x0, all, mul #3 ** st4q {z0\.q - z3\.q}, p0, \[x0\] ** ret */ TEST_STORE (st4q_vnum_f32_m3, svfloat32x4_t, float32_t, svst4q_vnum_f32 (p0, x0, -3, z0), svst4q_vnum (p0, x0, -3, z0)) /* ** st4q_vnum_f32_m4: ** st4q {z0\.q - z3\.q}, p0, \[x0, #-4, mul vl\] ** ret */ TEST_STORE (st4q_vnum_f32_m4, svfloat32x4_t, float32_t, svst4q_vnum_f32 (p0, x0, -4, z0), svst4q_vnum (p0, x0, -4, z0)) /* ** st4q_vnum_f32_m32: ** st4q {z0\.q - z3\.q}, p0, \[x0, #-32, mul vl\] ** ret */ TEST_STORE (st4q_vnum_f32_m32, svfloat32x4_t, float32_t, svst4q_vnum_f32 (p0, x0, -32, z0), svst4q_vnum (p0, x0, -32, z0)) /* ** st4q_vnum_f32_m36: ** [^{]* ** st4q {z0\.q - z3\.q}, p0, \[x[0-9]+\] ** ret */ TEST_STORE (st4q_vnum_f32_m36, svfloat32x4_t, float32_t, svst4q_vnum_f32 (p0, x0, -36, z0), svst4q_vnum (p0, x0, -36, z0)) /* ** st4q_vnum_f32_x1: ** cntb (x[0-9]+) ** ( ** madd (x[0-9]+), (?:x1, \1|\1, x1), x0 ** st4q {z0\.q - z3\.q}, p0, \[\2\] ** | ** mul (x[0-9]+), (?:x1, \1|\1, x1) ** st4q {z0\.q - z3\.q}, p0, \[x0, \3\] ** ) ** ret */ TEST_STORE (st4q_vnum_f32_x1, svfloat32x4_t, float32_t, svst4q_vnum_f32 (p0, x0, x1, z0), svst4q_vnum (p0, x0, x1, z0))