Christophe Lyon via Gcc-patches <gcc-patches@gcc.gnu.org> writes:
> @@ -31086,36 +31087,20 @@ arm_expand_vector_compare (rtx target, rtx_code
> code, rtx op0, rtx op1,
> case NE:
> if (TARGET_HAVE_MVE)
> {
> - rtx vpr_p0;
> - if (vcond_mve)
> - vpr_p0 = target;
> - else
> - vpr_p0 = gen_reg_rtx (arm_mode_to_pred_mode (cmp_mode));
> -
> switch (GET_MODE_CLASS (cmp_mode))
> {
> case MODE_VECTOR_INT:
> - emit_insn (gen_mve_vcmpq (code, cmp_mode, vpr_p0, op0, force_reg
> (cmp_mode, op1)));
> + emit_insn (gen_mve_vcmpq (code, cmp_mode, target, op0, force_reg
> (cmp_mode, op1)));
Pre-existing nit: long line. Same for later calls in the same function.
Richard
> break;
> case MODE_VECTOR_FLOAT:
> if (TARGET_HAVE_MVE_FLOAT)
> - emit_insn (gen_mve_vcmpq_f (code, cmp_mode, vpr_p0, op0,
> force_reg (cmp_mode, op1)));
> + emit_insn (gen_mve_vcmpq_f (code, cmp_mode, target, op0,
> force_reg (cmp_mode, op1)));
> else
> gcc_unreachable ();
> break;
> default:
> gcc_unreachable ();
> }
> -
> - /* If we are not expanding a vcond, build the result here. */
> - if (!vcond_mve)
> - {
> - rtx zero = gen_reg_rtx (cmp_result_mode);
> - rtx one = gen_reg_rtx (cmp_result_mode);
> - emit_move_insn (zero, CONST0_RTX (cmp_result_mode));
> - emit_move_insn (one, CONST1_RTX (cmp_result_mode));
> - emit_insn (gen_mve_vpselq (VPSELQ_S, cmp_result_mode, target,
> one, zero, vpr_p0));
> - }
> }
> else
> emit_insn (gen_neon_vc (code, cmp_mode, target, op0, op1));
> @@ -31127,23 +31112,7 @@ arm_expand_vector_compare (rtx target, rtx_code
> code, rtx op0, rtx op1,
> case GEU:
> case GTU:
> if (TARGET_HAVE_MVE)
> - {
> - rtx vpr_p0;
> - if (vcond_mve)
> - vpr_p0 = target;
> - else
> - vpr_p0 = gen_reg_rtx (arm_mode_to_pred_mode (cmp_mode));
> -
> - emit_insn (gen_mve_vcmpq (code, cmp_mode, vpr_p0, op0, force_reg
> (cmp_mode, op1)));
> - if (!vcond_mve)
> - {
> - rtx zero = gen_reg_rtx (cmp_result_mode);
> - rtx one = gen_reg_rtx (cmp_result_mode);
> - emit_move_insn (zero, CONST0_RTX (cmp_result_mode));
> - emit_move_insn (one, CONST1_RTX (cmp_result_mode));
> - emit_insn (gen_mve_vpselq (VPSELQ_S, cmp_result_mode, target,
> one, zero, vpr_p0));
> - }
> - }
> + emit_insn (gen_mve_vcmpq (code, cmp_mode, target, op0, force_reg
> (cmp_mode, op1)));
> else
> emit_insn (gen_neon_vc (code, cmp_mode, target,
> op0, force_reg (cmp_mode, op1)));
> @@ -31154,23 +31123,7 @@ arm_expand_vector_compare (rtx target, rtx_code
> code, rtx op0, rtx op1,
> case LEU:
> case LTU:
> if (TARGET_HAVE_MVE)
> - {
> - rtx vpr_p0;
> - if (vcond_mve)
> - vpr_p0 = target;
> - else
> - vpr_p0 = gen_reg_rtx (arm_mode_to_pred_mode (cmp_mode));
> -
> - emit_insn (gen_mve_vcmpq (swap_condition (code), cmp_mode, vpr_p0,
> force_reg (cmp_mode, op1), op0));
> - if (!vcond_mve)
> - {
> - rtx zero = gen_reg_rtx (cmp_result_mode);
> - rtx one = gen_reg_rtx (cmp_result_mode);
> - emit_move_insn (zero, CONST0_RTX (cmp_result_mode));
> - emit_move_insn (one, CONST1_RTX (cmp_result_mode));
> - emit_insn (gen_mve_vpselq (VPSELQ_S, cmp_result_mode, target,
> one, zero, vpr_p0));
> - }
> - }
> + emit_insn (gen_mve_vcmpq (swap_condition (code), cmp_mode, target,
> force_reg (cmp_mode, op1), op0));
> else
> emit_insn (gen_neon_vc (swap_condition (code), cmp_mode,
> target, force_reg (cmp_mode, op1), op0));
> @@ -31185,8 +31138,8 @@ arm_expand_vector_compare (rtx target, rtx_code code,
> rtx op0, rtx op1,
> rtx gt_res = gen_reg_rtx (cmp_result_mode);
> rtx alt_res = gen_reg_rtx (cmp_result_mode);
> rtx_code alt_code = (code == LTGT ? LT : LE);
> - if (arm_expand_vector_compare (gt_res, GT, op0, op1, true, vcond_mve)
> - || arm_expand_vector_compare (alt_res, alt_code, op0, op1, true,
> vcond_mve))
> + if (arm_expand_vector_compare (gt_res, GT, op0, op1, true)
> + || arm_expand_vector_compare (alt_res, alt_code, op0, op1, true))
> gcc_unreachable ();
> emit_insn (gen_rtx_SET (target, gen_rtx_IOR (cmp_result_mode,
> gt_res, alt_res)));
> @@ -31206,19 +31159,15 @@ arm_expand_vcond (rtx *operands, machine_mode
> cmp_result_mode)
> {
> /* When expanding for MVE, we do not want to emit a (useless) vpsel in
> arm_expand_vector_compare, and another one here. */
> - bool vcond_mve=false;
> rtx mask;
>
> if (TARGET_HAVE_MVE)
> - {
> - vcond_mve=true;
> - mask = gen_reg_rtx (arm_mode_to_pred_mode (cmp_result_mode));
> - }
> + mask = gen_reg_rtx (arm_mode_to_pred_mode (cmp_result_mode).require ());
> else
> mask = gen_reg_rtx (cmp_result_mode);
>
> bool inverted = arm_expand_vector_compare (mask, GET_CODE (operands[3]),
> - operands[4], operands[5], true,
> vcond_mve);
> + operands[4], operands[5], true);
> if (inverted)
> std::swap (operands[1], operands[2]);
> if (TARGET_NEON)
> @@ -31226,20 +31175,20 @@ arm_expand_vcond (rtx *operands, machine_mode
> cmp_result_mode)
> mask, operands[1], operands[2]));
> else
> {
> - machine_mode cmp_mode = GET_MODE (operands[4]);
> - rtx vpr_p0 = mask;
> - rtx zero = gen_reg_rtx (cmp_mode);
> - rtx one = gen_reg_rtx (cmp_mode);
> - emit_move_insn (zero, CONST0_RTX (cmp_mode));
> - emit_move_insn (one, CONST1_RTX (cmp_mode));
> + machine_mode cmp_mode = GET_MODE (operands[0]);
> +
> switch (GET_MODE_CLASS (cmp_mode))
> {
> case MODE_VECTOR_INT:
> - emit_insn (gen_mve_vpselq (VPSELQ_S, cmp_result_mode, operands[0],
> one, zero, vpr_p0));
> + emit_insn (gen_mve_vpselq (VPSELQ_S, cmp_mode, operands[0],
> + operands[1], operands[2], mask));
> break;
> case MODE_VECTOR_FLOAT:
> if (TARGET_HAVE_MVE_FLOAT)
> - emit_insn (gen_mve_vpselq_f (cmp_mode, operands[0], one, zero,
> vpr_p0));
> + emit_insn (gen_mve_vpselq_f (cmp_mode, operands[0],
> + operands[1], operands[2], mask));
> + else
> + gcc_unreachable ();
> break;
> default:
> gcc_unreachable ();
> @@ -34149,4 +34098,15 @@ arm_mode_base_reg_class (machine_mode mode)
>
> struct gcc_target targetm = TARGET_INITIALIZER;
>
> +/* Implement TARGET_VECTORIZE_GET_MASK_MODE. */
> +
> +opt_machine_mode
> +arm_get_mask_mode (machine_mode mode)
> +{
> + if (TARGET_HAVE_MVE)
> + return arm_mode_to_pred_mode (mode);
> +
> + return default_get_mask_mode (mode);
> +}
> +
> #include "gt-arm.h"
> diff --git a/gcc/config/arm/mve.md b/gcc/config/arm/mve.md
> index 9da78657798..fb25cac1cfd 100644
> --- a/gcc/config/arm/mve.md
> +++ b/gcc/config/arm/mve.md
> @@ -10540,3 +10540,57 @@ (define_insn "*mve_mov<mode>"
> vmsr%?\t P0, %1
> vmrs%?\t %0, P0"
> )
> +
> +;; Expanders for vec_cmp and vcond
> +
> +(define_expand "vec_cmp<mode><MVE_vpred>"
> + [(set (match_operand:<MVE_VPRED> 0 "s_register_operand")
> + (match_operator:<MVE_VPRED> 1 "comparison_operator"
> + [(match_operand:MVE_VLD_ST 2 "s_register_operand")
> + (match_operand:MVE_VLD_ST 3 "reg_or_zero_operand")]))]
> + "TARGET_HAVE_MVE
> + && (!<Is_float_mode> || flag_unsafe_math_optimizations)"
> +{
> + arm_expand_vector_compare (operands[0], GET_CODE (operands[1]),
> + operands[2], operands[3], false);
> + DONE;
> +})
> +
> +(define_expand "vec_cmpu<mode><MVE_vpred>"
> + [(set (match_operand:<MVE_VPRED> 0 "s_register_operand")
> + (match_operator:<MVE_VPRED> 1 "comparison_operator"
> + [(match_operand:MVE_2 2 "s_register_operand")
> + (match_operand:MVE_2 3 "reg_or_zero_operand")]))]
> + "TARGET_HAVE_MVE"
> +{
> + arm_expand_vector_compare (operands[0], GET_CODE (operands[1]),
> + operands[2], operands[3], false);
> + DONE;
> +})
> +
> +(define_expand "vcond_mask_<mode><MVE_vpred>"
> + [(set (match_operand:MVE_VLD_ST 0 "s_register_operand")
> + (if_then_else:MVE_VLD_ST
> + (match_operand:<MVE_VPRED> 3 "s_register_operand")
> + (match_operand:MVE_VLD_ST 1 "s_register_operand")
> + (match_operand:MVE_VLD_ST 2 "s_register_operand")))]
> + "TARGET_HAVE_MVE"
> +{
> + switch (GET_MODE_CLASS (<MODE>mode))
> + {
> + case MODE_VECTOR_INT:
> + emit_insn (gen_mve_vpselq (VPSELQ_S, <MODE>mode, operands[0],
> + operands[1], operands[2], operands[3]));
> + break;
> + case MODE_VECTOR_FLOAT:
> + if (TARGET_HAVE_MVE_FLOAT)
> + emit_insn (gen_mve_vpselq_f (<MODE>mode, operands[0],
> + operands[1], operands[2], operands[3]));
> + else
> + gcc_unreachable ();
> + break;
> + default:
> + gcc_unreachable ();
> + }
> + DONE;
> +})
> diff --git a/gcc/config/arm/neon.md b/gcc/config/arm/neon.md
> index 8b0a396947c..28310d93a4e 100644
> --- a/gcc/config/arm/neon.md
> +++ b/gcc/config/arm/neon.md
> @@ -1394,6 +1394,45 @@ (define_insn "*us_sub<mode>_neon"
> [(set_attr "type" "neon_qsub<q>")]
> )
>
> +(define_expand "vec_cmp<mode><v_cmp_result>"
> + [(set (match_operand:<V_cmp_result> 0 "s_register_operand")
> + (match_operator:<V_cmp_result> 1 "comparison_operator"
> + [(match_operand:VDQWH 2 "s_register_operand")
> + (match_operand:VDQWH 3 "reg_or_zero_operand")]))]
> + "TARGET_NEON
> + && (!<Is_float_mode> || flag_unsafe_math_optimizations)"
> +{
> + arm_expand_vector_compare (operands[0], GET_CODE (operands[1]),
> + operands[2], operands[3], false);
> + DONE;
> +})
> +
> +(define_expand "vec_cmpu<mode><mode>"
> + [(set (match_operand:VDQIW 0 "s_register_operand")
> + (match_operator:VDQIW 1 "comparison_operator"
> + [(match_operand:VDQIW 2 "s_register_operand")
> + (match_operand:VDQIW 3 "reg_or_zero_operand")]))]
> + "TARGET_NEON"
> +{
> + arm_expand_vector_compare (operands[0], GET_CODE (operands[1]),
> + operands[2], operands[3], false);
> + DONE;
> +})
> +
> +(define_expand "vcond_mask_<mode><v_cmp_result>"
> + [(set (match_operand:VDQWH 0 "s_register_operand")
> + (if_then_else:VDQWH
> + (match_operand:<V_cmp_result> 3 "s_register_operand")
> + (match_operand:VDQWH 1 "s_register_operand")
> + (match_operand:VDQWH 2 "s_register_operand")))]
> + "TARGET_NEON
> + && (!<Is_float_mode> || flag_unsafe_math_optimizations)"
> +{
> + emit_insn (gen_neon_vbsl<mode> (operands[0], operands[3], operands[1],
> + operands[2]));
> + DONE;
> +})
> +
> ;; Patterns for builtins.
>
> ; good for plain vadd, vaddq.
> diff --git a/gcc/config/arm/vec-common.md b/gcc/config/arm/vec-common.md
> index 68de4f0f943..9b461a76155 100644
> --- a/gcc/config/arm/vec-common.md
> +++ b/gcc/config/arm/vec-common.md
> @@ -363,33 +363,6 @@ (define_expand "vlshr<mode>3"
> }
> })
>
> -(define_expand "vec_cmp<mode><v_cmp_result>"
> - [(set (match_operand:<V_cmp_result> 0 "s_register_operand")
> - (match_operator:<V_cmp_result> 1 "comparison_operator"
> - [(match_operand:VDQWH 2 "s_register_operand")
> - (match_operand:VDQWH 3 "reg_or_zero_operand")]))]
> - "ARM_HAVE_<MODE>_ARITH
> - && !TARGET_REALLY_IWMMXT
> - && (!<Is_float_mode> || flag_unsafe_math_optimizations)"
> -{
> - arm_expand_vector_compare (operands[0], GET_CODE (operands[1]),
> - operands[2], operands[3], false, false);
> - DONE;
> -})
> -
> -(define_expand "vec_cmpu<mode><mode>"
> - [(set (match_operand:VDQIW 0 "s_register_operand")
> - (match_operator:VDQIW 1 "comparison_operator"
> - [(match_operand:VDQIW 2 "s_register_operand")
> - (match_operand:VDQIW 3 "reg_or_zero_operand")]))]
> - "ARM_HAVE_<MODE>_ARITH
> - && !TARGET_REALLY_IWMMXT"
> -{
> - arm_expand_vector_compare (operands[0], GET_CODE (operands[1]),
> - operands[2], operands[3], false, false);
> - DONE;
> -})
> -
> ;; Conditional instructions. These are comparisons with conditional moves
> for
> ;; vectors. They perform the assignment:
> ;;
> @@ -461,31 +434,6 @@ (define_expand "vcondu<mode><v_cmp_result>"
> DONE;
> })
>
> -(define_expand "vcond_mask_<mode><v_cmp_result>"
> - [(set (match_operand:VDQWH 0 "s_register_operand")
> - (if_then_else:VDQWH
> - (match_operand:<V_cmp_result> 3 "s_register_operand")
> - (match_operand:VDQWH 1 "s_register_operand")
> - (match_operand:VDQWH 2 "s_register_operand")))]
> - "ARM_HAVE_<MODE>_ARITH
> - && !TARGET_REALLY_IWMMXT
> - && (!<Is_float_mode> || flag_unsafe_math_optimizations)"
> -{
> - if (TARGET_NEON)
> - {
> - emit_insn (gen_neon_vbsl (<MODE>mode, operands[0], operands[3],
> - operands[1], operands[2]));
> - }
> - else if (TARGET_HAVE_MVE)
> - {
> - emit_insn (gen_mve_vpselq (VPSELQ_S, <MODE>mode, operands[0],
> - operands[1], operands[2], operands[3]));
> - }
> - else
> - gcc_unreachable ();
> - DONE;
> -})
> -
> (define_expand "vec_load_lanesoi<mode>"
> [(set (match_operand:OI 0 "s_register_operand")
> (unspec:OI [(match_operand:OI 1 "neon_struct_operand")
