https://gcc.gnu.org/g:8b33f39446c9a1cda460d9fef7243a4d20289f45
commit 8b33f39446c9a1cda460d9fef7243a4d20289f45 Author: Michael Meissner <[email protected]> Date: Tue Nov 11 23:31:28 2025 -0500 Add infrastructure for _Float16 and __bfloat16 types. This patch adds the infrastructure for adding 16-bit floating point types in the next patch. Two new types that will be added: _Float16 (HFmode): ================== This is the IEEE 754-2008 16-bit floating point. It has 1 sign bit, 5 exponent bits, 10 explicit mantassia bits (the 11th bit is implied with normalization). The PowerPC ISA 3.0 (power9) has instructions to convert between the scalar representations of _Float16 and float types. The PowerPC ISA 3.1 (power10 and power11) has instructions for converting between the even elements of _Float16 vectors and float vectors. In addition, the MMA subsystem has support for _Float16 vector processing. __bfloat16 (BFmode): ==================== This is the brain 16-bit floating point created by the Google Brain project. It has 1 sign bit, 8 exponent bits, 7 explicit mantissa bits (the 8th bit is implied with normalization). The 16 bits in the __bfloat16 format is the same as the upper 16 bits in the normal IEEE 754 32-bit floating point format. he PowerPC ISA 3.1 (power10 and power11) has instructions for converting between the even elements of _bfloat16 vectors and float vectors. In addition, the MMA subsystem has support for _bfloat16 vector processing. This patch adds new modes that will be used in the future. The V8HFmode and V8BFmodes are treated as normal vector modes. This patch does not add loads and stores for BFmode and HFmode. These will be added in the next patch. BFmode -- 16-bit mode for __bfloat16 support HFmode -- 16-bit mode for _Float16 support V8BFmode -- 128-bit vector mode __bfloat16 V8HFmode -- 128-bit vector mode _Float16 V4BFmode -- 64-bit vector mode __bfloat16 used in some insns V4HFmode -- 64-bit vector mode _Float16 used in some insns 2025-11-10 Michael Meissner <[email protected]> gcc/ * config/rs6000/altivec.md (VM): Add support for V8HFmode and V8BFmode. (VM2): Likewise. (VI_char): Likewise. (VI_scalar): Likewise. (VI_unit): Likewise. (VP_small): Likewise. (VP_small_lc): Likewise. (VU_char): Likewise. * config/rs6000/rs6000-modes.def (HFmode): Add new mode. (BFmode): Likewise. (V8BFmode): Likewise. (V8HFmode): Likewise. * config/rs6000/rs6000-p8swap.cc (rs6000_gen_stvx): Remove #ifdef for HAVE_V8HFmode. Add support for V8BFmode. (rs6000_gen_lvx): Likewise. (replace_swapped_load_constant): Likewise. * config/rs6000/rs6000.cc (rs6000_debug_reg_global): Add support for V8HFmode and V8BFmode. (rs6000_init_hard_regno_mode_ok): Likewise. (output_vec_const_move): Likewise. (rs6000_expand_vector_init): Likewise. (reg_offset_addressing_ok_p): Likewise. (rs6000_const_vec): Likewise. (rs6000_emit_move): Likewise. * config/rs6000/rs6000.h (ALTIVEC_VECTOR_MODE): Likewise. * config/rs6000/rs6000.md (FMOVE128_GPR): Likewise. (wd): Likewise. (du_or_d): Likewise. (BOOL_128): Likewise. (BOOL_REGS_OUTPUT): Likewise. (BOOL_REGS_OP1): Likewise. (BOOL_REGS_OP2): Likewise. (BOOL_REGS_UNARY): Likewise. (RELOAD): Likewise. * config/rs6000/vector.md (VEC_L): Likewise. (VEC_M): Likewise. (VEC_E): Likewise. (VEC_base): Likewise. (VEC_base_l): Likewise. * config/rs6000/vsx.md (VECTOR_16BIT): New mode iterator. (VSX_L): Add support for V8HFmode and V8BFmode. (VSX_M): Likewise. (VSX_XXBR): Likewise. (VSm): Likewise. (VSr): Likewise. (VSisa): Likewise. (??r): Likewise. (nW): Likewise. (VSc): Likewise. (VM3): Likewise. (VM3_char): Likewise. (vsx_le_perm_load_<mode>): Rename from vsx_le_perm_load_v8hi and add V8HFmode and V8BFmode. (vsx_le_perm_store_<mode>): Rename from vsx_le_perm_store_v8hi and add V8HFmode and V8BFmode. (splitter for vsx_le_perm_store_<mode>): Likewise. (vsx_ld_elemrev_<mode>): Rename from vsx_ld_elemrev_v8hi and add V8HFmode and V8BFmode support. (vsx_ld_elemrev_<mode>_internal): Rename from vsx_ld_elemrev_v8hi_internal and add V8HFmode and V8BFmode support. (vsx_st_elemrev_<mode>): Rename from vsx_st_elemrev_v8hi and add V8HFmode and V8BFmode support. (vsx_st_elemrev_<mode>_internal): Rename from vsx_st_elemrev_v8hi_internal and add V8HFmode and V8BFmode support. (xxswapd_<mode>): Rename from xxswapd_v8hi and add V8HFmode and V8BFmode support. (vsx_lxvd2x8_le_<MODE>): Rename from vsx_lxvd2x8_le_V8HI and add V8HFmode and V8BFmode support. (vsx_stxvd2x8_le_<MODE>): Rename from vsx_stxvd2x8_le_V8HI and add V8HFmode and V8BFmode support. (vsx_extract_<mode>_store_p9): Add V8HFmode and V8BFmode. (vsx_extract_<mode>_p8): Likewise. Diff: --- gcc/config/rs6000/altivec.md | 34 +++++++-- gcc/config/rs6000/rs6000-modes.def | 9 +++ gcc/config/rs6000/rs6000-p8swap.cc | 14 ++-- gcc/config/rs6000/rs6000.cc | 32 +++++++- gcc/config/rs6000/rs6000.h | 2 + gcc/config/rs6000/rs6000.md | 20 ++++- gcc/config/rs6000/vector.md | 37 ++++++++- gcc/config/rs6000/vsx.md | 151 ++++++++++++++++++++++--------------- 8 files changed, 217 insertions(+), 82 deletions(-) diff --git a/gcc/config/rs6000/altivec.md b/gcc/config/rs6000/altivec.md index fa3368079ada..d821960cb5f5 100644 --- a/gcc/config/rs6000/altivec.md +++ b/gcc/config/rs6000/altivec.md @@ -191,6 +191,8 @@ ;; otherwise handled by altivec (v2df, v2di, ti) (define_mode_iterator VM [V4SI V8HI + V8BF + V8HF V16QI V4SF V2DF @@ -203,6 +205,8 @@ ;; Like VM, except don't do TImode (define_mode_iterator VM2 [V4SI V8HI + V8BF + V8HF V16QI V4SF V2DF @@ -222,18 +226,38 @@ V1TI TI]) -(define_mode_attr VI_char [(V2DI "d") (V4SI "w") (V8HI "h") (V16QI "b")]) -(define_mode_attr VI_scalar [(V2DI "DI") (V4SI "SI") (V8HI "HI") (V16QI "QI")]) +(define_mode_attr VI_char [(V2DI "d") + (V4SI "w") + (V8HI "h") + (V8BF "h") + (V8HF "h") + (V16QI "b")]) +(define_mode_attr VI_scalar [(V2DI "DI") + (V4SI "SI") + (V8HI "HI") + (V8BF "BF") + (V8HF "HF") + (V16QI "QI")]) (define_mode_attr VI_unit [(V16QI "VECTOR_UNIT_ALTIVEC_P (V16QImode)") (V8HI "VECTOR_UNIT_ALTIVEC_P (V8HImode)") + (V8BF "VECTOR_UNIT_ALTIVEC_P (V8BFmode)") + (V8HF "VECTOR_UNIT_ALTIVEC_P (V8HFmode)") (V4SI "VECTOR_UNIT_ALTIVEC_P (V4SImode)") (V2DI "VECTOR_UNIT_P8_VECTOR_P (V2DImode)")]) ;; Vector pack/unpack (define_mode_iterator VP [V2DI V4SI V8HI]) -(define_mode_attr VP_small [(V2DI "V4SI") (V4SI "V8HI") (V8HI "V16QI")]) -(define_mode_attr VP_small_lc [(V2DI "v4si") (V4SI "v8hi") (V8HI "v16qi")]) -(define_mode_attr VU_char [(V2DI "w") (V4SI "h") (V8HI "b")]) +(define_mode_attr VP_small [(V2DI "V4SI") + (V4SI "V8HI") + (V8HI "V16QI")]) +(define_mode_attr VP_small_lc [(V2DI "v4si") + (V4SI "v8hi") + (V8HI "v16qi")]) +(define_mode_attr VU_char [(V2DI "w") + (V4SI "h") + (V8HI "b") + (V8BF "b") + (V8HF "b")]) ;; Vector negate (define_mode_iterator VNEG [V4SI V2DI]) diff --git a/gcc/config/rs6000/rs6000-modes.def b/gcc/config/rs6000/rs6000-modes.def index 9a8b505ab6a7..ab1f258ed1a8 100644 --- a/gcc/config/rs6000/rs6000-modes.def +++ b/gcc/config/rs6000/rs6000-modes.def @@ -45,6 +45,13 @@ FLOAT_MODE (TF, 16, ieee_quad_format); /* IBM 128-bit floating point. */ FLOAT_MODE (IF, 16, ibm_extended_format); +/* Explicit IEEE 16-bit floating point. */ +FLOAT_MODE (HF, 2, ieee_half_format); + +/* Explicit bfloat16 floating point. */ +FLOAT_MODE (BF, 2, arm_bfloat_half_format); +ADJUST_FLOAT_FORMAT (BF, &arm_bfloat_half_format); + /* Add any extra modes needed to represent the condition code. For the RS/6000, we need separate modes when unsigned (logical) comparisons @@ -70,6 +77,8 @@ VECTOR_MODES (FLOAT, 32); /* V16HF V8SF V4DF */ /* Half VMX/VSX vector (for internal use) */ VECTOR_MODE (FLOAT, SF, 2); /* V2SF */ VECTOR_MODE (INT, SI, 2); /* V2SI */ +VECTOR_MODE (FLOAT, BF, 4); /* V4BF */ +VECTOR_MODE (FLOAT, HF, 4); /* V4HF */ /* Replacement for TImode that only is allowed in GPRs. We also use PTImode for quad memory atomic operations to force getting an even/odd register diff --git a/gcc/config/rs6000/rs6000-p8swap.cc b/gcc/config/rs6000/rs6000-p8swap.cc index e92f01031270..7ba50f11a7bf 100644 --- a/gcc/config/rs6000/rs6000-p8swap.cc +++ b/gcc/config/rs6000/rs6000-p8swap.cc @@ -1598,10 +1598,10 @@ rs6000_gen_stvx (enum machine_mode mode, rtx dest_exp, rtx src_exp) stvx = gen_altivec_stvx_v16qi (src_exp, dest_exp); else if (mode == V8HImode) stvx = gen_altivec_stvx_v8hi (src_exp, dest_exp); -#ifdef HAVE_V8HFmode else if (mode == V8HFmode) stvx = gen_altivec_stvx_v8hf (src_exp, dest_exp); -#endif + else if (mode == V8BFmode) + stvx = gen_altivec_stvx_v8bf (src_exp, dest_exp); else if (mode == V4SImode) stvx = gen_altivec_stvx_v4si (src_exp, dest_exp); else if (mode == V4SFmode) @@ -1722,10 +1722,10 @@ rs6000_gen_lvx (enum machine_mode mode, rtx dest_exp, rtx src_exp) lvx = gen_altivec_lvx_v16qi (dest_exp, src_exp); else if (mode == V8HImode) lvx = gen_altivec_lvx_v8hi (dest_exp, src_exp); -#ifdef HAVE_V8HFmode else if (mode == V8HFmode) lvx = gen_altivec_lvx_v8hf (dest_exp, src_exp); -#endif + else if (mode == V8BFmode) + lvx = gen_altivec_lvx_v8bf (dest_exp, src_exp); else if (mode == V4SImode) lvx = gen_altivec_lvx_v4si (dest_exp, src_exp); else if (mode == V4SFmode) @@ -1930,11 +1930,7 @@ replace_swapped_load_constant (swap_web_entry *insn_entry, rtx swap_insn) rtx new_const_vector = gen_rtx_CONST_VECTOR (mode, XVEC (vals, 0)); new_mem = force_const_mem (mode, new_const_vector); } - else if ((mode == V8HImode) -#ifdef HAVE_V8HFmode - || (mode == V8HFmode) -#endif - ) + else if (mode == V8HImode || mode == V8HFmode || mode == V8BFmode) { rtx vals = gen_rtx_PARALLEL (mode, rtvec_alloc (8)); int i; diff --git a/gcc/config/rs6000/rs6000.cc b/gcc/config/rs6000/rs6000.cc index 635f05d0d02c..052e95d5a395 100644 --- a/gcc/config/rs6000/rs6000.cc +++ b/gcc/config/rs6000/rs6000.cc @@ -2307,6 +2307,8 @@ rs6000_debug_reg_global (void) V8SImode, V4DImode, V2TImode, + V8BFmode, + V8HFmode, V4SFmode, V2DFmode, V8SFmode, @@ -2952,18 +2954,24 @@ rs6000_init_hard_regno_mode_ok (bool global_init_p) rs6000_vector_unit[V16QImode] = VECTOR_ALTIVEC; rs6000_vector_align[V4SImode] = align32; rs6000_vector_align[V8HImode] = align32; + rs6000_vector_align[V8HFmode] = align32; + rs6000_vector_align[V8BFmode] = align32; rs6000_vector_align[V16QImode] = align32; if (TARGET_VSX) { rs6000_vector_mem[V4SImode] = VECTOR_VSX; rs6000_vector_mem[V8HImode] = VECTOR_VSX; + rs6000_vector_mem[V8HFmode] = VECTOR_VSX; + rs6000_vector_mem[V8BFmode] = VECTOR_VSX; rs6000_vector_mem[V16QImode] = VECTOR_VSX; } else { rs6000_vector_mem[V4SImode] = VECTOR_ALTIVEC; rs6000_vector_mem[V8HImode] = VECTOR_ALTIVEC; + rs6000_vector_mem[V8HFmode] = VECTOR_ALTIVEC; + rs6000_vector_mem[V8BFmode] = VECTOR_ALTIVEC; rs6000_vector_mem[V16QImode] = VECTOR_ALTIVEC; } } @@ -3078,6 +3086,10 @@ rs6000_init_hard_regno_mode_ok (bool global_init_p) reg_addr[V16QImode].reload_load = CODE_FOR_reload_v16qi_di_load; reg_addr[V8HImode].reload_store = CODE_FOR_reload_v8hi_di_store; reg_addr[V8HImode].reload_load = CODE_FOR_reload_v8hi_di_load; + reg_addr[V8BFmode].reload_store = CODE_FOR_reload_v8bf_di_store; + reg_addr[V8BFmode].reload_load = CODE_FOR_reload_v8bf_di_load; + reg_addr[V8HFmode].reload_store = CODE_FOR_reload_v8hf_di_store; + reg_addr[V8HFmode].reload_load = CODE_FOR_reload_v8hf_di_load; reg_addr[V4SImode].reload_store = CODE_FOR_reload_v4si_di_store; reg_addr[V4SImode].reload_load = CODE_FOR_reload_v4si_di_load; reg_addr[V2DImode].reload_store = CODE_FOR_reload_v2di_di_store; @@ -3129,6 +3141,8 @@ rs6000_init_hard_regno_mode_ok (bool global_init_p) reg_addr[V2DImode].reload_gpr_vsx = CODE_FOR_reload_gpr_from_vsxv2di; reg_addr[V4SFmode].reload_gpr_vsx = CODE_FOR_reload_gpr_from_vsxv4sf; reg_addr[V4SImode].reload_gpr_vsx = CODE_FOR_reload_gpr_from_vsxv4si; + reg_addr[V8BFmode].reload_gpr_vsx = CODE_FOR_reload_gpr_from_vsxv8bf; + reg_addr[V8HFmode].reload_gpr_vsx = CODE_FOR_reload_gpr_from_vsxv8hf; reg_addr[V8HImode].reload_gpr_vsx = CODE_FOR_reload_gpr_from_vsxv8hi; reg_addr[V16QImode].reload_gpr_vsx = CODE_FOR_reload_gpr_from_vsxv16qi; reg_addr[SFmode].reload_gpr_vsx = CODE_FOR_reload_gpr_from_vsxsf; @@ -3139,6 +3153,8 @@ rs6000_init_hard_regno_mode_ok (bool global_init_p) reg_addr[V2DImode].reload_vsx_gpr = CODE_FOR_reload_vsx_from_gprv2di; reg_addr[V4SFmode].reload_vsx_gpr = CODE_FOR_reload_vsx_from_gprv4sf; reg_addr[V4SImode].reload_vsx_gpr = CODE_FOR_reload_vsx_from_gprv4si; + reg_addr[V8BFmode].reload_vsx_gpr = CODE_FOR_reload_vsx_from_gprv8bf; + reg_addr[V8HFmode].reload_vsx_gpr = CODE_FOR_reload_vsx_from_gprv8hf; reg_addr[V8HImode].reload_vsx_gpr = CODE_FOR_reload_vsx_from_gprv8hi; reg_addr[V16QImode].reload_vsx_gpr = CODE_FOR_reload_vsx_from_gprv16qi; reg_addr[SFmode].reload_vsx_gpr = CODE_FOR_reload_vsx_from_gprsf; @@ -3176,6 +3192,10 @@ rs6000_init_hard_regno_mode_ok (bool global_init_p) reg_addr[V2DImode].reload_load = CODE_FOR_reload_v2di_si_load; reg_addr[V1TImode].reload_store = CODE_FOR_reload_v1ti_si_store; reg_addr[V1TImode].reload_load = CODE_FOR_reload_v1ti_si_load; + reg_addr[V8BFmode].reload_store = CODE_FOR_reload_v8bf_si_store; + reg_addr[V8BFmode].reload_load = CODE_FOR_reload_v8bf_si_load; + reg_addr[V8HFmode].reload_store = CODE_FOR_reload_v8hf_si_store; + reg_addr[V8HFmode].reload_load = CODE_FOR_reload_v8hf_si_load; reg_addr[V4SFmode].reload_store = CODE_FOR_reload_v4sf_si_store; reg_addr[V4SFmode].reload_load = CODE_FOR_reload_v4sf_si_load; reg_addr[V2DFmode].reload_store = CODE_FOR_reload_v2df_si_store; @@ -6873,6 +6893,8 @@ output_vec_const_move (rtx *operands) return "vspltisw %0,%1"; case E_V8HImode: + case E_V8HFmode: + case E_V8BFmode: return "vspltish %0,%1"; case E_V16QImode: @@ -7164,7 +7186,9 @@ rs6000_expand_vector_init (rtx target, rtx vals) return; } - if (TARGET_DIRECT_MOVE && (mode == V16QImode || mode == V8HImode)) + if (TARGET_DIRECT_MOVE + && (mode == V16QImode || mode == V8HImode || mode == V8HFmode + || mode == V8BFmode)) { rtx op[16]; /* Force the values into word_mode registers. */ @@ -8739,6 +8763,8 @@ reg_offset_addressing_ok_p (machine_mode mode) { case E_V16QImode: case E_V8HImode: + case E_V8HFmode: + case E_V8BFmode: case E_V4SFmode: case E_V4SImode: case E_V2DFmode: @@ -10847,6 +10873,8 @@ rs6000_const_vec (machine_mode mode) subparts = 4; break; case E_V8HImode: + case E_V8HFmode: + case E_V8BFmode: subparts = 8; break; case E_V16QImode: @@ -11302,6 +11330,8 @@ rs6000_emit_move (rtx dest, rtx source, machine_mode mode) case E_V16QImode: case E_V8HImode: + case E_V8HFmode: + case E_V8BFmode: case E_V4SFmode: case E_V4SImode: case E_V2DFmode: diff --git a/gcc/config/rs6000/rs6000.h b/gcc/config/rs6000/rs6000.h index cae8f269cf1b..abbc9c1c4cfe 100644 --- a/gcc/config/rs6000/rs6000.h +++ b/gcc/config/rs6000/rs6000.h @@ -991,6 +991,8 @@ enum data_align { align_abi, align_opt, align_both }; #define ALTIVEC_VECTOR_MODE(MODE) \ ((MODE) == V16QImode \ || (MODE) == V8HImode \ + || (MODE) == V8HFmode \ + || (MODE) == V8BFmode \ || (MODE) == V4SFmode \ || (MODE) == V4SImode \ || VECTOR_ALIGNMENT_P (MODE)) diff --git a/gcc/config/rs6000/rs6000.md b/gcc/config/rs6000/rs6000.md index 0717e86e9d6b..b188a0e6910f 100644 --- a/gcc/config/rs6000/rs6000.md +++ b/gcc/config/rs6000/rs6000.md @@ -554,6 +554,8 @@ (define_mode_iterator FMOVE128_GPR [TI V16QI V8HI + V8BF + V8HF V4SI V4SF V2DI @@ -718,6 +720,8 @@ (DI "d") (V16QI "b") (V8HI "h") + (V8BF "h") + (V8HF "h") (V4SI "w") (V2DI "d") (V1TI "q") @@ -730,6 +734,8 @@ (DI "d") (V16QI "du") (V8HI "du") + (V8BF "du") + (V8HF "du") (V4SI "du") (V2DI "d")]) @@ -779,6 +785,8 @@ PTI (V16QI "TARGET_ALTIVEC") (V8HI "TARGET_ALTIVEC") + (V8BF "TARGET_ALTIVEC") + (V8HF "TARGET_ALTIVEC") (V4SI "TARGET_ALTIVEC") (V4SF "TARGET_ALTIVEC") (V2DI "TARGET_ALTIVEC") @@ -796,6 +804,8 @@ (PTI "&r,r,r") (V16QI "wa,v,&?r,?r,?r") (V8HI "wa,v,&?r,?r,?r") + (V8BF "wa,v,&?r,?r,?r") + (V8HF "wa,v,&?r,?r,?r") (V4SI "wa,v,&?r,?r,?r") (V4SF "wa,v,&?r,?r,?r") (V2DI "wa,v,&?r,?r,?r") @@ -807,6 +817,8 @@ (PTI "r,0,r") (V16QI "wa,v,r,0,r") (V8HI "wa,v,r,0,r") + (V8BF "wa,v,r,0,r") + (V8HF "wa,v,r,0,r") (V4SI "wa,v,r,0,r") (V4SF "wa,v,r,0,r") (V2DI "wa,v,r,0,r") @@ -818,6 +830,8 @@ (PTI "r,r,0") (V16QI "wa,v,r,r,0") (V8HI "wa,v,r,r,0") + (V8BF "wa,v,r,r,0") + (V8HF "wa,v,r,r,0") (V4SI "wa,v,r,r,0") (V4SF "wa,v,r,r,0") (V2DI "wa,v,r,r,0") @@ -831,6 +845,8 @@ (PTI "r,0,0") (V16QI "wa,v,r,0,0") (V8HI "wa,v,r,0,0") + (V8BF "wa,v,r,0,0") + (V8HF "wa,v,r,0,0") (V4SI "wa,v,r,0,0") (V4SF "wa,v,r,0,0") (V2DI "wa,v,r,0,0") @@ -839,8 +855,8 @@ ;; Reload iterator for creating the function to allocate a base register to ;; supplement addressing modes. -(define_mode_iterator RELOAD [V16QI V8HI V4SI V2DI V4SF V2DF V1TI - SF SD SI DF DD DI TI PTI KF IF TF +(define_mode_iterator RELOAD [V16QI V8HI V8BF V8HF V4SI V2DI V4SF V2DF V1TI + SF SD SI DF DD DI TI PTI KF IF TF HF BF OO XO]) ;; Iterate over smin, smax diff --git a/gcc/config/rs6000/vector.md b/gcc/config/rs6000/vector.md index f5797387ca79..0b9727cca35c 100644 --- a/gcc/config/rs6000/vector.md +++ b/gcc/config/rs6000/vector.md @@ -50,11 +50,31 @@ (define_mode_iterator VEC_K [V16QI V8HI V4SI V4SF]) ;; Vector logical modes -(define_mode_iterator VEC_L [V16QI V8HI V4SI V2DI V4SF V2DF V1TI TI KF TF]) +(define_mode_iterator VEC_L [V16QI + V8HI + V8BF + V8HF + V4SI + V2DI + V4SF + V2DF + V1TI + TI + KF + TF]) ;; Vector modes for moves. Don't do TImode or TFmode here, since their ;; moves are handled elsewhere. -(define_mode_iterator VEC_M [V16QI V8HI V4SI V2DI V4SF V2DF V1TI KF]) +(define_mode_iterator VEC_M [V16QI + V8HI + V4SI + V2DI + V8BF + V8HF + V4SF + V2DF + V1TI + KF]) ;; Vector modes for types that don't need a realignment under VSX (define_mode_iterator VEC_N [V4SI V4SF V2DI V2DF V1TI KF TF]) @@ -63,7 +83,14 @@ (define_mode_iterator VEC_C [V16QI V8HI V4SI V2DI V4SF V2DF V1TI]) ;; Vector init/extract modes -(define_mode_iterator VEC_E [V16QI V8HI V4SI V2DI V4SF V2DF]) +(define_mode_iterator VEC_E [V16QI + V8HI + V4SI + V2DI + V8BF + V8HF + V4SF + V2DF]) ;; Vector modes for 64-bit base types (define_mode_iterator VEC_64 [V2DI V2DF]) @@ -76,6 +103,8 @@ (V8HI "HI") (V4SI "SI") (V2DI "DI") + (V8BF "BF") + (V8HF "HF") (V4SF "SF") (V2DF "DF") (V1TI "TI") @@ -86,6 +115,8 @@ (V8HI "hi") (V4SI "si") (V2DI "di") + (V8BF "bf") + (V8HF "hf") (V4SF "sf") (V2DF "df") (V1TI "ti") diff --git a/gcc/config/rs6000/vsx.md b/gcc/config/rs6000/vsx.md index dd3573b80868..3198802dabb9 100644 --- a/gcc/config/rs6000/vsx.md +++ b/gcc/config/rs6000/vsx.md @@ -46,9 +46,14 @@ ;; Iterator for vector floating point types supported by VSX (define_mode_iterator VSX_F [V4SF V2DF]) +;; Iterator for 8 element vectors +(define_mode_iterator VECTOR_16BIT [V8HI V8BF V8HF]) + ;; Iterator for logical types supported by VSX (define_mode_iterator VSX_L [V16QI V8HI + V8BF + V8HF V4SI V2DI V4SF @@ -61,6 +66,8 @@ ;; Iterator for memory moves. (define_mode_iterator VSX_M [V16QI V8HI + V8BF + V8HF V4SI V2DI V4SF @@ -71,6 +78,8 @@ TI]) (define_mode_attr VSX_XXBR [(V8HI "h") + (V8BF "h") + (V8HF "h") (V4SI "w") (V4SF "w") (V2DF "d") @@ -80,6 +89,8 @@ ;; Map into the appropriate load/store name based on the type (define_mode_attr VSm [(V16QI "vw4") (V8HI "vw4") + (V8BF "vw4") + (V8HF "vw4") (V4SI "vw4") (V4SF "vw4") (V2DF "vd2") @@ -93,6 +104,8 @@ ;; Map the register class used (define_mode_attr VSr [(V16QI "v") (V8HI "v") + (V8BF "v") + (V8HF "v") (V4SI "v") (V4SF "wa") (V2DI "wa") @@ -108,6 +121,8 @@ ;; What value we need in the "isa" field, to make the IEEE QP float work. (define_mode_attr VSisa [(V16QI "*") (V8HI "*") + (V8BF "p10") + (V8HF "p9v") (V4SI "*") (V4SF "*") (V2DI "*") @@ -124,6 +139,8 @@ ;; integer modes. (define_mode_attr ??r [(V16QI "??r") (V8HI "??r") + (V8BF "??r") + (V8HF "??r") (V4SI "??r") (V4SF "??r") (V2DI "??r") @@ -136,6 +153,8 @@ ;; A mode attribute used for 128-bit constant values. (define_mode_attr nW [(V16QI "W") (V8HI "W") + (V8BF "W") + (V8HF "W") (V4SI "W") (V4SF "W") (V2DI "W") @@ -163,6 +182,8 @@ ;; operation (define_mode_attr VSv [(V16QI "v") (V8HI "v") + (V8BF "v") + (V8HF "v") (V4SI "v") (V4SF "v") (V2DI "v") @@ -396,6 +417,8 @@ ;; Like VM2 in altivec.md, just do char, short, int, long, float and double (define_mode_iterator VM3 [V4SI V8HI + V8BF + V8HF V16QI V4SF V2DF @@ -407,6 +430,8 @@ (define_mode_attr VM3_char [(V2DI "d") (V4SI "w") (V8HI "h") + (V8BF "h") + (V8HF "h") (V16QI "b") (V2DF "d") (V4SF "w")]) @@ -541,21 +566,21 @@ [(set_attr "type" "vecload") (set_attr "length" "8")]) -(define_insn_and_split "*vsx_le_perm_load_v8hi" - [(set (match_operand:V8HI 0 "vsx_register_operand" "=wa") - (match_operand:V8HI 1 "indexed_or_indirect_operand" "Z"))] +(define_insn_and_split "*vsx_le_perm_load_<mode>" + [(set (match_operand:VECTOR_16BIT 0 "vsx_register_operand" "=wa") + (match_operand:VECTOR_16BIT 1 "indexed_or_indirect_operand" "Z"))] "!BYTES_BIG_ENDIAN && TARGET_VSX && !TARGET_P9_VECTOR" "#" "&& 1" [(set (match_dup 2) - (vec_select:V8HI + (vec_select:VECTOR_16BIT (match_dup 1) (parallel [(const_int 4) (const_int 5) (const_int 6) (const_int 7) (const_int 0) (const_int 1) (const_int 2) (const_int 3)]))) (set (match_dup 0) - (vec_select:V8HI + (vec_select:VECTOR_16BIT (match_dup 2) (parallel [(const_int 4) (const_int 5) (const_int 6) (const_int 7) @@ -802,27 +827,27 @@ (const_int 0) (const_int 1)])))] "") -(define_insn "*vsx_le_perm_store_v8hi" - [(set (match_operand:V8HI 0 "indexed_or_indirect_operand" "=Z") - (match_operand:V8HI 1 "vsx_register_operand" "wa"))] +(define_insn "*vsx_le_perm_store_<mode>" + [(set (match_operand:VECTOR_16BIT 0 "indexed_or_indirect_operand" "=Z") + (match_operand:VECTOR_16BIT 1 "vsx_register_operand" "wa"))] "!BYTES_BIG_ENDIAN && TARGET_VSX && !TARGET_P9_VECTOR" "#" [(set_attr "type" "vecstore") (set_attr "length" "12")]) (define_split - [(set (match_operand:V8HI 0 "indexed_or_indirect_operand") - (match_operand:V8HI 1 "vsx_register_operand"))] + [(set (match_operand:VECTOR_16BIT 0 "indexed_or_indirect_operand") + (match_operand:VECTOR_16BIT 1 "vsx_register_operand"))] "!BYTES_BIG_ENDIAN && TARGET_VSX && !TARGET_P9_VECTOR && !reload_completed" [(set (match_dup 2) - (vec_select:V8HI + (vec_select:VECTOR_16BIT (match_dup 1) (parallel [(const_int 4) (const_int 5) (const_int 6) (const_int 7) (const_int 0) (const_int 1) (const_int 2) (const_int 3)]))) (set (match_dup 0) - (vec_select:V8HI + (vec_select:VECTOR_16BIT (match_dup 2) (parallel [(const_int 4) (const_int 5) (const_int 6) (const_int 7) @@ -861,25 +886,25 @@ ;; The post-reload split requires that we re-permute the source ;; register in case it is still live. (define_split - [(set (match_operand:V8HI 0 "indexed_or_indirect_operand") - (match_operand:V8HI 1 "vsx_register_operand"))] + [(set (match_operand:VECTOR_16BIT 0 "indexed_or_indirect_operand") + (match_operand:VECTOR_16BIT 1 "vsx_register_operand"))] "!BYTES_BIG_ENDIAN && TARGET_VSX && !TARGET_P9_VECTOR && reload_completed" [(set (match_dup 1) - (vec_select:V8HI + (vec_select:VECTOR_16BIT (match_dup 1) (parallel [(const_int 4) (const_int 5) (const_int 6) (const_int 7) (const_int 0) (const_int 1) (const_int 2) (const_int 3)]))) (set (match_dup 0) - (vec_select:V8HI + (vec_select:VECTOR_16BIT (match_dup 1) (parallel [(const_int 4) (const_int 5) (const_int 6) (const_int 7) (const_int 0) (const_int 1) (const_int 2) (const_int 3)]))) (set (match_dup 1) - (vec_select:V8HI + (vec_select:VECTOR_16BIT (match_dup 1) (parallel [(const_int 4) (const_int 5) (const_int 6) (const_int 7) @@ -1434,15 +1459,15 @@ "lxvw4x %x0,%y1" [(set_attr "type" "vecload")]) -(define_expand "vsx_ld_elemrev_v8hi" - [(set (match_operand:V8HI 0 "vsx_register_operand" "=wa") - (vec_select:V8HI - (match_operand:V8HI 1 "memory_operand" "Z") +(define_expand "vsx_ld_elemrev_<mode>" + [(set (match_operand:VECTOR_16BIT 0 "vsx_register_operand" "=wa") + (vec_select:VECTOR_16BIT + (match_operand:VECTOR_16BIT 1 "memory_operand" "Z") (parallel [(const_int 7) (const_int 6) (const_int 5) (const_int 4) (const_int 3) (const_int 2) (const_int 1) (const_int 0)])))] - "VECTOR_MEM_VSX_P (V8HImode) && !BYTES_BIG_ENDIAN" + "VECTOR_MEM_VSX_P (<MODE>mode) && !BYTES_BIG_ENDIAN" { if (!TARGET_P9_VECTOR) { @@ -1452,9 +1477,9 @@ unsigned int reorder[16] = {13,12,15,14,9,8,11,10,5,4,7,6,1,0,3,2}; int i; - subreg = simplify_gen_subreg (V4SImode, operands[1], V8HImode, 0); + subreg = simplify_gen_subreg (V4SImode, operands[1], <MODE>mode, 0); emit_insn (gen_vsx_ld_elemrev_v4si (tmp, subreg)); - subreg2 = simplify_gen_subreg (V8HImode, tmp, V4SImode, 0); + subreg2 = simplify_gen_subreg (<MODE>mode, tmp, V4SImode, 0); for (i = 0; i < 16; ++i) perm[i] = GEN_INT (reorder[i]); @@ -1462,21 +1487,21 @@ pcv = force_reg (V16QImode, gen_rtx_CONST_VECTOR (V16QImode, gen_rtvec_v (16, perm))); - emit_insn (gen_altivec_vperm_v8hi_direct (operands[0], subreg2, - subreg2, pcv)); + emit_insn (gen_altivec_vperm_<mode>_direct (operands[0], subreg2, + subreg2, pcv)); DONE; } }) -(define_insn "*vsx_ld_elemrev_v8hi_internal" - [(set (match_operand:V8HI 0 "vsx_register_operand" "=wa") - (vec_select:V8HI - (match_operand:V8HI 1 "memory_operand" "Z") +(define_insn "*vsx_ld_elemrev_<mode>_internal" + [(set (match_operand:VECTOR_16BIT 0 "vsx_register_operand" "=wa") + (vec_select:VECTOR_16BIT + (match_operand:VECTOR_16BIT 1 "memory_operand" "Z") (parallel [(const_int 7) (const_int 6) (const_int 5) (const_int 4) (const_int 3) (const_int 2) (const_int 1) (const_int 0)])))] - "VECTOR_MEM_VSX_P (V8HImode) && !BYTES_BIG_ENDIAN && TARGET_P9_VECTOR" + "VECTOR_MEM_VSX_P (<MODE>mode) && !BYTES_BIG_ENDIAN && TARGET_P9_VECTOR" "lxvh8x %x0,%y1" [(set_attr "type" "vecload")]) @@ -1584,20 +1609,20 @@ "stxvw4x %x1,%y0" [(set_attr "type" "vecstore")]) -(define_expand "vsx_st_elemrev_v8hi" - [(set (match_operand:V8HI 0 "memory_operand" "=Z") - (vec_select:V8HI - (match_operand:V8HI 1 "vsx_register_operand" "wa") +(define_expand "vsx_st_elemrev_<mode>" + [(set (match_operand:VECTOR_16BIT 0 "memory_operand" "=Z") + (vec_select:VECTOR_16BIT + (match_operand:VECTOR_16BIT 1 "vsx_register_operand" "wa") (parallel [(const_int 7) (const_int 6) (const_int 5) (const_int 4) (const_int 3) (const_int 2) (const_int 1) (const_int 0)])))] - "VECTOR_MEM_VSX_P (V8HImode) && !BYTES_BIG_ENDIAN" + "VECTOR_MEM_VSX_P (<MODE>mode) && !BYTES_BIG_ENDIAN" { if (!TARGET_P9_VECTOR) { rtx mem_subreg, subreg, perm[16], pcv; - rtx tmp = gen_reg_rtx (V8HImode); + rtx tmp = gen_reg_rtx (<MODE>mode); /* 2 is leftmost element in register */ unsigned int reorder[16] = {13,12,15,14,9,8,11,10,5,4,7,6,1,0,3,2}; int i; @@ -1608,10 +1633,10 @@ pcv = force_reg (V16QImode, gen_rtx_CONST_VECTOR (V16QImode, gen_rtvec_v (16, perm))); - emit_insn (gen_altivec_vperm_v8hi_direct (tmp, operands[1], - operands[1], pcv)); - subreg = simplify_gen_subreg (V4SImode, tmp, V8HImode, 0); - mem_subreg = simplify_gen_subreg (V4SImode, operands[0], V8HImode, 0); + emit_insn (gen_altivec_vperm_<mode>_direct (tmp, operands[1], + operands[1], pcv)); + subreg = simplify_gen_subreg (V4SImode, tmp, <MODE>mode, 0); + mem_subreg = simplify_gen_subreg (V4SImode, operands[0], <MODE>mode, 0); emit_insn (gen_vsx_st_elemrev_v4si (mem_subreg, subreg)); DONE; } @@ -1626,15 +1651,15 @@ "stxvd2x %x1,%y0" [(set_attr "type" "vecstore")]) -(define_insn "*vsx_st_elemrev_v8hi_internal" - [(set (match_operand:V8HI 0 "memory_operand" "=Z") - (vec_select:V8HI - (match_operand:V8HI 1 "vsx_register_operand" "wa") +(define_insn "*vsx_st_elemrev_<mode>_internal" + [(set (match_operand:VECTOR_16BIT 0 "memory_operand" "=Z") + (vec_select:VECTOR_16BIT + (match_operand:VECTOR_16BIT 1 "vsx_register_operand" "wa") (parallel [(const_int 7) (const_int 6) (const_int 5) (const_int 4) (const_int 3) (const_int 2) (const_int 1) (const_int 0)])))] - "VECTOR_MEM_VSX_P (V8HImode) && !BYTES_BIG_ENDIAN && TARGET_P9_VECTOR" + "VECTOR_MEM_VSX_P (<MODE>mode) && !BYTES_BIG_ENDIAN && TARGET_P9_VECTOR" "stxvh8x %x1,%y0" [(set_attr "type" "vecstore")]) @@ -3299,10 +3324,10 @@ "xxpermdi %x0,%x1,%x1,2" [(set_attr "type" "vecperm")]) -(define_insn "xxswapd_v8hi" - [(set (match_operand:V8HI 0 "vsx_register_operand" "=wa") - (vec_select:V8HI - (match_operand:V8HI 1 "vsx_register_operand" "wa") +(define_insn "xxswapd_<mode>" + [(set (match_operand:VECTOR_16BIT 0 "vsx_register_operand" "=wa") + (vec_select:VECTOR_16BIT + (match_operand:VECTOR_16BIT 1 "vsx_register_operand" "wa") (parallel [(const_int 4) (const_int 5) (const_int 6) (const_int 7) (const_int 0) (const_int 1) @@ -3402,15 +3427,15 @@ "lxvd2x %x0,%y1" [(set_attr "type" "vecload")]) -(define_insn "*vsx_lxvd2x8_le_V8HI" - [(set (match_operand:V8HI 0 "vsx_register_operand" "=wa") - (vec_select:V8HI - (match_operand:V8HI 1 "memory_operand" "Z") +(define_insn "*vsx_lxvd2x8_le_<MODE>" + [(set (match_operand:VECTOR_16BIT 0 "vsx_register_operand" "=wa") + (vec_select:VECTOR_16BIT + (match_operand:VECTOR_16BIT 1 "memory_operand" "Z") (parallel [(const_int 4) (const_int 5) (const_int 6) (const_int 7) (const_int 0) (const_int 1) (const_int 2) (const_int 3)])))] - "!BYTES_BIG_ENDIAN && VECTOR_MEM_VSX_P (V8HImode) && !TARGET_P9_VECTOR" + "!BYTES_BIG_ENDIAN && VECTOR_MEM_VSX_P (<MODE>mode) && !TARGET_P9_VECTOR" "lxvd2x %x0,%y1" [(set_attr "type" "vecload")]) @@ -3478,15 +3503,15 @@ [(set_attr "type" "vecstore") (set_attr "length" "8")]) -(define_insn "*vsx_stxvd2x8_le_V8HI" - [(set (match_operand:V8HI 0 "memory_operand" "=Z") - (vec_select:V8HI - (match_operand:V8HI 1 "vsx_register_operand" "wa") +(define_insn "*vsx_stxvd2x8_le_<MODE>" + [(set (match_operand:VECTOR_16BIT 0 "memory_operand" "=Z") + (vec_select:VECTOR_16BIT + (match_operand:VECTOR_16BIT 1 "vsx_register_operand" "wa") (parallel [(const_int 4) (const_int 5) (const_int 6) (const_int 7) (const_int 0) (const_int 1) (const_int 2) (const_int 3)])))] - "!BYTES_BIG_ENDIAN && VECTOR_MEM_VSX_P (V8HImode) && !TARGET_P9_VECTOR" + "!BYTES_BIG_ENDIAN && VECTOR_MEM_VSX_P (<MODE>mode) && !TARGET_P9_VECTOR" "stxvd2x %x1,%y0" [(set_attr "type" "vecstore")]) @@ -4060,7 +4085,8 @@ if (which_alternative == 0 && ((<MODE>mode == V16QImode && INTVAL (operands[2]) == (BYTES_BIG_ENDIAN ? 7 : 8)) - || (<MODE>mode == V8HImode + || ((<MODE>mode == V8HImode || <MODE>mode == V8HFmode + || <MODE>mode == V8BFmode) && INTVAL (operands[2]) == (BYTES_BIG_ENDIAN ? 3 : 4)))) { enum machine_mode dest_mode = GET_MODE (operands[0]); @@ -4139,7 +4165,8 @@ else vec_tmp = src; } - else if (<MODE>mode == V8HImode) + else if (<MODE>mode == V8HImode || <MODE>mode == V8HFmode + || <MODE>mode == V8BFmode) { if (value != 3) emit_insn (gen_altivec_vsplth_direct (vec_tmp, src, element));
