X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;f=ir%2Fbe%2Fia32%2Fia32_spec.pl;h=a49559d035db91642954e250cbcca48f8a852916;hb=4137609950786f314fa679f833379ec63444c160;hp=adce379250755f5798eb32689e1c65f79f06583c;hpb=881891c80282c687aae7079eb62d6d69116ce2f0;p=libfirm diff --git a/ir/be/ia32/ia32_spec.pl b/ir/be/ia32/ia32_spec.pl index adce37925..a49559d03 100644 --- a/ir/be/ia32/ia32_spec.pl +++ b/ir/be/ia32/ia32_spec.pl @@ -4,7 +4,6 @@ use File::Basename; -$new_emit_syntax = 1; my $myname = $0; # the cpu architecture (ia32, ia64, mips, sparc, ppc, ...) @@ -17,7 +16,7 @@ $arch = "ia32"; # # => { # op_flags => "N|L|C|X|I|F|Y|H|c|K", -# irn_flags => "R|N|I|S" +# irn_flags => "R|N" # arity => "0|1|2|3 ... |variable|dynamic|any", # state => "floats|pinned|mem_pinned|exc_pinned", # args => [ @@ -28,11 +27,16 @@ $arch = "ia32"; # comment => "any comment for constructor", # reg_req => { in => [ "reg_class|register" ], out => [ "reg_class|register|in_rX" ] }, # cmp_attr => "c source code for comparing node attributes", +# outs => { "out1", "out2" } # optional, creates pn_op_out1, ... consts +# ins => { "in1", "in2" } # optional, creates n_op_in1, ... consts +# mode => "mode_Iu" # optional, predefines the mode # emit => "emit code with templates", -# attr => "attitional attribute arguments for constructor" -# init_attr => "emit attribute initialization template" -# rd_constructor => "c source code which constructs an ir_node" +# attr => "additional attribute arguments for constructor", +# init_attr => "emit attribute initialization template", +# hash_func => "name of the hash function for this operation", +# latency => "latency of this operation (can be float)" # attr_type => "name of the attribute struct", +# modified_flags => [ "CF", ... ] # optional, list of modified flags # }, # # ... # (all nodes you need to describe) @@ -51,13 +55,13 @@ $arch = "ia32"; # H irop_flag_highlevel # c irop_flag_constlike # K irop_flag_keep +# NB irop_flag_dump_noblock +# NI irop_flag_dump_noinput # # irn_flags: special node flags, OPTIONAL (default is 0) # following irn_flags are supported: # R rematerializeable # N not spillable -# I ignore for register allocation -# S modifies stack pointer # # state: state of the operation, OPTIONAL (default is "floats") # @@ -72,10 +76,7 @@ $arch = "ia32"; # # outs: if a node defines more than one output, the names of the projections # nodes having outs having automatically the mode mode_T -# One can also annotate some flags for each out, additional to irn_flags. -# They are separated from name with a colon ':', and concatenated by pipe '|' -# Only I and S are available at the moment (same meaning as in irn_flags). -# example: [ "frame:I", "stack:I|S", "M" ] +# example: [ "frame", "stack", "M" ] # # comment: OPTIONAL comment for the node constructor # @@ -100,7 +101,7 @@ $arch = "ia32"; # 0 - no special type # 1 - caller save (register must be saved by the caller of a function) # 2 - callee save (register must be saved by the called function) -# 4 - ignore (do not assign this register) +# 4 - ignore (do not automatically assign this register) # 8 - emitter can choose an arbitrary register of this class # 16 - the register is a virtual one # 32 - register represents a state @@ -144,14 +145,14 @@ $arch = "ia32"; { mode => "mode_E" } ], vfp => [ - { name => "vf0", type => 1 | 16 }, - { name => "vf1", type => 1 | 16 }, - { name => "vf2", type => 1 | 16 }, - { name => "vf3", type => 1 | 16 }, - { name => "vf4", type => 1 | 16 }, - { name => "vf5", type => 1 | 16 }, - { name => "vf6", type => 1 | 16 }, - { name => "vf7", type => 1 | 16 }, + { name => "vf0", type => 1 }, + { name => "vf1", type => 1 }, + { name => "vf2", type => 1 }, + { name => "vf3", type => 1 }, + { name => "vf4", type => 1 }, + { name => "vf5", type => 1 }, + { name => "vf6", type => 1 }, + { name => "vf7", type => 1 }, { name => "vfp_NOREG", type => 4 | 8 | 16 }, # we need a dummy register for NoReg nodes { name => "vfp_UKNWN", type => 4 | 8 | 16 }, # we need a dummy register for Unknown nodes { mode => "mode_E" } @@ -194,63 +195,57 @@ $arch = "ia32"; S1 => "${arch}_emit_source_register(node, 1);", S2 => "${arch}_emit_source_register(node, 2);", S3 => "${arch}_emit_source_register(node, 3);", + SB0 => "${arch}_emit_8bit_source_register_or_immediate(node, 0);", SB1 => "${arch}_emit_8bit_source_register_or_immediate(node, 1);", SB2 => "${arch}_emit_8bit_source_register_or_immediate(node, 2);", SB3 => "${arch}_emit_8bit_source_register_or_immediate(node, 3);", + SH0 => "${arch}_emit_8bit_high_source_register(node, 0);", + SS0 => "${arch}_emit_16bit_source_register_or_immediate(node, 0);", + SI0 => "${arch}_emit_source_register_or_immediate(node, 0);", SI1 => "${arch}_emit_source_register_or_immediate(node, 1);", SI3 => "${arch}_emit_source_register_or_immediate(node, 3);", D0 => "${arch}_emit_dest_register(node, 0);", D1 => "${arch}_emit_dest_register(node, 1);", + DS0 => "${arch}_emit_dest_register_size(node, 0);", DB0 => "${arch}_emit_8bit_dest_register(node, 0);", X0 => "${arch}_emit_x87_register(node, 0);", X1 => "${arch}_emit_x87_register(node, 1);", - SE => "${arch}_emit_extend_suffix(get_ia32_ls_mode(node));", - ME => "if(get_mode_size_bits(get_ia32_ls_mode(node)) != 32)\n - ia32_emit_mode_suffix(node);", + EX => "${arch}_emit_extend_suffix(node);", M => "${arch}_emit_mode_suffix(node);", XM => "${arch}_emit_x87_mode_suffix(node);", XXM => "${arch}_emit_xmm_mode_suffix(node);", XSD => "${arch}_emit_xmm_mode_suffix_s(node);", AM => "${arch}_emit_am(node);", - unop3 => "${arch}_emit_unop(node, 3);", - unop4 => "${arch}_emit_unop(node, 4);", + unop3 => "${arch}_emit_unop(node, n_ia32_unary_op);", + unop4 => "${arch}_emit_unop(node, n_ia32_binary_right);", binop => "${arch}_emit_binop(node);", x87_binop => "${arch}_emit_x87_binop(node);", CMP0 => "${arch}_emit_cmp_suffix_node(node, 0);", CMP3 => "${arch}_emit_cmp_suffix_node(node, 3);", ); -#--------------------------------------------------# -# _ # -# (_) # -# _ __ _____ __ _ _ __ ___ _ __ ___ # -# | '_ \ / _ \ \ /\ / / | | '__| / _ \| '_ \/ __| # -# | | | | __/\ V V / | | | | (_) | |_) \__ \ # -# |_| |_|\___| \_/\_/ |_|_| \___/| .__/|___/ # -# | | # -# |_| # -#--------------------------------------------------# + + $default_op_attr_type = "ia32_op_attr_t"; $default_attr_type = "ia32_attr_t"; $default_copy_attr = "ia32_copy_attr"; sub ia32_custom_init_attr { - my $node = shift; - my $name = shift; - my $res = ""; + my $constr = shift; + my $node = shift; + my $name = shift; + my $res = ""; if(defined($node->{modified_flags})) { - $res .= "\tset_ia32_flags(res, get_ia32_flags(res) | arch_irn_flags_modify_flags);\n"; + $res .= "\tarch_irn_add_flags(res, arch_irn_flags_modify_flags);\n"; } if(defined($node->{am})) { my $am = $node->{am}; if($am eq "source,unary") { - $res .= "\tset_ia32_am_support(res, ia32_am_Source, ia32_am_unary);"; + $res .= "\tset_ia32_am_support(res, ia32_am_unary);"; } elsif($am eq "source,binary") { - $res .= "\tset_ia32_am_support(res, ia32_am_Source, ia32_am_binary);"; - } elsif($am eq "source,ternary") { - $res .= "\tset_ia32_am_support(res, ia32_am_Source, ia32_am_ternary);"; + $res .= "\tset_ia32_am_support(res, ia32_am_binary);"; } elsif($am eq "none") { # nothing to do } else { @@ -268,54 +263,64 @@ sub ia32_custom_init_attr { $custom_init_attr_func = \&ia32_custom_init_attr; %init_attr = ( - ia32_attr_t => "\tinit_ia32_attributes(res, flags, in_reqs, out_reqs, exec_units, n_res);", - ia32_x87_attr_t => - "\tinit_ia32_attributes(res, flags, in_reqs, out_reqs, exec_units, n_res);\n". - "\tinit_ia32_x87_attributes(res);", ia32_asm_attr_t => - "\tinit_ia32_attributes(res, flags, in_reqs, out_reqs, exec_units, n_res);\n". + "\tinit_ia32_attributes(res, flags, in_reqs, exec_units, n_res);\n". "\tinit_ia32_x87_attributes(res);". "\tinit_ia32_asm_attributes(res);", - ia32_immediate_attr_t => - "\tinit_ia32_attributes(res, flags, in_reqs, out_reqs, exec_units, n_res);\n". - "\tinit_ia32_immediate_attributes(res, symconst, symconst_sign, offset);", - ia32_copyb_attr_t => - "\tinit_ia32_attributes(res, flags, in_reqs, out_reqs, exec_units, n_res);\n". - "\tinit_ia32_copyb_attributes(res, size);", + ia32_attr_t => + "\tinit_ia32_attributes(res, flags, in_reqs, exec_units, n_res);", + ia32_call_attr_t => + "\tinit_ia32_attributes(res, flags, in_reqs, exec_units, n_res);\n". + "\tinit_ia32_call_attributes(res, pop, call_tp);", ia32_condcode_attr_t => - "\tinit_ia32_attributes(res, flags, in_reqs, out_reqs, exec_units, n_res);\n". + "\tinit_ia32_attributes(res, flags, in_reqs, exec_units, n_res);\n". "\tinit_ia32_condcode_attributes(res, pnc);", + ia32_copyb_attr_t => + "\tinit_ia32_attributes(res, flags, in_reqs, exec_units, n_res);\n". + "\tinit_ia32_copyb_attributes(res, size);", + ia32_immediate_attr_t => + "\tinit_ia32_attributes(res, flags, in_reqs, exec_units, n_res);\n". + "\tinit_ia32_immediate_attributes(res, symconst, symconst_sign, no_pic_adjust, offset);", + ia32_x87_attr_t => + "\tinit_ia32_attributes(res, flags, in_reqs, exec_units, n_res);\n". + "\tinit_ia32_x87_attributes(res);", + ia32_climbframe_attr_t => + "\tinit_ia32_attributes(res, flags, in_reqs, exec_units, n_res);\n". + "\tinit_ia32_climbframe_attributes(res, count);", ); %compare_attr = ( - ia32_attr_t => "ia32_compare_nodes_attr", - ia32_x87_attr_t => "ia32_compare_x87_attr", ia32_asm_attr_t => "ia32_compare_asm_attr", - ia32_immediate_attr_t => "ia32_compare_immediate_attr", - ia32_copyb_attr_t => "ia32_compare_copyb_attr", - ia32_condcode_attr_t => "ia32_compare_condcode_attr", + ia32_attr_t => "ia32_compare_nodes_attr", + ia32_call_attr_t => "ia32_compare_call_attr", + ia32_condcode_attr_t => "ia32_compare_condcode_attr", + ia32_copyb_attr_t => "ia32_compare_copyb_attr", + ia32_immediate_attr_t => "ia32_compare_immediate_attr", + ia32_x87_attr_t => "ia32_compare_x87_attr", + ia32_climbframe_attr_t => "ia32_compare_climbframe_attr", ); %operands = ( ); -$mode_xmm = "mode_E"; -$mode_gp = "mode_Iu"; -$mode_flags = "mode_Iu"; -$mode_fpcw = "mode_fpcw"; -$status_flags = [ "CF", "PF", "AF", "ZF", "SF", "OF" ]; -$fpcw_flags = [ "FP_IM", "FP_DM", "FP_ZM", "FP_OM", "FP_UM", "FP_PM", - "FP_PC0", "FP_PC1", "FP_RC0", "FP_RC1", "FP_X" ]; +$mode_xmm = "mode_E"; +$mode_gp = "mode_Iu"; +$mode_flags = "mode_Iu"; +$mode_fpcw = "mode_fpcw"; +$status_flags = [ "CF", "PF", "AF", "ZF", "SF", "OF" ]; +$status_flags_wo_cf = [ "PF", "AF", "ZF", "SF", "OF" ]; +$fpcw_flags = [ "FP_IM", "FP_DM", "FP_ZM", "FP_OM", "FP_UM", "FP_PM", + "FP_PC0", "FP_PC1", "FP_RC0", "FP_RC1", "FP_X" ]; %nodes = ( Immediate => { state => "pinned", op_flags => "c", - irn_flags => "I", - reg_req => { out => [ "gp_NOREG" ] }, - attr => "ir_entity *symconst, int symconst_sign, long offset", + reg_req => { out => [ "gp_NOREG:I" ] }, + attr => "ir_entity *symconst, int symconst_sign, int no_pic_adjust, long offset", attr_type => "ia32_immediate_attr_t", + hash_func => "ia32_hash_Immediate", latency => 0, mode => $mode_gp, }, @@ -329,9 +334,10 @@ Asm => { init_attr => "attr->asm_text = asm_text;\n". "\tattr->register_map = register_map;\n", latency => 10, - modified_flags => 1, + modified_flags => $status_flags, }, +# "allocates" a free register ProduceVal => { op_flags => "c", irn_flags => "R", @@ -343,19 +349,6 @@ ProduceVal => { cmp_attr => "return 1;", }, -#-----------------------------------------------------------------# -# _ _ _ # -# (_) | | | | # -# _ _ __ | |_ ___ __ _ ___ _ __ _ __ ___ __| | ___ ___ # -# | | '_ \| __/ _ \/ _` |/ _ \ '__| | '_ \ / _ \ / _` |/ _ \/ __| # -# | | | | | || __/ (_| | __/ | | | | | (_) | (_| | __/\__ \ # -# |_|_| |_|\__\___|\__, |\___|_| |_| |_|\___/ \__,_|\___||___/ # -# __/ | # -# |___/ # -#-----------------------------------------------------------------# - -# commutative operations - Add => { irn_flags => "R", state => "exc_pinned", @@ -422,13 +415,13 @@ l_Adc => { Mul => { # we should not rematrialize this node. It produces 2 results and has - # very strict constrains + # very strict constraints state => "exc_pinned", reg_req => { in => [ "gp", "gp", "none", "eax", "gp" ], - out => [ "eax", "edx", "none" ] }, - ins => [ "base", "index", "mem", "val_high", "val_low" ], + out => [ "eax", "flags", "none", "edx" ] }, + ins => [ "base", "index", "mem", "left", "right" ], emit => '. mul%M %unop4', - outs => [ "res_low", "res_high", "M" ], + outs => [ "res_low", "flags", "M", "res_high" ], am => "source,binary", latency => 10, units => [ "GP" ], @@ -437,11 +430,13 @@ Mul => { l_Mul => { # we should not rematrialize this node. It produces 2 results and has - # very strict constrains + # very strict constraints op_flags => "C", cmp_attr => "return 1;", - outs => [ "EAX", "EDX", "M" ], - arity => 2 + reg_req => { in => [ "none", "none" ], + out => [ "none", "none", "none", "none" ] }, + ins => [ "left", "right" ], + outs => [ "res_low", "flags", "M", "res_high" ], }, IMul => { @@ -464,10 +459,10 @@ IMul1OP => { irn_flags => "R", state => "exc_pinned", reg_req => { in => [ "gp", "gp", "none", "eax", "gp" ], - out => [ "eax", "edx", "none" ] }, - ins => [ "base", "index", "mem", "val_high", "val_low" ], + out => [ "eax", "flags", "none", "edx" ] }, + ins => [ "base", "index", "mem", "left", "right" ], emit => '. imul%M %unop4', - outs => [ "res_low", "res_high", "M" ], + outs => [ "res_low", "flags", "M", "res_high" ], am => "source,binary", latency => 5, units => [ "GP" ], @@ -477,8 +472,10 @@ IMul1OP => { l_IMul => { op_flags => "C", cmp_attr => "return 1;", - outs => [ "res_low", "res_high", "M" ], - arity => 2 + reg_req => { in => [ "none", "none" ], + out => [ "none", "none", "none", "none" ] }, + ins => [ "left", "right" ], + outs => [ "res_low", "flags", "M", "res_high" ], }, And => { @@ -488,7 +485,6 @@ And => { out => [ "in_r4 in_r5", "flags", "none" ] }, ins => [ "base", "index", "mem", "left", "right" ], outs => [ "res", "flags", "M" ], - op_modes => "commutative | am | immediate | mode_neutral", am => "source,binary", emit => '. and%M %binop', units => [ "GP" ], @@ -575,6 +571,18 @@ Xor => { modified_flags => $status_flags }, +Xor0 => { + op_flags => "c", + irn_flags => "R", + reg_req => { out => [ "gp", "flags" ] }, + outs => [ "res", "flags" ], + emit => ". xor%M %D0, %D0", + units => [ "GP" ], + latency => 1, + mode => $mode_gp, + modified_flags => $status_flags +}, + XorMem => { irn_flags => "R", state => "exc_pinned", @@ -599,8 +607,6 @@ XorMem8Bit => { modified_flags => $status_flags }, -# not commutative operations - Sub => { irn_flags => "R", state => "exc_pinned", @@ -654,6 +660,17 @@ Sbb => { modified_flags => $status_flags }, +Sbb0 => { + irn_flags => "R", + reg_req => { in => [ "flags" ], out => [ "gp", "flags" ] }, + outs => [ "res", "flags" ], + emit => ". sbb%M %D0, %D0", + units => [ "GP" ], + latency => 1, + mode => $mode_gp, + modified_flags => $status_flags +}, + l_Sub => { reg_req => { in => [ "none", "none" ], out => [ "none" ] }, ins => [ "minuend", "subtrahend" ], @@ -671,7 +688,7 @@ IDiv => { out => [ "eax", "flags", "none", "edx", "none" ] }, ins => [ "base", "index", "mem", "divisor", "dividend_low", "dividend_high" ], outs => [ "div_res", "flags", "M", "mod_res", "X_exc" ], - am => "source,ternary", + am => "source,unary", emit => ". idiv%M %unop3", latency => 25, units => [ "GP" ], @@ -685,7 +702,7 @@ Div => { out => [ "eax", "flags", "none", "edx", "none" ] }, ins => [ "base", "index", "mem", "divisor", "dividend_low", "dividend_high" ], outs => [ "div_res", "flags", "M", "mod_res", "X_exc" ], - am => "source,ternary", + am => "source,unary", emit => ". div%M %unop3", latency => 25, units => [ "GP" ], @@ -698,7 +715,7 @@ Shl => { out => [ "in_r1 !in_r2", "flags" ] }, ins => [ "val", "count" ], outs => [ "res", "flags" ], - emit => '. shl %SB1, %S0', + emit => '. shl%M %SB1, %S0', units => [ "GP" ], latency => 1, mode => $mode_gp, @@ -719,8 +736,8 @@ ShlMem => { l_ShlDep => { cmp_attr => "return 1;", + reg_req => { in => [ "none", "none", "none" ], out => [ "none" ] }, ins => [ "val", "count", "dep" ], - arity => 3 }, ShlD => { @@ -738,8 +755,8 @@ ShlD => { l_ShlD => { cmp_attr => "return 1;", + reg_req => { in => [ "none", "none", "none" ], out => [ "none" ] }, ins => [ "val_high", "val_low", "count" ], - arity => 3, }, Shr => { @@ -748,7 +765,7 @@ Shr => { out => [ "in_r1 !in_r2", "flags" ] }, ins => [ "val", "count" ], outs => [ "res", "flags" ], - emit => '. shr %SB1, %S0', + emit => '. shr%M %SB1, %S0', units => [ "GP" ], mode => $mode_gp, latency => 1, @@ -769,8 +786,8 @@ ShrMem => { l_ShrDep => { cmp_attr => "return 1;", + reg_req => { in => [ "none", "none", "none" ], out => [ "none" ] }, ins => [ "val", "count", "dep" ], - arity => 3 }, ShrD => { @@ -788,7 +805,7 @@ ShrD => { l_ShrD => { cmp_attr => "return 1;", - arity => 3, + reg_req => { in => [ "none", "none", "none" ], out => [ "none" ] }, ins => [ "val_high", "val_low", "count" ], }, @@ -798,7 +815,7 @@ Sar => { out => [ "in_r1 !in_r2", "flags" ] }, ins => [ "val", "count" ], outs => [ "res", "flags" ], - emit => '. sar %SB1, %S0', + emit => '. sar%M %SB1, %S0', units => [ "GP" ], latency => 1, mode => $mode_gp, @@ -820,7 +837,7 @@ SarMem => { l_SarDep => { cmp_attr => "return 1;", ins => [ "val", "count", "dep" ], - arity => 3 + reg_req => { in => [ "none", "none", "none" ], out => [ "none" ] }, }, Ror => { @@ -829,7 +846,7 @@ Ror => { out => [ "in_r1 !in_r2", "flags" ] }, ins => [ "val", "count" ], outs => [ "res", "flags" ], - emit => '. ror %SB1, %S0', + emit => '. ror%M %SB1, %S0', units => [ "GP" ], latency => 1, mode => $mode_gp, @@ -854,7 +871,7 @@ Rol => { out => [ "in_r1 !in_r2", "flags" ] }, ins => [ "val", "count" ], outs => [ "res", "flags" ], - emit => '. rol %SB1, %S0', + emit => '. rol%M %SB1, %S0', units => [ "GP" ], latency => 1, mode => $mode_gp, @@ -873,13 +890,11 @@ RolMem => { modified_flags => $status_flags }, -# unary operations - Neg => { irn_flags => "R", reg_req => { in => [ "gp" ], out => [ "in_r1", "flags" ] }, - emit => '. neg %S0', + emit => '. neg%M %S0', ins => [ "val" ], outs => [ "res", "flags" ], units => [ "GP" ], @@ -902,7 +917,7 @@ NegMem => { Minus64Bit => { irn_flags => "R", - reg_req => { in => [ "gp", "gp" ], out => [ "in_r1", "gp" ] }, + reg_req => { in => [ "gp", "gp" ], out => [ "in_r1", "in_r2" ] }, outs => [ "low_res", "high_res" ], units => [ "GP" ], latency => 3, @@ -916,11 +931,11 @@ Inc => { out => [ "in_r1", "flags" ] }, ins => [ "val" ], outs => [ "res", "flags" ], - emit => '. inc %S0', + emit => '. inc%M %S0', units => [ "GP" ], mode => $mode_gp, latency => 1, - modified_flags => [ "OF", "SF", "ZF", "AF", "PF" ] + modified_flags => $status_flags_wo_cf }, IncMem => { @@ -932,7 +947,7 @@ IncMem => { units => [ "GP" ], mode => "mode_M", latency => 1, - modified_flags => [ "OF", "SF", "ZF", "AF", "PF" ] + modified_flags => $status_flags_wo_cf }, Dec => { @@ -941,11 +956,11 @@ Dec => { out => [ "in_r1", "flags" ] }, ins => [ "val" ], outs => [ "res", "flags" ], - emit => '. dec %S0', + emit => '. dec%M %S0', units => [ "GP" ], mode => $mode_gp, latency => 1, - modified_flags => [ "OF", "SF", "ZF", "AF", "PF" ] + modified_flags => $status_flags_wo_cf }, DecMem => { @@ -957,7 +972,7 @@ DecMem => { units => [ "GP" ], mode => "mode_M", latency => 1, - modified_flags => [ "OF", "SF", "ZF", "AF", "PF" ] + modified_flags => $status_flags_wo_cf }, Not => { @@ -966,7 +981,7 @@ Not => { out => [ "in_r1", "flags" ] }, ins => [ "val" ], outs => [ "res", "flags" ], - emit => '. not %S0', + emit => '. not%M %S0', units => [ "GP" ], latency => 1, mode => $mode_gp, @@ -986,8 +1001,8 @@ NotMem => { }, Cmc => { - reg_req => { in => [ "flags" ], out => [ "flags" ] }, - emit => '.cmc', + reg_req => { in => [ "flags" ], out => [ "flags" ] }, + emit => '.cmc', units => [ "GP" ], latency => 1, mode => $mode_flags, @@ -995,22 +1010,21 @@ Cmc => { }, Stc => { - reg_req => { out => [ "flags" ] }, - emit => '.stc', + reg_req => { out => [ "flags" ] }, + emit => '.stc', units => [ "GP" ], latency => 1, mode => $mode_flags, modified_flags => $status_flags }, -# other operations - Cmp => { irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "gp", "gp" ] , out => [ "flags" ] }, + reg_req => { in => [ "gp", "gp", "none", "gp", "gp" ], + out => [ "flags", "none", "none" ] }, ins => [ "base", "index", "mem", "left", "right" ], - outs => [ "eflags" ], + outs => [ "eflags", "unused", "M" ], am => "source,binary", emit => '. cmp%M %binop', attr => "int ins_permuted, int cmp_unsigned", @@ -1025,9 +1039,10 @@ Cmp => { Cmp8Bit => { irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "eax ebx ecx edx", "eax ebx ecx edx" ] , out => [ "flags" ] }, + reg_req => { in => [ "gp", "gp", "none", "eax ebx ecx edx", "eax ebx ecx edx" ] , + out => [ "flags", "none", "none" ] }, ins => [ "base", "index", "mem", "left", "right" ], - outs => [ "eflags" ], + outs => [ "eflags", "unused", "M" ], am => "source,binary", emit => '. cmpb %binop', attr => "int ins_permuted, int cmp_unsigned", @@ -1042,9 +1057,10 @@ Cmp8Bit => { Test => { irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "gp", "gp" ] , out => [ "flags" ] }, + reg_req => { in => [ "gp", "gp", "none", "gp", "gp" ] , + out => [ "flags", "none", "none" ] }, ins => [ "base", "index", "mem", "left", "right" ], - outs => [ "eflags" ], + outs => [ "eflags", "unused", "M" ], am => "source,binary", emit => '. test%M %binop', attr => "int ins_permuted, int cmp_unsigned", @@ -1059,9 +1075,10 @@ Test => { Test8Bit => { irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "eax ebx ecx edx", "eax ebx ecx edx" ] , out => [ "flags" ] }, + reg_req => { in => [ "gp", "gp", "none", "eax ebx ecx edx", "eax ebx ecx edx" ] , + out => [ "flags", "none", "none" ] }, ins => [ "base", "index", "mem", "left", "right" ], - outs => [ "eflags" ], + outs => [ "eflags", "unused", "M" ], am => "source,binary", emit => '. testb %binop', attr => "int ins_permuted, int cmp_unsigned", @@ -1073,42 +1090,42 @@ Test8Bit => { modified_flags => $status_flags }, -Set => { +Setcc => { #irn_flags => "R", reg_req => { in => [ "eflags" ], out => [ "eax ebx ecx edx" ] }, ins => [ "eflags" ], + outs => [ "res" ], attr_type => "ia32_condcode_attr_t", - attr => "pn_Cmp pnc, int ins_permuted", - init_attr => "attr->attr.data.ins_permuted = ins_permuted;\n". - "\tset_ia32_ls_mode(res, mode_Bu);\n", - emit => '. set%CMP0 %DB0', + attr => "pn_Cmp pnc", + init_attr => "set_ia32_ls_mode(res, mode_Bu);\n", latency => 1, units => [ "GP" ], mode => $mode_gp, }, -SetMem => { +SetccMem => { #irn_flags => "R", state => "exc_pinned", reg_req => { in => [ "gp", "gp", "none", "eflags" ], out => [ "none" ] }, ins => [ "base", "index", "mem","eflags" ], attr_type => "ia32_condcode_attr_t", - attr => "pn_Cmp pnc, int ins_permuted", - init_attr => "attr->attr.data.ins_permuted = ins_permuted;\n". - "\tset_ia32_ls_mode(res, mode_Bu);\n", + attr => "pn_Cmp pnc", + init_attr => "set_ia32_ls_mode(res, mode_Bu);\n", emit => '. set%CMP3 %AM', latency => 1, units => [ "GP" ], mode => 'mode_M', }, -CMov => { +CMovcc => { #irn_flags => "R", # (note: leave the false,true order intact to make it compatible with other # ia32_binary ops) state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "gp", "gp", "eflags" ], out => [ "in_r4 in_r5" ] }, + reg_req => { in => [ "gp", "gp", "none", "gp", "gp", "eflags" ], + out => [ "in_r4 in_r5", "flags", "none" ] }, ins => [ "base", "index", "mem", "val_false", "val_true", "eflags" ], + outs => [ "res", "flags", "M" ], am => "source,binary", attr_type => "ia32_condcode_attr_t", attr => "int ins_permuted, pn_Cmp pnc", @@ -1133,13 +1150,24 @@ Jcc => { SwitchJmp => { state => "pinned", op_flags => "L|X|Y", - reg_req => { in => [ "gp" ], out => [ "none" ] }, + reg_req => { in => [ "gp" ] }, mode => "mode_T", attr_type => "ia32_condcode_attr_t", - attr => "pn_Cmp pnc", + attr => "long pnc", latency => 3, units => [ "BRANCH" ], modified_flags => $status_flags, + init_attr => "info->out_infos = NULL;", # XXX ugly hack for out requirements +}, + +Jmp => { + state => "pinned", + irn_flags => "J", + op_flags => "X", + reg_req => { out => [ "none" ] }, + latency => 1, + units => [ "BRANCH" ], + mode => "mode_X", }, IJmp => { @@ -1152,6 +1180,7 @@ IJmp => { latency => 1, units => [ "BRANCH" ], mode => "mode_X", + init_attr => "info->out_infos = NULL;", # XXX ugly hack for out requirements }, Const => { @@ -1159,7 +1188,7 @@ Const => { irn_flags => "R", reg_req => { out => [ "gp" ] }, units => [ "GP" ], - attr => "ir_entity *symconst, int symconst_sign, long offset", + attr => "ir_entity *symconst, int symconst_sign, int no_pic_adjust, long offset", attr_type => "ia32_immediate_attr_t", latency => 1, mode => $mode_gp, @@ -1176,9 +1205,8 @@ GetEIP => { Unknown_GP => { state => "pinned", - op_flags => "c", - irn_flags => "I", - reg_req => { out => [ "gp_UKNWN" ] }, + op_flags => "c|NB", + reg_req => { out => [ "gp_UKNWN:I" ] }, units => [], emit => "", latency => 0, @@ -1187,9 +1215,8 @@ Unknown_GP => { Unknown_VFP => { state => "pinned", - op_flags => "c", - irn_flags => "I", - reg_req => { out => [ "vfp_UKNWN" ] }, + op_flags => "c|NB", + reg_req => { out => [ "vfp_UKNWN:I" ] }, units => [], emit => "", mode => "mode_E", @@ -1199,9 +1226,8 @@ Unknown_VFP => { Unknown_XMM => { state => "pinned", - op_flags => "c", - irn_flags => "I", - reg_req => { out => [ "xmm_UKNWN" ] }, + op_flags => "c|NB", + reg_req => { out => [ "xmm_UKNWN:I" ] }, units => [], emit => "", latency => 0, @@ -1210,9 +1236,8 @@ Unknown_XMM => { NoReg_GP => { state => "pinned", - op_flags => "c", - irn_flags => "I", - reg_req => { out => [ "gp_NOREG" ] }, + op_flags => "c|NB|NI", + reg_req => { out => [ "gp_NOREG:I" ] }, units => [], emit => "", latency => 0, @@ -1221,9 +1246,8 @@ NoReg_GP => { NoReg_VFP => { state => "pinned", - op_flags => "c", - irn_flags => "I", - reg_req => { out => [ "vfp_NOREG" ] }, + op_flags => "c|NB|NI", + reg_req => { out => [ "vfp_NOREG:I" ] }, units => [], emit => "", mode => "mode_E", @@ -1233,9 +1257,8 @@ NoReg_VFP => { NoReg_XMM => { state => "pinned", - op_flags => "c", - irn_flags => "I", - reg_req => { out => [ "xmm_NOREG" ] }, + op_flags => "c|NB|NI", + reg_req => { out => [ "xmm_NOREG:I" ] }, units => [], emit => "", latency => 0, @@ -1245,8 +1268,7 @@ NoReg_XMM => { ChangeCW => { state => "pinned", op_flags => "c", - irn_flags => "I", - reg_req => { out => [ "fp_cw" ] }, + reg_req => { out => [ "fpcw:I" ] }, mode => $mode_fpcw, latency => 3, units => [ "GP" ], @@ -1256,7 +1278,7 @@ ChangeCW => { FldCW => { op_flags => "L|F", state => "pinned", - reg_req => { in => [ "gp", "gp", "none" ], out => [ "fp_cw" ] }, + reg_req => { in => [ "gp", "gp", "none" ], out => [ "fpcw:I" ] }, ins => [ "base", "index", "mem" ], latency => 5, emit => ". fldcw %AM", @@ -1304,29 +1326,15 @@ Cltd => { Load => { op_flags => "L|F", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none" ], out => [ "gp", "none", "none" ] }, + reg_req => { in => [ "gp", "gp", "none" ], + out => [ "gp", "none", "none", "none" ] }, ins => [ "base", "index", "mem" ], - outs => [ "res", "M", "X_exc" ], + outs => [ "res", "unused", "M", "X_exc" ], latency => 0, - emit => ". mov%SE%ME%.l %AM, %D0", + emit => ". mov%EX%.l %AM, %D0", units => [ "GP" ], }, -l_Load => { - op_flags => "L|F", - cmp_attr => "return 1;", - outs => [ "res", "M" ], - arity => 2, -}, - -l_Store => { - op_flags => "L|F", - cmp_attr => "return 1;", - state => "exc_pinned", - arity => 3, - mode => "mode_M", -}, - Store => { op_flags => "L|F", state => "exc_pinned", @@ -1366,20 +1374,30 @@ Lea => { Push => { state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "gp", "esp" ], out => [ "esp", "none" ] }, + reg_req => { in => [ "gp", "gp", "none", "gp", "esp" ], out => [ "esp:I|S", "none" ] }, ins => [ "base", "index", "mem", "val", "stack" ], emit => '. push%M %unop3', - outs => [ "stack:I|S", "M" ], - am => "source,binary", + outs => [ "stack", "M" ], + am => "source,unary", latency => 2, units => [ "GP" ], }, Pop => { state => "exc_pinned", - reg_req => { in => [ "none", "esp" ], out => [ "gp", "none", "none", "esp" ] }, + reg_req => { in => [ "none", "esp" ], out => [ "gp", "none", "none", "esp:I|S" ] }, + ins => [ "mem", "stack" ], + outs => [ "res", "M", "unused", "stack" ], + emit => '. pop%M %D0', + latency => 3, # Pop is more expensive than Push on Athlon + units => [ "GP" ], +}, + +PopEbp => { + state => "exc_pinned", + reg_req => { in => [ "none", "esp" ], out => [ "ebp:I", "none", "none", "esp:I|S" ] }, ins => [ "mem", "stack" ], - outs => [ "res", "M", "unused", "stack:I|S" ], + outs => [ "res", "M", "unused", "stack" ], emit => '. pop%M %D0', latency => 3, # Pop is more expensive than Push on Athlon units => [ "GP" ], @@ -1387,53 +1405,51 @@ Pop => { PopMem => { state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "esp" ], out => [ "none", "none", "none", "esp" ] }, + reg_req => { in => [ "gp", "gp", "none", "esp" ], out => [ "none", "none", "none", "esp:I|S" ] }, ins => [ "base", "index", "mem", "stack" ], - outs => [ "unused0", "M", "unused1", "stack:I|S" ], + outs => [ "unused0", "M", "unused1", "stack" ], emit => '. pop%M %AM', latency => 3, # Pop is more expensive than Push on Athlon units => [ "GP" ], }, Enter => { - reg_req => { in => [ "esp" ], out => [ "ebp", "esp", "none" ] }, + reg_req => { in => [ "esp" ], out => [ "ebp", "esp:I|S", "none" ] }, emit => '. enter', - outs => [ "frame:I", "stack:I|S", "M" ], + outs => [ "frame", "stack", "M" ], latency => 15, units => [ "GP" ], }, Leave => { - reg_req => { in => [ "esp", "ebp" ], out => [ "ebp", "esp" ] }, + reg_req => { in => [ "ebp" ], out => [ "ebp:I", "esp:I|S" ] }, emit => '. leave', - outs => [ "frame:I", "stack:I|S" ], + outs => [ "frame", "stack" ], latency => 3, units => [ "GP" ], }, AddSP => { - irn_flags => "I", state => "pinned", - reg_req => { in => [ "gp", "gp", "none", "esp", "gp" ], out => [ "in_r4", "none" ] }, + reg_req => { in => [ "gp", "gp", "none", "esp", "gp" ], out => [ "esp:I|S", "none" ] }, ins => [ "base", "index", "mem", "stack", "size" ], am => "source,binary", emit => '. addl %binop', latency => 1, - outs => [ "stack:I|S", "M" ], + outs => [ "stack", "M" ], units => [ "GP" ], modified_flags => $status_flags }, SubSP => { -#irn_flags => "I", state => "pinned", - reg_req => { in => [ "gp", "gp", "none", "esp", "gp" ], out => [ "in_r4", "gp", "none" ] }, + reg_req => { in => [ "gp", "gp", "none", "esp", "gp" ], out => [ "esp:I|S", "gp", "none" ] }, ins => [ "base", "index", "mem", "stack", "size" ], am => "source,binary", emit => ". subl %binop\n". ". movl %%esp, %D1", latency => 2, - outs => [ "stack:I|S", "addr", "M" ], + outs => [ "stack", "addr", "M" ], units => [ "GP" ], modified_flags => $status_flags }, @@ -1453,6 +1469,9 @@ LdTls => { latency => 1, }, +# +# BT supports source address mode, but this is unused yet +# Bt => { irn_flags => "R", state => "exc_pinned", @@ -1465,14 +1484,242 @@ Bt => { modified_flags => $status_flags # only CF is set, but the other flags are undefined }, -#-----------------------------------------------------------------------------# -# _____ _____ ______ __ _ _ _ # -# / ____/ ____| ____| / _| | | | | | # -# | (___| (___ | |__ | |_| | ___ __ _| |_ _ __ ___ __| | ___ ___ # -# \___ \\___ \| __| | _| |/ _ \ / _` | __| | '_ \ / _ \ / _` |/ _ \/ __| # -# ____) |___) | |____ | | | | (_) | (_| | |_ | | | | (_) | (_| | __/\__ \ # -# |_____/_____/|______| |_| |_|\___/ \__,_|\__| |_| |_|\___/ \__,_|\___||___/ # -#-----------------------------------------------------------------------------# +Bsf => { + irn_flags => "R", + state => "exc_pinned", + reg_req => { in => [ "gp", "gp", "none", "gp" ], + out => [ "gp", "flags", "none" ] }, + ins => [ "base", "index", "mem", "operand" ], + outs => [ "res", "flags", "M" ], + am => "source,binary", + emit => '. bsf%M %unop3, %D0', + units => [ "GP" ], + latency => 1, + mode => $mode_gp, + modified_flags => $status_flags +}, + +Bsr => { + irn_flags => "R", + state => "exc_pinned", + reg_req => { in => [ "gp", "gp", "none", "gp" ], + out => [ "gp", "flags", "none" ] }, + ins => [ "base", "index", "mem", "operand" ], + outs => [ "res", "flags", "M" ], + am => "source,binary", + emit => '. bsr%M %unop3, %D0', + units => [ "GP" ], + latency => 1, + mode => $mode_gp, + modified_flags => $status_flags +}, + +# +# SSE4.2 or SSE4a popcnt instruction +# +Popcnt => { + irn_flags => "R", + state => "exc_pinned", + reg_req => { in => [ "gp", "gp", "none", "gp" ], + out => [ "gp", "flags", "none" ] }, + ins => [ "base", "index", "mem", "operand" ], + outs => [ "res", "flags", "M" ], + am => "source,binary", + emit => '. popcnt%M %unop3, %D0', + units => [ "GP" ], + latency => 1, + mode => $mode_gp, + modified_flags => $status_flags +}, + +Call => { + state => "exc_pinned", + reg_req => { + in => [ "gp", "gp", "none", "gp", "esp", "fpcw", "eax", "ecx", "edx" ], + out => [ "esp:I|S", "fpcw:I", "none", "eax", "ecx", "edx", "vf0", "vf1", "vf2", "vf3", "vf4", "vf5", "vf6", "vf7", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7" ] + }, + ins => [ "base", "index", "mem", "addr", "stack", "fpcw", "eax", "ecx", "edx" ], + outs => [ "stack", "fpcw", "M", "eax", "ecx", "edx", "vf0", "vf1", "vf2", "vf3", "vf4", "vf5", "vf6", "vf7", "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7" ], + attr_type => "ia32_call_attr_t", + attr => "unsigned pop, ir_type *call_tp", + am => "source,unary", + units => [ "BRANCH" ], + latency => 4, # random number + modified_flags => $status_flags +}, + +# +# a Helper node for frame-climbing, needed for __builtin_(frame|return)_address +# +# PS: try gcc __builtin_frame_address(100000) :-) +# +ClimbFrame => { + reg_req => { in => [ "gp", "gp", "gp"], out => [ "in_r3" ] }, + ins => [ "frame", "cnt", "tmp" ], + outs => [ "res" ], + latency => 4, # random number + attr_type => "ia32_climbframe_attr_t", + attr => "unsigned count", + units => [ "GP" ], + mode => $mode_gp +}, + +# +# bswap +# +Bswap => { + irn_flags => "R", + reg_req => { in => [ "gp" ], + out => [ "in_r1" ] }, + emit => '. bswap%M %S0', + ins => [ "val" ], + units => [ "GP" ], + latency => 1, + mode => $mode_gp, +}, + +# +# bswap16, use xchg here +# +Bswap16 => { + irn_flags => "R", + reg_req => { in => [ "eax ebx ecx edx" ], + out => [ "in_r1" ] }, + emit => '. xchg %SB0, %SH0', + ins => [ "val" ], + units => [ "GP" ], + latency => 1, + mode => $mode_gp, +}, + +# +# BreakPoint +# +Breakpoint => { + state => "pinned", + reg_req => { in => [ "none" ], out => [ "none" ] }, + ins => [ "mem" ], + latency => 0, + emit => ". int3", + units => [ "GP" ], + mode => mode_M, +}, + +# +# Undefined Instruction on ALL x86 CPU's +# +UD2 => { + state => "pinned", + reg_req => { in => [ "none" ], out => [ "none" ] }, + ins => [ "mem" ], + latency => 0, + emit => ". .value 0x0b0f", + units => [ "GP" ], + mode => mode_M, +}, + +# +# outport +# +Outport => { + irn_flags => "R", + state => "pinned", + reg_req => { in => [ "edx", "eax", "none" ], out => [ "none" ] }, + ins => [ "port", "value", "mem" ], + emit => '. out%M %SS0, %SI1', + units => [ "GP" ], + latency => 1, + mode => mode_M, + modified_flags => $status_flags +}, + +# +# inport +# +Inport => { + irn_flags => "R", + state => "pinned", + reg_req => { in => [ "edx", "none" ], out => [ "eax", "none" ] }, + ins => [ "port", "mem" ], + outs => [ "res", "M" ], + emit => '. in%M %DS0, %SS0', + units => [ "GP" ], + latency => 1, + mode => mode_T, + modified_flags => $status_flags +}, + +# +# Intel style prefetching +# +Prefetch0 => { + op_flags => "L|F", + state => "exc_pinned", + reg_req => { in => [ "gp", "gp", "none" ], out => [ "none" ] }, + ins => [ "base", "index", "mem" ], + outs => [ "M" ], + latency => 0, + emit => ". prefetcht0 %AM", + units => [ "GP" ], +}, + +Prefetch1 => { + op_flags => "L|F", + state => "exc_pinned", + reg_req => { in => [ "gp", "gp", "none" ], out => [ "none" ] }, + ins => [ "base", "index", "mem" ], + outs => [ "M" ], + latency => 0, + emit => ". prefetcht1 %AM", + units => [ "GP" ], +}, + +Prefetch2 => { + op_flags => "L|F", + state => "exc_pinned", + reg_req => { in => [ "gp", "gp", "none" ], out => [ "none" ] }, + ins => [ "base", "index", "mem" ], + outs => [ "M" ], + latency => 0, + emit => ". prefetcht2 %AM", + units => [ "GP" ], +}, + +PrefetchNTA => { + op_flags => "L|F", + state => "exc_pinned", + reg_req => { in => [ "gp", "gp", "none" ], out => [ "none" ] }, + ins => [ "base", "index", "mem" ], + outs => [ "M" ], + latency => 0, + emit => ". prefetchnta %AM", + units => [ "GP" ], +}, + +# +# 3DNow! prefetch instructions +# +Prefetch => { + op_flags => "L|F", + state => "exc_pinned", + reg_req => { in => [ "gp", "gp", "none" ], out => [ "none" ] }, + ins => [ "base", "index", "mem" ], + outs => [ "M" ], + latency => 0, + emit => ". prefetch %AM", + units => [ "GP" ], +}, + +PrefetchW => { + op_flags => "L|F", + state => "exc_pinned", + reg_req => { in => [ "gp", "gp", "none" ], out => [ "none" ] }, + ins => [ "base", "index", "mem" ], + outs => [ "M" ], + latency => 0, + emit => ". prefetchw %AM", + units => [ "GP" ], +}, # produces a 0/+0.0 xZero => { @@ -1543,13 +1790,13 @@ xMovd => { mode => $mode_xmm }, -# commutative operations - xAdd => { irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], out => [ "in_r4 in_r5" ] }, + reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], + out => [ "in_r4 in_r5", "flags", "none" ] }, ins => [ "base", "index", "mem", "left", "right" ], + outs => [ "res", "flags", "M" ], am => "source,binary", emit => '. add%XXM %binop', latency => 4, @@ -1560,8 +1807,10 @@ xAdd => { xMul => { irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], out => [ "in_r4 in_r5" ] }, + reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], + out => [ "in_r4 in_r5", "flags", "none" ] }, ins => [ "base", "index", "mem", "left", "right" ], + outs => [ "res", "flags", "M" ], am => "source,binary", emit => '. mul%XXM %binop', latency => 4, @@ -1572,8 +1821,10 @@ xMul => { xMax => { irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], out => [ "in_r4 in_r5" ] }, + reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], + out => [ "in_r4 in_r5", "flags", "none" ] }, ins => [ "base", "index", "mem", "left", "right" ], + outs => [ "res", "flags", "M" ], am => "source,binary", emit => '. max%XXM %binop', latency => 2, @@ -1584,8 +1835,10 @@ xMax => { xMin => { irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], out => [ "in_r4 in_r5" ] }, + reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], + out => [ "in_r4 in_r5", "flags", "none" ] }, ins => [ "base", "index", "mem", "left", "right" ], + outs => [ "res", "flags", "M" ], am => "source,binary", emit => '. min%XXM %binop', latency => 2, @@ -1596,8 +1849,10 @@ xMin => { xAnd => { irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], out => [ "in_r4 in_r5" ] }, + reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], + out => [ "in_r4 in_r5", "flags", "none" ] }, ins => [ "base", "index", "mem", "left", "right" ], + outs => [ "res", "flags", "M" ], am => "source,binary", emit => '. andp%XSD %binop', latency => 3, @@ -1608,8 +1863,10 @@ xAnd => { xOr => { irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], out => [ "in_r4 in_r5" ] }, + reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], + out => [ "in_r4 in_r5", "flags", "none" ] }, ins => [ "base", "index", "mem", "left", "right" ], + outs => [ "res", "flags", "M" ], am => "source,binary", emit => '. orp%XSD %binop', latency => 3, @@ -1620,8 +1877,10 @@ xOr => { xXor => { irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], out => [ "in_r4 in_r5" ] }, + reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], + out => [ "in_r4 in_r5", "flags", "none" ] }, ins => [ "base", "index", "mem", "left", "right" ], + outs => [ "res", "flags", "M" ], am => "source,binary", emit => '. xorp%XSD %binop', latency => 3, @@ -1629,13 +1888,13 @@ xXor => { mode => $mode_xmm }, -# not commutative operations - xAndNot => { irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], out => [ "in_r4 !in_r5" ] }, + reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], + out => [ "in_r4 !in_r5", "flags", "none" ] }, ins => [ "base", "index", "mem", "left", "right" ], + outs => [ "res", "flags", "M" ], am => "source,binary", emit => '. andnp%XSD %binop', latency => 3, @@ -1646,8 +1905,10 @@ xAndNot => { xSub => { irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], out => [ "in_r4" ] }, + reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], + out => [ "in_r4", "flags", "none" ] }, ins => [ "base", "index", "mem", "minuend", "subtrahend" ], + outs => [ "res", "flags", "M" ], am => "source,binary", emit => '. sub%XXM %binop', latency => 4, @@ -1658,21 +1919,21 @@ xSub => { xDiv => { irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], out => [ "in_r4 !in_r5", "none" ] }, + reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], + out => [ "in_r4 !in_r5", "flags", "none" ] }, ins => [ "base", "index", "mem", "dividend", "divisor" ], + outs => [ "res", "flags", "M" ], am => "source,binary", - outs => [ "res", "M" ], emit => '. div%XXM %binop', latency => 16, units => [ "SSE" ], }, -# other operations - Ucomi => { irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], out => [ "eflags" ] }, + reg_req => { in => [ "gp", "gp", "none", "xmm", "xmm" ], + out => [ "eflags" ] }, ins => [ "base", "index", "mem", "left", "right" ], outs => [ "flags" ], am => "source,binary", @@ -1685,14 +1946,13 @@ Ucomi => { modified_flags => 1, }, -# Load / Store - xLoad => { op_flags => "L|F", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none" ], out => [ "xmm", "none", "none" ] }, + reg_req => { in => [ "gp", "gp", "none" ], + out => [ "xmm", "none", "none", "none" ] }, ins => [ "base", "index", "mem" ], - outs => [ "res", "M", "X_exc" ], + outs => [ "res", "unused", "M", "X_exc" ], emit => '. mov%XXM %AM, %D0', attr => "ir_mode *load_mode", init_attr => "attr->ls_mode = load_mode;", @@ -1715,8 +1975,9 @@ xStore => { xStoreSimple => { op_flags => "L|F", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "xmm" ] }, + reg_req => { in => [ "gp", "gp", "none", "xmm" ], out => [ "none" ] }, ins => [ "base", "index", "mem", "val" ], + outs => [ "M" ], emit => '. mov%XXM %S3, %AM', latency => 0, units => [ "SSE" ], @@ -1752,6 +2013,7 @@ l_LLtoFloat => { op_flags => "L|F", cmp_attr => "return 1;", ins => [ "val_high", "val_low" ], + reg_req => { in => [ "none", "none" ], out => [ "none" ] } }, l_FloattoLL => { @@ -1759,10 +2021,9 @@ l_FloattoLL => { cmp_attr => "return 1;", ins => [ "val" ], outs => [ "res_high", "res_low" ], + reg_req => { in => [ "none" ], out => [ "none", "none" ] } }, -# CopyB - CopyB => { op_flags => "F|H", state => "pinned", @@ -1789,12 +2050,23 @@ CopyB_i => { # modified_flags => [ "DF" ] }, -# Conversions +Cwtl => { + state => "exc_pinned", + reg_req => { in => [ "eax" ], out => [ "eax" ] }, + ins => [ "val" ], + outs => [ "res" ], + emit => '. cwtl', + units => [ "GP" ], + latency => 1, + mode => $mode_gp, +}, Conv_I2I => { state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "gp" ], out => [ "gp", "none" ] }, + reg_req => { in => [ "gp", "gp", "none", "gp" ], + out => [ "gp", "none", "none" ] }, ins => [ "base", "index", "mem", "val" ], + outs => [ "res", "flags", "M" ], am => "source,unary", units => [ "GP" ], latency => 1, @@ -1805,8 +2077,10 @@ Conv_I2I => { Conv_I2I8Bit => { state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "eax ebx ecx edx" ], out => [ "gp", "none" ] }, + reg_req => { in => [ "gp", "gp", "none", "eax ebx ecx edx" ], + out => [ "gp", "none", "none" ] }, ins => [ "base", "index", "mem", "val" ], + outs => [ "res", "flags", "M" ], am => "source,unary", units => [ "GP" ], latency => 1, @@ -1845,28 +2119,16 @@ Conv_FP2FP => { mode => $mode_xmm, }, -#----------------------------------------------------------# -# _ _ _ __ _ _ # -# (_) | | | | / _| | | | # -# __ ___ _ __| |_ _ _ __ _| | | |_| | ___ __ _| |_ # -# \ \ / / | '__| __| | | |/ _` | | | _| |/ _ \ / _` | __| # -# \ V /| | | | |_| |_| | (_| | | | | | | (_) | (_| | |_ # -# \_/ |_|_| \__|\__,_|\__,_|_| |_| |_|\___/ \__,_|\__| # -# | | # -# _ __ ___ __| | ___ ___ # -# | '_ \ / _ \ / _` |/ _ \/ __| # -# | | | | (_) | (_| | __/\__ \ # -# |_| |_|\___/ \__,_|\___||___/ # -#----------------------------------------------------------# - # rematerialisation disabled for all float nodes for now, because the fpcw # handler runs before spilling and we might end up with wrong fpcw then vfadd => { # irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "vfp", "vfp", "fpcw" ], out => [ "vfp" ] }, + reg_req => { in => [ "gp", "gp", "none", "vfp", "vfp", "fpcw" ], + out => [ "vfp", "none", "none" ] }, ins => [ "base", "index", "mem", "left", "right", "fpcw" ], + outs => [ "res", "dummy", "M" ], am => "source,binary", latency => 4, units => [ "VFP" ], @@ -1877,8 +2139,10 @@ vfadd => { vfmul => { # irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "vfp", "vfp", "fpcw" ], out => [ "vfp" ] }, + reg_req => { in => [ "gp", "gp", "none", "vfp", "vfp", "fpcw" ], + out => [ "vfp", "none", "none" ] }, ins => [ "base", "index", "mem", "left", "right", "fpcw" ], + outs => [ "res", "dummy", "M" ], am => "source,binary", latency => 4, units => [ "VFP" ], @@ -1889,8 +2153,10 @@ vfmul => { vfsub => { # irn_flags => "R", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "vfp", "vfp", "fpcw" ], out => [ "vfp" ] }, + reg_req => { in => [ "gp", "gp", "none", "vfp", "vfp", "fpcw" ], + out => [ "vfp", "none", "none" ] }, ins => [ "base", "index", "mem", "minuend", "subtrahend", "fpcw" ], + outs => [ "res", "dummy", "M" ], am => "source,binary", latency => 4, units => [ "VFP" ], @@ -1900,10 +2166,11 @@ vfsub => { vfdiv => { state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "vfp", "vfp", "fpcw" ], out => [ "vfp", "none" ] }, + reg_req => { in => [ "gp", "gp", "none", "vfp", "vfp", "fpcw" ], + out => [ "vfp", "none", "none" ] }, ins => [ "base", "index", "mem", "dividend", "divisor", "fpcw" ], + outs => [ "res", "dummy", "M" ], am => "source,binary", - outs => [ "res", "M" ], latency => 20, units => [ "VFP" ], attr_type => "ia32_x87_attr_t", @@ -1938,15 +2205,14 @@ vfchs => { attr_type => "ia32_x87_attr_t", }, -# virtual Load and Store - vfld => { irn_flags => "R", op_flags => "L|F", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none" ], out => [ "vfp", "none", "none" ] }, + reg_req => { in => [ "gp", "gp", "none" ], + out => [ "vfp", "none", "none", "none" ] }, ins => [ "base", "index", "mem" ], - outs => [ "res", "M", "X_exc" ], + outs => [ "res", "unused", "M", "X_exc" ], attr => "ir_mode *load_mode", init_attr => "attr->attr.ls_mode = load_mode;", latency => 2, @@ -1958,7 +2224,8 @@ vfst => { irn_flags => "R", op_flags => "L|F", state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "vfp" ], out => [ "none", "none" ] }, + reg_req => { in => [ "gp", "gp", "none", "vfp" ], + out => [ "none", "none" ] }, ins => [ "base", "index", "mem", "val" ], outs => [ "M", "X_exc" ], attr => "ir_mode *store_mode", @@ -1969,28 +2236,22 @@ vfst => { attr_type => "ia32_x87_attr_t", }, -# Conversions - vfild => { state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none" ], out => [ "vfp", "none" ] }, - outs => [ "res", "M" ], + reg_req => { in => [ "gp", "gp", "none" ], + out => [ "vfp", "none", "none" ] }, + outs => [ "res", "unused", "M" ], ins => [ "base", "index", "mem" ], latency => 4, units => [ "VFP" ], attr_type => "ia32_x87_attr_t", }, -l_vfild => { - cmp_attr => "return 1;", - outs => [ "res", "M" ], - arity => 2, -}, - vfist => { state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "vfp", "fpcw" ] }, + reg_req => { in => [ "gp", "gp", "none", "vfp", "fpcw" ], out => [ "none" ] }, ins => [ "base", "index", "mem", "val", "fpcw" ], + outs => [ "M" ], latency => 4, units => [ "VFP" ], mode => "mode_M", @@ -2000,27 +2261,18 @@ vfist => { # SSE3 fisttp instruction vfisttp => { state => "exc_pinned", - reg_req => { in => [ "gp", "gp", "none", "vfp" ] }, + reg_req => { in => [ "gp", "gp", "none", "vfp" ], out => [ "in_r4", "none" ]}, ins => [ "base", "index", "mem", "val" ], + outs => [ "res", "M" ], latency => 4, units => [ "VFP" ], - mode => "mode_M", attr_type => "ia32_x87_attr_t", }, -l_vfist => { - cmp_attr => "return 1;", - state => "exc_pinned", - arity => 3, - mode => "mode_M", -}, - - -# constants - vfldz => { irn_flags => "R", reg_req => { out => [ "vfp" ] }, + outs => [ "res" ], latency => 4, units => [ "VFP" ], mode => "mode_E", @@ -2030,6 +2282,7 @@ vfldz => { vfld1 => { irn_flags => "R", reg_req => { out => [ "vfp" ] }, + outs => [ "res" ], latency => 4, units => [ "VFP" ], mode => "mode_E", @@ -2039,6 +2292,7 @@ vfld1 => { vfldpi => { irn_flags => "R", reg_req => { out => [ "vfp" ] }, + outs => [ "res" ], latency => 4, units => [ "VFP" ], mode => "mode_E", @@ -2048,6 +2302,7 @@ vfldpi => { vfldln2 => { irn_flags => "R", reg_req => { out => [ "vfp" ] }, + outs => [ "res" ], latency => 4, units => [ "VFP" ], mode => "mode_E", @@ -2057,6 +2312,7 @@ vfldln2 => { vfldlg2 => { irn_flags => "R", reg_req => { out => [ "vfp" ] }, + outs => [ "res" ], latency => 4, units => [ "VFP" ], mode => "mode_E", @@ -2066,6 +2322,7 @@ vfldlg2 => { vfldl2t => { irn_flags => "R", reg_req => { out => [ "vfp" ] }, + outs => [ "res" ], latency => 4, units => [ "VFP" ], mode => "mode_E", @@ -2075,16 +2332,15 @@ vfldl2t => { vfldl2e => { irn_flags => "R", reg_req => { out => [ "vfp" ] }, + outs => [ "res" ], latency => 4, units => [ "VFP" ], mode => "mode_E", attr_type => "ia32_x87_attr_t", }, -# other - vFucomFnstsw => { -# we can't allow to rematerialize this node so we don't have +# we can't allow to rematerialize this node so we don't # accidently produce Phi(Fucom, Fucom(ins_permuted)) # irn_flags => "R", reg_req => { in => [ "vfp", "vfp" ], out => [ "eax" ] }, @@ -2135,245 +2391,208 @@ Sahf => { mode => $mode_flags, }, -#------------------------------------------------------------------------# -# ___ _____ __ _ _ _ # -# __ _( _ )___ | / _| | ___ __ _| |_ _ __ ___ __| | ___ ___ # -# \ \/ / _ \ / / | |_| |/ _ \ / _` | __| | '_ \ / _ \ / _` |/ _ \/ __| # -# > < (_) |/ / | _| | (_) | (_| | |_ | | | | (_) | (_| | __/\__ \ # -# /_/\_\___//_/ |_| |_|\___/ \__,_|\__| |_| |_|\___/ \__,_|\___||___/ # -#------------------------------------------------------------------------# - -# Note: gas is strangely buggy: fdivrp and fdivp as well as fsubrp and fsubp -# are swapped, we work this around in the emitter... - fadd => { state => "exc_pinned", - rd_constructor => "NONE", - reg_req => { }, emit => '. fadd%XM %x87_binop', latency => 4, attr_type => "ia32_x87_attr_t", + constructors => {}, }, faddp => { state => "exc_pinned", - rd_constructor => "NONE", - reg_req => { }, emit => '. faddp%XM %x87_binop', latency => 4, attr_type => "ia32_x87_attr_t", + constructors => {}, }, fmul => { state => "exc_pinned", - rd_constructor => "NONE", - reg_req => { }, emit => '. fmul%XM %x87_binop', latency => 4, attr_type => "ia32_x87_attr_t", + constructors => {}, }, fmulp => { state => "exc_pinned", - rd_constructor => "NONE", - reg_req => { }, emit => '. fmulp%XM %x87_binop',, latency => 4, attr_type => "ia32_x87_attr_t", + constructors => {}, }, fsub => { state => "exc_pinned", - rd_constructor => "NONE", - reg_req => { }, emit => '. fsub%XM %x87_binop', latency => 4, attr_type => "ia32_x87_attr_t", + constructors => {}, }, +# Note: gas is strangely buggy: fdivrp and fdivp as well as fsubrp and fsubp +# are swapped, we work this around in the emitter... + fsubp => { state => "exc_pinned", - rd_constructor => "NONE", - reg_req => { }, # see note about gas bugs emit => '. fsubrp%XM %x87_binop', latency => 4, attr_type => "ia32_x87_attr_t", + constructors => {}, }, fsubr => { state => "exc_pinned", - rd_constructor => "NONE", irn_flags => "R", - reg_req => { }, emit => '. fsubr%XM %x87_binop', latency => 4, attr_type => "ia32_x87_attr_t", + constructors => {}, }, fsubrp => { state => "exc_pinned", - rd_constructor => "NONE", irn_flags => "R", - reg_req => { }, -# see note about gas bugs +# see note about gas bugs before fsubp emit => '. fsubp%XM %x87_binop', latency => 4, attr_type => "ia32_x87_attr_t", + constructors => {}, }, fprem => { - rd_constructor => "NONE", - reg_req => { }, emit => '. fprem1', latency => 20, attr_type => "ia32_x87_attr_t", + constructors => {}, }, # this node is just here, to keep the simulator running # we can omit this when a fprem simulation function exists fpremp => { - rd_constructor => "NONE", - reg_req => { }, emit => '. fprem1\n'. '. fstp %X0', latency => 20, attr_type => "ia32_x87_attr_t", + constructors => {}, }, fdiv => { state => "exc_pinned", - rd_constructor => "NONE", - reg_req => { }, emit => '. fdiv%XM %x87_binop', latency => 20, attr_type => "ia32_x87_attr_t", + constructors => {}, }, fdivp => { state => "exc_pinned", - rd_constructor => "NONE", - reg_req => { }, -# see note about gas bugs +# see note about gas bugs before fsubp emit => '. fdivrp%XM %x87_binop', latency => 20, attr_type => "ia32_x87_attr_t", + constructors => {}, }, fdivr => { state => "exc_pinned", - rd_constructor => "NONE", - reg_req => { }, emit => '. fdivr%XM %x87_binop', latency => 20, attr_type => "ia32_x87_attr_t", + constructors => {}, }, fdivrp => { state => "exc_pinned", - rd_constructor => "NONE", - reg_req => { }, -# see note about gas bugs +# see note about gas bugs before fsubp emit => '. fdivp%XM %x87_binop', latency => 20, attr_type => "ia32_x87_attr_t", + constructors => {}, }, fabs => { - rd_constructor => "NONE", - reg_req => { }, emit => '. fabs', latency => 4, attr_type => "ia32_x87_attr_t", + constructors => {}, }, fchs => { op_flags => "R|K", - rd_constructor => "NONE", - reg_req => { }, emit => '. fchs', latency => 4, attr_type => "ia32_x87_attr_t", + constructors => {}, }, -# x87 Load and Store - fld => { - rd_constructor => "NONE", op_flags => "R|L|F", state => "exc_pinned", - reg_req => { }, emit => '. fld%XM %AM', attr_type => "ia32_x87_attr_t", latency => 2, + constructors => {}, }, fst => { - rd_constructor => "NONE", op_flags => "R|L|F", state => "exc_pinned", - reg_req => { }, emit => '. fst%XM %AM', mode => "mode_M", attr_type => "ia32_x87_attr_t", latency => 2, + constructors => {}, }, fstp => { - rd_constructor => "NONE", op_flags => "R|L|F", state => "exc_pinned", - reg_req => { }, emit => '. fstp%XM %AM', mode => "mode_M", attr_type => "ia32_x87_attr_t", latency => 2, + constructors => {}, }, -# Conversions - fild => { state => "exc_pinned", - rd_constructor => "NONE", - reg_req => { }, - emit => '. fild%M %AM', + emit => '. fild%XM %AM', attr_type => "ia32_x87_attr_t", latency => 2, + constructors => {}, }, fist => { state => "exc_pinned", - rd_constructor => "NONE", - reg_req => { }, - emit => '. fist%M %AM', + emit => '. fist%XM %AM', mode => "mode_M", attr_type => "ia32_x87_attr_t", latency => 2, + constructors => {}, }, fistp => { state => "exc_pinned", - rd_constructor => "NONE", - reg_req => { }, - emit => '. fistp%M %AM', + emit => '. fistp%XM %AM', mode => "mode_M", attr_type => "ia32_x87_attr_t", latency => 2, + constructors => {}, }, -# SSE3 firsttp instruction +# SSE3 fisttp instruction fisttp => { state => "exc_pinned", - rd_constructor => "NONE", - reg_req => { }, - emit => '. fist%M %AM', + emit => '. fisttp%XM %AM', mode => "mode_M", attr_type => "ia32_x87_attr_t", latency => 2, + constructors => {}, }, -# constants - fldz => { op_flags => "R|c|K", irn_flags => "R", @@ -2443,7 +2662,7 @@ fldl2e => { fxch => { op_flags => "R|K", - reg_req => { }, + reg_req => { out => [ "none" ] }, cmp_attr => "return 1;", emit => '. fxch %X0', attr_type => "ia32_x87_attr_t", @@ -2453,7 +2672,7 @@ fxch => { fpush => { op_flags => "R|K", - reg_req => {}, + reg_req => { out => [ "none" ] }, cmp_attr => "return 1;", emit => '. fld %X0', attr_type => "ia32_x87_attr_t", @@ -2471,7 +2690,7 @@ fpushCopy => { fpop => { op_flags => "K", - reg_req => { }, + reg_req => { out => [ "none" ] }, cmp_attr => "return 1;", emit => '. fstp %X0', attr_type => "ia32_x87_attr_t", @@ -2481,7 +2700,7 @@ fpop => { ffreep => { op_flags => "K", - reg_req => { }, + reg_req => { out => [ "none" ] }, cmp_attr => "return 1;", emit => '. ffreep %X0', attr_type => "ia32_x87_attr_t", @@ -2491,7 +2710,7 @@ ffreep => { emms => { op_flags => "K", - reg_req => { }, + reg_req => { out => [ "none" ] }, cmp_attr => "return 1;", emit => '. emms', attr_type => "ia32_x87_attr_t", @@ -2501,7 +2720,7 @@ emms => { femms => { op_flags => "K", - reg_req => { }, + reg_req => { out => [ "none" ] }, cmp_attr => "return 1;", emit => '. femms', attr_type => "ia32_x87_attr_t", @@ -2509,8 +2728,6 @@ femms => { latency => 3, }, -# compare - FucomFnstsw => { reg_req => { }, emit => ". fucom %X1\n". @@ -2557,17 +2774,6 @@ FtstFnstsw => { latency => 2, }, - -# -------------------------------------------------------------------------------- # -# ____ ____ _____ _ _ # -# / ___/ ___|| ____| __ _____ ___| |_ ___ _ __ _ __ ___ __| | ___ ___ # -# \___ \___ \| _| \ \ / / _ \/ __| __/ _ \| '__| | '_ \ / _ \ / _` |/ _ \/ __| # -# ___) |__) | |___ \ V / __/ (__| || (_) | | | | | | (_) | (_| | __/\__ \ # -# |____/____/|_____| \_/ \___|\___|\__\___/|_| |_| |_|\___/ \__,_|\___||___/ # -# # -# -------------------------------------------------------------------------------- # - - # Spilling and reloading of SSE registers, hardcoded, not generated # xxLoad => { @@ -2593,14 +2799,6 @@ xxStore => { ); # end of %nodes -# Include the generated SIMD node specification written by the SIMD optimization -$my_script_name = dirname($myname) . "/../ia32/ia32_simd_spec.pl"; -unless ($return = do $my_script_name) { - warn "couldn't parse $my_script_name: $@" if $@; - warn "couldn't do $my_script_name: $!" unless defined $return; - warn "couldn't run $my_script_name" unless $return; -} - # Transform some attributes foreach my $op (keys(%nodes)) { my $node = $nodes{$op};