2014-05-23 Segher Boessenkool <segher@kernel.crashing.org>
+ * config/rs6000/rs6000.md (type): Delete "idiv", "ldiv". Add
+ "div".
+ (bits): New mode_attr.
+ (idiv_ldiv): Delete mode_attr.
+ (udiv<mode>3, *div<mode>3, div<div_extend>_<mode>): Adjust.
+ * config/rs6000/rs6000.c (rs6000_adjust_cost, is_cracked_insn,
+ rs6000_adjust_priority, is_nonpipeline_insn,
+ insn_must_be_first_in_group, insn_must_be_last_in_group): Adjust.
+
+ * config/rs6000/40x.md (ppc403-idiv): Adjust.
+ * config/rs6000/440.md (ppc440-idiv): Adjust.
+ * config/rs6000/476.md (ppc476-idiv): Adjust.
+ * config/rs6000/601.md (ppc601-idiv): Adjust.
+ * config/rs6000/603.md (ppc603-idiv): Adjust.
+ * config/rs6000/6xx.md (ppc604-idiv, ppc620-idiv, ppc630-idiv,
+ ppc620-ldiv): Adjust.
+ * config/rs6000/7450.md (ppc7450-idiv): Adjust.
+ * config/rs6000/7xx.md (ppc750-idiv): Adjust.
+ * config/rs6000/8540.md (ppc8540_divide): Adjust.
+ * config/rs6000/a2.md (ppca2-idiv, ppca2-ldiv): Adjust.
+ * config/rs6000/cell.md (cell-idiv, cell-ldiv): Adjust.
+ * config/rs6000/e300c2c3.md (ppce300c3_divide): Adjust.
+ * config/rs6000/e500mc.md (e500mc_divide): Adjust.
+ * config/rs6000/e500mc64.md (e500mc64_divide): Adjust.
+ * config/rs6000/e5500.md (e5500_divide, e5500_divide_d): Adjust.
+ * config/rs6000/e6500.md (e6500_divide, e6500_divide_d): Adjust.
+ * config/rs6000/mpc.md (mpccore-idiv): Adjust.
+ * config/rs6000/power4.md (power4-idiv, power4-ldiv): Adjust.
+ * config/rs6000/power5.md (power5-idiv, power5-ldiv): Adjust.
+ * config/rs6000/power6.md (power6-idiv, power6-ldiv): Adjust.
+ * config/rs6000/power7.md (power7-idiv, power7-ldiv): Adjust.
+ * config/rs6000/power8.md (power8-idiv, power8-ldiv): Adjust.
+ * config/rs6000/rs64.md (rs64a-idiv, rs64a-ldiv): Adjust.
+ * config/rs6000/titan.md (titan_fxu_div): Adjust.
+
+2014-05-23 Segher Boessenkool <segher@kernel.crashing.org>
+
* config/rs6000/rs6000.md (type): Delete "insert_word",
"insert_dword". Add "insert".
(size): Update comment.
"iu_40x")
(define_insn_reservation "ppc403-idiv" 33
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
(eq_attr "cpu" "ppc403,ppc405"))
"iu_40x*33")
"ppc440_issue,ppc440_i_pipe")
(define_insn_reservation "ppc440-idiv" 34
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
(eq_attr "cpu" "ppc440"))
"ppc440_issue,ppc440_i_pipe*33")
ppc476_i_pipe")
(define_insn_reservation "ppc476-idiv" 11
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
(eq_attr "cpu" "ppc476"))
"ppc476_issue,\
ppc476_i_pipe*11")
"iu_ppc601*5")
(define_insn_reservation "ppc601-idiv" 36
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
(eq_attr "cpu" "ppc601"))
"iu_ppc601*36")
"iu_603*2")
(define_insn_reservation "ppc603-idiv" 37
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
(eq_attr "cpu" "ppc603"))
"iu_603*37")
"mciu_6xx*5")
(define_insn_reservation "ppc604-idiv" 20
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
(eq_attr "cpu" "ppc604,ppc604e"))
"mciu_6xx*19")
(define_insn_reservation "ppc620-idiv" 37
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "32")
(eq_attr "cpu" "ppc620"))
"mciu_6xx*36")
(define_insn_reservation "ppc630-idiv" 21
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "32")
(eq_attr "cpu" "ppc630"))
"mciu_6xx*20")
(define_insn_reservation "ppc620-ldiv" 37
- (and (eq_attr "type" "ldiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "64")
(eq_attr "cpu" "ppc620,ppc630"))
"mciu_6xx*36")
"ppc7450_du,mciu_7450")
(define_insn_reservation "ppc7450-idiv" 23
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
(eq_attr "cpu" "ppc7450"))
"ppc7450_du,mciu_7450*23")
"ppc750_du,iu1_7xx")
(define_insn_reservation "ppc750-idiv" 19
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
(eq_attr "cpu" "ppc750,ppc7400"))
"ppc750_du,iu1_7xx*19")
;; reservation of miu_stage3 here because we use the average latency
;; time.
(define_insn_reservation "ppc8540_divide" 14
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
(eq_attr "cpu" "ppc8540,ppc8548"))
"ppc8540_decode,ppc8540_issue+ppc8540_mu_stage0+ppc8540_mu_div,\
ppc8540_mu_div*13")
;; D.4.9
(define_insn_reservation "ppca2-idiv" 32
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "32")
(eq_attr "cpu" "ppca2"))
"mult*32")
(define_insn_reservation "ppca2-ldiv" 65
- (and (eq_attr "type" "ldiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "64")
(eq_attr "cpu" "ppca2"))
"mult*65")
;; divide
(define_insn_reservation "cell-idiv" 32
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "32")
(eq_attr "cpu" "cell"))
"slot1,nonpipeline,nonpipeline*30")
(define_insn_reservation "cell-ldiv" 64
- (and (eq_attr "type" "ldiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "64")
(eq_attr "cpu" "cell"))
"slot1,nonpipeline,nonpipeline*62")
;; Divide. We use the average latency time here. We omit reserving a
;; retire unit because of the result automata will be huge.
(define_insn_reservation "ppce300c3_divide" 20
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
(ior (eq_attr "cpu" "ppce300c2") (eq_attr "cpu" "ppce300c3")))
"ppce300c3_decode,ppce300c3_issue+ppce300c3_iu_stage0+ppce300c3_mu_div,\
ppce300c3_mu_div*19")
;; Divide. We use the average latency time here.
(define_insn_reservation "e500mc_divide" 14
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
(eq_attr "cpu" "ppce500mc"))
"e500mc_decode,e500mc_issue+e500mc_mu_stage0+e500mc_mu_div,\
e500mc_mu_div*13")
;; Divide. We use the average latency time here.
(define_insn_reservation "e500mc64_divide" 14
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
(eq_attr "cpu" "ppce500mc64"))
"e500mc64_decode,e500mc64_issue+e500mc64_mu_stage0+e500mc64_mu_div,\
e500mc64_mu_div*13")
;; CFX - Divide.
(define_insn_reservation "e5500_divide" 16
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "32")
(eq_attr "cpu" "ppce5500"))
"e5500_decode,e5500_cfx_stage0+e5500_cfx_div,\
e5500_cfx_div*15")
(define_insn_reservation "e5500_divide_d" 26
- (and (eq_attr "type" "ldiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "64")
(eq_attr "cpu" "ppce5500"))
"e5500_decode,e5500_cfx_stage0+e5500_cfx_div,\
e5500_cfx_div*25")
;; CFX - Divide.
(define_insn_reservation "e6500_divide" 16
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "32")
(eq_attr "cpu" "ppce6500"))
"e6500_decode,e6500_cfx_stage0+e6500_cfx_div,\
e6500_cfx_div*15")
(define_insn_reservation "e6500_divide_d" 26
- (and (eq_attr "type" "ldiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "64")
(eq_attr "cpu" "ppce6500"))
"e6500_decode,e6500_cfx_stage0+e6500_cfx_div,\
e6500_cfx_div*25")
; Divide latency varies greatly from 2-11, use 6 as average
(define_insn_reservation "mpccore-idiv" 6
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
(eq_attr "cpu" "mpccore"))
"mciu_mpc*6")
; SPR move only executes in first IU.
; Integer division only executes in second IU.
(define_insn_reservation "power4-idiv" 36
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "32")
(eq_attr "cpu" "power4"))
"du1_power4+du2_power4,iu2_power4*35")
(define_insn_reservation "power4-ldiv" 68
- (and (eq_attr "type" "ldiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "64")
(eq_attr "cpu" "power4"))
"du1_power4+du2_power4,iu2_power4*67")
; SPR move only executes in first IU.
; Integer division only executes in second IU.
(define_insn_reservation "power5-idiv" 36
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "32")
(eq_attr "cpu" "power5"))
"du1_power5+du2_power5,iu2_power5*35")
(define_insn_reservation "power5-ldiv" 68
- (and (eq_attr "type" "ldiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "64")
(eq_attr "cpu" "power5"))
"du1_power5+du2_power5,iu2_power5*67")
"store_data_bypass_p")
(define_insn_reservation "power6-idiv" 44
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "32")
(eq_attr "cpu" "power6"))
"(iu1_power6*44+iu2_power6*44+fpu1_power6*44)\
|(iu1_power6*44+iu2_power6*44+fpu2_power6*44)");
; "store_data_bypass_p")
(define_insn_reservation "power6-ldiv" 56
- (and (eq_attr "type" "ldiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "64")
(eq_attr "cpu" "power6"))
"(iu1_power6*56+iu2_power6*56+fpu1_power6*56)\
|(iu1_power6*56+iu2_power6*56+fpu2_power6*56)");
"DU2F_power7,FXU_power7,nothing*3,FXU_power7")
(define_insn_reservation "power7-idiv" 36
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "32")
(eq_attr "cpu" "power7"))
"DU2F_power7,iu1_power7*36|iu2_power7*36")
(define_insn_reservation "power7-ldiv" 68
- (and (eq_attr "type" "ldiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "64")
(eq_attr "cpu" "power7"))
"DU2F_power7,iu1_power7*68|iu2_power7*68")
; FXU divides are not pipelined
(define_insn_reservation "power8-idiv" 37
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "32")
(eq_attr "cpu" "power8"))
"DU_any_power8,fxu0_power8*37|fxu1_power8*37")
(define_insn_reservation "power8-ldiv" 68
- (and (eq_attr "type" "ldiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "64")
(eq_attr "cpu" "power8"))
"DU_any_power8,fxu0_power8*68|fxu1_power8*68")
return 17;
break;
}
- case TYPE_IDIV:
+ case TYPE_DIV:
{
if (! store_data_bypass_p (dep_insn, insn))
- return 45;
- break;
- }
- case TYPE_LDIV:
- {
- if (! store_data_bypass_p (dep_insn, insn))
- return 57;
+ return get_attr_size (dep_insn) == SIZE_32 ? 45 : 57;
break;
}
default:
return 17;
break;
}
- case TYPE_IDIV:
- {
- if (set_to_load_agen (dep_insn, insn))
- return 45;
- break;
- }
- case TYPE_LDIV:
+ case TYPE_DIV:
{
if (set_to_load_agen (dep_insn, insn))
- return 57;
+ return get_attr_size (dep_insn) == SIZE_32 ? 45 : 57;
break;
}
default:
|| type == TYPE_COMPARE || type == TYPE_DELAYED_COMPARE
|| (type == TYPE_MUL
&& get_attr_dot (insn) == DOT_YES)
- || type == TYPE_IDIV || type == TYPE_LDIV
+ || type == TYPE_DIV
|| (type == TYPE_INSERT
&& get_attr_size (insn) == SIZE_32))
return true;
break;
case TYPE_MUL:
- case TYPE_IDIV:
+ case TYPE_DIV:
fprintf (stderr, "priority was %#x (%d) before adjustment\n",
priority, priority);
if (priority >= 0 && priority < 0x01000000)
type = get_attr_type (insn);
if (type == TYPE_MUL
- || type == TYPE_IDIV
- || type == TYPE_LDIV
+ || type == TYPE_DIV
|| type == TYPE_SDIV
|| type == TYPE_DDIV
|| type == TYPE_SSQRT
case TYPE_CR_LOGICAL:
case TYPE_MTJMPR:
case TYPE_MFJMPR:
- case TYPE_IDIV:
- case TYPE_LDIV:
+ case TYPE_DIV:
case TYPE_LOAD_L:
case TYPE_STORE_C:
case TYPE_ISYNC:
case TYPE_VAR_SHIFT_ROTATE:
case TYPE_TRAP:
case TYPE_MUL:
- case TYPE_IDIV:
case TYPE_INSERT:
case TYPE_DELAYED_COMPARE:
case TYPE_FPCOMPARE:
case TYPE_LOAD_L:
case TYPE_STORE_C:
return true;
+ case TYPE_DIV:
+ if (get_attr_size (insn) == SIZE_32)
+ return true;
+ else
+ break;
case TYPE_LOAD:
case TYPE_STORE:
case TYPE_FPLOAD:
case TYPE_MFCR:
case TYPE_MFCRF:
case TYPE_MTCR:
- case TYPE_IDIV:
- case TYPE_LDIV:
+ case TYPE_DIV:
case TYPE_COMPARE:
case TYPE_DELAYED_COMPARE:
case TYPE_VAR_DELAYED_COMPARE:
case TYPE_VAR_SHIFT_ROTATE:
case TYPE_TRAP:
case TYPE_MUL:
- case TYPE_IDIV:
case TYPE_DELAYED_COMPARE:
case TYPE_FPCOMPARE:
case TYPE_MFCR:
case TYPE_LOAD_L:
case TYPE_STORE_C:
return true;
+ case TYPE_DIV:
+ if (get_attr_size (insn) == SIZE_32)
+ return true;
+ else
+ break;
default:
break;
}
(define_attr "type"
"integer,two,three,
shift,var_shift_rotate,insert,
- mul,halfmul,idiv,ldiv,
+ mul,halfmul,div,
exts,cntlz,popcnt,isel,
load,store,fpload,fpstore,vecload,vecstore,
cmp,
(V4SI "w")
(V2DI "d")])
+;; How many bits in this mode?
+(define_mode_attr bits [(QI "8") (HI "16") (SI "32") (DI "64")])
+
; DImode bits
(define_mode_attr dbits [(QI "56") (HI "48") (SI "32")])
(V2DI "X,X,X,X,X")
(V2DF "X,X,X,X,X")
(V1TI "X,X,X,X,X")])
-
-;; Mode attribute to give the correct type for integer divides
-(define_mode_attr idiv_ldiv [(SI "idiv")
- (DI "ldiv")])
-
\f
;; Start with fixed-point load and store insns. Here we put only the more
;; complex forms. Basic data transfer is done later.
(match_operand:GPR 2 "gpc_reg_operand" "r")))]
""
"div<wd>u %0,%1,%2"
- [(set_attr "type" "<idiv_ldiv>")])
+ [(set_attr "type" "div")
+ (set_attr "size" "<bits>")])
;; For powers of two we can do srai/aze for divide and then adjust for
(match_operand:GPR 2 "gpc_reg_operand" "r")))]
""
"div<wd> %0,%1,%2"
- [(set_attr "type" "<idiv_ldiv>")])
+ [(set_attr "type" "div")
+ (set_attr "size" "<bits>")])
(define_expand "mod<mode>3"
[(use (match_operand:GPR 0 "gpc_reg_operand" ""))
UNSPEC_DIV_EXTEND))]
"TARGET_POPCNTD"
"div<wd><div_extend> %0,%1,%2"
- [(set_attr "type" "<idiv_ldiv>")])
+ [(set_attr "type" "div")
+ (set_attr "size" "<bits>")])
\f
;; Pack/unpack 128-bit floating point types that take 2 scalar registers
"mciu_rs64*34")
(define_insn_reservation "rs64a-idiv" 66
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "32")
(eq_attr "cpu" "rs64a"))
"mciu_rs64*66")
(define_insn_reservation "rs64a-ldiv" 66
- (and (eq_attr "type" "ldiv")
+ (and (eq_attr "type" "div")
+ (eq_attr "size" "64")
(eq_attr "cpu" "rs64a"))
"mciu_rs64*66")
;; through its latency and initial disptach bottlenecks (i.e. issue
;; slots and fxu scheduler availability)
(define_insn_reservation "titan_fxu_div" 34
- (and (eq_attr "type" "idiv")
+ (and (eq_attr "type" "div")
(eq_attr "cpu" "titan"))
"titan_issue,titan_fxu_sh")