xref: /freebsd/contrib/llvm-project/llvm/lib/Target/PowerPC/PPCScheduleP7.td (revision 7fdf597e96a02165cfe22ff357b857d5fa15ed8a)
1//===-- PPCScheduleP7.td - PPC P7 Scheduling Definitions ---*- tablegen -*-===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8//
9// This file defines the SchedModel for the POWER7 processor.
10//
11//===----------------------------------------------------------------------===//
12
13def P7Model : SchedMachineModel {
14  let IssueWidth = 6;  // 4 (non-branch) instructions are dispatched per cycle.
15                       // Note that the dispatch bundle size is 6 (including
16                       // branches), but the total internal issue bandwidth per
17                       // cycle (from all queues) is 8.
18
19  let LoadLatency = 3; // Optimistic load latency assuming bypass.
20                       // This is overriden by OperandCycles if the
21                       // Itineraries are queried instead.
22  let MispredictPenalty = 16;
23
24  let MicroOpBufferSize = 44;
25
26  // Try to make sure we have at least 10 dispatch groups in a loop.
27  let LoopMicroOpBufferSize = 40;
28
29  let CompleteModel = 0;
30
31  let UnsupportedFeatures = [HasSPE, PrefixInstrs, MMA,
32                             PairedVectorMemops, IsISA3_0, IsISA2_07,
33                             PCRelativeMemops, IsISA3_1, IsISAFuture];
34}
35
36let SchedModel = P7Model in {
37  def P7_LSU_FXU: ProcResource<4>;
38  def P7_LSU: ProcResource<2> {
39    let Super = P7_LSU_FXU;
40  }
41  def P7_FXU: ProcResource<2> {
42    let Super = P7_LSU_FXU;
43  }
44  // Implemented as two 2-way SIMD operations for double- and single-precision.
45  def P7_FPU: ProcResource<4>;
46  // Scalar binary floating point instructions can only use two FPUs.
47  def P7_ScalarFPU: ProcResource<2> {
48    let Super = P7_FPU;
49  }
50  def P7_VectorFPU: ProcResource<2> {
51    let Super = P7_FPU;
52  }
53  // Executing simple FX, complex FX, permute and 4-way SIMD single-precision FP ops
54  def P7_VMX: ProcResource<1>;
55  def P7_VPM: ProcResource<1> {
56    let Super = P7_VMX;
57    let BufferSize = 1;
58  }
59  def P7_VXS: ProcResource<1> {
60    let Super = P7_VMX;
61  }
62  def P7_DFU: ProcResource<1>;
63  def P7_BRU: ProcResource<1>;
64  def P7_CRU: ProcResource<1>;
65
66  def P7_PORT_LS : ProcResource<2>;
67  def P7_PORT_FX : ProcResource<2>;
68  def P7_PORT_FP : ProcResource<2>;
69  def P7_PORT_BR : ProcResource<1>;
70  def P7_PORT_CR : ProcResource<1>;
71
72  def P7_DISP_LS : SchedWriteRes<[P7_PORT_LS]>;
73  def P7_DISP_FX : SchedWriteRes<[P7_PORT_FX]>;
74  def P7_DISP_FP : SchedWriteRes<[P7_PORT_FP]>;
75  def P7_DISP_BR : SchedWriteRes<[P7_PORT_BR]>;
76  def P7_DISP_CR : SchedWriteRes<[P7_PORT_CR]>;
77
78  def P7_BRU_NONE : SchedWriteRes<[P7_BRU]>;
79  def P7_BRU_3C : SchedWriteRes<[P7_BRU]> { let Latency = 3; }
80  def P7_BRU_4C : SchedWriteRes<[P7_BRU]> { let Latency = 4; }
81  def P7_CRU_NONE : SchedWriteRes<[P7_CRU]>;
82  def P7_CRU_3C : SchedWriteRes<[P7_CRU]> { let Latency = 3; }
83  def P7_CRU_6C : SchedWriteRes<[P7_CRU]> { let Latency = 6; }
84  def P7_LSU_NONE : SchedWriteRes<[P7_LSU]>;
85  def P7_LSU_2C : SchedWriteRes<[P7_LSU]> { let Latency = 2; }
86  def P7_LSU_3C : SchedWriteRes<[P7_LSU]> { let Latency = 3; }
87  def P7_LSU_4C : SchedWriteRes<[P7_LSU]> { let Latency = 4; }
88  def P7_FXU_NONE : SchedWriteRes<[P7_FXU]>;
89  def P7_FXU_2C : SchedWriteRes<[P7_FXU]> { let Latency = 2; }
90  def P7_FXU_3C : SchedWriteRes<[P7_FXU]> { let Latency = 3; }
91  def P7_FXU_4C : SchedWriteRes<[P7_FXU]> { let Latency = 4; }
92  def P7_FXU_5C : SchedWriteRes<[P7_FXU]> { let Latency = 5; }
93  def P7_FXU_38C : SchedWriteRes<[P7_FXU]> { let Latency = 38; }
94  def P7_FXU_69C : SchedWriteRes<[P7_FXU]> { let Latency = 69; }
95  def P7_LSU_FXU_2C : SchedWriteRes<[P7_LSU_FXU]> { let Latency = 2; }
96  def P7_FPU_NONE : SchedWriteRes<[P7_FPU]>;
97  def P7_VectorFPU_6C : SchedWriteRes<[P7_VectorFPU]> { let Latency = 6; }
98  def P7_VectorFPU_25C : SchedWriteRes<[P7_VectorFPU]> { let Latency = 25; }
99  def P7_VectorFPU_30C : SchedWriteRes<[P7_VectorFPU]> { let Latency = 30; }
100  def P7_VectorFPU_31C : SchedWriteRes<[P7_VectorFPU]> { let Latency = 31; }
101  def P7_VectorFPU_42C : SchedWriteRes<[P7_VectorFPU]> { let Latency = 42; }
102  def P7_ScalarFPU_6C : SchedWriteRes<[P7_ScalarFPU]> { let Latency = 6; }
103  def P7_ScalarFPU_8C : SchedWriteRes<[P7_ScalarFPU]> { let Latency = 8; }
104  def P7_ScalarFPU_27C : SchedWriteRes<[P7_ScalarFPU]> { let Latency = 27; }
105  def P7_ScalarFPU_31C : SchedWriteRes<[P7_ScalarFPU]> { let Latency = 31; }
106  def P7_ScalarFPU_32C : SchedWriteRes<[P7_ScalarFPU]> { let Latency = 32; }
107  def P7_ScalarFPU_33C : SchedWriteRes<[P7_ScalarFPU]> { let Latency = 33; }
108  def P7_ScalarFPU_42C : SchedWriteRes<[P7_ScalarFPU]> { let Latency = 42; }
109  def P7_ScalarFPU_44C : SchedWriteRes<[P7_ScalarFPU]> { let Latency = 44; }
110  def P7_VXS_2C : SchedWriteRes<[P7_VXS]> { let Latency = 2; }
111  def P7_VPM_3C : SchedWriteRes<[P7_VPM]> { let Latency = 3; }
112
113  // Instruction of BRU pipeline
114
115  def : InstRW<[P7_BRU_NONE, P7_DISP_BR],
116    (instregex "^B(L)?(A)?(8)?(_NOP|_NOTOC)?(_TLS|_RM)?(_)?$")>;
117
118  def : InstRW<[P7_BRU_3C, P7_DISP_BR], (instrs
119    BDZLRLp, BDZLRm, BDZLRp, BDZLm, BDZLp, BDZm, BDZp,
120    BDNZ, BDNZ8, BDNZA, BDNZAm, BDNZAp, BDNZL, BDNZLA, BDNZLAm, BDNZLAp, BDNZLR,
121    BDNZLR8, BDNZLRL, BDNZLRLm, BDNZLRLp, BDNZLRm, BDNZLRp, BDNZLm, BDNZLp,
122    BDNZm, BDNZp, BDZ, BDZ8, BDZA, BDZAm, BDZAp, BDZL, BDZLA, BDZLAm, BDZLAp,
123    BDZLR, BDZLR8, BDZLRL, BDZLRLm, BLR, BLR8, BLRL, BCL, BCLR, BCLRL, BCLRLn,
124    BCLRn, BCLalways, BCLn, BCTR, BCTR8, BCTRL, BCTRL8, BCTRL8_LDinto_toc,
125    BCTRL8_LDinto_toc_RM, BCTRL8_RM, BCTRL_LWZinto_toc, BCTRL_LWZinto_toc_RM,
126    BCTRL_RM, BCn, BC, BCC, BCCA, BCCCTR, BCCCTR8, BCCCTRL, BCCCTRL8, BCCL,
127    BCCLA, BCCLR, BCCLRL, BCCTR, BCCTR8, BCCTR8n, BCCTRL, BCCTRL8,
128    BCCCTR, BCCCTR8, BCCCTRL, BCCCTRL8, BCCL, BCCLA, BCCLR, BCCLRL, BCCTR,
129    BCCTR8, BCCTR8n, BCCTRL, BCCTRL8, BCCTRL8n, BCCTRLn, BCCTRn, gBC, gBCA,
130    gBCAat, gBCCTR, gBCCTRL, gBCL, gBCLA, gBCLAat, gBCLR, gBCLRL, gBCLat, gBCat,
131    MFCTR, MFCTR8, MFLR, MFLR8
132  )>;
133
134  def : InstRW<[P7_BRU_4C], (instrs MTLR, MTLR8, MTCTR, MTCTR8, MTCTR8loop, MTCTRloop)>;
135
136  // Instructions of CRU pipeline
137
138  def : InstRW<[P7_CRU_NONE], (instrs MFCR, MFCR8)>;
139  def : InstRW<[P7_CRU_3C], (instrs MCRF)>;
140  def : InstRW<[P7_CRU_6C, P7_DISP_CR], (instrs
141    CR6SET, CR6UNSET, CRSET, CRUNSET,
142    CRAND, CRANDC, CREQV, CRNAND, CRNOR, CRNOT, CROR, CRORC
143  )>;
144
145  // Instructions of LSU and FXU pipelines
146
147  def : InstRW<[P7_LSU_NONE, P7_DISP_LS], (instrs LMW, LWARX, LWARXL, LDARX, LDARXL)>;
148  def : InstRW<[P7_LSU_2C, P7_DISP_LS], (instrs LHBRX, LHBRX8, LWBRX, LWBRX8)>;
149  def : InstRW<[P7_LSU_3C], (instrs MFSR, MFSRIN)>;
150
151  def : InstRW<[P7_LSU_3C, P7_DISP_LS], (instrs
152    LFS, LFSX, LFSXTLS, LFSXTLS_, LFD, LFDX, LFDXTLS, LFDXTLS_, LXSDX, LXVD2X,
153    LXVW4X, LXVDSX
154  )>;
155
156  def : InstRW<[P7_LSU_3C, P7_FXU_3C, P7_DISP_LS], (instrs
157    LFSU, LFSUX, LFDU, LFDUX)>;
158
159  def : InstRW<[P7_LSU_NONE, P7_FPU_NONE, P7_DISP_LS], (instrs
160    STXSDX, STXVD2X, STXVW4X)>;
161
162  def : InstRW<[P7_LSU_4C, P7_FXU_4C, P7_DISP_LS], (instrs
163    LBARX, LBZCIX, LDBRX, LDCIX, LFIWAX, LFIWZX, LHARX, LHZCIX, LSWI, LVEBX,
164    LVEHX, LVEWX, LVSL, LVSR, LVX, LVXL, LWZCIX,
165    STFD, STFDU, STFDUX, STFDX, STFIWX, STFS, STFSU, STFSUX, STFSX,
166    STHCIX, STSWI, STVEBX, STVEHX, STVEWX, STVX, STVXL, STWCIX,
167    LHA, LHA8, LHAX, LHAX8, LWA, LWAX, LWAX_32, LWA_32, LHAU, LHAU8,
168    LHAUX, LHAUX8, LWAUX
169  )>;
170
171  def : InstRW<[P7_LSU_NONE, P7_FXU_NONE, P7_DISP_LS], (instrs
172    STB, STB8, STH, STH8, STW, STW8, STD, STBX, STBX8, STHX, STHX8, STWX,
173    STWX8, STDX, STHBRX, STWBRX, STMW, STWCX, STDCX, STDU, STHU, STHU8,
174    STBU, STBU8, STWU, STWU8, STDUX, STWUX, STWUX8, STHUX, STHUX8, STBUX, STBUX8
175  )>;
176
177  def : InstRW<[P7_LSU_2C, P7_FXU_2C, P7_DISP_LS], (instrs
178    LWZU, LWZU8, LHZU, LHZU8, LBZU, LBZU8, LDU,
179    LWZUX, LWZUX8, LHZUX, LHZUX8, LBZUX, LBZUX8, LDUX
180  )>;
181
182  def : InstRW<[P7_LSU_FXU_2C, P7_DISP_FX], (instrs
183    (instregex "^(ADD|L)I(S)?(8)?$"),
184    (instregex "^(ADD|SUBF)(4|8)(TLS)?(_)?(_rec)?$"),
185    (instregex "^(X)?ORI(S)?(8)?$"),
186    (instregex "^(X)OR(8)?(_rec)?$"),
187    ADDIC, ADDIC8, SUBFIC, SUBFIC8, SUBFZE, SUBFZE8,
188    ADDE, ADDE8, ADDME, ADDME8, SUBFME, SUBFME8,
189    NEG, NEG8, NEG8_rec, NEG_rec, NEG8O, NEGO,
190    ANDI_rec, ANDIS_rec, AND, AND8, AND_rec, AND8_rec,
191    NAND, NAND8, NAND_rec, NAND8_rec, NOR, NOR8, NOR_rec, NOR8_rec,
192    EQV, EQV8, EQV_rec, EQV8_rec, ANDC, ANDC8, ANDC_rec, ANDC8_rec,
193    ORC, ORC8, ORC_rec, ORC8_rec
194  )>;
195
196  def : InstRW<[P7_FXU_2C, P7_DISP_FX], (instrs
197    CMPD, CMPDI, CMPLD, CMPLDI, CMPLW, CMPLWI, CMPW, CMPWI,
198    EXTSB8_32_64, EXTSB8_rec, EXTSH8_32_64, EXTSH8_rec, EXTSW_32,
199    EXTSW_32_64, EXTSW_32_64_rec, POPCNTB, POPCNTB8, POPCNTD, POPCNTW,
200    ADDPCIS, ANDI8_rec, ANDIS8_rec, SUBFUS, SUBFUS_rec,
201    ADD4O, ADD8O, ADDC, ADDC8, SUBFO, SUBF8O, SUBFC, SUBFC8,
202    ADDIC_rec, ADDE8_rec, ADDE_rec, SUBFE8_rec, SUBFE_rec,
203    ADDME8_rec, ADDME_rec, SUBFME8_rec, SUBFME_rec, ADDZE8_rec, ADDZE_rec,
204    SUBFZE_rec, SUBFZE8_rec, ADD8O_rec, SUBFO_rec, SUBF8O_rec, ADD4O_rec,
205    ADD8O_rec, SUBF8O_rec, SUBFO_rec, ADDE8O, ADDEO, SUBFE8O, SUBFEO, ADDME8O,
206    ADDMEO, SUBFME8O, SUBFMEO, ADDZE8O, ADDZEO, SUBFZE8O, SUBFZEO, NEG8O_rec,
207    NEGO_rec, ADDEO, ADDE8O, SUBFEO, SUBFE8O, ADDMEO, SUBFMEO, SUBFME8O, ADDME8O,
208    ADDZEO, ADDZE8O, SUBFZEO, SUBFZE8O, NEG8O_rec, NEGO_rec,
209    ADDE8O_rec, ADDEO_rec, ADDMEO_rec, ADDME8O_rec, SUBFMEO_rec, SUBFME8O_rec,
210    ADDZEO_rec, ADDZE8O_rec, SUBFZEO_rec, SUBFZE8O_rec,
211    ADDC8_rec, ADDC_rec, ADDCO, ADDCO_rec, ADDC8O, ADDC8O_rec,
212    SUBFC8_rec, SUBFC_rec, SUBFCO, SUBFC8O,  SUBFCO_rec, SUBFC8O_rec,
213    EXTSB, EXTSB8, EXTSB_rec, EXTSH, EXTSH8, EXTSH_rec, EXTSW, EXTSW_rec,
214    RLDICL, RLDICL_rec, RLDICR, RLDICR_rec, RLDIC, RLDIC_rec,
215    RLWINM, RLWINM8, RLWINM_rec, RLDCL, RLDCL_rec, RLDCR, RLDCR_rec,
216    RLWNM, RLWNM8, RLWNM_rec, RLDIMI, RLDIMI_rec,
217    RLDICL_32, RLDICL_32_64, RLDICL_32_rec, RLDICR_32, RLWINM8_rec, RLWNM8_rec,
218    SLD, SLD_rec, SLW, SLW8, SLW_rec, SLW8_rec, SRD, SRD_rec, SRW, SRW8, SRW_rec,
219    SRW8_rec, SRADI, SRADI_rec, SRAWI, SRAWI_rec, SRAD, SRAD_rec, SRAW, SRAW_rec,
220    SRADI_32, SUBFE, SUBFE8, SUBFE8O_rec, SUBFEO_rec
221  )>;
222
223  def : InstRW<[P7_FXU_3C, P7_DISP_FX], (instregex "^CNT(L|T)Z(D|W)(8)?(M)?(_rec)?$")>;
224
225  def : InstRW<[P7_FXU_5C, P7_DISP_FX], (instrs
226    MULLI, MULLI8, MULLW, MULHW, MULHWU, MULLD, MULHD, MULHDU, MULLWO, MULLDO,
227    MULLW_rec, MULLD_rec, MULHD_rec, MULHW_rec, MULHDU_rec, MULHWU_rec, MULLWO_rec,
228    MULLDO_rec
229  )>;
230
231  def : InstRW<[P7_FXU_38C, P7_DISP_FX], (instrs
232    DIVDE, DIVDEO, DIVDEO_rec, DIVDEU, DIVDEUO, DIVDEUO_rec, DIVDEU_rec, DIVDE_rec,
233    DIVWE, DIVWEO, DIVWEO_rec, DIVWEU, DIVWEUO, DIVWEUO_rec, DIVWEU_rec, DIVWE_rec,
234    DIVW, DIVWU, DIVWU_rec, DIVWO, DIVWO_rec, DIVWUO, DIVWUO_rec, DIVW_rec
235  )>;
236
237  def : InstRW<[P7_FXU_69C, P7_DISP_FX], (instrs
238    DIVD, DIVDU, DIVDO, DIVDO_rec, DIVDUO, DIVDUO_rec, DIVDU_rec, DIVD_rec)>;
239
240  // Instructions of FPU and VMX pipeline
241
242  def : InstRW<[P7_ScalarFPU_6C, P7_DISP_FP], (instrs
243    (instregex "^F(N)?(M)?(R|ADD|SUB|ABS|NEG|NABS|UL)(D|S)?(_rec)?$"),
244    (instregex "^FC(T|F)I(D|W)(U)?(S)?(Z)?(_rec)?$"),
245    (instregex "^XS(N)?M(SUB|ADD)(A|M)(D|S)P$"),
246    (instregex "^XS(NEG|ABS|NABS|ADD|SUB|MUL)(D|S)P(s)?$"),
247    FRE, FRES_rec, FRE_rec, FRSP_rec, FTDIV, FTSQRT,
248    FRSP, FRES, FRSQRTE, FRSQRTES, FRSQRTES_rec, FRSQRTE_rec, FSELD, FSELS,
249    FSELD_rec, FSELS_rec, FCPSGND, FCPSGND_rec, FCPSGNS, FCPSGNS_rec,
250    FRIMD, FRIMD_rec, FRIMS, FRIMS_rec, FRIND, FRIND_rec, FRINS, FRINS_rec,
251    FRIPD, FRIPD_rec, FRIPS, FRIPS_rec, FRIZD, FRIZD_rec, FRIZS, FRIZS_rec,
252    XSCPSGNDP, XSCVDPSP, XSCVDPSXDS, XSCVDPSXDSs, XSCVDPSXWS, XSCVDPSXWSs,
253    XSCVDPUXDS, XSCVDPUXDSs, XSCVDPUXWS, XSCVDPUXWSs, XSCVSPDP, XSCVSXDDP,
254    XSCVUXDDP, XSMAXDP, XSMINDP, XSRDPI, XSRDPIC, XSRDPIM, XSRDPIP, XSRDPIZ,
255    XSREDP, XSRSQRTEDP, XSTDIVDP, XSTSQRTDP, XSCMPODP, XSCMPUDP
256  )>;
257
258  def : InstRW<[P7_VectorFPU_6C, P7_DISP_FP], (instrs
259    (instregex "^XV(N)?(M)?(ADD|SUB)(A|M)?(D|S)P$"),
260    (instregex "^XV(MAX|MIN|MUL|NEG|ABS|ADD|NABS)(D|S)P$"),
261    XVCMPEQDP, XVCMPEQDP_rec, XVCMPGEDP, XVCMPGEDP_rec, XVCMPGTDP, XVCMPGTDP_rec,
262    XVCPSGNDP, XVCVDPSXDS, XVCVDPSXWS, XVCVDPUXDS, XVCVDPUXWS, XVCVSPSXDS,
263    XVCVSPSXWS, XVCVSPUXDS, XVCVSPUXWS, XVCVSXDDP, XVCVSXWDP, XVCVUXDDP,
264    XVCVUXWDP, XVRDPI, XVRDPIC, XVRDPIM, XVRDPIP, XVRDPIZ, XVREDP,
265    XVRSPI, XVRSPIC, XVRSPIM, XVRSPIP, XVRSPIZ, XVRSQRTEDP, XVTDIVDP,
266    XVTSQRTDP
267  )>;
268
269  // TODO: Altivec instructions are not listed in Book IV.
270  def : InstRW<[P7_VPM_3C, P7_DISP_FP], (instrs
271    (instregex "^VPK(S|U)(H|W)(S|U)(S|M)$"),
272    (instregex "^VUPK(H|L)(S|P)(X|B|H)$"),
273    VPERM, XXMRGHW, XXMRGLW, XXPERMDI, XXPERMDIs, XXSLDWI, XXSLDWIs,
274    VSPLTB, VSPLTBs, VSPLTH, VSPLTHs, VSPLTISB, VSPLTISH, VSPLTISW, VSPLTW,
275    XXSPLTW, XXSPLTWs, VSEL, XXSEL, VPKPX
276  )>;
277
278  def : InstRW<[P7_VXS_2C, P7_DISP_FP], (instrs
279    (instregex "^VADD(U|S)(B|H|W)(S|M)$"),
280    (instregex "^V(MAX|MIN)(S|U)(B|H|W)$"),
281    (instregex "^V(MRG)(L|H)(B|H|W)$"),
282    XXLORf, XXLXORdpz, XXLXORspz, XXLXORz, XVRSQRTESP, XVRESP,
283    XVTDIVSP, XVTSQRTSP, XVCMPEQSP, XVCMPEQSP_rec, XVCMPGESP, XVCMPGESP_rec,
284    XVCMPGTSP, XVCMPGTSP_rec, XVCVSXDSP, XVCVSXWSP, XVCVUXDSP, XVCVUXWSP,
285    XVCPSGNSP, XVCVDPSP, VADDCUW, VADDFP, VAND, VANDC, VAVGSB, VAVGSH,
286    VAVGSW, VAVGUB, VAVGUH, VAVGUW, VCFSX, VCFUX, VCMPBFP, VCMPBFP_rec,
287    VCMPEQFP, VCMPEQFP_rec, VCMPEQUB, VCMPEQUB_rec, VCMPEQUH, VCMPEQUH_rec,
288    VCMPEQUW, VCMPEQUW_rec, VCMPGEFP, VCMPGEFP_rec, VCMPGTFP, VCMPGTFP_rec,
289    VCMPGTSB, VCMPGTSB_rec, VCMPGTSH, VCMPGTSH_rec, VCMPGTSW, VCMPGTSW_rec,
290    VCMPGTUB, VCMPGTUB_rec, VCMPGTUH, VCMPGTUH_rec, VCMPGTUW, VCMPGTUW_rec,
291    VCTSXS, VCTUXS, VEXPTEFP, VLOGEFP, VNOR, VOR,
292    VMADDFP, VMHADDSHS, VMHRADDSHS, VMLADDUHM, VNMSUBFP, VMAXFP, VMINFP,
293    VMSUMMBM, VMSUMSHM, VMSUMSHS, VMSUMUBM, VMSUMUDM, VMSUMUHM, VMSUMUHS,
294    VMULESB, VMULESH, VMULEUB, VMULEUH, VMULOSB, VMULOSH, VMULOUB, VMULOUH,
295    VREFP, VRFIM, VRFIN, VRFIP, VRFIZ, VRLB, VRLH, VRLW, VRSQRTEFP,
296    VSR, VSRAB, VSRAH, VSRAW, VSRB, VSRH, VSRO, VSRW, VSUBCUW, VSL, VSLB,
297    VSLDOI, VSLH, VSLO, VSLW, VSUBFP, VSUBSBS, VSUBSHS, VSUBSWS, VSUBUBM,
298    VSUBUBS, VSUBUHM, VSUBUHS, VSUBUWM, VSUBUWS, VSUM2SWS, VSUM4SBS, VSUM4SHS,
299    VSUM4UBS, VSUMSWS, VXOR, XXLAND, XXLANDC, XXLNOR, XXLOR, XXLXOR
300  )>;
301
302  def : InstRW<[P7_ScalarFPU_8C, P7_DISP_FP],
303    (instrs FCMPOD, FCMPOS, FCMPUD, FCMPUS)>;
304  def : InstRW<[P7_ScalarFPU_27C, P7_DISP_FP], (instrs FDIVS, FDIVS_rec)>;
305  def : InstRW<[P7_ScalarFPU_31C, P7_DISP_FP], (instrs XSDIVDP)>;
306  def : InstRW<[P7_ScalarFPU_32C, P7_DISP_FP], (instrs FSQRTS, XSSQRTSP, FSQRTS_rec)>;
307  def : InstRW<[P7_ScalarFPU_33C, P7_DISP_FP], (instrs FDIV, FDIV_rec)>;
308  def : InstRW<[P7_ScalarFPU_42C, P7_DISP_FP], (instrs XSSQRTDP)>;
309  def : InstRW<[P7_ScalarFPU_44C, P7_DISP_FP], (instrs FSQRT, FSQRT_rec)>;
310
311  def : InstRW<[P7_VectorFPU_25C, P7_DISP_FP], (instrs XVDIVSP)>;
312  def : InstRW<[P7_VectorFPU_30C, P7_DISP_FP], (instrs XVSQRTSP)>;
313  def : InstRW<[P7_VectorFPU_31C, P7_DISP_FP], (instrs XVDIVDP)>;
314  def : InstRW<[P7_VectorFPU_42C, P7_DISP_FP], (instrs XVSQRTDP)>;
315}
316