• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1//===-- X86InstrShiftRotate.td - Shift and Rotate Instrs ---*- tablegen -*-===//
2//
3//                     The LLVM Compiler Infrastructure
4//
5// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
7//
8//===----------------------------------------------------------------------===//
9//
10// This file describes the shift and rotate instructions.
11//
12//===----------------------------------------------------------------------===//
13
14// FIXME: Someone needs to smear multipattern goodness all over this file.
15
16let Defs = [EFLAGS] in {
17
18let Constraints = "$src1 = $dst", SchedRW = [WriteShift] in {
19let Uses = [CL] in {
20def SHL8rCL  : I<0xD2, MRM4r, (outs GR8 :$dst), (ins GR8 :$src1),
21                 "shl{b}\t{%cl, $dst|$dst, cl}",
22                 [(set GR8:$dst, (shl GR8:$src1, CL))]>;
23def SHL16rCL : I<0xD3, MRM4r, (outs GR16:$dst), (ins GR16:$src1),
24                 "shl{w}\t{%cl, $dst|$dst, cl}",
25                 [(set GR16:$dst, (shl GR16:$src1, CL))]>, OpSize16;
26def SHL32rCL : I<0xD3, MRM4r, (outs GR32:$dst), (ins GR32:$src1),
27                 "shl{l}\t{%cl, $dst|$dst, cl}",
28                 [(set GR32:$dst, (shl GR32:$src1, CL))]>, OpSize32;
29def SHL64rCL : RI<0xD3, MRM4r, (outs GR64:$dst), (ins GR64:$src1),
30                  "shl{q}\t{%cl, $dst|$dst, cl}",
31                  [(set GR64:$dst, (shl GR64:$src1, CL))]>;
32} // Uses = [CL]
33
34def SHL8ri   : Ii8<0xC0, MRM4r, (outs GR8 :$dst), (ins GR8 :$src1, u8imm:$src2),
35                   "shl{b}\t{$src2, $dst|$dst, $src2}",
36                   [(set GR8:$dst, (shl GR8:$src1, (i8 imm:$src2)))]>;
37
38let isConvertibleToThreeAddress = 1 in {   // Can transform into LEA.
39def SHL16ri  : Ii8<0xC1, MRM4r, (outs GR16:$dst), (ins GR16:$src1, u8imm:$src2),
40                   "shl{w}\t{$src2, $dst|$dst, $src2}",
41                   [(set GR16:$dst, (shl GR16:$src1, (i8 imm:$src2)))]>,
42                   OpSize16;
43def SHL32ri  : Ii8<0xC1, MRM4r, (outs GR32:$dst), (ins GR32:$src1, u8imm:$src2),
44                   "shl{l}\t{$src2, $dst|$dst, $src2}",
45                   [(set GR32:$dst, (shl GR32:$src1, (i8 imm:$src2)))]>,
46                   OpSize32;
47def SHL64ri  : RIi8<0xC1, MRM4r, (outs GR64:$dst),
48                    (ins GR64:$src1, u8imm:$src2),
49                    "shl{q}\t{$src2, $dst|$dst, $src2}",
50                    [(set GR64:$dst, (shl GR64:$src1, (i8 imm:$src2)))]>;
51} // isConvertibleToThreeAddress = 1
52
53// NOTE: We don't include patterns for shifts of a register by one, because
54// 'add reg,reg' is cheaper (and we have a Pat pattern for shift-by-one).
55let hasSideEffects = 0 in {
56def SHL8r1   : I<0xD0, MRM4r, (outs GR8:$dst), (ins GR8:$src1),
57                 "shl{b}\t$dst", []>;
58def SHL16r1  : I<0xD1, MRM4r, (outs GR16:$dst), (ins GR16:$src1),
59                 "shl{w}\t$dst", []>, OpSize16;
60def SHL32r1  : I<0xD1, MRM4r, (outs GR32:$dst), (ins GR32:$src1),
61                 "shl{l}\t$dst", []>, OpSize32;
62def SHL64r1  : RI<0xD1, MRM4r, (outs GR64:$dst), (ins GR64:$src1),
63                 "shl{q}\t$dst", []>;
64} // hasSideEffects = 0
65} // Constraints = "$src = $dst", SchedRW
66
67
68let SchedRW = [WriteShiftLd, WriteRMW] in {
69// FIXME: Why do we need an explicit "Uses = [CL]" when the instr has a pattern
70// using CL?
71let Uses = [CL] in {
72def SHL8mCL  : I<0xD2, MRM4m, (outs), (ins i8mem :$dst),
73                 "shl{b}\t{%cl, $dst|$dst, cl}",
74                 [(store (shl (loadi8 addr:$dst), CL), addr:$dst)]>;
75def SHL16mCL : I<0xD3, MRM4m, (outs), (ins i16mem:$dst),
76                 "shl{w}\t{%cl, $dst|$dst, cl}",
77                 [(store (shl (loadi16 addr:$dst), CL), addr:$dst)]>,
78                 OpSize16;
79def SHL32mCL : I<0xD3, MRM4m, (outs), (ins i32mem:$dst),
80                 "shl{l}\t{%cl, $dst|$dst, cl}",
81                 [(store (shl (loadi32 addr:$dst), CL), addr:$dst)]>,
82                 OpSize32;
83def SHL64mCL : RI<0xD3, MRM4m, (outs), (ins i64mem:$dst),
84                  "shl{q}\t{%cl, $dst|$dst, cl}",
85                  [(store (shl (loadi64 addr:$dst), CL), addr:$dst)]>,
86                  Requires<[In64BitMode]>;
87}
88def SHL8mi   : Ii8<0xC0, MRM4m, (outs), (ins i8mem :$dst, u8imm:$src),
89                   "shl{b}\t{$src, $dst|$dst, $src}",
90                [(store (shl (loadi8 addr:$dst), (i8 imm:$src)), addr:$dst)]>;
91def SHL16mi  : Ii8<0xC1, MRM4m, (outs), (ins i16mem:$dst, u8imm:$src),
92                   "shl{w}\t{$src, $dst|$dst, $src}",
93               [(store (shl (loadi16 addr:$dst), (i8 imm:$src)), addr:$dst)]>,
94               OpSize16;
95def SHL32mi  : Ii8<0xC1, MRM4m, (outs), (ins i32mem:$dst, u8imm:$src),
96                   "shl{l}\t{$src, $dst|$dst, $src}",
97               [(store (shl (loadi32 addr:$dst), (i8 imm:$src)), addr:$dst)]>,
98               OpSize32;
99def SHL64mi : RIi8<0xC1, MRM4m, (outs), (ins i64mem:$dst, u8imm:$src),
100                  "shl{q}\t{$src, $dst|$dst, $src}",
101                  [(store (shl (loadi64 addr:$dst), (i8 imm:$src)), addr:$dst)]>,
102                  Requires<[In64BitMode]>;
103
104// Shift by 1
105def SHL8m1   : I<0xD0, MRM4m, (outs), (ins i8mem :$dst),
106                 "shl{b}\t$dst",
107                [(store (shl (loadi8 addr:$dst), (i8 1)), addr:$dst)]>;
108def SHL16m1  : I<0xD1, MRM4m, (outs), (ins i16mem:$dst),
109                 "shl{w}\t$dst",
110                 [(store (shl (loadi16 addr:$dst), (i8 1)), addr:$dst)]>,
111                 OpSize16;
112def SHL32m1  : I<0xD1, MRM4m, (outs), (ins i32mem:$dst),
113                 "shl{l}\t$dst",
114                 [(store (shl (loadi32 addr:$dst), (i8 1)), addr:$dst)]>,
115                 OpSize32;
116def SHL64m1 : RI<0xD1, MRM4m, (outs), (ins i64mem:$dst),
117                  "shl{q}\t$dst",
118                 [(store (shl (loadi64 addr:$dst), (i8 1)), addr:$dst)]>,
119                 Requires<[In64BitMode]>;
120} // SchedRW
121
122let Constraints = "$src1 = $dst", SchedRW = [WriteShift] in {
123let Uses = [CL] in {
124def SHR8rCL  : I<0xD2, MRM5r, (outs GR8 :$dst), (ins GR8 :$src1),
125                 "shr{b}\t{%cl, $dst|$dst, cl}",
126                 [(set GR8:$dst, (srl GR8:$src1, CL))]>;
127def SHR16rCL : I<0xD3, MRM5r, (outs GR16:$dst), (ins GR16:$src1),
128                 "shr{w}\t{%cl, $dst|$dst, cl}",
129                 [(set GR16:$dst, (srl GR16:$src1, CL))]>, OpSize16;
130def SHR32rCL : I<0xD3, MRM5r, (outs GR32:$dst), (ins GR32:$src1),
131                 "shr{l}\t{%cl, $dst|$dst, cl}",
132                 [(set GR32:$dst, (srl GR32:$src1, CL))]>, OpSize32;
133def SHR64rCL : RI<0xD3, MRM5r, (outs GR64:$dst), (ins GR64:$src1),
134                  "shr{q}\t{%cl, $dst|$dst, cl}",
135                  [(set GR64:$dst, (srl GR64:$src1, CL))]>;
136}
137
138def SHR8ri   : Ii8<0xC0, MRM5r, (outs GR8:$dst), (ins GR8:$src1, u8imm:$src2),
139                   "shr{b}\t{$src2, $dst|$dst, $src2}",
140                   [(set GR8:$dst, (srl GR8:$src1, (i8 imm:$src2)))]>;
141def SHR16ri  : Ii8<0xC1, MRM5r, (outs GR16:$dst), (ins GR16:$src1, u8imm:$src2),
142                   "shr{w}\t{$src2, $dst|$dst, $src2}",
143                   [(set GR16:$dst, (srl GR16:$src1, (i8 imm:$src2)))]>,
144                   OpSize16;
145def SHR32ri  : Ii8<0xC1, MRM5r, (outs GR32:$dst), (ins GR32:$src1, u8imm:$src2),
146                   "shr{l}\t{$src2, $dst|$dst, $src2}",
147                   [(set GR32:$dst, (srl GR32:$src1, (i8 imm:$src2)))]>,
148                   OpSize32;
149def SHR64ri : RIi8<0xC1, MRM5r, (outs GR64:$dst), (ins GR64:$src1, u8imm:$src2),
150                  "shr{q}\t{$src2, $dst|$dst, $src2}",
151                  [(set GR64:$dst, (srl GR64:$src1, (i8 imm:$src2)))]>;
152
153// Shift right by 1
154def SHR8r1   : I<0xD0, MRM5r, (outs GR8:$dst), (ins GR8:$src1),
155                 "shr{b}\t$dst",
156                 [(set GR8:$dst, (srl GR8:$src1, (i8 1)))]>;
157def SHR16r1  : I<0xD1, MRM5r, (outs GR16:$dst), (ins GR16:$src1),
158                 "shr{w}\t$dst",
159                 [(set GR16:$dst, (srl GR16:$src1, (i8 1)))]>, OpSize16;
160def SHR32r1  : I<0xD1, MRM5r, (outs GR32:$dst), (ins GR32:$src1),
161                 "shr{l}\t$dst",
162                 [(set GR32:$dst, (srl GR32:$src1, (i8 1)))]>, OpSize32;
163def SHR64r1  : RI<0xD1, MRM5r, (outs GR64:$dst), (ins GR64:$src1),
164                 "shr{q}\t$dst",
165                 [(set GR64:$dst, (srl GR64:$src1, (i8 1)))]>;
166} // Constraints = "$src = $dst", SchedRW
167
168
169let SchedRW = [WriteShiftLd, WriteRMW] in {
170let Uses = [CL] in {
171def SHR8mCL  : I<0xD2, MRM5m, (outs), (ins i8mem :$dst),
172                 "shr{b}\t{%cl, $dst|$dst, cl}",
173                 [(store (srl (loadi8 addr:$dst), CL), addr:$dst)]>;
174def SHR16mCL : I<0xD3, MRM5m, (outs), (ins i16mem:$dst),
175                 "shr{w}\t{%cl, $dst|$dst, cl}",
176                 [(store (srl (loadi16 addr:$dst), CL), addr:$dst)]>,
177                 OpSize16;
178def SHR32mCL : I<0xD3, MRM5m, (outs), (ins i32mem:$dst),
179                 "shr{l}\t{%cl, $dst|$dst, cl}",
180                 [(store (srl (loadi32 addr:$dst), CL), addr:$dst)]>,
181                 OpSize32;
182def SHR64mCL : RI<0xD3, MRM5m, (outs), (ins i64mem:$dst),
183                  "shr{q}\t{%cl, $dst|$dst, cl}",
184                  [(store (srl (loadi64 addr:$dst), CL), addr:$dst)]>,
185                  Requires<[In64BitMode]>;
186}
187def SHR8mi   : Ii8<0xC0, MRM5m, (outs), (ins i8mem :$dst, u8imm:$src),
188                   "shr{b}\t{$src, $dst|$dst, $src}",
189                [(store (srl (loadi8 addr:$dst), (i8 imm:$src)), addr:$dst)]>;
190def SHR16mi  : Ii8<0xC1, MRM5m, (outs), (ins i16mem:$dst, u8imm:$src),
191                   "shr{w}\t{$src, $dst|$dst, $src}",
192               [(store (srl (loadi16 addr:$dst), (i8 imm:$src)), addr:$dst)]>,
193               OpSize16;
194def SHR32mi  : Ii8<0xC1, MRM5m, (outs), (ins i32mem:$dst, u8imm:$src),
195                   "shr{l}\t{$src, $dst|$dst, $src}",
196               [(store (srl (loadi32 addr:$dst), (i8 imm:$src)), addr:$dst)]>,
197               OpSize32;
198def SHR64mi : RIi8<0xC1, MRM5m, (outs), (ins i64mem:$dst, u8imm:$src),
199                  "shr{q}\t{$src, $dst|$dst, $src}",
200                 [(store (srl (loadi64 addr:$dst), (i8 imm:$src)), addr:$dst)]>,
201                 Requires<[In64BitMode]>;
202
203// Shift by 1
204def SHR8m1   : I<0xD0, MRM5m, (outs), (ins i8mem :$dst),
205                 "shr{b}\t$dst",
206                 [(store (srl (loadi8 addr:$dst), (i8 1)), addr:$dst)]>;
207def SHR16m1  : I<0xD1, MRM5m, (outs), (ins i16mem:$dst),
208                 "shr{w}\t$dst",
209                 [(store (srl (loadi16 addr:$dst), (i8 1)), addr:$dst)]>,
210                 OpSize16;
211def SHR32m1  : I<0xD1, MRM5m, (outs), (ins i32mem:$dst),
212                 "shr{l}\t$dst",
213                 [(store (srl (loadi32 addr:$dst), (i8 1)), addr:$dst)]>,
214                 OpSize32;
215def SHR64m1 : RI<0xD1, MRM5m, (outs), (ins i64mem:$dst),
216                  "shr{q}\t$dst",
217                 [(store (srl (loadi64 addr:$dst), (i8 1)), addr:$dst)]>,
218                 Requires<[In64BitMode]>;
219} // SchedRW
220
221let Constraints = "$src1 = $dst", SchedRW = [WriteShift] in {
222let Uses = [CL] in {
223def SAR8rCL  : I<0xD2, MRM7r, (outs GR8 :$dst), (ins GR8 :$src1),
224                 "sar{b}\t{%cl, $dst|$dst, cl}",
225                 [(set GR8:$dst, (sra GR8:$src1, CL))]>;
226def SAR16rCL : I<0xD3, MRM7r, (outs GR16:$dst), (ins GR16:$src1),
227                 "sar{w}\t{%cl, $dst|$dst, cl}",
228                 [(set GR16:$dst, (sra GR16:$src1, CL))]>,
229                 OpSize16;
230def SAR32rCL : I<0xD3, MRM7r, (outs GR32:$dst), (ins GR32:$src1),
231                 "sar{l}\t{%cl, $dst|$dst, cl}",
232                 [(set GR32:$dst, (sra GR32:$src1, CL))]>,
233                 OpSize32;
234def SAR64rCL : RI<0xD3, MRM7r, (outs GR64:$dst), (ins GR64:$src1),
235                 "sar{q}\t{%cl, $dst|$dst, cl}",
236                 [(set GR64:$dst, (sra GR64:$src1, CL))]>;
237}
238
239def SAR8ri   : Ii8<0xC0, MRM7r, (outs GR8 :$dst), (ins GR8 :$src1, u8imm:$src2),
240                   "sar{b}\t{$src2, $dst|$dst, $src2}",
241                   [(set GR8:$dst, (sra GR8:$src1, (i8 imm:$src2)))]>;
242def SAR16ri  : Ii8<0xC1, MRM7r, (outs GR16:$dst), (ins GR16:$src1, u8imm:$src2),
243                   "sar{w}\t{$src2, $dst|$dst, $src2}",
244                   [(set GR16:$dst, (sra GR16:$src1, (i8 imm:$src2)))]>,
245                   OpSize16;
246def SAR32ri  : Ii8<0xC1, MRM7r, (outs GR32:$dst), (ins GR32:$src1, u8imm:$src2),
247                   "sar{l}\t{$src2, $dst|$dst, $src2}",
248                   [(set GR32:$dst, (sra GR32:$src1, (i8 imm:$src2)))]>,
249                   OpSize32;
250def SAR64ri  : RIi8<0xC1, MRM7r, (outs GR64:$dst),
251                    (ins GR64:$src1, u8imm:$src2),
252                    "sar{q}\t{$src2, $dst|$dst, $src2}",
253                    [(set GR64:$dst, (sra GR64:$src1, (i8 imm:$src2)))]>;
254
255// Shift by 1
256def SAR8r1   : I<0xD0, MRM7r, (outs GR8 :$dst), (ins GR8 :$src1),
257                 "sar{b}\t$dst",
258                 [(set GR8:$dst, (sra GR8:$src1, (i8 1)))]>;
259def SAR16r1  : I<0xD1, MRM7r, (outs GR16:$dst), (ins GR16:$src1),
260                 "sar{w}\t$dst",
261                 [(set GR16:$dst, (sra GR16:$src1, (i8 1)))]>, OpSize16;
262def SAR32r1  : I<0xD1, MRM7r, (outs GR32:$dst), (ins GR32:$src1),
263                 "sar{l}\t$dst",
264                 [(set GR32:$dst, (sra GR32:$src1, (i8 1)))]>, OpSize32;
265def SAR64r1  : RI<0xD1, MRM7r, (outs GR64:$dst), (ins GR64:$src1),
266                  "sar{q}\t$dst",
267                  [(set GR64:$dst, (sra GR64:$src1, (i8 1)))]>;
268} // Constraints = "$src = $dst", SchedRW
269
270
271let SchedRW = [WriteShiftLd, WriteRMW] in {
272let Uses = [CL] in {
273def SAR8mCL  : I<0xD2, MRM7m, (outs), (ins i8mem :$dst),
274                 "sar{b}\t{%cl, $dst|$dst, cl}",
275                 [(store (sra (loadi8 addr:$dst), CL), addr:$dst)]>;
276def SAR16mCL : I<0xD3, MRM7m, (outs), (ins i16mem:$dst),
277                 "sar{w}\t{%cl, $dst|$dst, cl}",
278                 [(store (sra (loadi16 addr:$dst), CL), addr:$dst)]>,
279                 OpSize16;
280def SAR32mCL : I<0xD3, MRM7m, (outs), (ins i32mem:$dst),
281                 "sar{l}\t{%cl, $dst|$dst, cl}",
282                 [(store (sra (loadi32 addr:$dst), CL), addr:$dst)]>,
283                 OpSize32;
284def SAR64mCL : RI<0xD3, MRM7m, (outs), (ins i64mem:$dst),
285                 "sar{q}\t{%cl, $dst|$dst, cl}",
286                 [(store (sra (loadi64 addr:$dst), CL), addr:$dst)]>,
287                 Requires<[In64BitMode]>;
288}
289def SAR8mi   : Ii8<0xC0, MRM7m, (outs), (ins i8mem :$dst, u8imm:$src),
290                   "sar{b}\t{$src, $dst|$dst, $src}",
291                [(store (sra (loadi8 addr:$dst), (i8 imm:$src)), addr:$dst)]>;
292def SAR16mi  : Ii8<0xC1, MRM7m, (outs), (ins i16mem:$dst, u8imm:$src),
293                   "sar{w}\t{$src, $dst|$dst, $src}",
294               [(store (sra (loadi16 addr:$dst), (i8 imm:$src)), addr:$dst)]>,
295               OpSize16;
296def SAR32mi  : Ii8<0xC1, MRM7m, (outs), (ins i32mem:$dst, u8imm:$src),
297                   "sar{l}\t{$src, $dst|$dst, $src}",
298               [(store (sra (loadi32 addr:$dst), (i8 imm:$src)), addr:$dst)]>,
299               OpSize32;
300def SAR64mi  : RIi8<0xC1, MRM7m, (outs), (ins i64mem:$dst, u8imm:$src),
301                    "sar{q}\t{$src, $dst|$dst, $src}",
302                 [(store (sra (loadi64 addr:$dst), (i8 imm:$src)), addr:$dst)]>,
303                 Requires<[In64BitMode]>;
304
305// Shift by 1
306def SAR8m1   : I<0xD0, MRM7m, (outs), (ins i8mem :$dst),
307                 "sar{b}\t$dst",
308                [(store (sra (loadi8 addr:$dst), (i8 1)), addr:$dst)]>;
309def SAR16m1  : I<0xD1, MRM7m, (outs), (ins i16mem:$dst),
310                 "sar{w}\t$dst",
311               [(store (sra (loadi16 addr:$dst), (i8 1)), addr:$dst)]>,
312               OpSize16;
313def SAR32m1  : I<0xD1, MRM7m, (outs), (ins i32mem:$dst),
314                 "sar{l}\t$dst",
315               [(store (sra (loadi32 addr:$dst), (i8 1)), addr:$dst)]>,
316               OpSize32;
317def SAR64m1 : RI<0xD1, MRM7m, (outs), (ins i64mem:$dst),
318                  "sar{q}\t$dst",
319                 [(store (sra (loadi64 addr:$dst), (i8 1)), addr:$dst)]>,
320                 Requires<[In64BitMode]>;
321} // SchedRW
322
323//===----------------------------------------------------------------------===//
324// Rotate instructions
325//===----------------------------------------------------------------------===//
326
327let hasSideEffects = 0 in {
328let Constraints = "$src1 = $dst", SchedRW = [WriteShift] in {
329
330let Uses = [CL, EFLAGS] in {
331def RCL8rCL : I<0xD2, MRM2r, (outs GR8:$dst), (ins GR8:$src1),
332                "rcl{b}\t{%cl, $dst|$dst, cl}", []>;
333def RCL16rCL : I<0xD3, MRM2r, (outs GR16:$dst), (ins GR16:$src1),
334                 "rcl{w}\t{%cl, $dst|$dst, cl}", []>, OpSize16;
335def RCL32rCL : I<0xD3, MRM2r, (outs GR32:$dst), (ins GR32:$src1),
336                 "rcl{l}\t{%cl, $dst|$dst, cl}", []>, OpSize32;
337def RCL64rCL : RI<0xD3, MRM2r, (outs GR64:$dst), (ins GR64:$src1),
338                  "rcl{q}\t{%cl, $dst|$dst, cl}", []>;
339} // Uses = [CL, EFLAGS]
340
341let Uses = [EFLAGS] in {
342def RCL8r1 : I<0xD0, MRM2r, (outs GR8:$dst), (ins GR8:$src1),
343               "rcl{b}\t$dst", []>;
344def RCL8ri : Ii8<0xC0, MRM2r, (outs GR8:$dst), (ins GR8:$src1, u8imm:$cnt),
345                 "rcl{b}\t{$cnt, $dst|$dst, $cnt}", []>;
346def RCL16r1 : I<0xD1, MRM2r, (outs GR16:$dst), (ins GR16:$src1),
347                "rcl{w}\t$dst", []>, OpSize16;
348def RCL16ri : Ii8<0xC1, MRM2r, (outs GR16:$dst), (ins GR16:$src1, u8imm:$cnt),
349                  "rcl{w}\t{$cnt, $dst|$dst, $cnt}", []>, OpSize16;
350def RCL32r1 : I<0xD1, MRM2r, (outs GR32:$dst), (ins GR32:$src1),
351                "rcl{l}\t$dst", []>, OpSize32;
352def RCL32ri : Ii8<0xC1, MRM2r, (outs GR32:$dst), (ins GR32:$src1, u8imm:$cnt),
353                  "rcl{l}\t{$cnt, $dst|$dst, $cnt}", []>, OpSize32;
354def RCL64r1 : RI<0xD1, MRM2r, (outs GR64:$dst), (ins GR64:$src1),
355                 "rcl{q}\t$dst", []>;
356def RCL64ri : RIi8<0xC1, MRM2r, (outs GR64:$dst), (ins GR64:$src1, u8imm:$cnt),
357                   "rcl{q}\t{$cnt, $dst|$dst, $cnt}", []>;
358} // Uses = [EFLAGS]
359
360let Uses = [CL, EFLAGS] in {
361def RCR8rCL : I<0xD2, MRM3r, (outs GR8:$dst), (ins GR8:$src1),
362                "rcr{b}\t{%cl, $dst|$dst, cl}", []>;
363def RCR16rCL : I<0xD3, MRM3r, (outs GR16:$dst), (ins GR16:$src1),
364                 "rcr{w}\t{%cl, $dst|$dst, cl}", []>, OpSize16;
365def RCR32rCL : I<0xD3, MRM3r, (outs GR32:$dst), (ins GR32:$src1),
366                 "rcr{l}\t{%cl, $dst|$dst, cl}", []>, OpSize32;
367def RCR64rCL : RI<0xD3, MRM3r, (outs GR64:$dst), (ins GR64:$src1),
368                  "rcr{q}\t{%cl, $dst|$dst, cl}", []>;
369} // Uses = [CL, EFLAGS]
370
371let Uses = [EFLAGS] in {
372def RCR8r1 : I<0xD0, MRM3r, (outs GR8:$dst), (ins GR8:$src1),
373               "rcr{b}\t$dst", []>;
374def RCR8ri : Ii8<0xC0, MRM3r, (outs GR8:$dst), (ins GR8:$src1, u8imm:$cnt),
375                 "rcr{b}\t{$cnt, $dst|$dst, $cnt}", []>;
376def RCR16r1 : I<0xD1, MRM3r, (outs GR16:$dst), (ins GR16:$src1),
377                "rcr{w}\t$dst", []>, OpSize16;
378def RCR16ri : Ii8<0xC1, MRM3r, (outs GR16:$dst), (ins GR16:$src1, u8imm:$cnt),
379                  "rcr{w}\t{$cnt, $dst|$dst, $cnt}", []>, OpSize16;
380def RCR32r1 : I<0xD1, MRM3r, (outs GR32:$dst), (ins GR32:$src1),
381                "rcr{l}\t$dst", []>, OpSize32;
382def RCR32ri : Ii8<0xC1, MRM3r, (outs GR32:$dst), (ins GR32:$src1, u8imm:$cnt),
383                  "rcr{l}\t{$cnt, $dst|$dst, $cnt}", []>, OpSize32;
384def RCR64r1 : RI<0xD1, MRM3r, (outs GR64:$dst), (ins GR64:$src1),
385                 "rcr{q}\t$dst", []>;
386def RCR64ri : RIi8<0xC1, MRM3r, (outs GR64:$dst), (ins GR64:$src1, u8imm:$cnt),
387                   "rcr{q}\t{$cnt, $dst|$dst, $cnt}", []>;
388} // Uses = [EFLAGS]
389
390} // Constraints = "$src = $dst"
391
392let SchedRW = [WriteShiftLd, WriteRMW], mayStore = 1 in {
393let Uses = [EFLAGS] in {
394def RCL8m1 : I<0xD0, MRM2m, (outs), (ins i8mem:$dst),
395               "rcl{b}\t$dst", []>;
396def RCL8mi : Ii8<0xC0, MRM2m, (outs), (ins i8mem:$dst, u8imm:$cnt),
397                 "rcl{b}\t{$cnt, $dst|$dst, $cnt}", []>;
398def RCL16m1 : I<0xD1, MRM2m, (outs), (ins i16mem:$dst),
399                "rcl{w}\t$dst", []>, OpSize16;
400def RCL16mi : Ii8<0xC1, MRM2m, (outs), (ins i16mem:$dst, u8imm:$cnt),
401                  "rcl{w}\t{$cnt, $dst|$dst, $cnt}", []>, OpSize16;
402def RCL32m1 : I<0xD1, MRM2m, (outs), (ins i32mem:$dst),
403                "rcl{l}\t$dst", []>, OpSize32;
404def RCL32mi : Ii8<0xC1, MRM2m, (outs), (ins i32mem:$dst, u8imm:$cnt),
405                  "rcl{l}\t{$cnt, $dst|$dst, $cnt}", []>, OpSize32;
406def RCL64m1 : RI<0xD1, MRM2m, (outs), (ins i64mem:$dst),
407                 "rcl{q}\t$dst", []>, Requires<[In64BitMode]>;
408def RCL64mi : RIi8<0xC1, MRM2m, (outs), (ins i64mem:$dst, u8imm:$cnt),
409                   "rcl{q}\t{$cnt, $dst|$dst, $cnt}", []>,
410                   Requires<[In64BitMode]>;
411
412def RCR8m1 : I<0xD0, MRM3m, (outs), (ins i8mem:$dst),
413               "rcr{b}\t$dst", []>;
414def RCR8mi : Ii8<0xC0, MRM3m, (outs), (ins i8mem:$dst, u8imm:$cnt),
415                 "rcr{b}\t{$cnt, $dst|$dst, $cnt}", []>;
416def RCR16m1 : I<0xD1, MRM3m, (outs), (ins i16mem:$dst),
417                "rcr{w}\t$dst", []>, OpSize16;
418def RCR16mi : Ii8<0xC1, MRM3m, (outs), (ins i16mem:$dst, u8imm:$cnt),
419                  "rcr{w}\t{$cnt, $dst|$dst, $cnt}", []>, OpSize16;
420def RCR32m1 : I<0xD1, MRM3m, (outs), (ins i32mem:$dst),
421                "rcr{l}\t$dst", []>, OpSize32;
422def RCR32mi : Ii8<0xC1, MRM3m, (outs), (ins i32mem:$dst, u8imm:$cnt),
423                  "rcr{l}\t{$cnt, $dst|$dst, $cnt}", []>, OpSize32;
424def RCR64m1 : RI<0xD1, MRM3m, (outs), (ins i64mem:$dst),
425                 "rcr{q}\t$dst", []>, Requires<[In64BitMode]>;
426def RCR64mi : RIi8<0xC1, MRM3m, (outs), (ins i64mem:$dst, u8imm:$cnt),
427                   "rcr{q}\t{$cnt, $dst|$dst, $cnt}", []>,
428                   Requires<[In64BitMode]>;
429} // Uses = [EFLAGS]
430
431let Uses = [CL, EFLAGS] in {
432def RCL8mCL : I<0xD2, MRM2m, (outs), (ins i8mem:$dst),
433                "rcl{b}\t{%cl, $dst|$dst, cl}", []>;
434def RCL16mCL : I<0xD3, MRM2m, (outs), (ins i16mem:$dst),
435                 "rcl{w}\t{%cl, $dst|$dst, cl}", []>, OpSize16;
436def RCL32mCL : I<0xD3, MRM2m, (outs), (ins i32mem:$dst),
437                 "rcl{l}\t{%cl, $dst|$dst, cl}", []>, OpSize32;
438def RCL64mCL : RI<0xD3, MRM2m, (outs), (ins i64mem:$dst),
439                  "rcl{q}\t{%cl, $dst|$dst, cl}", []>,
440                  Requires<[In64BitMode]>;
441
442def RCR8mCL : I<0xD2, MRM3m, (outs), (ins i8mem:$dst),
443                "rcr{b}\t{%cl, $dst|$dst, cl}", []>;
444def RCR16mCL : I<0xD3, MRM3m, (outs), (ins i16mem:$dst),
445                 "rcr{w}\t{%cl, $dst|$dst, cl}", []>, OpSize16;
446def RCR32mCL : I<0xD3, MRM3m, (outs), (ins i32mem:$dst),
447                 "rcr{l}\t{%cl, $dst|$dst, cl}", []>, OpSize32;
448def RCR64mCL : RI<0xD3, MRM3m, (outs), (ins i64mem:$dst),
449                  "rcr{q}\t{%cl, $dst|$dst, cl}", []>,
450                  Requires<[In64BitMode]>;
451} // Uses = [CL, EFLAGS]
452} // SchedRW
453} // hasSideEffects = 0
454
455let Constraints = "$src1 = $dst", SchedRW = [WriteShift] in {
456// FIXME: provide shorter instructions when imm8 == 1
457let Uses = [CL] in {
458def ROL8rCL  : I<0xD2, MRM0r, (outs GR8 :$dst), (ins GR8 :$src1),
459                 "rol{b}\t{%cl, $dst|$dst, cl}",
460                 [(set GR8:$dst, (rotl GR8:$src1, CL))]>;
461def ROL16rCL : I<0xD3, MRM0r, (outs GR16:$dst), (ins GR16:$src1),
462                 "rol{w}\t{%cl, $dst|$dst, cl}",
463                 [(set GR16:$dst, (rotl GR16:$src1, CL))]>, OpSize16;
464def ROL32rCL : I<0xD3, MRM0r, (outs GR32:$dst), (ins GR32:$src1),
465                 "rol{l}\t{%cl, $dst|$dst, cl}",
466                 [(set GR32:$dst, (rotl GR32:$src1, CL))]>, OpSize32;
467def ROL64rCL : RI<0xD3, MRM0r, (outs GR64:$dst), (ins GR64:$src1),
468                  "rol{q}\t{%cl, $dst|$dst, cl}",
469                  [(set GR64:$dst, (rotl GR64:$src1, CL))]>;
470}
471
472def ROL8ri   : Ii8<0xC0, MRM0r, (outs GR8 :$dst), (ins GR8 :$src1, u8imm:$src2),
473                   "rol{b}\t{$src2, $dst|$dst, $src2}",
474                   [(set GR8:$dst, (rotl GR8:$src1, (i8 imm:$src2)))]>;
475def ROL16ri  : Ii8<0xC1, MRM0r, (outs GR16:$dst), (ins GR16:$src1, u8imm:$src2),
476                   "rol{w}\t{$src2, $dst|$dst, $src2}",
477                   [(set GR16:$dst, (rotl GR16:$src1, (i8 imm:$src2)))]>, OpSize16;
478def ROL32ri  : Ii8<0xC1, MRM0r, (outs GR32:$dst), (ins GR32:$src1, u8imm:$src2),
479                   "rol{l}\t{$src2, $dst|$dst, $src2}",
480                   [(set GR32:$dst, (rotl GR32:$src1, (i8 imm:$src2)))]>, OpSize32;
481def ROL64ri  : RIi8<0xC1, MRM0r, (outs GR64:$dst),
482                    (ins GR64:$src1, u8imm:$src2),
483                    "rol{q}\t{$src2, $dst|$dst, $src2}",
484                    [(set GR64:$dst, (rotl GR64:$src1, (i8 imm:$src2)))]>;
485
486// Rotate by 1
487def ROL8r1   : I<0xD0, MRM0r, (outs GR8 :$dst), (ins GR8 :$src1),
488                 "rol{b}\t$dst",
489                 [(set GR8:$dst, (rotl GR8:$src1, (i8 1)))]>;
490def ROL16r1  : I<0xD1, MRM0r, (outs GR16:$dst), (ins GR16:$src1),
491                 "rol{w}\t$dst",
492                 [(set GR16:$dst, (rotl GR16:$src1, (i8 1)))]>, OpSize16;
493def ROL32r1  : I<0xD1, MRM0r, (outs GR32:$dst), (ins GR32:$src1),
494                 "rol{l}\t$dst",
495                 [(set GR32:$dst, (rotl GR32:$src1, (i8 1)))]>, OpSize32;
496def ROL64r1  : RI<0xD1, MRM0r, (outs GR64:$dst), (ins GR64:$src1),
497                  "rol{q}\t$dst",
498                  [(set GR64:$dst, (rotl GR64:$src1, (i8 1)))]>;
499} // Constraints = "$src = $dst", SchedRW
500
501let SchedRW = [WriteShiftLd, WriteRMW] in {
502let Uses = [CL] in {
503def ROL8mCL  : I<0xD2, MRM0m, (outs), (ins i8mem :$dst),
504                 "rol{b}\t{%cl, $dst|$dst, cl}",
505                 [(store (rotl (loadi8 addr:$dst), CL), addr:$dst)]>;
506def ROL16mCL : I<0xD3, MRM0m, (outs), (ins i16mem:$dst),
507                 "rol{w}\t{%cl, $dst|$dst, cl}",
508                 [(store (rotl (loadi16 addr:$dst), CL), addr:$dst)]>, OpSize16;
509def ROL32mCL : I<0xD3, MRM0m, (outs), (ins i32mem:$dst),
510                 "rol{l}\t{%cl, $dst|$dst, cl}",
511                 [(store (rotl (loadi32 addr:$dst), CL), addr:$dst)]>, OpSize32;
512def ROL64mCL :  RI<0xD3, MRM0m, (outs), (ins i64mem:$dst),
513                   "rol{q}\t{%cl, $dst|$dst, cl}",
514                   [(store (rotl (loadi64 addr:$dst), CL), addr:$dst)]>,
515                   Requires<[In64BitMode]>;
516}
517def ROL8mi   : Ii8<0xC0, MRM0m, (outs), (ins i8mem :$dst, u8imm:$src1),
518                   "rol{b}\t{$src1, $dst|$dst, $src1}",
519               [(store (rotl (loadi8 addr:$dst), (i8 imm:$src1)), addr:$dst)]>;
520def ROL16mi  : Ii8<0xC1, MRM0m, (outs), (ins i16mem:$dst, u8imm:$src1),
521                   "rol{w}\t{$src1, $dst|$dst, $src1}",
522              [(store (rotl (loadi16 addr:$dst), (i8 imm:$src1)), addr:$dst)]>,
523              OpSize16;
524def ROL32mi  : Ii8<0xC1, MRM0m, (outs), (ins i32mem:$dst, u8imm:$src1),
525                   "rol{l}\t{$src1, $dst|$dst, $src1}",
526              [(store (rotl (loadi32 addr:$dst), (i8 imm:$src1)), addr:$dst)]>,
527              OpSize32;
528def ROL64mi  : RIi8<0xC1, MRM0m, (outs), (ins i64mem:$dst, u8imm:$src1),
529                    "rol{q}\t{$src1, $dst|$dst, $src1}",
530                [(store (rotl (loadi64 addr:$dst), (i8 imm:$src1)), addr:$dst)]>,
531                Requires<[In64BitMode]>;
532
533// Rotate by 1
534def ROL8m1   : I<0xD0, MRM0m, (outs), (ins i8mem :$dst),
535                 "rol{b}\t$dst",
536                 [(store (rotl (loadi8 addr:$dst), (i8 1)), addr:$dst)]>;
537def ROL16m1  : I<0xD1, MRM0m, (outs), (ins i16mem:$dst),
538                 "rol{w}\t$dst",
539                 [(store (rotl (loadi16 addr:$dst), (i8 1)), addr:$dst)]>,
540                 OpSize16;
541def ROL32m1  : I<0xD1, MRM0m, (outs), (ins i32mem:$dst),
542                 "rol{l}\t$dst",
543                 [(store (rotl (loadi32 addr:$dst), (i8 1)), addr:$dst)]>,
544                 OpSize32;
545def ROL64m1  : RI<0xD1, MRM0m, (outs), (ins i64mem:$dst),
546                 "rol{q}\t$dst",
547                 [(store (rotl (loadi64 addr:$dst), (i8 1)), addr:$dst)]>,
548                 Requires<[In64BitMode]>;
549} // SchedRW
550
551let Constraints = "$src1 = $dst", SchedRW = [WriteShift] in {
552let Uses = [CL] in {
553def ROR8rCL  : I<0xD2, MRM1r, (outs GR8 :$dst), (ins GR8 :$src1),
554                 "ror{b}\t{%cl, $dst|$dst, cl}",
555                 [(set GR8:$dst, (rotr GR8:$src1, CL))]>;
556def ROR16rCL : I<0xD3, MRM1r, (outs GR16:$dst), (ins GR16:$src1),
557                 "ror{w}\t{%cl, $dst|$dst, cl}",
558                 [(set GR16:$dst, (rotr GR16:$src1, CL))]>, OpSize16;
559def ROR32rCL : I<0xD3, MRM1r, (outs GR32:$dst), (ins GR32:$src1),
560                 "ror{l}\t{%cl, $dst|$dst, cl}",
561                 [(set GR32:$dst, (rotr GR32:$src1, CL))]>, OpSize32;
562def ROR64rCL : RI<0xD3, MRM1r, (outs GR64:$dst), (ins GR64:$src1),
563                  "ror{q}\t{%cl, $dst|$dst, cl}",
564                  [(set GR64:$dst, (rotr GR64:$src1, CL))]>;
565}
566
567def ROR8ri   : Ii8<0xC0, MRM1r, (outs GR8 :$dst), (ins GR8 :$src1, u8imm:$src2),
568                   "ror{b}\t{$src2, $dst|$dst, $src2}",
569                   [(set GR8:$dst, (rotr GR8:$src1, (i8 relocImm:$src2)))]>;
570def ROR16ri  : Ii8<0xC1, MRM1r, (outs GR16:$dst), (ins GR16:$src1, u8imm:$src2),
571                   "ror{w}\t{$src2, $dst|$dst, $src2}",
572                   [(set GR16:$dst, (rotr GR16:$src1, (i8 relocImm:$src2)))]>,
573                   OpSize16;
574def ROR32ri  : Ii8<0xC1, MRM1r, (outs GR32:$dst), (ins GR32:$src1, u8imm:$src2),
575                   "ror{l}\t{$src2, $dst|$dst, $src2}",
576                   [(set GR32:$dst, (rotr GR32:$src1, (i8 relocImm:$src2)))]>,
577                   OpSize32;
578def ROR64ri  : RIi8<0xC1, MRM1r, (outs GR64:$dst),
579                    (ins GR64:$src1, u8imm:$src2),
580                    "ror{q}\t{$src2, $dst|$dst, $src2}",
581                    [(set GR64:$dst, (rotr GR64:$src1, (i8 relocImm:$src2)))]>;
582
583// Rotate by 1
584def ROR8r1   : I<0xD0, MRM1r, (outs GR8 :$dst), (ins GR8 :$src1),
585                 "ror{b}\t$dst",
586                 [(set GR8:$dst, (rotl GR8:$src1, (i8 7)))]>;
587def ROR16r1  : I<0xD1, MRM1r, (outs GR16:$dst), (ins GR16:$src1),
588                 "ror{w}\t$dst",
589                 [(set GR16:$dst, (rotl GR16:$src1, (i8 15)))]>, OpSize16;
590def ROR32r1  : I<0xD1, MRM1r, (outs GR32:$dst), (ins GR32:$src1),
591                 "ror{l}\t$dst",
592                 [(set GR32:$dst, (rotl GR32:$src1, (i8 31)))]>, OpSize32;
593def ROR64r1  : RI<0xD1, MRM1r, (outs GR64:$dst), (ins GR64:$src1),
594                  "ror{q}\t$dst",
595                  [(set GR64:$dst, (rotl GR64:$src1, (i8 63)))]>;
596} // Constraints = "$src = $dst", SchedRW
597
598let SchedRW = [WriteShiftLd, WriteRMW] in {
599let Uses = [CL] in {
600def ROR8mCL  : I<0xD2, MRM1m, (outs), (ins i8mem :$dst),
601                 "ror{b}\t{%cl, $dst|$dst, cl}",
602                 [(store (rotr (loadi8 addr:$dst), CL), addr:$dst)]>;
603def ROR16mCL : I<0xD3, MRM1m, (outs), (ins i16mem:$dst),
604                 "ror{w}\t{%cl, $dst|$dst, cl}",
605                 [(store (rotr (loadi16 addr:$dst), CL), addr:$dst)]>, OpSize16;
606def ROR32mCL : I<0xD3, MRM1m, (outs), (ins i32mem:$dst),
607                 "ror{l}\t{%cl, $dst|$dst, cl}",
608                 [(store (rotr (loadi32 addr:$dst), CL), addr:$dst)]>, OpSize32;
609def ROR64mCL : RI<0xD3, MRM1m, (outs), (ins i64mem:$dst),
610                  "ror{q}\t{%cl, $dst|$dst, cl}",
611                  [(store (rotr (loadi64 addr:$dst), CL), addr:$dst)]>,
612                  Requires<[In64BitMode]>;
613}
614def ROR8mi   : Ii8<0xC0, MRM1m, (outs), (ins i8mem :$dst, u8imm:$src),
615                   "ror{b}\t{$src, $dst|$dst, $src}",
616                   [(store (rotr (loadi8 addr:$dst), (i8 imm:$src)), addr:$dst)]>;
617def ROR16mi  : Ii8<0xC1, MRM1m, (outs), (ins i16mem:$dst, u8imm:$src),
618                   "ror{w}\t{$src, $dst|$dst, $src}",
619                   [(store (rotr (loadi16 addr:$dst), (i8 imm:$src)), addr:$dst)]>,
620                   OpSize16;
621def ROR32mi  : Ii8<0xC1, MRM1m, (outs), (ins i32mem:$dst, u8imm:$src),
622                   "ror{l}\t{$src, $dst|$dst, $src}",
623                   [(store (rotr (loadi32 addr:$dst), (i8 imm:$src)), addr:$dst)]>,
624                   OpSize32;
625def ROR64mi  : RIi8<0xC1, MRM1m, (outs), (ins i64mem:$dst, u8imm:$src),
626                    "ror{q}\t{$src, $dst|$dst, $src}",
627                    [(store (rotr (loadi64 addr:$dst), (i8 imm:$src)), addr:$dst)]>,
628                    Requires<[In64BitMode]>;
629
630// Rotate by 1
631def ROR8m1   : I<0xD0, MRM1m, (outs), (ins i8mem :$dst),
632                 "ror{b}\t$dst",
633                 [(store (rotl (loadi8 addr:$dst), (i8 7)), addr:$dst)]>;
634def ROR16m1  : I<0xD1, MRM1m, (outs), (ins i16mem:$dst),
635                 "ror{w}\t$dst",
636                 [(store (rotl (loadi16 addr:$dst), (i8 15)), addr:$dst)]>,
637                 OpSize16;
638def ROR32m1  : I<0xD1, MRM1m, (outs), (ins i32mem:$dst),
639                 "ror{l}\t$dst",
640                 [(store (rotl (loadi32 addr:$dst), (i8 31)), addr:$dst)]>,
641                 OpSize32;
642def ROR64m1  : RI<0xD1, MRM1m, (outs), (ins i64mem:$dst),
643                 "ror{q}\t$dst",
644                 [(store (rotl (loadi64 addr:$dst), (i8 63)), addr:$dst)]>,
645                 Requires<[In64BitMode]>;
646} // SchedRW
647
648
649//===----------------------------------------------------------------------===//
650// Double shift instructions (generalizations of rotate)
651//===----------------------------------------------------------------------===//
652
653let Constraints = "$src1 = $dst" in {
654
655let Uses = [CL], SchedRW = [WriteSHDrrcl] in {
656def SHLD16rrCL : I<0xA5, MRMDestReg, (outs GR16:$dst),
657                   (ins GR16:$src1, GR16:$src2),
658                   "shld{w}\t{%cl, $src2, $dst|$dst, $src2, cl}",
659                   [(set GR16:$dst, (X86shld GR16:$src1, GR16:$src2, CL))]>,
660                   TB, OpSize16;
661def SHRD16rrCL : I<0xAD, MRMDestReg, (outs GR16:$dst),
662                   (ins GR16:$src1, GR16:$src2),
663                   "shrd{w}\t{%cl, $src2, $dst|$dst, $src2, cl}",
664                   [(set GR16:$dst, (X86shrd GR16:$src1, GR16:$src2, CL))]>,
665                   TB, OpSize16;
666def SHLD32rrCL : I<0xA5, MRMDestReg, (outs GR32:$dst),
667                   (ins GR32:$src1, GR32:$src2),
668                   "shld{l}\t{%cl, $src2, $dst|$dst, $src2, cl}",
669                   [(set GR32:$dst, (X86shld GR32:$src1, GR32:$src2, CL))]>,
670                   TB, OpSize32;
671def SHRD32rrCL : I<0xAD, MRMDestReg, (outs GR32:$dst),
672                   (ins GR32:$src1, GR32:$src2),
673                   "shrd{l}\t{%cl, $src2, $dst|$dst, $src2, cl}",
674                   [(set GR32:$dst, (X86shrd GR32:$src1, GR32:$src2, CL))]>,
675                   TB, OpSize32;
676def SHLD64rrCL : RI<0xA5, MRMDestReg, (outs GR64:$dst),
677                    (ins GR64:$src1, GR64:$src2),
678                    "shld{q}\t{%cl, $src2, $dst|$dst, $src2, cl}",
679                    [(set GR64:$dst, (X86shld GR64:$src1, GR64:$src2, CL))]>,
680                    TB;
681def SHRD64rrCL : RI<0xAD, MRMDestReg, (outs GR64:$dst),
682                    (ins GR64:$src1, GR64:$src2),
683                    "shrd{q}\t{%cl, $src2, $dst|$dst, $src2, cl}",
684                    [(set GR64:$dst, (X86shrd GR64:$src1, GR64:$src2, CL))]>,
685                    TB;
686} // SchedRW
687
688let isCommutable = 1, SchedRW = [WriteSHDrri] in {  // These instructions commute to each other.
689def SHLD16rri8 : Ii8<0xA4, MRMDestReg,
690                     (outs GR16:$dst),
691                     (ins GR16:$src1, GR16:$src2, u8imm:$src3),
692                     "shld{w}\t{$src3, $src2, $dst|$dst, $src2, $src3}",
693                     [(set GR16:$dst, (X86shld GR16:$src1, GR16:$src2,
694                                      (i8 imm:$src3)))]>,
695                     TB, OpSize16;
696def SHRD16rri8 : Ii8<0xAC, MRMDestReg,
697                     (outs GR16:$dst),
698                     (ins GR16:$src1, GR16:$src2, u8imm:$src3),
699                     "shrd{w}\t{$src3, $src2, $dst|$dst, $src2, $src3}",
700                     [(set GR16:$dst, (X86shrd GR16:$src1, GR16:$src2,
701                                      (i8 imm:$src3)))]>,
702                     TB, OpSize16;
703def SHLD32rri8 : Ii8<0xA4, MRMDestReg,
704                     (outs GR32:$dst),
705                     (ins GR32:$src1, GR32:$src2, u8imm:$src3),
706                     "shld{l}\t{$src3, $src2, $dst|$dst, $src2, $src3}",
707                     [(set GR32:$dst, (X86shld GR32:$src1, GR32:$src2,
708                                      (i8 imm:$src3)))]>,
709                 TB, OpSize32;
710def SHRD32rri8 : Ii8<0xAC, MRMDestReg,
711                     (outs GR32:$dst),
712                     (ins GR32:$src1, GR32:$src2, u8imm:$src3),
713                     "shrd{l}\t{$src3, $src2, $dst|$dst, $src2, $src3}",
714                     [(set GR32:$dst, (X86shrd GR32:$src1, GR32:$src2,
715                                      (i8 imm:$src3)))]>,
716                 TB, OpSize32;
717def SHLD64rri8 : RIi8<0xA4, MRMDestReg,
718                      (outs GR64:$dst),
719                      (ins GR64:$src1, GR64:$src2, u8imm:$src3),
720                      "shld{q}\t{$src3, $src2, $dst|$dst, $src2, $src3}",
721                      [(set GR64:$dst, (X86shld GR64:$src1, GR64:$src2,
722                                       (i8 imm:$src3)))]>,
723                 TB;
724def SHRD64rri8 : RIi8<0xAC, MRMDestReg,
725                      (outs GR64:$dst),
726                      (ins GR64:$src1, GR64:$src2, u8imm:$src3),
727                      "shrd{q}\t{$src3, $src2, $dst|$dst, $src2, $src3}",
728                      [(set GR64:$dst, (X86shrd GR64:$src1, GR64:$src2,
729                                       (i8 imm:$src3)))]>,
730                 TB;
731} // SchedRW
732} // Constraints = "$src = $dst"
733
734let Uses = [CL], SchedRW = [WriteSHDmrcl] in {
735def SHLD16mrCL : I<0xA5, MRMDestMem, (outs), (ins i16mem:$dst, GR16:$src2),
736                   "shld{w}\t{%cl, $src2, $dst|$dst, $src2, cl}",
737                   [(store (X86shld (loadi16 addr:$dst), GR16:$src2, CL),
738                     addr:$dst)]>, TB, OpSize16;
739def SHRD16mrCL : I<0xAD, MRMDestMem, (outs), (ins i16mem:$dst, GR16:$src2),
740                  "shrd{w}\t{%cl, $src2, $dst|$dst, $src2, cl}",
741                  [(store (X86shrd (loadi16 addr:$dst), GR16:$src2, CL),
742                    addr:$dst)]>, TB, OpSize16;
743
744def SHLD32mrCL : I<0xA5, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src2),
745                   "shld{l}\t{%cl, $src2, $dst|$dst, $src2, cl}",
746                   [(store (X86shld (loadi32 addr:$dst), GR32:$src2, CL),
747                     addr:$dst)]>, TB, OpSize32;
748def SHRD32mrCL : I<0xAD, MRMDestMem, (outs), (ins i32mem:$dst, GR32:$src2),
749                  "shrd{l}\t{%cl, $src2, $dst|$dst, $src2, cl}",
750                  [(store (X86shrd (loadi32 addr:$dst), GR32:$src2, CL),
751                    addr:$dst)]>, TB, OpSize32;
752
753def SHLD64mrCL : RI<0xA5, MRMDestMem, (outs), (ins i64mem:$dst, GR64:$src2),
754                    "shld{q}\t{%cl, $src2, $dst|$dst, $src2, cl}",
755                    [(store (X86shld (loadi64 addr:$dst), GR64:$src2, CL),
756                      addr:$dst)]>, TB;
757def SHRD64mrCL : RI<0xAD, MRMDestMem, (outs), (ins i64mem:$dst, GR64:$src2),
758                    "shrd{q}\t{%cl, $src2, $dst|$dst, $src2, cl}",
759                    [(store (X86shrd (loadi64 addr:$dst), GR64:$src2, CL),
760                      addr:$dst)]>, TB;
761} // SchedRW
762
763let SchedRW = [WriteSHDmri] in {
764def SHLD16mri8 : Ii8<0xA4, MRMDestMem,
765                    (outs), (ins i16mem:$dst, GR16:$src2, u8imm:$src3),
766                    "shld{w}\t{$src3, $src2, $dst|$dst, $src2, $src3}",
767                    [(store (X86shld (loadi16 addr:$dst), GR16:$src2,
768                                      (i8 imm:$src3)), addr:$dst)]>,
769                    TB, OpSize16;
770def SHRD16mri8 : Ii8<0xAC, MRMDestMem,
771                     (outs), (ins i16mem:$dst, GR16:$src2, u8imm:$src3),
772                     "shrd{w}\t{$src3, $src2, $dst|$dst, $src2, $src3}",
773                    [(store (X86shrd (loadi16 addr:$dst), GR16:$src2,
774                                      (i8 imm:$src3)), addr:$dst)]>,
775                     TB, OpSize16;
776
777def SHLD32mri8 : Ii8<0xA4, MRMDestMem,
778                    (outs), (ins i32mem:$dst, GR32:$src2, u8imm:$src3),
779                    "shld{l}\t{$src3, $src2, $dst|$dst, $src2, $src3}",
780                    [(store (X86shld (loadi32 addr:$dst), GR32:$src2,
781                                      (i8 imm:$src3)), addr:$dst)]>,
782                    TB, OpSize32;
783def SHRD32mri8 : Ii8<0xAC, MRMDestMem,
784                     (outs), (ins i32mem:$dst, GR32:$src2, u8imm:$src3),
785                     "shrd{l}\t{$src3, $src2, $dst|$dst, $src2, $src3}",
786                     [(store (X86shrd (loadi32 addr:$dst), GR32:$src2,
787                                       (i8 imm:$src3)), addr:$dst)]>,
788                     TB, OpSize32;
789
790def SHLD64mri8 : RIi8<0xA4, MRMDestMem,
791                      (outs), (ins i64mem:$dst, GR64:$src2, u8imm:$src3),
792                      "shld{q}\t{$src3, $src2, $dst|$dst, $src2, $src3}",
793                      [(store (X86shld (loadi64 addr:$dst), GR64:$src2,
794                                       (i8 imm:$src3)), addr:$dst)]>,
795                 TB;
796def SHRD64mri8 : RIi8<0xAC, MRMDestMem,
797                      (outs), (ins i64mem:$dst, GR64:$src2, u8imm:$src3),
798                      "shrd{q}\t{$src3, $src2, $dst|$dst, $src2, $src3}",
799                      [(store (X86shrd (loadi64 addr:$dst), GR64:$src2,
800                                       (i8 imm:$src3)), addr:$dst)]>,
801                 TB;
802} // SchedRW
803
804} // Defs = [EFLAGS]
805
806// Sandy Bridge and newer Intel processors support faster rotates using
807// SHLD to avoid a partial flag update on the normal rotate instructions.
808let Predicates = [HasFastSHLDRotate], AddedComplexity = 5 in {
809  def : Pat<(rotl GR32:$src, (i8 imm:$shamt)),
810            (SHLD32rri8 GR32:$src, GR32:$src, imm:$shamt)>;
811  def : Pat<(rotl GR64:$src, (i8 imm:$shamt)),
812            (SHLD64rri8 GR64:$src, GR64:$src, imm:$shamt)>;
813}
814
815def ROT32L2R_imm8  : SDNodeXForm<imm, [{
816  // Convert a ROTL shamt to a ROTR shamt on 32-bit integer.
817  return getI8Imm(32 - N->getZExtValue(), SDLoc(N));
818}]>;
819
820def ROT64L2R_imm8  : SDNodeXForm<imm, [{
821  // Convert a ROTL shamt to a ROTR shamt on 64-bit integer.
822  return getI8Imm(64 - N->getZExtValue(), SDLoc(N));
823}]>;
824
825multiclass bmi_rotate<string asm, RegisterClass RC, X86MemOperand x86memop> {
826let hasSideEffects = 0 in {
827  def ri : Ii8<0xF0, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, u8imm:$src2),
828               !strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
829               []>, TAXD, VEX, Sched<[WriteShift]>;
830  let mayLoad = 1 in
831  def mi : Ii8<0xF0, MRMSrcMem, (outs RC:$dst),
832               (ins x86memop:$src1, u8imm:$src2),
833               !strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
834               []>, TAXD, VEX, Sched<[WriteShiftLd]>;
835}
836}
837
838multiclass bmi_shift<string asm, RegisterClass RC, X86MemOperand x86memop> {
839let hasSideEffects = 0 in {
840  def rr : I<0xF7, MRMSrcReg4VOp3, (outs RC:$dst), (ins RC:$src1, RC:$src2),
841             !strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>,
842             VEX, Sched<[WriteShift]>;
843  let mayLoad = 1 in
844  def rm : I<0xF7, MRMSrcMem4VOp3,
845             (outs RC:$dst), (ins x86memop:$src1, RC:$src2),
846             !strconcat(asm, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"), []>,
847             VEX, Sched<[WriteShiftLd,
848                         // x86memop:$src1
849                         ReadDefault, ReadDefault, ReadDefault, ReadDefault,
850                         ReadDefault,
851                         // RC:$src2
852                         ReadAfterLd]>;
853}
854}
855
856let Predicates = [HasBMI2] in {
857  defm RORX32 : bmi_rotate<"rorx{l}", GR32, i32mem>;
858  defm RORX64 : bmi_rotate<"rorx{q}", GR64, i64mem>, VEX_W;
859  defm SARX32 : bmi_shift<"sarx{l}", GR32, i32mem>, T8XS;
860  defm SARX64 : bmi_shift<"sarx{q}", GR64, i64mem>, T8XS, VEX_W;
861  defm SHRX32 : bmi_shift<"shrx{l}", GR32, i32mem>, T8XD;
862  defm SHRX64 : bmi_shift<"shrx{q}", GR64, i64mem>, T8XD, VEX_W;
863  defm SHLX32 : bmi_shift<"shlx{l}", GR32, i32mem>, T8PD;
864  defm SHLX64 : bmi_shift<"shlx{q}", GR64, i64mem>, T8PD, VEX_W;
865
866  // Prefer RORX which is non-destructive and doesn't update EFLAGS.
867  let AddedComplexity = 10 in {
868    def : Pat<(rotl GR32:$src, (i8 imm:$shamt)),
869              (RORX32ri GR32:$src, (ROT32L2R_imm8 imm:$shamt))>;
870    def : Pat<(rotl GR64:$src, (i8 imm:$shamt)),
871              (RORX64ri GR64:$src, (ROT64L2R_imm8 imm:$shamt))>;
872  }
873
874  def : Pat<(rotl (loadi32 addr:$src), (i8 imm:$shamt)),
875            (RORX32mi addr:$src, (ROT32L2R_imm8 imm:$shamt))>;
876  def : Pat<(rotl (loadi64 addr:$src), (i8 imm:$shamt)),
877            (RORX64mi addr:$src, (ROT64L2R_imm8 imm:$shamt))>;
878
879  // Prefer SARX/SHRX/SHLX over SAR/SHR/SHL with variable shift BUT not
880  // immedidate shift, i.e. the following code is considered better
881  //
882  //  mov %edi, %esi
883  //  shl $imm, %esi
884  //  ... %edi, ...
885  //
886  // than
887  //
888  //  movb $imm, %sil
889  //  shlx %sil, %edi, %esi
890  //  ... %edi, ...
891  //
892  let AddedComplexity = 1 in {
893    def : Pat<(sra GR32:$src1, GR8:$src2),
894              (SARX32rr GR32:$src1,
895                        (INSERT_SUBREG
896                          (i32 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>;
897    def : Pat<(sra GR64:$src1, GR8:$src2),
898              (SARX64rr GR64:$src1,
899                        (INSERT_SUBREG
900                          (i64 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>;
901
902    def : Pat<(srl GR32:$src1, GR8:$src2),
903              (SHRX32rr GR32:$src1,
904                        (INSERT_SUBREG
905                          (i32 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>;
906    def : Pat<(srl GR64:$src1, GR8:$src2),
907              (SHRX64rr GR64:$src1,
908                        (INSERT_SUBREG
909                          (i64 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>;
910
911    def : Pat<(shl GR32:$src1, GR8:$src2),
912              (SHLX32rr GR32:$src1,
913                        (INSERT_SUBREG
914                          (i32 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>;
915    def : Pat<(shl GR64:$src1, GR8:$src2),
916              (SHLX64rr GR64:$src1,
917                        (INSERT_SUBREG
918                          (i64 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>;
919  }
920
921  // We prefer to use
922  //  mov (%ecx), %esi
923  //  shl $imm, $esi
924  //
925  // over
926  //
927  //  movb $imm, %al
928  //  shlx %al, (%ecx), %esi
929  //
930  // This priority is enforced by IsProfitableToFoldLoad.
931  def : Pat<(sra (loadi32 addr:$src1), GR8:$src2),
932            (SARX32rm addr:$src1,
933                      (INSERT_SUBREG
934                        (i32 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>;
935  def : Pat<(sra (loadi64 addr:$src1), GR8:$src2),
936            (SARX64rm addr:$src1,
937                      (INSERT_SUBREG
938                        (i64 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>;
939
940  def : Pat<(srl (loadi32 addr:$src1), GR8:$src2),
941            (SHRX32rm addr:$src1,
942                      (INSERT_SUBREG
943                        (i32 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>;
944  def : Pat<(srl (loadi64 addr:$src1), GR8:$src2),
945            (SHRX64rm addr:$src1,
946                      (INSERT_SUBREG
947                        (i64 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>;
948
949  def : Pat<(shl (loadi32 addr:$src1), GR8:$src2),
950            (SHLX32rm addr:$src1,
951                      (INSERT_SUBREG
952                        (i32 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>;
953  def : Pat<(shl (loadi64 addr:$src1), GR8:$src2),
954            (SHLX64rm addr:$src1,
955                      (INSERT_SUBREG
956                        (i64 (IMPLICIT_DEF)), GR8:$src2, sub_8bit))>;
957}
958