Gitiles
Code Review Sign In
gerrit-public.fairphone.software / toolchain / llvm-project / 15748f492123053edda8ab18140630ad9efa7f76 / . / llvm / test / CodeGen / PowerPC / build-vector-tests.ll
blob: 7babd3fd58260d3b8dc5e0a400509cfc73dd14de [file] [log] [blame]
Nemanja Ivanovic15748f42016-12-06 11:47:14 +0000[diff] [blame^]1; RUN: llc -mcpu=pwr9 -ppc-vsr-nums-as-vr -mtriple=powerpc64-unknown-unknown \
2; RUN: < %s | FileCheck %s -check-prefix=P9BE -implicit-check-not frsp
3; RUN: llc -mcpu=pwr9 -ppc-vsr-nums-as-vr -mtriple=powerpc64le-unknown-unknown \
4; RUN: < %s | FileCheck %s -check-prefix=P9LE -implicit-check-not frsp
5; RUN: llc -mcpu=pwr8 -ppc-vsr-nums-as-vr -mtriple=powerpc64-unknown-unknown \
6; RUN: < %s | FileCheck %s -check-prefix=P8BE -implicit-check-not frsp
7; RUN: llc -mcpu=pwr8 -ppc-vsr-nums-as-vr -mtriple=powerpc64le-unknown-unknown \
8; RUN: < %s | FileCheck %s -check-prefix=P8LE -implicit-check-not frsp
9
10; This test case comes from the following C test case (included as it may be
11; slightly more readable than the LLVM IR.
12
13;/* This test case provides various ways of building vectors to ensure we
14; produce optimal code for all cases. The cases are (for each type):
15; - All zeros
16; - All ones
17; - Splat of a constant
18; - From different values already in registers
19; - From different constants
20; - From different values in memory
21; - Splat of a value in register
22; - Splat of a value in memory
23; - Inserting element into existing vector
24; - Inserting element from existing vector into existing vector
25;
26; With conversions (float <-> int)
27; - Splat of a constant
28; - From different values already in registers
29; - From different constants
30; - From different values in memory
31; - Splat of a value in register
32; - Splat of a value in memory
33; - Inserting element into existing vector
34; - Inserting element from existing vector into existing vector
35;*/
36;
37;/*=================================== int ===================================*/
38;// P8: xxlxor //
39;// P9: xxlxor //
40;vector int allZeroi() { //
41; return (vector int)0; //
42;} //
43;// P8: vspltisb -1 //
44;// P9: xxspltisb 255 //
45;vector int allOnei() { //
46; return (vector int)-1; //
47;} //
48;// P8: vspltisw 1 //
49;// P9: vspltisw 1 //
50;vector int spltConst1i() { //
51; return (vector int)1; //
52;} //
53;// P8: vspltisw -15; vsrw //
54;// P9: vspltisw -15; vsrw //
55;vector int spltConst16ki() { //
56; return (vector int)((1<<15) - 1); //
57;} //
58;// P8: vspltisw -16; vsrw //
59;// P9: vspltisw -16; vsrw //
60;vector int spltConst32ki() { //
61; return (vector int)((1<<16) - 1); //
62;} //
63;// P8: 4 x mtvsrwz, 2 x xxmrgh, vmrgow //
64;// P9: 2 x mtvsrdd, vmrgow //
65;vector int fromRegsi(int a, int b, int c, int d) { //
66; return (vector int){ a, b, c, d }; //
67;} //
68;// P8: lxvd2x, xxswapd //
69;// P9: lxvx (or even lxv) //
70;vector int fromDiffConstsi() { //
71; return (vector int) { 242, -113, 889, 19 }; //
72;} //
73;// P8: lxvd2x, xxswapd //
74;// P9: lxvx //
75;vector int fromDiffMemConsAi(int *arr) { //
76; return (vector int) { arr[0], arr[1], arr[2], arr[3] }; //
77;} //
78;// P8: 2 x lxvd2x, 2 x xxswapd, vperm //
79;// P9: 2 x lxvx, vperm //
80;vector int fromDiffMemConsDi(int *arr) { //
81; return (vector int) { arr[3], arr[2], arr[1], arr[0] }; //
82;} //
83;// P8: sldi 2, lxvd2x, xxswapd //
84;// P9: sldi 2, lxvx //
85;vector int fromDiffMemVarAi(int *arr, int elem) { //
86; return (vector int) { arr[elem], arr[elem+1], arr[elem+2], arr[elem+3] }; //
87;} //
88;// P8: sldi 2, 2 x lxvd2x, 2 x xxswapd, vperm //
89;// P9: sldi 2, 2 x lxvx, vperm //
90;vector int fromDiffMemVarDi(int *arr, int elem) { //
91; return (vector int) { arr[elem], arr[elem-1], arr[elem-2], arr[elem-3] }; //
92;} //
93;// P8: 4 x lwz, 4 x mtvsrwz, 2 x xxmrghd, vmrgow //
94;// P9: 4 x lwz, 2 x mtvsrdd, vmrgow //
95;vector int fromRandMemConsi(int *arr) { //
96; return (vector int) { arr[4], arr[18], arr[2], arr[88] }; //
97;} //
98;// P8: sldi 2, 4 x lwz, 4 x mtvsrwz, 2 x xxmrghd, vmrgow //
99;// P9: sldi 2, add, 4 x lwz, 2 x mtvsrdd, vmrgow //
100;vector int fromRandMemVari(int *arr, int elem) { //
101; return (vector int) { arr[elem+4], arr[elem+1], arr[elem+2], arr[elem+8] };//
102;} //
103;// P8: mtvsrwz, xxspltw //
104;// P9: mtvsrws //
105;vector int spltRegVali(int val) { //
106; return (vector int) val; //
107;} //
108;// P8: lxsiwax, xxspltw //
109;// P9: lxvwsx //
110;vector int spltMemVali(int *ptr) { //
111; return (vector int)*ptr; //
112;} //
113;// P8: vspltisw //
114;// P9: vspltisw //
115;vector int spltCnstConvftoi() { //
116; return (vector int) 4.74f; //
117;} //
118;// P8: 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspsxws //
119;// P9: 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvdpsxws //
120;vector int fromRegsConvftoi(float a, float b, float c, float d) { //
121; return (vector int) { a, b, c, d }; //
122;} //
123;// P8: lxvd2x, xxswapd //
124;// P9: lxvx (even lxv) //
125;vector int fromDiffConstsConvftoi() { //
126; return (vector int) { 24.46f, 234.f, 988.19f, 422.39f }; //
127;} //
128;// P8: lxvd2x, xxswapd, xvcvspsxws //
129;// P9: lxvx, xvcvspsxws //
130;vector int fromDiffMemConsAConvftoi(float *ptr) { //
131; return (vector int) { ptr[0], ptr[1], ptr[2], ptr[3] }; //
132;} //
133;// P8: 2 x lxvd2x, 2 x xxswapd, vperm, xvcvspsxws //
134;// P9: 2 x lxvx, vperm, xvcvspsxws //
135;vector int fromDiffMemConsDConvftoi(float *ptr) { //
136; return (vector int) { ptr[3], ptr[2], ptr[1], ptr[0] }; //
137;} //
138;// P8: 4 x lxsspx, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspsxws //
139;// P9: 4 x lxssp, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspsxws //
140;// Note: if the consecutive loads learns to handle pre-inc, this can be: //
141;// sldi 2, load, xvcvspuxws //
142;vector int fromDiffMemVarAConvftoi(float *arr, int elem) { //
143; return (vector int) { arr[elem], arr[elem+1], arr[elem+2], arr[elem+3] }; //
144;} //
145;// P8: 4 x lxsspx, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspsxws //
146;// P9: 4 x lxssp, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspsxws //
147;// Note: if the consecutive loads learns to handle pre-inc, this can be: //
148;// sldi 2, 2 x load, vperm, xvcvspuxws //
149;vector int fromDiffMemVarDConvftoi(float *arr, int elem) { //
150; return (vector int) { arr[elem], arr[elem-1], arr[elem-2], arr[elem-3] }; //
151;} //
152;// P8: xscvdpsxws, xxspltw //
153;// P9: xscvdpsxws, xxspltw //
154;vector int spltRegValConvftoi(float val) { //
155; return (vector int) val; //
156;} //
157;// P8: lxsspx, xscvdpsxws, xxspltw //
158;// P9: lxvwsx, xvcvspsxws //
159;vector int spltMemValConvftoi(float *ptr) { //
160; return (vector int)*ptr; //
161;} //
162;// P8: vspltisw //
163;// P9: vspltisw //
164;vector int spltCnstConvdtoi() { //
165; return (vector int) 4.74; //
166;} //
167;// P8: 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspsxws //
168;// P9: 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspsxws //
169;vector int fromRegsConvdtoi(double a, double b, double c, double d) { //
170; return (vector int) { a, b, c, d }; //
171;} //
172;// P8: lxvd2x, xxswapd //
173;// P9: lxvx (even lxv) //
174;vector int fromDiffConstsConvdtoi() { //
175; return (vector int) { 24.46, 234., 988.19, 422.39 }; //
176;} //
177;// P8: 2 x lxvd2x, 2 x xxswapd, xxmrgld, xxmrghd, 2 x xvcvdpsp, vmrgew, //
178;// xvcvspsxws //
179;// P9: 2 x lxvx, 2 x xxswapd, xxmrgld, xxmrghd, 2 x xvcvdpsp, vmrgew, //
180;// xvcvspsxws //
181;vector int fromDiffMemConsAConvdtoi(double *ptr) { //
182; return (vector int) { ptr[0], ptr[1], ptr[2], ptr[3] }; //
183;} //
184;// P8: 4 x lxsdx, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspsxws //
185;// P9: 4 x lfd, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspsxws //
186;vector int fromDiffMemConsDConvdtoi(double *ptr) { //
187; return (vector int) { ptr[3], ptr[2], ptr[1], ptr[0] }; //
188;} //
189;// P8: lfdux, 3 x lxsdx, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspsxws //
190;// P9: lfdux, 3 x lfd, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspsxws //
191;vector int fromDiffMemVarAConvdtoi(double *arr, int elem) { //
192; return (vector int) { arr[elem], arr[elem+1], arr[elem+2], arr[elem+3] }; //
193;} //
194;// P8: lfdux, 3 x lxsdx, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspsxws //
195;// P9: lfdux, 3 x lfd, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspsxws //
196;vector int fromDiffMemVarDConvdtoi(double *arr, int elem) { //
197; return (vector int) { arr[elem], arr[elem-1], arr[elem-2], arr[elem-3] }; //
198;} //
199;// P8: xscvdpsxws, xxspltw //
200;// P9: xscvdpsxws, xxspltw //
201;vector int spltRegValConvdtoi(double val) { //
202; return (vector int) val; //
203;} //
204;// P8: lxsdx, xscvdpsxws, xxspltw //
205;// P9: lxssp, xscvdpsxws, xxspltw //
206;vector int spltMemValConvdtoi(double *ptr) { //
207; return (vector int)*ptr; //
208;} //
209;/*=================================== int ===================================*/
210;/*=============================== unsigned int ==============================*/
211;// P8: xxlxor //
212;// P9: xxlxor //
213;vector unsigned int allZeroui() { //
214; return (vector unsigned int)0; //
215;} //
216;// P8: vspltisb -1 //
217;// P9: xxspltisb 255 //
218;vector unsigned int allOneui() { //
219; return (vector unsigned int)-1; //
220;} //
221;// P8: vspltisw 1 //
222;// P9: vspltisw 1 //
223;vector unsigned int spltConst1ui() { //
224; return (vector unsigned int)1; //
225;} //
226;// P8: vspltisw -15; vsrw //
227;// P9: vspltisw -15; vsrw //
228;vector unsigned int spltConst16kui() { //
229; return (vector unsigned int)((1<<15) - 1); //
230;} //
231;// P8: vspltisw -16; vsrw //
232;// P9: vspltisw -16; vsrw //
233;vector unsigned int spltConst32kui() { //
234; return (vector unsigned int)((1<<16) - 1); //
235;} //
236;// P8: 4 x mtvsrwz, 2 x xxmrghd, vmrgow //
237;// P9: 2 x mtvsrdd, vmrgow //
238;vector unsigned int fromRegsui(unsigned int a, unsigned int b, //
239; unsigned int c, unsigned int d) { //
240; return (vector unsigned int){ a, b, c, d }; //
241;} //
242;// P8: lxvd2x, xxswapd //
243;// P9: lxvx (or even lxv) //
244;vector unsigned int fromDiffConstsui() { //
245; return (vector unsigned int) { 242, -113, 889, 19 }; //
246;} //
247;// P8: lxvd2x, xxswapd //
248;// P9: lxvx //
249;vector unsigned int fromDiffMemConsAui(unsigned int *arr) { //
250; return (vector unsigned int) { arr[0], arr[1], arr[2], arr[3] }; //
251;} //
252;// P8: 2 x lxvd2x, 2 x xxswapd, vperm //
253;// P9: 2 x lxvx, vperm //
254;vector unsigned int fromDiffMemConsDui(unsigned int *arr) { //
255; return (vector unsigned int) { arr[3], arr[2], arr[1], arr[0] }; //
256;} //
257;// P8: sldi 2, lxvd2x, xxswapd //
258;// P9: sldi 2, lxvx //
259;vector unsigned int fromDiffMemVarAui(unsigned int *arr, int elem) { //
260; return (vector unsigned int) { arr[elem], arr[elem+1], //
261; arr[elem+2], arr[elem+3] }; //
262;} //
263;// P8: sldi 2, 2 x lxvd2x, 2 x xxswapd, vperm //
264;// P9: sldi 2, 2 x lxvx, vperm //
265;vector unsigned int fromDiffMemVarDui(unsigned int *arr, int elem) { //
266; return (vector unsigned int) { arr[elem], arr[elem-1], //
267; arr[elem-2], arr[elem-3] }; //
268;} //
269;// P8: 4 x lwz, 4 x mtvsrwz, 2 x xxmrghd, vmrgow //
270;// P9: 4 x lwz, 2 x mtvsrdd, vmrgow //
271;vector unsigned int fromRandMemConsui(unsigned int *arr) { //
272; return (vector unsigned int) { arr[4], arr[18], arr[2], arr[88] }; //
273;} //
274;// P8: sldi 2, 4 x lwz, 4 x mtvsrwz, 2 x xxmrghd, vmrgow //
275;// P9: sldi 2, add, 4 x lwz, 2 x mtvsrdd, vmrgow //
276;vector unsigned int fromRandMemVarui(unsigned int *arr, int elem) { //
277; return (vector unsigned int) { arr[elem+4], arr[elem+1], //
278; arr[elem+2], arr[elem+8] }; //
279;} //
280;// P8: mtvsrwz, xxspltw //
281;// P9: mtvsrws //
282;vector unsigned int spltRegValui(unsigned int val) { //
283; return (vector unsigned int) val; //
284;} //
285;// P8: lxsiwax, xxspltw //
286;// P9: lxvwsx //
287;vector unsigned int spltMemValui(unsigned int *ptr) { //
288; return (vector unsigned int)*ptr; //
289;} //
290;// P8: vspltisw //
291;// P9: vspltisw //
292;vector unsigned int spltCnstConvftoui() { //
293; return (vector unsigned int) 4.74f; //
294;} //
295;// P8: 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspuxws //
296;// P9: 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspuxws //
297;vector unsigned int fromRegsConvftoui(float a, float b, float c, float d) { //
298; return (vector unsigned int) { a, b, c, d }; //
299;} //
300;// P8: lxvd2x, xxswapd //
301;// P9: lxvx (even lxv) //
302;vector unsigned int fromDiffConstsConvftoui() { //
303; return (vector unsigned int) { 24.46f, 234.f, 988.19f, 422.39f }; //
304;} //
305;// P8: lxvd2x, xxswapd, xvcvspuxws //
306;// P9: lxvx, xvcvspuxws //
307;vector unsigned int fromDiffMemConsAConvftoui(float *ptr) { //
308; return (vector unsigned int) { ptr[0], ptr[1], ptr[2], ptr[3] }; //
309;} //
310;// P8: 2 x lxvd2x, 2 x xxswapd, vperm, xvcvspuxws //
311;// P9: 2 x lxvx, vperm, xvcvspuxws //
312;vector unsigned int fromDiffMemConsDConvftoui(float *ptr) { //
313; return (vector unsigned int) { ptr[3], ptr[2], ptr[1], ptr[0] }; //
314;} //
315;// P8: lfsux, 3 x lxsspx, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspuxws //
316;// P9: lfsux, 3 x lfs, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspuxws //
317;// Note: if the consecutive loads learns to handle pre-inc, this can be: //
318;// sldi 2, load, xvcvspuxws //
319;vector unsigned int fromDiffMemVarAConvftoui(float *arr, int elem) { //
320; return (vector unsigned int) { arr[elem], arr[elem+1], //
321; arr[elem+2], arr[elem+3] }; //
322;} //
323;// P8: lfsux, 3 x lxsspx, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspuxws //
324;// P9: lfsux, 3 x lfs, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspuxws //
325;// Note: if the consecutive loads learns to handle pre-inc, this can be: //
326;// sldi 2, 2 x load, vperm, xvcvspuxws //
327;vector unsigned int fromDiffMemVarDConvftoui(float *arr, int elem) { //
328; return (vector unsigned int) { arr[elem], arr[elem-1], //
329; arr[elem-2], arr[elem-3] }; //
330;} //
331;// P8: xscvdpuxws, xxspltw //
332;// P9: xscvdpuxws, xxspltw //
333;vector unsigned int spltRegValConvftoui(float val) { //
334; return (vector unsigned int) val; //
335;} //
336;// P8: lxsspx, xscvdpuxws, xxspltw //
337;// P9: lxvwsx, xvcvspuxws //
338;vector unsigned int spltMemValConvftoui(float *ptr) { //
339; return (vector unsigned int)*ptr; //
340;} //
341;// P8: vspltisw //
342;// P9: vspltisw //
343;vector unsigned int spltCnstConvdtoui() { //
344; return (vector unsigned int) 4.74; //
345;} //
346;// P8: 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspuxws //
347;// P9: 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspuxws //
348;vector unsigned int fromRegsConvdtoui(double a, double b, //
349; double c, double d) { //
350; return (vector unsigned int) { a, b, c, d }; //
351;} //
352;// P8: lxvd2x, xxswapd //
353;// P9: lxvx (even lxv) //
354;vector unsigned int fromDiffConstsConvdtoui() { //
355; return (vector unsigned int) { 24.46, 234., 988.19, 422.39 }; //
356;} //
357;// P8: 2 x lxvd2x, 2 x xxswapd, xxmrgld, xxmrghd, 2 x xvcvdpsp, vmrgew, //
358;// xvcvspuxws //
359;// P9: 2 x lxvx, xxmrgld, xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspuxws //
360;vector unsigned int fromDiffMemConsAConvdtoui(double *ptr) { //
361; return (vector unsigned int) { ptr[0], ptr[1], ptr[2], ptr[3] }; //
362;} //
363;// P8: 4 x lxsdx, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspuxws //
364;// P9: 4 x lfd, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspuxws //
365;vector unsigned int fromDiffMemConsDConvdtoui(double *ptr) { //
366; return (vector unsigned int) { ptr[3], ptr[2], ptr[1], ptr[0] }; //
367;} //
368;// P8: lfdux, 3 x lxsdx, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspuxws //
369;// P9: lfdux, 3 x lfd, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspuxws //
370;vector unsigned int fromDiffMemVarAConvdtoui(double *arr, int elem) { //
371; return (vector unsigned int) { arr[elem], arr[elem+1], //
372; arr[elem+2], arr[elem+3] }; //
373;} //
374;// P8: lfdux, 3 x lxsdx, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspuxws //
375;// P9: lfdux, 3 x lfd, 2 x xxmrghd, 2 x xvcvdpsp, vmrgew, xvcvspuxws //
376;vector unsigned int fromDiffMemVarDConvdtoui(double *arr, int elem) { //
377; return (vector unsigned int) { arr[elem], arr[elem-1], //
378; arr[elem-2], arr[elem-3] }; //
379;} //
380;// P8: xscvdpuxws, xxspltw //
381;// P9: xscvdpuxws, xxspltw //
382;vector unsigned int spltRegValConvdtoui(double val) { //
383; return (vector unsigned int) val; //
384;} //
385;// P8: lxsspx, xscvdpuxws, xxspltw //
386;// P9: lfd, xscvdpuxws, xxspltw //
387;vector unsigned int spltMemValConvdtoui(double *ptr) { //
388; return (vector unsigned int)*ptr; //
389;} //
390;/*=============================== unsigned int ==============================*/
391;/*=============================== long long =================================*/
392;// P8: xxlxor //
393;// P9: xxlxor //
394;vector long long allZeroll() { //
395; return (vector long long)0; //
396;} //
397;// P8: vspltisb -1 //
398;// P9: xxspltisb 255 //
399;vector long long allOnell() { //
400; return (vector long long)-1; //
401;} //
402;// P8: constant pool load (possible: vmrgew (xxlxor), (vspltisw)) //
403;// P9: constant pool load (possible: vmrgew (xxlxor), (vspltisw)) //
404;vector long long spltConst1ll() { //
405; return (vector long long)1; //
406;} //
407;// P8: constant pool load (possible: vmrgew (xxlxor), (vspltisw, vsrw)) //
408;// P9: constant pool load (possible: vmrgew (xxlxor), (vspltisw, vsrw)) //
409;vector long long spltConst16kll() { //
410; return (vector long long)((1<<15) - 1); //
411;} //
412;// P8: constant pool load (possible: vmrgew (xxlxor), (vspltisw, vsrw)) //
413;// P9: constant pool load (possible: vmrgew (xxlxor), (vspltisw, vsrw)) //
414;vector long long spltConst32kll() { //
415; return (vector long long)((1<<16) - 1); //
416;} //
417;// P8: 2 x mtvsrd, xxmrghd //
418;// P9: mtvsrdd //
419;vector long long fromRegsll(long long a, long long b) { //
420; return (vector long long){ a, b }; //
421;} //
422;// P8: lxvd2x, xxswapd //
423;// P9: lxvx (or even lxv) //
424;vector long long fromDiffConstsll() { //
425; return (vector long long) { 242, -113 }; //
426;} //
427;// P8: lxvd2x, xxswapd //
428;// P9: lxvx //
429;vector long long fromDiffMemConsAll(long long *arr) { //
430; return (vector long long) { arr[0], arr[1] }; //
431;} //
432;// P8: lxvd2x //
433;// P9: lxvx, xxswapd (maybe just use lxvd2x) //
434;vector long long fromDiffMemConsDll(long long *arr) { //
435; return (vector long long) { arr[3], arr[2] }; //
436;} //
437;// P8: sldi 3, lxvd2x, xxswapd //
438;// P9: sldi 3, lxvx //
439;vector long long fromDiffMemVarAll(long long *arr, int elem) { //
440; return (vector long long) { arr[elem], arr[elem+1] }; //
441;} //
442;// P8: sldi 3, lxvd2x //
443;// P9: sldi 3, lxvx, xxswapd (maybe just use lxvd2x) //
444;vector long long fromDiffMemVarDll(long long *arr, int elem) { //
445; return (vector long long) { arr[elem], arr[elem-1] }; //
446;} //
447;// P8: 2 x ld, 2 x mtvsrd, xxmrghd //
448;// P9: 2 x ld, mtvsrdd //
449;vector long long fromRandMemConsll(long long *arr) { //
450; return (vector long long) { arr[4], arr[18] }; //
451;} //
452;// P8: sldi 3, add, 2 x ld, 2 x mtvsrd, xxmrghd //
453;// P9: sldi 3, add, 2 x ld, mtvsrdd //
454;vector long long fromRandMemVarll(long long *arr, int elem) { //
455; return (vector long long) { arr[elem+4], arr[elem+1] }; //
456;} //
457;// P8: mtvsrd, xxspltd //
458;// P9: mtvsrdd //
459;vector long long spltRegValll(long long val) { //
460; return (vector long long) val; //
461;} //
462;// P8: lxvdsx //
463;// P9: lxvdsx //
464;vector long long spltMemValll(long long *ptr) { //
465; return (vector long long)*ptr; //
466;} //
467;// P8: constant pool load (possible: vmrgew (xxlxor), (vspltisw)) //
468;// P9: constant pool load (possible: vmrgew (xxlxor), (vspltisw)) //
469;vector long long spltCnstConvftoll() { //
470; return (vector long long) 4.74f; //
471;} //
472;// P8: xxmrghd, xvcvdpsxds //
473;// P9: xxmrghd, xvcvdpsxds //
474;vector long long fromRegsConvftoll(float a, float b) { //
475; return (vector long long) { a, b }; //
476;} //
477;// P8: lxvd2x, xxswapd //
478;// P9: lxvx (even lxv) //
479;vector long long fromDiffConstsConvftoll() { //
480; return (vector long long) { 24.46f, 234.f }; //
481;} //
482;// P8: 2 x lxsspx, xxmrghd, xvcvdpsxds //
483;// P9: 2 x lxssp, xxmrghd, xvcvdpsxds //
484;vector long long fromDiffMemConsAConvftoll(float *ptr) { //
485; return (vector long long) { ptr[0], ptr[1] }; //
486;} //
487;// P8: 2 x lxsspx, xxmrghd, xvcvdpsxds //
488;// P9: 2 x lxssp, xxmrghd, xvcvdpsxds //
489;vector long long fromDiffMemConsDConvftoll(float *ptr) { //
490; return (vector long long) { ptr[3], ptr[2] }; //
491;} //
492;// P8: sldi 2, lfsux, lxsspx, xxmrghd, xvcvdpsxds //
493;// P9: sldi 2, lfsux, lfs, xxmrghd, xvcvdpsxds //
494;vector long long fromDiffMemVarAConvftoll(float *arr, int elem) { //
495; return (vector long long) { arr[elem], arr[elem+1] }; //
496;} //
497;// P8: sldi 2, lfsux, lxsspx, xxmrghd, xvcvdpsxds //
498;// P9: sldi 2, lfsux, lfs, xxmrghd, xvcvdpsxds //
499;vector long long fromDiffMemVarDConvftoll(float *arr, int elem) { //
500; return (vector long long) { arr[elem], arr[elem-1] }; //
501;} //
502;// P8: xscvdpsxds, xxspltd //
503;// P9: xscvdpsxds, xxspltd //
504;vector long long spltRegValConvftoll(float val) { //
505; return (vector long long) val; //
506;} //
507;// P8: lxsspx, xscvdpsxds, xxspltd //
508;// P9: lfs, xscvdpsxds, xxspltd //
509;vector long long spltMemValConvftoll(float *ptr) { //
510; return (vector long long)*ptr; //
511;} //
512;// P8: constant pool load (possible: vmrgew (xxlxor), (vspltisw)) //
513;// P9: constant pool load (possible: vmrgew (xxlxor), (vspltisw)) //
514;vector long long spltCnstConvdtoll() { //
515; return (vector long long) 4.74; //
516;} //
517;// P8: xxmrghd, xvcvdpsxds //
518;// P9: xxmrghd, xvcvdpsxds //
519;vector long long fromRegsConvdtoll(double a, double b) { //
520; return (vector long long) { a, b }; //
521;} //
522;// P8: lxvd2x, xxswapd //
523;// P9: lxvx (even lxv) //
524;vector long long fromDiffConstsConvdtoll() { //
525; return (vector long long) { 24.46, 234. }; //
526;} //
527;// P8: lxvd2x, xxswapd, xvcvdpsxds //
528;// P9: lxvx, xvcvdpsxds //
529;vector long long fromDiffMemConsAConvdtoll(double *ptr) { //
530; return (vector long long) { ptr[0], ptr[1] }; //
531;} //
532;// P8: lxvd2x, xvcvdpsxds //
533;// P9: lxvx, xxswapd, xvcvdpsxds //
534;vector long long fromDiffMemConsDConvdtoll(double *ptr) { //
535; return (vector long long) { ptr[3], ptr[2] }; //
536;} //
537;// P8: sldi 3, lxvd2x, xxswapd, xvcvdpsxds //
538;// P9: sldi 3, lxvx, xvcvdpsxds //
539;vector long long fromDiffMemVarAConvdtoll(double *arr, int elem) { //
540; return (vector long long) { arr[elem], arr[elem+1] }; //
541;} //
542;// P8: sldi 3, lxvd2x, xvcvdpsxds //
543;// P9: sldi 3, lxvx, xxswapd, xvcvdpsxds //
544;vector long long fromDiffMemVarDConvdtoll(double *arr, int elem) { //
545; return (vector long long) { arr[elem], arr[elem-1] }; //
546;} //
547;// P8: xscvdpsxds, xxspltd //
548;// P9: xscvdpsxds, xxspltd //
549;vector long long spltRegValConvdtoll(double val) { //
550; return (vector long long) val; //
551;} //
552;// P8: lxvdsx, xvcvdpsxds //
553;// P9: lxvdsx, xvcvdpsxds //
554;vector long long spltMemValConvdtoll(double *ptr) { //
555; return (vector long long)*ptr; //
556;} //
557;/*=============================== long long =================================*/
558;/*========================== unsigned long long =============================*/
559;// P8: xxlxor //
560;// P9: xxlxor //
561;vector unsigned long long allZeroull() { //
562; return (vector unsigned long long)0; //
563;} //
564;// P8: vspltisb -1 //
565;// P9: xxspltisb 255 //
566;vector unsigned long long allOneull() { //
567; return (vector unsigned long long)-1; //
568;} //
569;// P8: constant pool load (possible: vmrgew (xxlxor), (vspltisw)) //
570;// P9: constant pool load (possible: vmrgew (xxlxor), (vspltisw)) //
571;vector unsigned long long spltConst1ull() { //
572; return (vector unsigned long long)1; //
573;} //
574;// P8: constant pool load (possible: vmrgew (xxlxor), (vspltisw, vsrw)) //
575;// P9: constant pool load (possible: vmrgew (xxlxor), (vspltisw, vsrw)) //
576;vector unsigned long long spltConst16kull() { //
577; return (vector unsigned long long)((1<<15) - 1); //
578;} //
579;// P8: constant pool load (possible: vmrgew (xxlxor), (vspltisw, vsrw)) //
580;// P9: constant pool load (possible: vmrgew (xxlxor), (vspltisw, vsrw)) //
581;vector unsigned long long spltConst32kull() { //
582; return (vector unsigned long long)((1<<16) - 1); //
583;} //
584;// P8: 2 x mtvsrd, xxmrghd //
585;// P9: mtvsrdd //
586;vector unsigned long long fromRegsull(unsigned long long a, //
587; unsigned long long b) { //
588; return (vector unsigned long long){ a, b }; //
589;} //
590;// P8: lxvd2x, xxswapd //
591;// P9: lxvx (or even lxv) //
592;vector unsigned long long fromDiffConstsull() { //
593; return (vector unsigned long long) { 242, -113 }; //
594;} //
595;// P8: lxvd2x, xxswapd //
596;// P9: lxvx //
597;vector unsigned long long fromDiffMemConsAull(unsigned long long *arr) { //
598; return (vector unsigned long long) { arr[0], arr[1] }; //
599;} //
600;// P8: lxvd2x //
601;// P9: lxvx, xxswapd (maybe just use lxvd2x) //
602;vector unsigned long long fromDiffMemConsDull(unsigned long long *arr) { //
603; return (vector unsigned long long) { arr[3], arr[2] }; //
604;} //
605;// P8: sldi 3, lxvd2x, xxswapd //
606;// P9: sldi 3, lxvx //
607;vector unsigned long long fromDiffMemVarAull(unsigned long long *arr, //
608; int elem) { //
609; return (vector unsigned long long) { arr[elem], arr[elem+1] }; //
610;} //
611;// P8: sldi 3, lxvd2x //
612;// P9: sldi 3, lxvx, xxswapd (maybe just use lxvd2x) //
613;vector unsigned long long fromDiffMemVarDull(unsigned long long *arr, //
614; int elem) { //
615; return (vector unsigned long long) { arr[elem], arr[elem-1] }; //
616;} //
617;// P8: 2 x ld, 2 x mtvsrd, xxmrghd //
618;// P9: 2 x ld, mtvsrdd //
619;vector unsigned long long fromRandMemConsull(unsigned long long *arr) { //
620; return (vector unsigned long long) { arr[4], arr[18] }; //
621;} //
622;// P8: sldi 3, add, 2 x ld, 2 x mtvsrd, xxmrghd //
623;// P9: sldi 3, add, 2 x ld, mtvsrdd //
624;vector unsigned long long fromRandMemVarull(unsigned long long *arr, //
625; int elem) { //
626; return (vector unsigned long long) { arr[elem+4], arr[elem+1] }; //
627;} //
628;// P8: mtvsrd, xxspltd //
629;// P9: mtvsrdd //
630;vector unsigned long long spltRegValull(unsigned long long val) { //
631; return (vector unsigned long long) val; //
632;} //
633;// P8: lxvdsx //
634;// P9: lxvdsx //
635;vector unsigned long long spltMemValull(unsigned long long *ptr) { //
636; return (vector unsigned long long)*ptr; //
637;} //
638;// P8: constant pool load (possible: vmrgew (xxlxor), (vspltisw)) //
639;// P9: constant pool load (possible: vmrgew (xxlxor), (vspltisw)) //
640;vector unsigned long long spltCnstConvftoull() { //
641; return (vector unsigned long long) 4.74f; //
642;} //
643;// P8: xxmrghd, xvcvdpuxds //
644;// P9: xxmrghd, xvcvdpuxds //
645;vector unsigned long long fromRegsConvftoull(float a, float b) { //
646; return (vector unsigned long long) { a, b }; //
647;} //
648;// P8: lxvd2x, xxswapd //
649;// P9: lxvx (even lxv) //
650;vector unsigned long long fromDiffConstsConvftoull() { //
651; return (vector unsigned long long) { 24.46f, 234.f }; //
652;} //
653;// P8: 2 x lxsspx, xxmrghd, xvcvdpuxds //
654;// P9: 2 x lxssp, xxmrghd, xvcvdpuxds //
655;vector unsigned long long fromDiffMemConsAConvftoull(float *ptr) { //
656; return (vector unsigned long long) { ptr[0], ptr[1] }; //
657;} //
658;// P8: 2 x lxsspx, xxmrghd, xvcvdpuxds //
659;// P9: 2 x lxssp, xxmrghd, xvcvdpuxds //
660;vector unsigned long long fromDiffMemConsDConvftoull(float *ptr) { //
661; return (vector unsigned long long) { ptr[3], ptr[2] }; //
662;} //
663;// P8: sldi 2, lfsux, lxsspx, xxmrghd, xvcvdpuxds //
664;// P9: sldi 2, lfsux, lfs, xxmrghd, xvcvdpuxds //
665;vector unsigned long long fromDiffMemVarAConvftoull(float *arr, int elem) { //
666; return (vector unsigned long long) { arr[elem], arr[elem+1] }; //
667;} //
668;// P8: sldi 2, lfsux, lxsspx, xxmrghd, xvcvdpuxds //
669;// P9: sldi 2, lfsux, lfs, xxmrghd, xvcvdpuxds //
670;vector unsigned long long fromDiffMemVarDConvftoull(float *arr, int elem) { //
671; return (vector unsigned long long) { arr[elem], arr[elem-1] }; //
672;} //
673;// P8: xscvdpuxds, xxspltd //
674;// P9: xscvdpuxds, xxspltd //
675;vector unsigned long long spltRegValConvftoull(float val) { //
676; return (vector unsigned long long) val; //
677;} //
678;// P8: lxsspx, xscvdpuxds, xxspltd //
679;// P9: lfs, xscvdpuxds, xxspltd //
680;vector unsigned long long spltMemValConvftoull(float *ptr) { //
681; return (vector unsigned long long)*ptr; //
682;} //
683;// P8: constant pool load (possible: vmrgew (xxlxor), (vspltisw)) //
684;// P9: constant pool load (possible: vmrgew (xxlxor), (vspltisw)) //
685;vector unsigned long long spltCnstConvdtoull() { //
686; return (vector unsigned long long) 4.74; //
687;} //
688;// P8: xxmrghd, xvcvdpuxds //
689;// P9: xxmrghd, xvcvdpuxds //
690;vector unsigned long long fromRegsConvdtoull(double a, double b) { //
691; return (vector unsigned long long) { a, b }; //
692;} //
693;// P8: lxvd2x, xxswapd //
694;// P9: lxvx (even lxv) //
695;vector unsigned long long fromDiffConstsConvdtoull() { //
696; return (vector unsigned long long) { 24.46, 234. }; //
697;} //
698;// P8: lxvd2x, xxswapd, xvcvdpuxds //
699;// P9: lxvx, xvcvdpuxds //
700;vector unsigned long long fromDiffMemConsAConvdtoull(double *ptr) { //
701; return (vector unsigned long long) { ptr[0], ptr[1] }; //
702;} //
703;// P8: lxvd2x, xvcvdpuxds //
704;// P9: lxvx, xxswapd, xvcvdpuxds //
705;vector unsigned long long fromDiffMemConsDConvdtoull(double *ptr) { //
706; return (vector unsigned long long) { ptr[3], ptr[2] }; //
707;} //
708;// P8: sldi 3, lxvd2x, xxswapd, xvcvdpuxds //
709;// P9: sldi 3, lxvx, xvcvdpuxds //
710;vector unsigned long long fromDiffMemVarAConvdtoull(double *arr, int elem) { //
711; return (vector unsigned long long) { arr[elem], arr[elem+1] }; //
712;} //
713;// P8: sldi 3, lxvd2x, xvcvdpuxds //
714;// P9: sldi 3, lxvx, xxswapd, xvcvdpuxds //
715;vector unsigned long long fromDiffMemVarDConvdtoull(double *arr, int elem) { //
716; return (vector unsigned long long) { arr[elem], arr[elem-1] }; //
717;} //
718;// P8: xscvdpuxds, xxspltd //
719;// P9: xscvdpuxds, xxspltd //
720;vector unsigned long long spltRegValConvdtoull(double val) { //
721; return (vector unsigned long long) val; //
722;} //
723;// P8: lxvdsx, xvcvdpuxds //
724;// P9: lxvdsx, xvcvdpuxds //
725;vector unsigned long long spltMemValConvdtoull(double *ptr) { //
726; return (vector unsigned long long)*ptr; //
727;} //
728;/*========================== unsigned long long ==============================*/
729
730; Function Attrs: norecurse nounwind readnone
731define <4 x i32> @allZeroi() {
732entry:
733 ret <4 x i32> zeroinitializer
734; P9BE-LABEL: allZeroi
735; P9LE-LABEL: allZeroi
736; P8BE-LABEL: allZeroi
737; P8LE-LABEL: allZeroi
738; P9BE: xxlxor v2, v2, v2
739; P9BE: blr
740; P9LE: xxlxor v2, v2, v2
741; P9LE: blr
742; P8BE: xxlxor v2, v2, v2
743; P8BE: blr
744; P8LE: xxlxor v2, v2, v2
745; P8LE: blr
746}
747
748; Function Attrs: norecurse nounwind readnone
749define <4 x i32> @allOnei() {
750entry:
751 ret <4 x i32> <i32 -1, i32 -1, i32 -1, i32 -1>
752; P9BE-LABEL: allOnei
753; P9LE-LABEL: allOnei
754; P8BE-LABEL: allOnei
755; P8LE-LABEL: allOnei
756; P9BE: xxspltib v2, 255
757; P9BE: blr
758; P9LE: xxspltib v2, 255
759; P9LE: blr
760; P8BE: vspltisb v2, -1
761; P8BE: blr
762; P8LE: vspltisb v2, -1
763; P8LE: blr
764}
765
766; Function Attrs: norecurse nounwind readnone
767define <4 x i32> @spltConst1i() {
768entry:
769 ret <4 x i32> <i32 1, i32 1, i32 1, i32 1>
770; P9BE-LABEL: spltConst1i
771; P9LE-LABEL: spltConst1i
772; P8BE-LABEL: spltConst1i
773; P8LE-LABEL: spltConst1i
774; P9BE: vspltisw v2, 1
775; P9BE: blr
776; P9LE: vspltisw v2, 1
777; P9LE: blr
778; P8BE: vspltisw v2, 1
779; P8BE: blr
780; P8LE: vspltisw v2, 1
781; P8LE: blr
782}
783
784; Function Attrs: norecurse nounwind readnone
785define <4 x i32> @spltConst16ki() {
786entry:
787 ret <4 x i32> <i32 32767, i32 32767, i32 32767, i32 32767>
788; P9BE-LABEL: spltConst16ki
789; P9LE-LABEL: spltConst16ki
790; P8BE-LABEL: spltConst16ki
791; P8LE-LABEL: spltConst16ki
792; P9BE: vspltisw v2, -15
793; P9BE: vsrw v2, v2, v2
794; P9BE: blr
795; P9LE: vspltisw v2, -15
796; P9LE: vsrw v2, v2, v2
797; P9LE: blr
798; P8BE: vspltisw v2, -15
799; P8BE: vsrw v2, v2, v2
800; P8BE: blr
801; P8LE: vspltisw v2, -15
802; P8LE: vsrw v2, v2, v2
803; P8LE: blr
804}
805
806; Function Attrs: norecurse nounwind readnone
807define <4 x i32> @spltConst32ki() {
808entry:
809 ret <4 x i32> <i32 65535, i32 65535, i32 65535, i32 65535>
810; P9BE-LABEL: spltConst32ki
811; P9LE-LABEL: spltConst32ki
812; P8BE-LABEL: spltConst32ki
813; P8LE-LABEL: spltConst32ki
814; P9BE: vspltisw v2, -16
815; P9BE: vsrw v2, v2, v2
816; P9BE: blr
817; P9LE: vspltisw v2, -16
818; P9LE: vsrw v2, v2, v2
819; P9LE: blr
820; P8BE: vspltisw v2, -16
821; P8BE: vsrw v2, v2, v2
822; P8BE: blr
823; P8LE: vspltisw v2, -16
824; P8LE: vsrw v2, v2, v2
825; P8LE: blr
826}
827
828; Function Attrs: norecurse nounwind readnone
829define <4 x i32> @fromRegsi(i32 signext %a, i32 signext %b, i32 signext %c, i32 signext %d) {
830entry:
831 %vecinit = insertelement <4 x i32> undef, i32 %a, i32 0
832 %vecinit1 = insertelement <4 x i32> %vecinit, i32 %b, i32 1
833 %vecinit2 = insertelement <4 x i32> %vecinit1, i32 %c, i32 2
834 %vecinit3 = insertelement <4 x i32> %vecinit2, i32 %d, i32 3
835 ret <4 x i32> %vecinit3
836; P9BE-LABEL: fromRegsi
837; P9LE-LABEL: fromRegsi
838; P8BE-LABEL: fromRegsi
839; P8LE-LABEL: fromRegsi
840; P9BE-DAG: mtvsrdd [[REG1:v[0-9]+]], r3, r5
841; P9BE-DAG: mtvsrdd [[REG2:v[0-9]+]], r4, r6
842; P9BE: vmrgow v2, [[REG1]], [[REG2]]
843; P9BE: blr
844; P9LE-DAG: mtvsrdd [[REG1:v[0-9]+]], r5, r3
845; P9LE-DAG: mtvsrdd [[REG2:v[0-9]+]], r6, r4
846; P9LE: vmrgow v2, [[REG2]], [[REG1]]
847; P9LE: blr
848; P8BE-DAG: mtvsrwz {{[vf]}}[[REG1:[0-9]+]], r3
849; P8BE-DAG: mtvsrwz {{[vf]}}[[REG2:[0-9]+]], r4
850; P8BE-DAG: mtvsrwz {{[vf]}}[[REG3:[0-9]+]], r5
851; P8BE-DAG: mtvsrwz {{[vf]}}[[REG4:[0-9]+]], r6
852; P8BE-DAG: xxmrghd [[REG5:v[0-9]+]], {{[v][s]*}}[[REG1]], {{[v][s]*}}[[REG3]]
853; P8BE-DAG: xxmrghd [[REG6:v[0-9]+]], {{[v][s]*}}[[REG2]], {{[v][s]*}}[[REG4]]
854; P8BE: vmrgow v2, [[REG5]], [[REG6]]
855; P8LE-DAG: mtvsrwz {{[vf]}}[[REG1:[0-9]+]], r3
856; P8LE-DAG: mtvsrwz {{[vf]}}[[REG2:[0-9]+]], r4
857; P8LE-DAG: mtvsrwz {{[vf]}}[[REG3:[0-9]+]], r5
858; P8LE-DAG: mtvsrwz {{[vf]}}[[REG4:[0-9]+]], r6
859; P8LE: xxmrghd [[REG5:v[0-9]+]], {{[v][s]*}}[[REG3]], {{[v][s]*}}[[REG1]]
860; P8LE: xxmrghd [[REG6:v[0-9]+]], {{[v][s]*}}[[REG4]], {{[v][s]*}}[[REG2]]
861; P8LE: vmrgow v2, [[REG6]], [[REG5]]
862}
863
864; Function Attrs: norecurse nounwind readnone
865define <4 x i32> @fromDiffConstsi() {
866entry:
867 ret <4 x i32> <i32 242, i32 -113, i32 889, i32 19>
868; P9BE-LABEL: fromDiffConstsi
869; P9LE-LABEL: fromDiffConstsi
870; P8BE-LABEL: fromDiffConstsi
871; P8LE-LABEL: fromDiffConstsi
872; P9BE: lxvx
873; P9BE: blr
874; P9LE: lxvx
875; P9LE: blr
876; P8BE: lxvw4x
877; P8BE: blr
878; P8LE: lxvd2x
879; P8LE: xxswapd
880; P8LE: blr
881}
882
883; Function Attrs: norecurse nounwind readonly
884define <4 x i32> @fromDiffMemConsAi(i32* nocapture readonly %arr) {
885entry:
886 %0 = load i32, i32* %arr, align 4
887 %vecinit = insertelement <4 x i32> undef, i32 %0, i32 0
888 %arrayidx1 = getelementptr inbounds i32, i32* %arr, i64 1
889 %1 = load i32, i32* %arrayidx1, align 4
890 %vecinit2 = insertelement <4 x i32> %vecinit, i32 %1, i32 1
891 %arrayidx3 = getelementptr inbounds i32, i32* %arr, i64 2
892 %2 = load i32, i32* %arrayidx3, align 4
893 %vecinit4 = insertelement <4 x i32> %vecinit2, i32 %2, i32 2
894 %arrayidx5 = getelementptr inbounds i32, i32* %arr, i64 3
895 %3 = load i32, i32* %arrayidx5, align 4
896 %vecinit6 = insertelement <4 x i32> %vecinit4, i32 %3, i32 3
897 ret <4 x i32> %vecinit6
898; P9BE-LABEL: fromDiffMemConsAi
899; P9LE-LABEL: fromDiffMemConsAi
900; P8BE-LABEL: fromDiffMemConsAi
901; P8LE-LABEL: fromDiffMemConsAi
902; P9BE: lxvx
903; P9BE: blr
904; P9LE: lxvx
905; P9LE: blr
906; P8BE: lxvw4x
907; P8BE: blr
908; P8LE: lxvd2x
909; P8LE: xxswapd
910; P8LE: blr
911}
912
913; Function Attrs: norecurse nounwind readonly
914define <4 x i32> @fromDiffMemConsDi(i32* nocapture readonly %arr) {
915entry:
916 %arrayidx = getelementptr inbounds i32, i32* %arr, i64 3
917 %0 = load i32, i32* %arrayidx, align 4
918 %vecinit = insertelement <4 x i32> undef, i32 %0, i32 0
919 %arrayidx1 = getelementptr inbounds i32, i32* %arr, i64 2
920 %1 = load i32, i32* %arrayidx1, align 4
921 %vecinit2 = insertelement <4 x i32> %vecinit, i32 %1, i32 1
922 %arrayidx3 = getelementptr inbounds i32, i32* %arr, i64 1
923 %2 = load i32, i32* %arrayidx3, align 4
924 %vecinit4 = insertelement <4 x i32> %vecinit2, i32 %2, i32 2
925 %3 = load i32, i32* %arr, align 4
926 %vecinit6 = insertelement <4 x i32> %vecinit4, i32 %3, i32 3
927 ret <4 x i32> %vecinit6
928; P9BE-LABEL: fromDiffMemConsDi
929; P9LE-LABEL: fromDiffMemConsDi
930; P8BE-LABEL: fromDiffMemConsDi
931; P8LE-LABEL: fromDiffMemConsDi
932; P9BE: lxvx
933; P9BE: lxvx
934; P9BE: vperm
935; P9BE: blr
936; P9LE: lxvx
937; P9LE: lxvx
938; P9LE: vperm
939; P9LE: blr
940; P8BE: lxvw4x
941; P8BE: lxvw4x
942; P8BE: vperm
943; P8BE: blr
944; P8LE: lxvd2x
945; P8LE-DAG: lxvd2x
946; P8LE-DAG: xxswapd
947; P8LE: xxswapd
948; P8LE: vperm
949; P8LE: blr
950}
951
952; Function Attrs: norecurse nounwind readonly
953define <4 x i32> @fromDiffMemVarAi(i32* nocapture readonly %arr, i32 signext %elem) {
954entry:
955 %idxprom = sext i32 %elem to i64
956 %arrayidx = getelementptr inbounds i32, i32* %arr, i64 %idxprom
957 %0 = load i32, i32* %arrayidx, align 4
958 %vecinit = insertelement <4 x i32> undef, i32 %0, i32 0
959 %add = add nsw i32 %elem, 1
960 %idxprom1 = sext i32 %add to i64
961 %arrayidx2 = getelementptr inbounds i32, i32* %arr, i64 %idxprom1
962 %1 = load i32, i32* %arrayidx2, align 4
963 %vecinit3 = insertelement <4 x i32> %vecinit, i32 %1, i32 1
964 %add4 = add nsw i32 %elem, 2
965 %idxprom5 = sext i32 %add4 to i64
966 %arrayidx6 = getelementptr inbounds i32, i32* %arr, i64 %idxprom5
967 %2 = load i32, i32* %arrayidx6, align 4
968 %vecinit7 = insertelement <4 x i32> %vecinit3, i32 %2, i32 2
969 %add8 = add nsw i32 %elem, 3
970 %idxprom9 = sext i32 %add8 to i64
971 %arrayidx10 = getelementptr inbounds i32, i32* %arr, i64 %idxprom9
972 %3 = load i32, i32* %arrayidx10, align 4
973 %vecinit11 = insertelement <4 x i32> %vecinit7, i32 %3, i32 3
974 ret <4 x i32> %vecinit11
975; P9BE-LABEL: fromDiffMemVarAi
976; P9LE-LABEL: fromDiffMemVarAi
977; P8BE-LABEL: fromDiffMemVarAi
978; P8LE-LABEL: fromDiffMemVarAi
979; P9BE: sldi r4, r4, 2
980; P9BE: lxvx v2, r3, r4
981; P9BE: blr
982; P9LE: sldi r4, r4, 2
983; P9LE: lxvx v2, r3, r4
984; P9LE: blr
985; P8BE: sldi r4, r4, 2
986; P8BE: lxvw4x {{[vs0-9]+}}, r3, r4
987; P8BE: blr
988; P8LE: sldi r4, r4, 2
989; P8LE: lxvd2x {{[vs0-9]+}}, r3, r4
990; P8LE: xxswapd
991; P8LE: blr
992}
993
994; Function Attrs: norecurse nounwind readonly
995define <4 x i32> @fromDiffMemVarDi(i32* nocapture readonly %arr, i32 signext %elem) {
996entry:
997 %idxprom = sext i32 %elem to i64
998 %arrayidx = getelementptr inbounds i32, i32* %arr, i64 %idxprom
999 %0 = load i32, i32* %arrayidx, align 4
1000 %vecinit = insertelement <4 x i32> undef, i32 %0, i32 0
1001 %sub = add nsw i32 %elem, -1
1002 %idxprom1 = sext i32 %sub to i64
1003 %arrayidx2 = getelementptr inbounds i32, i32* %arr, i64 %idxprom1
1004 %1 = load i32, i32* %arrayidx2, align 4
1005 %vecinit3 = insertelement <4 x i32> %vecinit, i32 %1, i32 1
1006 %sub4 = add nsw i32 %elem, -2
1007 %idxprom5 = sext i32 %sub4 to i64
1008 %arrayidx6 = getelementptr inbounds i32, i32* %arr, i64 %idxprom5
1009 %2 = load i32, i32* %arrayidx6, align 4
1010 %vecinit7 = insertelement <4 x i32> %vecinit3, i32 %2, i32 2
1011 %sub8 = add nsw i32 %elem, -3
1012 %idxprom9 = sext i32 %sub8 to i64
1013 %arrayidx10 = getelementptr inbounds i32, i32* %arr, i64 %idxprom9
1014 %3 = load i32, i32* %arrayidx10, align 4
1015 %vecinit11 = insertelement <4 x i32> %vecinit7, i32 %3, i32 3
1016 ret <4 x i32> %vecinit11
1017; P9BE-LABEL: fromDiffMemVarDi
1018; P9LE-LABEL: fromDiffMemVarDi
1019; P8BE-LABEL: fromDiffMemVarDi
1020; P8LE-LABEL: fromDiffMemVarDi
1021; P9BE: sldi r4, r4, 2
1022; P9BE-DAG: lxvx {{[vs0-9]+}}, r3, r4
1023; P9BE-DAG: lxvx
1024; P9BE: vperm
1025; P9BE: blr
1026; P9LE: sldi r4, r4, 2
1027; P9LE-DAG: lxvx {{[vs0-9]+}}, r3, r4
1028; P9LE-DAG: lxvx
1029; P9LE: vperm
1030; P9LE: blr
1031; P8BE: sldi r4, r4, 2
1032; P8BE-DAG: lxvw4x {{[vs0-9]+}}, r3, r4
1033; P8BE-DAG: lxvw4x
1034; P8BE: vperm
1035; P8BE: blr
1036; P8LE: sldi r4, r4, 2
1037; P8LE-DAG: lxvd2x
1038; P8LE-DAG: lxvd2x
1039; P8LE-DAG: xxswapd
1040; P8LE: xxswapd
1041; P8LE: vperm
1042; P8LE: blr
1043}
1044
1045; Function Attrs: norecurse nounwind readonly
1046define <4 x i32> @fromRandMemConsi(i32* nocapture readonly %arr) {
1047entry:
1048 %arrayidx = getelementptr inbounds i32, i32* %arr, i64 4
1049 %0 = load i32, i32* %arrayidx, align 4
1050 %vecinit = insertelement <4 x i32> undef, i32 %0, i32 0
1051 %arrayidx1 = getelementptr inbounds i32, i32* %arr, i64 18
1052 %1 = load i32, i32* %arrayidx1, align 4
1053 %vecinit2 = insertelement <4 x i32> %vecinit, i32 %1, i32 1
1054 %arrayidx3 = getelementptr inbounds i32, i32* %arr, i64 2
1055 %2 = load i32, i32* %arrayidx3, align 4
1056 %vecinit4 = insertelement <4 x i32> %vecinit2, i32 %2, i32 2
1057 %arrayidx5 = getelementptr inbounds i32, i32* %arr, i64 88
1058 %3 = load i32, i32* %arrayidx5, align 4
1059 %vecinit6 = insertelement <4 x i32> %vecinit4, i32 %3, i32 3
1060 ret <4 x i32> %vecinit6
1061; P9BE-LABEL: fromRandMemConsi
1062; P9LE-LABEL: fromRandMemConsi
1063; P8BE-LABEL: fromRandMemConsi
1064; P8LE-LABEL: fromRandMemConsi
1065; P9BE: lwz
1066; P9BE: lwz
1067; P9BE: lwz
1068; P9BE: lwz
1069; P9BE: mtvsrdd
1070; P9BE: mtvsrdd
1071; P9BE: vmrgow
1072; P9LE: lwz
1073; P9LE: lwz
1074; P9LE: lwz
1075; P9LE: lwz
1076; P9LE: mtvsrdd
1077; P9LE: mtvsrdd
1078; P9LE: vmrgow
1079; P8BE: lwz
1080; P8BE: lwz
1081; P8BE: lwz
1082; P8BE: lwz
1083; P8BE: mtvsrwz
1084; P8BE: mtvsrwz
1085; P8BE: mtvsrwz
1086; P8BE: mtvsrwz
1087; P8BE: xxmrghd
1088; P8BE: xxmrghd
1089; P8BE: vmrgow
1090; P8LE: lwz
1091; P8LE: lwz
1092; P8LE: lwz
1093; P8LE: lwz
1094; P8LE: mtvsrwz
1095; P8LE: mtvsrwz
1096; P8LE: mtvsrwz
1097; P8LE: mtvsrwz
1098; P8LE: xxmrghd
1099; P8LE: xxmrghd
1100; P8LE: vmrgow
1101}
1102
1103; Function Attrs: norecurse nounwind readonly
1104define <4 x i32> @fromRandMemVari(i32* nocapture readonly %arr, i32 signext %elem) {
1105entry:
1106 %add = add nsw i32 %elem, 4
1107 %idxprom = sext i32 %add to i64
1108 %arrayidx = getelementptr inbounds i32, i32* %arr, i64 %idxprom
1109 %0 = load i32, i32* %arrayidx, align 4
1110 %vecinit = insertelement <4 x i32> undef, i32 %0, i32 0
1111 %add1 = add nsw i32 %elem, 1
1112 %idxprom2 = sext i32 %add1 to i64
1113 %arrayidx3 = getelementptr inbounds i32, i32* %arr, i64 %idxprom2
1114 %1 = load i32, i32* %arrayidx3, align 4
1115 %vecinit4 = insertelement <4 x i32> %vecinit, i32 %1, i32 1
1116 %add5 = add nsw i32 %elem, 2
1117 %idxprom6 = sext i32 %add5 to i64
1118 %arrayidx7 = getelementptr inbounds i32, i32* %arr, i64 %idxprom6
1119 %2 = load i32, i32* %arrayidx7, align 4
1120 %vecinit8 = insertelement <4 x i32> %vecinit4, i32 %2, i32 2
1121 %add9 = add nsw i32 %elem, 8
1122 %idxprom10 = sext i32 %add9 to i64
1123 %arrayidx11 = getelementptr inbounds i32, i32* %arr, i64 %idxprom10
1124 %3 = load i32, i32* %arrayidx11, align 4
1125 %vecinit12 = insertelement <4 x i32> %vecinit8, i32 %3, i32 3
1126 ret <4 x i32> %vecinit12
1127; P9BE-LABEL: fromRandMemVari
1128; P9LE-LABEL: fromRandMemVari
1129; P8BE-LABEL: fromRandMemVari
1130; P8LE-LABEL: fromRandMemVari
1131; P9BE: sldi r4, r4, 2
1132; P9BE: lwz
1133; P9BE: lwz
1134; P9BE: lwz
1135; P9BE: lwz
1136; P9BE: mtvsrdd
1137; P9BE: mtvsrdd
1138; P9BE: vmrgow
1139; P9LE: sldi r4, r4, 2
1140; P9LE: lwz
1141; P9LE: lwz
1142; P9LE: lwz
1143; P9LE: lwz
1144; P9LE: mtvsrdd
1145; P9LE: mtvsrdd
1146; P9LE: vmrgow
1147; P8BE: sldi r4, r4, 2
1148; P8BE: lwz
1149; P8BE: lwz
1150; P8BE: lwz
1151; P8BE: lwz
1152; P8BE: mtvsrwz
1153; P8BE: mtvsrwz
1154; P8BE: mtvsrwz
1155; P8BE: mtvsrwz
1156; P8BE: xxmrghd
1157; P8BE: xxmrghd
1158; P8BE: vmrgow
1159; P8LE: sldi r4, r4, 2
1160; P8LE: lwz
1161; P8LE: lwz
1162; P8LE: lwz
1163; P8LE: lwz
1164; P8LE: mtvsrwz
1165; P8LE: mtvsrwz
1166; P8LE: mtvsrwz
1167; P8LE: mtvsrwz
1168; P8LE: xxmrghd
1169; P8LE: xxmrghd
1170; P8LE: vmrgow
1171}
1172
1173; Function Attrs: norecurse nounwind readnone
1174define <4 x i32> @spltRegVali(i32 signext %val) {
1175entry:
1176 %splat.splatinsert = insertelement <4 x i32> undef, i32 %val, i32 0
1177 %splat.splat = shufflevector <4 x i32> %splat.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
1178 ret <4 x i32> %splat.splat
1179; P9BE-LABEL: spltRegVali
1180; P9LE-LABEL: spltRegVali
1181; P8BE-LABEL: spltRegVali
1182; P8LE-LABEL: spltRegVali
1183; P9BE: mtvsrws v2, r3
1184; P9BE: blr
1185; P9LE: mtvsrws v2, r3
1186; P9LE: blr
1187; P8BE: mtvsrwz {{[vsf0-9]+}}, r3
1188; P8BE: xxspltw v2, {{[vsf0-9]+}}, 1
1189; P8BE: blr
1190; P8LE: mtvsrwz {{[vsf0-9]+}}, r3
1191; P8LE: xxspltw v2, {{[vsf0-9]+}}, 1
1192; P8LE: blr
1193}
1194
1195; Function Attrs: norecurse nounwind readonly
1196define <4 x i32> @spltMemVali(i32* nocapture readonly %ptr) {
1197entry:
1198 %0 = load i32, i32* %ptr, align 4
1199 %splat.splatinsert = insertelement <4 x i32> undef, i32 %0, i32 0
1200 %splat.splat = shufflevector <4 x i32> %splat.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
1201 ret <4 x i32> %splat.splat
1202; P9BE-LABEL: spltMemVali
1203; P9LE-LABEL: spltMemVali
1204; P8BE-LABEL: spltMemVali
1205; P8LE-LABEL: spltMemVali
1206; P9BE: lxvwsx v2, 0, r3
1207; P9BE: blr
1208; P9LE: lxvwsx v2, 0, r3
1209; P9LE: blr
1210; P8BE: lxsiwax {{[vsf0-9]+}}, 0, r3
1211; P8BE: xxspltw v2, {{[vsf0-9]+}}, 1
1212; P8BE: blr
1213; P8LE: lxsiwax {{[vsf0-9]+}}, 0, r3
1214; P8LE: xxspltw v2, {{[vsf0-9]+}}, 1
1215; P8LE: blr
1216}
1217
1218; Function Attrs: norecurse nounwind readnone
1219define <4 x i32> @spltCnstConvftoi() {
1220entry:
1221 ret <4 x i32> <i32 4, i32 4, i32 4, i32 4>
1222; P9BE-LABEL: spltCnstConvftoi
1223; P9LE-LABEL: spltCnstConvftoi
1224; P8BE-LABEL: spltCnstConvftoi
1225; P8LE-LABEL: spltCnstConvftoi
1226; P9BE: vspltisw v2, 4
1227; P9BE: blr
1228; P9LE: vspltisw v2, 4
1229; P9LE: blr
1230; P8BE: vspltisw v2, 4
1231; P8BE: blr
1232; P8LE: vspltisw v2, 4
1233; P8LE: blr
1234}
1235
1236; Function Attrs: norecurse nounwind readnone
1237define <4 x i32> @fromRegsConvftoi(float %a, float %b, float %c, float %d) {
1238entry:
1239 %conv = fptosi float %a to i32
1240 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
1241 %conv1 = fptosi float %b to i32
1242 %vecinit2 = insertelement <4 x i32> %vecinit, i32 %conv1, i32 1
1243 %conv3 = fptosi float %c to i32
1244 %vecinit4 = insertelement <4 x i32> %vecinit2, i32 %conv3, i32 2
1245 %conv5 = fptosi float %d to i32
1246 %vecinit6 = insertelement <4 x i32> %vecinit4, i32 %conv5, i32 3
1247 ret <4 x i32> %vecinit6
1248; P9BE-LABEL: fromRegsConvftoi
1249; P9LE-LABEL: fromRegsConvftoi
1250; P8BE-LABEL: fromRegsConvftoi
1251; P8LE-LABEL: fromRegsConvftoi
1252; P9BE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs1, vs3
1253; P9BE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs2, vs4
1254; P9BE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
1255; P9BE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
1256; P9BE: vmrgew v2, [[REG3]], [[REG4]]
1257; P9BE: xvcvspsxws v2, v2
1258; P9LE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs3, vs1
1259; P9LE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs4, vs2
1260; P9LE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
1261; P9LE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
1262; P9LE: vmrgew v2, [[REG4]], [[REG3]]
1263; P9LE: xvcvspsxws v2, v2
1264; P8BE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs1, vs3
1265; P8BE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs2, vs4
1266; P8BE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
1267; P8BE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
1268; P8BE: vmrgew v2, [[REG3]], [[REG4]]
1269; P8BE: xvcvspsxws v2, v2
1270; P8LE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs3, vs1
1271; P8LE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs4, vs2
1272; P8LE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
1273; P8LE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
1274; P8LE: vmrgew v2, [[REG4]], [[REG3]]
1275; P8LE: xvcvspsxws v2, v2
1276}
1277
1278; Function Attrs: norecurse nounwind readnone
1279define <4 x i32> @fromDiffConstsConvftoi() {
1280entry:
1281 ret <4 x i32> <i32 24, i32 234, i32 988, i32 422>
1282; P9BE-LABEL: fromDiffConstsConvftoi
1283; P9LE-LABEL: fromDiffConstsConvftoi
1284; P8BE-LABEL: fromDiffConstsConvftoi
1285; P8LE-LABEL: fromDiffConstsConvftoi
1286; P9BE: lxvx
1287; P9BE: blr
1288; P9LE: lxvx
1289; P9LE: blr
1290; P8BE: lxvw4x
1291; P8BE: blr
1292; P8LE: lxvd2x
1293; P8LE: xxswapd
1294; P8LE: blr
1295}
1296
1297; Function Attrs: norecurse nounwind readonly
1298define <4 x i32> @fromDiffMemConsAConvftoi(float* nocapture readonly %ptr) {
1299entry:
1300 %0 = bitcast float* %ptr to <4 x float>*
1301 %1 = load <4 x float>, <4 x float>* %0, align 4
1302 %2 = fptosi <4 x float> %1 to <4 x i32>
1303 ret <4 x i32> %2
1304; P9BE-LABEL: fromDiffMemConsAConvftoi
1305; P9LE-LABEL: fromDiffMemConsAConvftoi
1306; P8BE-LABEL: fromDiffMemConsAConvftoi
1307; P8LE-LABEL: fromDiffMemConsAConvftoi
1308; P9BE: lxvx [[REG1:[vs0-9]+]], 0, r3
1309; P9BE: xvcvspsxws v2, [[REG1]]
1310; P9BE: blr
1311; P9LE: lxvx [[REG1:[vs0-9]+]], 0, r3
1312; P9LE: xvcvspsxws v2, [[REG1]]
1313; P9LE: blr
1314; P8BE: lxvw4x [[REG1:[vs0-9]+]], 0, r3
1315; P8BE: xvcvspsxws v2, [[REG1]]
1316; P8BE: blr
1317; P8LE: lxvd2x [[REG1:[vs0-9]+]], 0, r3
1318; P8LE: xxswapd v2, [[REG1]]
1319; P8LE: xvcvspsxws v2, v2
1320; P8LE: blr
1321}
1322
1323; Function Attrs: norecurse nounwind readonly
1324define <4 x i32> @fromDiffMemConsDConvftoi(float* nocapture readonly %ptr) {
1325entry:
1326 %arrayidx = getelementptr inbounds float, float* %ptr, i64 3
1327 %0 = load float, float* %arrayidx, align 4
1328 %conv = fptosi float %0 to i32
1329 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
1330 %arrayidx1 = getelementptr inbounds float, float* %ptr, i64 2
1331 %1 = load float, float* %arrayidx1, align 4
1332 %conv2 = fptosi float %1 to i32
1333 %vecinit3 = insertelement <4 x i32> %vecinit, i32 %conv2, i32 1
1334 %arrayidx4 = getelementptr inbounds float, float* %ptr, i64 1
1335 %2 = load float, float* %arrayidx4, align 4
1336 %conv5 = fptosi float %2 to i32
1337 %vecinit6 = insertelement <4 x i32> %vecinit3, i32 %conv5, i32 2
1338 %3 = load float, float* %ptr, align 4
1339 %conv8 = fptosi float %3 to i32
1340 %vecinit9 = insertelement <4 x i32> %vecinit6, i32 %conv8, i32 3
1341 ret <4 x i32> %vecinit9
1342; P9BE-LABEL: fromDiffMemConsDConvftoi
1343; P9LE-LABEL: fromDiffMemConsDConvftoi
1344; P8BE-LABEL: fromDiffMemConsDConvftoi
1345; P8LE-LABEL: fromDiffMemConsDConvftoi
1346; P9BE: lxvx
1347; P9BE: lxvx
1348; P9BE: vperm
1349; P9BE: xvcvspsxws
1350; P9BE: blr
1351; P9LE: lxvx
1352; P9LE: lxvx
1353; P9LE: vperm
1354; P9LE: xvcvspsxws
1355; P9LE: blr
1356; P8BE: lxvw4x
1357; P8BE: lxvw4x
1358; P8BE: vperm
1359; P8BE: xvcvspsxws
1360; P8BE: blr
1361; P8LE: lxvd2x
1362; P8LE-DAG: lxvd2x
1363; P8LE-DAG: xxswapd
1364; P8LE: xxswapd
1365; P8LE: vperm
1366; P8LE: xvcvspsxws
1367; P8LE: blr
1368}
1369
1370; Function Attrs: norecurse nounwind readonly
1371define <4 x i32> @fromDiffMemVarAConvftoi(float* nocapture readonly %arr, i32 signext %elem) {
1372entry:
1373 %idxprom = sext i32 %elem to i64
1374 %arrayidx = getelementptr inbounds float, float* %arr, i64 %idxprom
1375 %0 = load float, float* %arrayidx, align 4
1376 %conv = fptosi float %0 to i32
1377 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
1378 %add = add nsw i32 %elem, 1
1379 %idxprom1 = sext i32 %add to i64
1380 %arrayidx2 = getelementptr inbounds float, float* %arr, i64 %idxprom1
1381 %1 = load float, float* %arrayidx2, align 4
1382 %conv3 = fptosi float %1 to i32
1383 %vecinit4 = insertelement <4 x i32> %vecinit, i32 %conv3, i32 1
1384 %add5 = add nsw i32 %elem, 2
1385 %idxprom6 = sext i32 %add5 to i64
1386 %arrayidx7 = getelementptr inbounds float, float* %arr, i64 %idxprom6
1387 %2 = load float, float* %arrayidx7, align 4
1388 %conv8 = fptosi float %2 to i32
1389 %vecinit9 = insertelement <4 x i32> %vecinit4, i32 %conv8, i32 2
1390 %add10 = add nsw i32 %elem, 3
1391 %idxprom11 = sext i32 %add10 to i64
1392 %arrayidx12 = getelementptr inbounds float, float* %arr, i64 %idxprom11
1393 %3 = load float, float* %arrayidx12, align 4
1394 %conv13 = fptosi float %3 to i32
1395 %vecinit14 = insertelement <4 x i32> %vecinit9, i32 %conv13, i32 3
1396 ret <4 x i32> %vecinit14
1397; P9BE-LABEL: fromDiffMemVarAConvftoi
1398; P9LE-LABEL: fromDiffMemVarAConvftoi
1399; P8BE-LABEL: fromDiffMemVarAConvftoi
1400; P8LE-LABEL: fromDiffMemVarAConvftoi
1401; FIXME: implement finding consecutive loads with pre-inc
1402; P9BE: lfsux
1403; P9LE: lfsux
1404; P8BE: lfsux
1405; P8LE: lfsux
1406}
1407
1408; Function Attrs: norecurse nounwind readonly
1409define <4 x i32> @fromDiffMemVarDConvftoi(float* nocapture readonly %arr, i32 signext %elem) {
1410entry:
1411 %idxprom = sext i32 %elem to i64
1412 %arrayidx = getelementptr inbounds float, float* %arr, i64 %idxprom
1413 %0 = load float, float* %arrayidx, align 4
1414 %conv = fptosi float %0 to i32
1415 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
1416 %sub = add nsw i32 %elem, -1
1417 %idxprom1 = sext i32 %sub to i64
1418 %arrayidx2 = getelementptr inbounds float, float* %arr, i64 %idxprom1
1419 %1 = load float, float* %arrayidx2, align 4
1420 %conv3 = fptosi float %1 to i32
1421 %vecinit4 = insertelement <4 x i32> %vecinit, i32 %conv3, i32 1
1422 %sub5 = add nsw i32 %elem, -2
1423 %idxprom6 = sext i32 %sub5 to i64
1424 %arrayidx7 = getelementptr inbounds float, float* %arr, i64 %idxprom6
1425 %2 = load float, float* %arrayidx7, align 4
1426 %conv8 = fptosi float %2 to i32
1427 %vecinit9 = insertelement <4 x i32> %vecinit4, i32 %conv8, i32 2
1428 %sub10 = add nsw i32 %elem, -3
1429 %idxprom11 = sext i32 %sub10 to i64
1430 %arrayidx12 = getelementptr inbounds float, float* %arr, i64 %idxprom11
1431 %3 = load float, float* %arrayidx12, align 4
1432 %conv13 = fptosi float %3 to i32
1433 %vecinit14 = insertelement <4 x i32> %vecinit9, i32 %conv13, i32 3
1434 ret <4 x i32> %vecinit14
1435; P9BE-LABEL: fromDiffMemVarDConvftoi
1436; P9LE-LABEL: fromDiffMemVarDConvftoi
1437; P8BE-LABEL: fromDiffMemVarDConvftoi
1438; P8LE-LABEL: fromDiffMemVarDConvftoi
1439; FIXME: implement finding consecutive loads with pre-inc
1440; P9BE: lfsux
1441; P9LE: lfsux
1442; P8BE: lfsux
1443; P8LE: lfsux
1444}
1445
1446; Function Attrs: norecurse nounwind readnone
1447define <4 x i32> @spltRegValConvftoi(float %val) {
1448entry:
1449 %conv = fptosi float %val to i32
1450 %splat.splatinsert = insertelement <4 x i32> undef, i32 %conv, i32 0
1451 %splat.splat = shufflevector <4 x i32> %splat.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
1452 ret <4 x i32> %splat.splat
1453; P9BE-LABEL: spltRegValConvftoi
1454; P9LE-LABEL: spltRegValConvftoi
1455; P8BE-LABEL: spltRegValConvftoi
1456; P8LE-LABEL: spltRegValConvftoi
1457; P9BE: xscvdpsxws f[[REG1:[0-9]+]], f1
1458; P9BE: xxspltw v2, vs[[REG1]], 1
1459; P9BE: blr
1460; P9LE: xscvdpsxws f[[REG1:[0-9]+]], f1
1461; P9LE: xxspltw v2, vs[[REG1]], 1
1462; P9LE: blr
1463; P8BE: xscvdpsxws f[[REG1:[0-9]+]], f1
1464; P8BE: xxspltw v2, vs[[REG1]], 1
1465; P8BE: blr
1466; P8LE: xscvdpsxws f[[REG1:[0-9]+]], f1
1467; P8LE: xxspltw v2, vs[[REG1]], 1
1468; P8LE: blr
1469}
1470
1471; Function Attrs: norecurse nounwind readonly
1472define <4 x i32> @spltMemValConvftoi(float* nocapture readonly %ptr) {
1473entry:
1474 %0 = load float, float* %ptr, align 4
1475 %conv = fptosi float %0 to i32
1476 %splat.splatinsert = insertelement <4 x i32> undef, i32 %conv, i32 0
1477 %splat.splat = shufflevector <4 x i32> %splat.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
1478 ret <4 x i32> %splat.splat
1479; P9BE-LABEL: spltMemValConvftoi
1480; P9LE-LABEL: spltMemValConvftoi
1481; P8BE-LABEL: spltMemValConvftoi
1482; P8LE-LABEL: spltMemValConvftoi
1483; P9BE: lxvwsx [[REG1:[vs0-9]+]], 0, r3
1484; P9BE: xvcvspsxws v2, [[REG1]]
1485; P9LE: [[REG1:[vs0-9]+]], 0, r3
1486; P9LE: xvcvspsxws v2, [[REG1]]
1487; P8BE: lxsspx [[REG1:f[0-9]+]], 0, r3
1488; P8BE: xscvdpsxws f[[REG2:[0-9]+]], [[REG1]]
1489; P8BE: xxspltw v2, vs[[REG2]], 1
1490; P8LE: lxsspx [[REG1:f[0-9]+]], 0, r3
1491; P8LE: xscvdpsxws f[[REG2:[vs0-9]+]], [[REG1]]
1492; P8LE: xxspltw v2, vs[[REG2]], 1
1493}
1494
1495; Function Attrs: norecurse nounwind readnone
1496define <4 x i32> @spltCnstConvdtoi() {
1497entry:
1498 ret <4 x i32> <i32 4, i32 4, i32 4, i32 4>
1499; P9BE-LABEL: spltCnstConvdtoi
1500; P9LE-LABEL: spltCnstConvdtoi
1501; P8BE-LABEL: spltCnstConvdtoi
1502; P8LE-LABEL: spltCnstConvdtoi
1503; P9BE: vspltisw v2, 4
1504; P9BE: blr
1505; P9LE: vspltisw v2, 4
1506; P9LE: blr
1507; P8BE: vspltisw v2, 4
1508; P8BE: blr
1509; P8LE: vspltisw v2, 4
1510; P8LE: blr
1511}
1512
1513; Function Attrs: norecurse nounwind readnone
1514define <4 x i32> @fromRegsConvdtoi(double %a, double %b, double %c, double %d) {
1515entry:
1516 %conv = fptosi double %a to i32
1517 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
1518 %conv1 = fptosi double %b to i32
1519 %vecinit2 = insertelement <4 x i32> %vecinit, i32 %conv1, i32 1
1520 %conv3 = fptosi double %c to i32
1521 %vecinit4 = insertelement <4 x i32> %vecinit2, i32 %conv3, i32 2
1522 %conv5 = fptosi double %d to i32
1523 %vecinit6 = insertelement <4 x i32> %vecinit4, i32 %conv5, i32 3
1524 ret <4 x i32> %vecinit6
1525; P9BE-LABEL: fromRegsConvdtoi
1526; P9LE-LABEL: fromRegsConvdtoi
1527; P8BE-LABEL: fromRegsConvdtoi
1528; P8LE-LABEL: fromRegsConvdtoi
1529; P9BE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs1, vs3
1530; P9BE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs2, vs4
1531; P9BE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
1532; P9BE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
1533; P9BE: vmrgew v2, [[REG3]], [[REG4]]
1534; P9BE: xvcvspsxws v2, v2
1535; P9LE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs3, vs1
1536; P9LE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs4, vs2
1537; P9LE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
1538; P9LE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
1539; P9LE: vmrgew v2, [[REG4]], [[REG3]]
1540; P9LE: xvcvspsxws v2, v2
1541; P8BE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs1, vs3
1542; P8BE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs2, vs4
1543; P8BE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
1544; P8BE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
1545; P8BE: vmrgew v2, [[REG3]], [[REG4]]
1546; P8BE: xvcvspsxws v2, v2
1547; P8LE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs3, vs1
1548; P8LE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs4, vs2
1549; P8LE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
1550; P8LE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
1551; P8LE: vmrgew v2, [[REG4]], [[REG3]]
1552; P8LE: xvcvspsxws v2, v2
1553}
1554
1555; Function Attrs: norecurse nounwind readnone
1556define <4 x i32> @fromDiffConstsConvdtoi() {
1557entry:
1558 ret <4 x i32> <i32 24, i32 234, i32 988, i32 422>
1559; P9BE-LABEL: fromDiffConstsConvdtoi
1560; P9LE-LABEL: fromDiffConstsConvdtoi
1561; P8BE-LABEL: fromDiffConstsConvdtoi
1562; P8LE-LABEL: fromDiffConstsConvdtoi
1563; P9BE: lxvx
1564; P9BE: blr
1565; P9LE: lxvx
1566; P9LE: blr
1567; P8BE: lxvw4x
1568; P8BE: blr
1569; P8LE: lxvd2x
1570; P8LE: xxswapd
1571; P8LE: blr
1572}
1573
1574; Function Attrs: norecurse nounwind readonly
1575define <4 x i32> @fromDiffMemConsAConvdtoi(double* nocapture readonly %ptr) {
1576entry:
1577 %0 = bitcast double* %ptr to <2 x double>*
1578 %1 = load <2 x double>, <2 x double>* %0, align 8
1579 %2 = fptosi <2 x double> %1 to <2 x i32>
1580 %arrayidx4 = getelementptr inbounds double, double* %ptr, i64 2
1581 %3 = bitcast double* %arrayidx4 to <2 x double>*
1582 %4 = load <2 x double>, <2 x double>* %3, align 8
1583 %5 = fptosi <2 x double> %4 to <2 x i32>
1584 %vecinit9 = shufflevector <2 x i32> %2, <2 x i32> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
1585 ret <4 x i32> %vecinit9
1586; P9BE-LABEL: fromDiffMemConsAConvdtoi
1587; P9LE-LABEL: fromDiffMemConsAConvdtoi
1588; P8BE-LABEL: fromDiffMemConsAConvdtoi
1589; P8LE-LABEL: fromDiffMemConsAConvdtoi
1590; P9BE: lxvx [[REG1:[vs0-9]+]], 0, r3
1591; P9BE: lxvx [[REG2:[vs0-9]+]], r3, r4
1592; P9BE-DAG: xxmrgld [[REG3:[vs0-9]+]], [[REG1]], [[REG2]]
1593; P9BE-DAG: xxmrghd [[REG4:[vs0-9]+]], [[REG1]], [[REG2]]
1594; P9BE-DAG: xvcvdpsp [[REG5:[vs0-9]+]], [[REG3]]
1595; P9BE-DAG: xvcvdpsp [[REG6:[vs0-9]+]], [[REG4]]
1596; P9BE: vmrgew v2, [[REG6]], [[REG5]]
1597; P9BE: xvcvspsxws v2, v2
1598; P9LE: lxvx [[REG1:[vs0-9]+]], 0, r3
1599; P9LE: lxvx [[REG2:[vs0-9]+]], r3, r4
1600; P9LE-DAG: xxmrgld [[REG3:[vs0-9]+]], [[REG2]], [[REG1]]
1601; P9LE-DAG: xxmrghd [[REG4:[vs0-9]+]], [[REG2]], [[REG1]]
1602; P9LE-DAG: xvcvdpsp [[REG5:[vs0-9]+]], [[REG3]]
1603; P9LE-DAG: xvcvdpsp [[REG6:[vs0-9]+]], [[REG4]]
1604; P9LE: vmrgew v2, [[REG6]], [[REG5]]
1605; P9LE: xvcvspsxws v2, v2
1606; P8BE: lxvd2x [[REG1:[vs0-9]+]], 0, r3
1607; P8BE: lxvd2x [[REG2:[vs0-9]+]], r3, r4
1608; P8BE-DAG: xxmrgld [[REG3:[vs0-9]+]], [[REG1]], [[REG2]]
1609; P8BE-DAG: xxmrghd [[REG4:[vs0-9]+]], [[REG1]], [[REG2]]
1610; P8BE-DAG: xvcvdpsp [[REG5:[vs0-9]+]], [[REG3]]
1611; P8BE-DAG: xvcvdpsp [[REG6:[vs0-9]+]], [[REG4]]
1612; P8BE: vmrgew v2, [[REG6]], [[REG5]]
1613; P8BE: xvcvspsxws v2, v2
1614; P8LE: lxvd2x [[REG1:[vs0-9]+]], 0, r3
1615; P8LE: lxvd2x [[REG2:[vs0-9]+]], r3, r4
1616; P8LE-DAG: xxswapd [[REG3:[vs0-9]+]], [[REG1]]
1617; P8LE-DAG: xxswapd [[REG4:[vs0-9]+]], [[REG2]]
1618; P8LE-DAG: xxmrgld [[REG5:[vs0-9]+]], [[REG4]], [[REG3]]
1619; P8LE-DAG: xxmrghd [[REG6:[vs0-9]+]], [[REG4]], [[REG3]]
1620; P8LE-DAG: xvcvdpsp [[REG7:[vs0-9]+]], [[REG5]]
1621; P8LE-DAG: xvcvdpsp [[REG8:[vs0-9]+]], [[REG6]]
1622; P8LE: vmrgew v2, [[REG8]], [[REG7]]
1623; P8LE: xvcvspsxws v2, v2
1624}
1625
1626; Function Attrs: norecurse nounwind readonly
1627define <4 x i32> @fromDiffMemConsDConvdtoi(double* nocapture readonly %ptr) {
1628entry:
1629 %arrayidx = getelementptr inbounds double, double* %ptr, i64 3
1630 %0 = load double, double* %arrayidx, align 8
1631 %conv = fptosi double %0 to i32
1632 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
1633 %arrayidx1 = getelementptr inbounds double, double* %ptr, i64 2
1634 %1 = load double, double* %arrayidx1, align 8
1635 %conv2 = fptosi double %1 to i32
1636 %vecinit3 = insertelement <4 x i32> %vecinit, i32 %conv2, i32 1
1637 %arrayidx4 = getelementptr inbounds double, double* %ptr, i64 1
1638 %2 = load double, double* %arrayidx4, align 8
1639 %conv5 = fptosi double %2 to i32
1640 %vecinit6 = insertelement <4 x i32> %vecinit3, i32 %conv5, i32 2
1641 %3 = load double, double* %ptr, align 8
1642 %conv8 = fptosi double %3 to i32
1643 %vecinit9 = insertelement <4 x i32> %vecinit6, i32 %conv8, i32 3
1644 ret <4 x i32> %vecinit9
1645; P9BE-LABEL: fromDiffMemConsDConvdtoi
1646; P9LE-LABEL: fromDiffMemConsDConvdtoi
1647; P8BE-LABEL: fromDiffMemConsDConvdtoi
1648; P8LE-LABEL: fromDiffMemConsDConvdtoi
1649; P9BE: lfd
1650; P9BE: lfd
1651; P9BE: lfd
1652; P9BE: lfd
1653; P9BE: xxmrghd
1654; P9BE: xxmrghd
1655; P9BE: xvcvdpsp
1656; P9BE: xvcvdpsp
1657; P9BE: vmrgew
1658; P9BE: xvcvspsxws v2
1659; P9LE: lfd
1660; P9LE: lfd
1661; P9LE: lfd
1662; P9LE: lfd
1663; P9LE: xxmrghd
1664; P9LE: xxmrghd
1665; P9LE: xvcvdpsp
1666; P9LE: xvcvdpsp
1667; P9LE: vmrgew
1668; P9LE: xvcvspsxws v2
1669; P8BE: lxsdx
1670; P8BE: lxsdx
1671; P8BE: lxsdx
1672; P8BE: lxsdx
1673; P8BE: xxmrghd
1674; P8BE: xxmrghd
1675; P8BE: xvcvdpsp
1676; P8BE: xvcvdpsp
1677; P8BE: vmrgew
1678; P8BE: xvcvspsxws v2
1679; P8LE: lxsdx
1680; P8LE: lxsdx
1681; P8LE: lxsdx
1682; P8LE: lxsdx
1683; P8LE: xxmrghd
1684; P8LE: xxmrghd
1685; P8LE: xvcvdpsp
1686; P8LE: xvcvdpsp
1687; P8LE: vmrgew
1688; P8LE: xvcvspsxws v2
1689}
1690
1691; Function Attrs: norecurse nounwind readonly
1692define <4 x i32> @fromDiffMemVarAConvdtoi(double* nocapture readonly %arr, i32 signext %elem) {
1693entry:
1694 %idxprom = sext i32 %elem to i64
1695 %arrayidx = getelementptr inbounds double, double* %arr, i64 %idxprom
1696 %0 = load double, double* %arrayidx, align 8
1697 %conv = fptosi double %0 to i32
1698 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
1699 %add = add nsw i32 %elem, 1
1700 %idxprom1 = sext i32 %add to i64
1701 %arrayidx2 = getelementptr inbounds double, double* %arr, i64 %idxprom1
1702 %1 = load double, double* %arrayidx2, align 8
1703 %conv3 = fptosi double %1 to i32
1704 %vecinit4 = insertelement <4 x i32> %vecinit, i32 %conv3, i32 1
1705 %add5 = add nsw i32 %elem, 2
1706 %idxprom6 = sext i32 %add5 to i64
1707 %arrayidx7 = getelementptr inbounds double, double* %arr, i64 %idxprom6
1708 %2 = load double, double* %arrayidx7, align 8
1709 %conv8 = fptosi double %2 to i32
1710 %vecinit9 = insertelement <4 x i32> %vecinit4, i32 %conv8, i32 2
1711 %add10 = add nsw i32 %elem, 3
1712 %idxprom11 = sext i32 %add10 to i64
1713 %arrayidx12 = getelementptr inbounds double, double* %arr, i64 %idxprom11
1714 %3 = load double, double* %arrayidx12, align 8
1715 %conv13 = fptosi double %3 to i32
1716 %vecinit14 = insertelement <4 x i32> %vecinit9, i32 %conv13, i32 3
1717 ret <4 x i32> %vecinit14
1718; P9BE-LABEL: fromDiffMemVarAConvdtoi
1719; P9LE-LABEL: fromDiffMemVarAConvdtoi
1720; P8BE-LABEL: fromDiffMemVarAConvdtoi
1721; P8LE-LABEL: fromDiffMemVarAConvdtoi
1722; P9BE: lfdux
1723; P9BE: lfd
1724; P9BE: lfd
1725; P9BE: lfd
1726; P9BE: xxmrghd
1727; P9BE: xxmrghd
1728; P9BE: xvcvdpsp
1729; P9BE: xvcvdpsp
1730; P9BE: vmrgew
1731; P9BE: xvcvspsxws v2
1732; P9LE: lfdux
1733; P9LE: lfd
1734; P9LE: lfd
1735; P9LE: lfd
1736; P9LE: xxmrghd
1737; P9LE: xxmrghd
1738; P9LE: xvcvdpsp
1739; P9LE: xvcvdpsp
1740; P9LE: vmrgew
1741; P9LE: xvcvspsxws v2
1742; P8BE: lfdux
1743; P8BE: lxsdx
1744; P8BE: lxsdx
1745; P8BE: lxsdx
1746; P8BE: xxmrghd
1747; P8BE: xxmrghd
1748; P8BE: xvcvdpsp
1749; P8BE: xvcvdpsp
1750; P8BE: vmrgew
1751; P8BE: xvcvspsxws v2
1752; P8LE: lfdux
1753; P8LE: lxsdx
1754; P8LE: lxsdx
1755; P8LE: lxsdx
1756; P8LE: xxmrghd
1757; P8LE: xxmrghd
1758; P8LE: xvcvdpsp
1759; P8LE: xvcvdpsp
1760; P8LE: vmrgew
1761; P8LE: xvcvspsxws v2
1762}
1763
1764; Function Attrs: norecurse nounwind readonly
1765define <4 x i32> @fromDiffMemVarDConvdtoi(double* nocapture readonly %arr, i32 signext %elem) {
1766entry:
1767 %idxprom = sext i32 %elem to i64
1768 %arrayidx = getelementptr inbounds double, double* %arr, i64 %idxprom
1769 %0 = load double, double* %arrayidx, align 8
1770 %conv = fptosi double %0 to i32
1771 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
1772 %sub = add nsw i32 %elem, -1
1773 %idxprom1 = sext i32 %sub to i64
1774 %arrayidx2 = getelementptr inbounds double, double* %arr, i64 %idxprom1
1775 %1 = load double, double* %arrayidx2, align 8
1776 %conv3 = fptosi double %1 to i32
1777 %vecinit4 = insertelement <4 x i32> %vecinit, i32 %conv3, i32 1
1778 %sub5 = add nsw i32 %elem, -2
1779 %idxprom6 = sext i32 %sub5 to i64
1780 %arrayidx7 = getelementptr inbounds double, double* %arr, i64 %idxprom6
1781 %2 = load double, double* %arrayidx7, align 8
1782 %conv8 = fptosi double %2 to i32
1783 %vecinit9 = insertelement <4 x i32> %vecinit4, i32 %conv8, i32 2
1784 %sub10 = add nsw i32 %elem, -3
1785 %idxprom11 = sext i32 %sub10 to i64
1786 %arrayidx12 = getelementptr inbounds double, double* %arr, i64 %idxprom11
1787 %3 = load double, double* %arrayidx12, align 8
1788 %conv13 = fptosi double %3 to i32
1789 %vecinit14 = insertelement <4 x i32> %vecinit9, i32 %conv13, i32 3
1790 ret <4 x i32> %vecinit14
1791; P9BE-LABEL: fromDiffMemVarDConvdtoi
1792; P9LE-LABEL: fromDiffMemVarDConvdtoi
1793; P8BE-LABEL: fromDiffMemVarDConvdtoi
1794; P8LE-LABEL: fromDiffMemVarDConvdtoi
1795; P9BE: lfdux
1796; P9BE: lfd
1797; P9BE: lfd
1798; P9BE: lfd
1799; P9BE: xxmrghd
1800; P9BE: xxmrghd
1801; P9BE: xvcvdpsp
1802; P9BE: xvcvdpsp
1803; P9BE: vmrgew
1804; P9BE: xvcvspsxws v2
1805; P9LE: lfdux
1806; P9LE: lfd
1807; P9LE: lfd
1808; P9LE: lfd
1809; P9LE: xxmrghd
1810; P9LE: xxmrghd
1811; P9LE: xvcvdpsp
1812; P9LE: xvcvdpsp
1813; P9LE: vmrgew
1814; P9LE: xvcvspsxws v2
1815; P8BE: lfdux
1816; P8BE: lxsdx
1817; P8BE: lxsdx
1818; P8BE: lxsdx
1819; P8BE: xxmrghd
1820; P8BE: xxmrghd
1821; P8BE: xvcvdpsp
1822; P8BE: xvcvdpsp
1823; P8BE: vmrgew
1824; P8BE: xvcvspsxws v2
1825; P8LE: lfdux
1826; P8LE: lxsdx
1827; P8LE: lxsdx
1828; P8LE: lxsdx
1829; P8LE: xxmrghd
1830; P8LE: xxmrghd
1831; P8LE: xvcvdpsp
1832; P8LE: xvcvdpsp
1833; P8LE: vmrgew
1834; P8LE: xvcvspsxws v2
1835}
1836
1837; Function Attrs: norecurse nounwind readnone
1838define <4 x i32> @spltRegValConvdtoi(double %val) {
1839entry:
1840 %conv = fptosi double %val to i32
1841 %splat.splatinsert = insertelement <4 x i32> undef, i32 %conv, i32 0
1842 %splat.splat = shufflevector <4 x i32> %splat.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
1843 ret <4 x i32> %splat.splat
1844; P9BE-LABEL: spltRegValConvdtoi
1845; P9LE-LABEL: spltRegValConvdtoi
1846; P8BE-LABEL: spltRegValConvdtoi
1847; P8LE-LABEL: spltRegValConvdtoi
1848; P9BE: xscvdpsxws
1849; P9BE: xxspltw
1850; P9BE: blr
1851; P9LE: xscvdpsxws
1852; P9LE: xxspltw
1853; P9LE: blr
1854; P8BE: xscvdpsxws
1855; P8BE: xxspltw
1856; P8BE: blr
1857; P8LE: xscvdpsxws
1858; P8LE: xxspltw
1859; P8LE: blr
1860}
1861
1862; Function Attrs: norecurse nounwind readonly
1863define <4 x i32> @spltMemValConvdtoi(double* nocapture readonly %ptr) {
1864entry:
1865 %0 = load double, double* %ptr, align 8
1866 %conv = fptosi double %0 to i32
1867 %splat.splatinsert = insertelement <4 x i32> undef, i32 %conv, i32 0
1868 %splat.splat = shufflevector <4 x i32> %splat.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
1869 ret <4 x i32> %splat.splat
1870; P9BE-LABEL: spltMemValConvdtoi
1871; P9LE-LABEL: spltMemValConvdtoi
1872; P8BE-LABEL: spltMemValConvdtoi
1873; P8LE-LABEL: spltMemValConvdtoi
1874; P9BE: lfd
1875; P9BE: xscvdpsxws
1876; P9BE: xxspltw
1877; P9BE: blr
1878; P9LE: lfd
1879; P9LE: xscvdpsxws
1880; P9LE: xxspltw
1881; P9LE: blr
1882; P8BE: lxsdx
1883; P8BE: xscvdpsxws
1884; P8BE: xxspltw
1885; P8BE: blr
1886; P8LE: lxsdx
1887; P8LE: xscvdpsxws
1888; P8LE: xxspltw
1889; P8LE: blr
1890}
1891; Function Attrs: norecurse nounwind readnone
1892define <4 x i32> @allZeroui() {
1893entry:
1894 ret <4 x i32> zeroinitializer
1895; P9BE-LABEL: allZeroui
1896; P9LE-LABEL: allZeroui
1897; P8BE-LABEL: allZeroui
1898; P8LE-LABEL: allZeroui
1899; P9BE: xxlxor v2, v2, v2
1900; P9BE: blr
1901; P9LE: xxlxor v2, v2, v2
1902; P9LE: blr
1903; P8BE: xxlxor v2, v2, v2
1904; P8BE: blr
1905; P8LE: xxlxor v2, v2, v2
1906; P8LE: blr
1907}
1908
1909; Function Attrs: norecurse nounwind readnone
1910define <4 x i32> @allOneui() {
1911entry:
1912 ret <4 x i32> <i32 -1, i32 -1, i32 -1, i32 -1>
1913; P9BE-LABEL: allOneui
1914; P9LE-LABEL: allOneui
1915; P8BE-LABEL: allOneui
1916; P8LE-LABEL: allOneui
1917; P9BE: xxspltib v2, 255
1918; P9BE: blr
1919; P9LE: xxspltib v2, 255
1920; P9LE: blr
1921; P8BE: vspltisb v2, -1
1922; P8BE: blr
1923; P8LE: vspltisb v2, -1
1924; P8LE: blr
1925}
1926
1927; Function Attrs: norecurse nounwind readnone
1928define <4 x i32> @spltConst1ui() {
1929entry:
1930 ret <4 x i32> <i32 1, i32 1, i32 1, i32 1>
1931; P9BE-LABEL: spltConst1ui
1932; P9LE-LABEL: spltConst1ui
1933; P8BE-LABEL: spltConst1ui
1934; P8LE-LABEL: spltConst1ui
1935; P9BE: vspltisw v2, 1
1936; P9BE: blr
1937; P9LE: vspltisw v2, 1
1938; P9LE: blr
1939; P8BE: vspltisw v2, 1
1940; P8BE: blr
1941; P8LE: vspltisw v2, 1
1942; P8LE: blr
1943}
1944
1945; Function Attrs: norecurse nounwind readnone
1946define <4 x i32> @spltConst16kui() {
1947entry:
1948 ret <4 x i32> <i32 32767, i32 32767, i32 32767, i32 32767>
1949; P9BE-LABEL: spltConst16kui
1950; P9LE-LABEL: spltConst16kui
1951; P8BE-LABEL: spltConst16kui
1952; P8LE-LABEL: spltConst16kui
1953; P9BE: vspltisw v2, -15
1954; P9BE: vsrw v2, v2, v2
1955; P9BE: blr
1956; P9LE: vspltisw v2, -15
1957; P9LE: vsrw v2, v2, v2
1958; P9LE: blr
1959; P8BE: vspltisw v2, -15
1960; P8BE: vsrw v2, v2, v2
1961; P8BE: blr
1962; P8LE: vspltisw v2, -15
1963; P8LE: vsrw v2, v2, v2
1964; P8LE: blr
1965}
1966
1967; Function Attrs: norecurse nounwind readnone
1968define <4 x i32> @spltConst32kui() {
1969entry:
1970 ret <4 x i32> <i32 65535, i32 65535, i32 65535, i32 65535>
1971; P9BE-LABEL: spltConst32kui
1972; P9LE-LABEL: spltConst32kui
1973; P8BE-LABEL: spltConst32kui
1974; P8LE-LABEL: spltConst32kui
1975; P9BE: vspltisw v2, -16
1976; P9BE: vsrw v2, v2, v2
1977; P9BE: blr
1978; P9LE: vspltisw v2, -16
1979; P9LE: vsrw v2, v2, v2
1980; P9LE: blr
1981; P8BE: vspltisw v2, -16
1982; P8BE: vsrw v2, v2, v2
1983; P8BE: blr
1984; P8LE: vspltisw v2, -16
1985; P8LE: vsrw v2, v2, v2
1986; P8LE: blr
1987}
1988
1989; Function Attrs: norecurse nounwind readnone
1990define <4 x i32> @fromRegsui(i32 zeroext %a, i32 zeroext %b, i32 zeroext %c, i32 zeroext %d) {
1991entry:
1992 %vecinit = insertelement <4 x i32> undef, i32 %a, i32 0
1993 %vecinit1 = insertelement <4 x i32> %vecinit, i32 %b, i32 1
1994 %vecinit2 = insertelement <4 x i32> %vecinit1, i32 %c, i32 2
1995 %vecinit3 = insertelement <4 x i32> %vecinit2, i32 %d, i32 3
1996 ret <4 x i32> %vecinit3
1997; P9BE-LABEL: fromRegsui
1998; P9LE-LABEL: fromRegsui
1999; P8BE-LABEL: fromRegsui
2000; P8LE-LABEL: fromRegsui
2001; P9BE-DAG: mtvsrdd [[REG1:v[0-9]+]], r3, r5
2002; P9BE-DAG: mtvsrdd [[REG2:v[0-9]+]], r4, r6
2003; P9BE: vmrgow v2, [[REG1]], [[REG2]]
2004; P9BE: blr
2005; P9LE-DAG: mtvsrdd [[REG1:v[0-9]+]], r5, r3
2006; P9LE-DAG: mtvsrdd [[REG2:v[0-9]+]], r6, r4
2007; P9LE: vmrgow v2, [[REG2]], [[REG1]]
2008; P9LE: blr
2009; P8BE-DAG: mtvsrwz {{[vf]}}[[REG1:[0-9]+]], r3
2010; P8BE-DAG: mtvsrwz {{[vf]}}[[REG2:[0-9]+]], r4
2011; P8BE-DAG: mtvsrwz {{[vf]}}[[REG3:[0-9]+]], r5
2012; P8BE-DAG: mtvsrwz {{[vf]}}[[REG4:[0-9]+]], r6
2013; P8BE-DAG: xxmrghd [[REG5:v[0-9]+]], {{[v][s]*}}[[REG1]], {{[v][s]*}}[[REG3]]
2014; P8BE-DAG: xxmrghd [[REG6:v[0-9]+]], {{[v][s]*}}[[REG2]], {{[v][s]*}}[[REG4]]
2015; P8BE: vmrgow v2, [[REG5]], [[REG6]]
2016; P8LE-DAG: mtvsrwz {{[vf]}}[[REG1:[0-9]+]], r3
2017; P8LE-DAG: mtvsrwz {{[vf]}}[[REG2:[0-9]+]], r4
2018; P8LE-DAG: mtvsrwz {{[vf]}}[[REG3:[0-9]+]], r5
2019; P8LE-DAG: mtvsrwz {{[vf]}}[[REG4:[0-9]+]], r6
2020; P8LE: xxmrghd [[REG5:v[0-9]+]], {{[v][s]*}}[[REG3]], {{[v][s]*}}[[REG1]]
2021; P8LE: xxmrghd [[REG6:v[0-9]+]], {{[v][s]*}}[[REG4]], {{[v][s]*}}[[REG2]]
2022; P8LE: vmrgow v2, [[REG6]], [[REG5]]
2023}
2024
2025; Function Attrs: norecurse nounwind readnone
2026define <4 x i32> @fromDiffConstsui() {
2027entry:
2028 ret <4 x i32> <i32 242, i32 -113, i32 889, i32 19>
2029; P9BE-LABEL: fromDiffConstsui
2030; P9LE-LABEL: fromDiffConstsui
2031; P8BE-LABEL: fromDiffConstsui
2032; P8LE-LABEL: fromDiffConstsui
2033; P9BE: lxvx
2034; P9BE: blr
2035; P9LE: lxvx
2036; P9LE: blr
2037; P8BE: lxvw4x
2038; P8BE: blr
2039; P8LE: lxvd2x
2040; P8LE: xxswapd
2041; P8LE: blr
2042}
2043
2044; Function Attrs: norecurse nounwind readonly
2045define <4 x i32> @fromDiffMemConsAui(i32* nocapture readonly %arr) {
2046entry:
2047 %0 = load i32, i32* %arr, align 4
2048 %vecinit = insertelement <4 x i32> undef, i32 %0, i32 0
2049 %arrayidx1 = getelementptr inbounds i32, i32* %arr, i64 1
2050 %1 = load i32, i32* %arrayidx1, align 4
2051 %vecinit2 = insertelement <4 x i32> %vecinit, i32 %1, i32 1
2052 %arrayidx3 = getelementptr inbounds i32, i32* %arr, i64 2
2053 %2 = load i32, i32* %arrayidx3, align 4
2054 %vecinit4 = insertelement <4 x i32> %vecinit2, i32 %2, i32 2
2055 %arrayidx5 = getelementptr inbounds i32, i32* %arr, i64 3
2056 %3 = load i32, i32* %arrayidx5, align 4
2057 %vecinit6 = insertelement <4 x i32> %vecinit4, i32 %3, i32 3
2058 ret <4 x i32> %vecinit6
2059; P9BE-LABEL: fromDiffMemConsAui
2060; P9LE-LABEL: fromDiffMemConsAui
2061; P8BE-LABEL: fromDiffMemConsAui
2062; P8LE-LABEL: fromDiffMemConsAui
2063; P9BE: lxvx
2064; P9BE: blr
2065; P9LE: lxvx
2066; P9LE: blr
2067; P8BE: lxvw4x
2068; P8BE: blr
2069; P8LE: lxvd2x
2070; P8LE: xxswapd
2071; P8LE: blr
2072}
2073
2074; Function Attrs: norecurse nounwind readonly
2075define <4 x i32> @fromDiffMemConsDui(i32* nocapture readonly %arr) {
2076entry:
2077 %arrayidx = getelementptr inbounds i32, i32* %arr, i64 3
2078 %0 = load i32, i32* %arrayidx, align 4
2079 %vecinit = insertelement <4 x i32> undef, i32 %0, i32 0
2080 %arrayidx1 = getelementptr inbounds i32, i32* %arr, i64 2
2081 %1 = load i32, i32* %arrayidx1, align 4
2082 %vecinit2 = insertelement <4 x i32> %vecinit, i32 %1, i32 1
2083 %arrayidx3 = getelementptr inbounds i32, i32* %arr, i64 1
2084 %2 = load i32, i32* %arrayidx3, align 4
2085 %vecinit4 = insertelement <4 x i32> %vecinit2, i32 %2, i32 2
2086 %3 = load i32, i32* %arr, align 4
2087 %vecinit6 = insertelement <4 x i32> %vecinit4, i32 %3, i32 3
2088 ret <4 x i32> %vecinit6
2089; P9BE-LABEL: fromDiffMemConsDui
2090; P9LE-LABEL: fromDiffMemConsDui
2091; P8BE-LABEL: fromDiffMemConsDui
2092; P8LE-LABEL: fromDiffMemConsDui
2093; P9BE: lxvx
2094; P9BE: lxvx
2095; P9BE: vperm
2096; P9BE: blr
2097; P9LE: lxvx
2098; P9LE: lxvx
2099; P9LE: vperm
2100; P9LE: blr
2101; P8BE: lxvw4x
2102; P8BE: lxvw4x
2103; P8BE: vperm
2104; P8BE: blr
2105; P8LE: lxvd2x
2106; P8LE-DAG: lxvd2x
2107; P8LE-DAG: xxswapd
2108; P8LE: xxswapd
2109; P8LE: vperm
2110; P8LE: blr
2111}
2112
2113; Function Attrs: norecurse nounwind readonly
2114define <4 x i32> @fromDiffMemVarAui(i32* nocapture readonly %arr, i32 signext %elem) {
2115entry:
2116 %idxprom = sext i32 %elem to i64
2117 %arrayidx = getelementptr inbounds i32, i32* %arr, i64 %idxprom
2118 %0 = load i32, i32* %arrayidx, align 4
2119 %vecinit = insertelement <4 x i32> undef, i32 %0, i32 0
2120 %add = add nsw i32 %elem, 1
2121 %idxprom1 = sext i32 %add to i64
2122 %arrayidx2 = getelementptr inbounds i32, i32* %arr, i64 %idxprom1
2123 %1 = load i32, i32* %arrayidx2, align 4
2124 %vecinit3 = insertelement <4 x i32> %vecinit, i32 %1, i32 1
2125 %add4 = add nsw i32 %elem, 2
2126 %idxprom5 = sext i32 %add4 to i64
2127 %arrayidx6 = getelementptr inbounds i32, i32* %arr, i64 %idxprom5
2128 %2 = load i32, i32* %arrayidx6, align 4
2129 %vecinit7 = insertelement <4 x i32> %vecinit3, i32 %2, i32 2
2130 %add8 = add nsw i32 %elem, 3
2131 %idxprom9 = sext i32 %add8 to i64
2132 %arrayidx10 = getelementptr inbounds i32, i32* %arr, i64 %idxprom9
2133 %3 = load i32, i32* %arrayidx10, align 4
2134 %vecinit11 = insertelement <4 x i32> %vecinit7, i32 %3, i32 3
2135 ret <4 x i32> %vecinit11
2136; P9BE-LABEL: fromDiffMemVarAui
2137; P9LE-LABEL: fromDiffMemVarAui
2138; P8BE-LABEL: fromDiffMemVarAui
2139; P8LE-LABEL: fromDiffMemVarAui
2140; P9BE: sldi r4, r4, 2
2141; P9BE: lxvx v2, r3, r4
2142; P9BE: blr
2143; P9LE: sldi r4, r4, 2
2144; P9LE: lxvx v2, r3, r4
2145; P9LE: blr
2146; P8BE: sldi r4, r4, 2
2147; P8BE: lxvw4x {{[vs0-9]+}}, r3, r4
2148; P8BE: blr
2149; P8LE: sldi r4, r4, 2
2150; P8LE: lxvd2x {{[vs0-9]+}}, r3, r4
2151; P8LE: xxswapd
2152; P8LE: blr
2153}
2154
2155; Function Attrs: norecurse nounwind readonly
2156define <4 x i32> @fromDiffMemVarDui(i32* nocapture readonly %arr, i32 signext %elem) {
2157entry:
2158 %idxprom = sext i32 %elem to i64
2159 %arrayidx = getelementptr inbounds i32, i32* %arr, i64 %idxprom
2160 %0 = load i32, i32* %arrayidx, align 4
2161 %vecinit = insertelement <4 x i32> undef, i32 %0, i32 0
2162 %sub = add nsw i32 %elem, -1
2163 %idxprom1 = sext i32 %sub to i64
2164 %arrayidx2 = getelementptr inbounds i32, i32* %arr, i64 %idxprom1
2165 %1 = load i32, i32* %arrayidx2, align 4
2166 %vecinit3 = insertelement <4 x i32> %vecinit, i32 %1, i32 1
2167 %sub4 = add nsw i32 %elem, -2
2168 %idxprom5 = sext i32 %sub4 to i64
2169 %arrayidx6 = getelementptr inbounds i32, i32* %arr, i64 %idxprom5
2170 %2 = load i32, i32* %arrayidx6, align 4
2171 %vecinit7 = insertelement <4 x i32> %vecinit3, i32 %2, i32 2
2172 %sub8 = add nsw i32 %elem, -3
2173 %idxprom9 = sext i32 %sub8 to i64
2174 %arrayidx10 = getelementptr inbounds i32, i32* %arr, i64 %idxprom9
2175 %3 = load i32, i32* %arrayidx10, align 4
2176 %vecinit11 = insertelement <4 x i32> %vecinit7, i32 %3, i32 3
2177 ret <4 x i32> %vecinit11
2178; P9BE-LABEL: fromDiffMemVarDui
2179; P9LE-LABEL: fromDiffMemVarDui
2180; P8BE-LABEL: fromDiffMemVarDui
2181; P8LE-LABEL: fromDiffMemVarDui
2182; P9BE: sldi r4, r4, 2
2183; P9BE-DAG: lxvx {{[vs0-9]+}}, r3, r4
2184; P9BE-DAG: lxvx
2185; P9BE: vperm
2186; P9BE: blr
2187; P9LE: sldi r4, r4, 2
2188; P9LE-DAG: lxvx {{[vs0-9]+}}, r3, r4
2189; P9LE-DAG: lxvx
2190; P9LE: vperm
2191; P9LE: blr
2192; P8BE: sldi r4, r4, 2
2193; P8BE-DAG: lxvw4x {{[vs0-9]+}}, r3, r4
2194; P8BE-DAG: lxvw4x
2195; P8BE: vperm
2196; P8BE: blr
2197; P8LE: sldi r4, r4, 2
2198; P8LE-DAG: lxvd2x
2199; P8LE-DAG: lxvd2x
2200; P8LE-DAG: xxswapd
2201; P8LE: xxswapd
2202; P8LE: vperm
2203; P8LE: blr
2204}
2205
2206; Function Attrs: norecurse nounwind readonly
2207define <4 x i32> @fromRandMemConsui(i32* nocapture readonly %arr) {
2208entry:
2209 %arrayidx = getelementptr inbounds i32, i32* %arr, i64 4
2210 %0 = load i32, i32* %arrayidx, align 4
2211 %vecinit = insertelement <4 x i32> undef, i32 %0, i32 0
2212 %arrayidx1 = getelementptr inbounds i32, i32* %arr, i64 18
2213 %1 = load i32, i32* %arrayidx1, align 4
2214 %vecinit2 = insertelement <4 x i32> %vecinit, i32 %1, i32 1
2215 %arrayidx3 = getelementptr inbounds i32, i32* %arr, i64 2
2216 %2 = load i32, i32* %arrayidx3, align 4
2217 %vecinit4 = insertelement <4 x i32> %vecinit2, i32 %2, i32 2
2218 %arrayidx5 = getelementptr inbounds i32, i32* %arr, i64 88
2219 %3 = load i32, i32* %arrayidx5, align 4
2220 %vecinit6 = insertelement <4 x i32> %vecinit4, i32 %3, i32 3
2221 ret <4 x i32> %vecinit6
2222; P9BE-LABEL: fromRandMemConsui
2223; P9LE-LABEL: fromRandMemConsui
2224; P8BE-LABEL: fromRandMemConsui
2225; P8LE-LABEL: fromRandMemConsui
2226; P9BE: lwz
2227; P9BE: lwz
2228; P9BE: lwz
2229; P9BE: lwz
2230; P9BE: mtvsrdd
2231; P9BE: mtvsrdd
2232; P9BE: vmrgow
2233; P9LE: lwz
2234; P9LE: lwz
2235; P9LE: lwz
2236; P9LE: lwz
2237; P9LE: mtvsrdd
2238; P9LE: mtvsrdd
2239; P9LE: vmrgow
2240; P8BE: lwz
2241; P8BE: lwz
2242; P8BE: lwz
2243; P8BE: lwz
2244; P8BE: mtvsrwz
2245; P8BE: mtvsrwz
2246; P8BE: mtvsrwz
2247; P8BE: mtvsrwz
2248; P8BE: xxmrghd
2249; P8BE: xxmrghd
2250; P8BE: vmrgow
2251; P8LE: lwz
2252; P8LE: lwz
2253; P8LE: lwz
2254; P8LE: lwz
2255; P8LE: mtvsrwz
2256; P8LE: mtvsrwz
2257; P8LE: mtvsrwz
2258; P8LE: mtvsrwz
2259; P8LE: xxmrghd
2260; P8LE: xxmrghd
2261; P8LE: vmrgow
2262}
2263
2264; Function Attrs: norecurse nounwind readonly
2265define <4 x i32> @fromRandMemVarui(i32* nocapture readonly %arr, i32 signext %elem) {
2266entry:
2267 %add = add nsw i32 %elem, 4
2268 %idxprom = sext i32 %add to i64
2269 %arrayidx = getelementptr inbounds i32, i32* %arr, i64 %idxprom
2270 %0 = load i32, i32* %arrayidx, align 4
2271 %vecinit = insertelement <4 x i32> undef, i32 %0, i32 0
2272 %add1 = add nsw i32 %elem, 1
2273 %idxprom2 = sext i32 %add1 to i64
2274 %arrayidx3 = getelementptr inbounds i32, i32* %arr, i64 %idxprom2
2275 %1 = load i32, i32* %arrayidx3, align 4
2276 %vecinit4 = insertelement <4 x i32> %vecinit, i32 %1, i32 1
2277 %add5 = add nsw i32 %elem, 2
2278 %idxprom6 = sext i32 %add5 to i64
2279 %arrayidx7 = getelementptr inbounds i32, i32* %arr, i64 %idxprom6
2280 %2 = load i32, i32* %arrayidx7, align 4
2281 %vecinit8 = insertelement <4 x i32> %vecinit4, i32 %2, i32 2
2282 %add9 = add nsw i32 %elem, 8
2283 %idxprom10 = sext i32 %add9 to i64
2284 %arrayidx11 = getelementptr inbounds i32, i32* %arr, i64 %idxprom10
2285 %3 = load i32, i32* %arrayidx11, align 4
2286 %vecinit12 = insertelement <4 x i32> %vecinit8, i32 %3, i32 3
2287 ret <4 x i32> %vecinit12
2288; P9BE-LABEL: fromRandMemVarui
2289; P9LE-LABEL: fromRandMemVarui
2290; P8BE-LABEL: fromRandMemVarui
2291; P8LE-LABEL: fromRandMemVarui
2292; P9BE: sldi r4, r4, 2
2293; P9BE: lwz
2294; P9BE: lwz
2295; P9BE: lwz
2296; P9BE: lwz
2297; P9BE: mtvsrdd
2298; P9BE: mtvsrdd
2299; P9BE: vmrgow
2300; P9LE: sldi r4, r4, 2
2301; P9LE: lwz
2302; P9LE: lwz
2303; P9LE: lwz
2304; P9LE: lwz
2305; P9LE: mtvsrdd
2306; P9LE: mtvsrdd
2307; P9LE: vmrgow
2308; P8BE: sldi r4, r4, 2
2309; P8BE: lwz
2310; P8BE: lwz
2311; P8BE: lwz
2312; P8BE: lwz
2313; P8BE: mtvsrwz
2314; P8BE: mtvsrwz
2315; P8BE: mtvsrwz
2316; P8BE: mtvsrwz
2317; P8BE: xxmrghd
2318; P8BE: xxmrghd
2319; P8BE: vmrgow
2320; P8LE: sldi r4, r4, 2
2321; P8LE: lwz
2322; P8LE: lwz
2323; P8LE: lwz
2324; P8LE: lwz
2325; P8LE: mtvsrwz
2326; P8LE: mtvsrwz
2327; P8LE: mtvsrwz
2328; P8LE: mtvsrwz
2329; P8LE: xxmrghd
2330; P8LE: xxmrghd
2331; P8LE: vmrgow
2332}
2333
2334; Function Attrs: norecurse nounwind readnone
2335define <4 x i32> @spltRegValui(i32 zeroext %val) {
2336entry:
2337 %splat.splatinsert = insertelement <4 x i32> undef, i32 %val, i32 0
2338 %splat.splat = shufflevector <4 x i32> %splat.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
2339 ret <4 x i32> %splat.splat
2340; P9BE-LABEL: spltRegValui
2341; P9LE-LABEL: spltRegValui
2342; P8BE-LABEL: spltRegValui
2343; P8LE-LABEL: spltRegValui
2344; P9BE: mtvsrws v2, r3
2345; P9BE: blr
2346; P9LE: mtvsrws v2, r3
2347; P9LE: blr
2348; P8BE: mtvsrwz {{[vsf0-9]+}}, r3
2349; P8BE: xxspltw v2, {{[vsf0-9]+}}, 1
2350; P8BE: blr
2351; P8LE: mtvsrwz {{[vsf0-9]+}}, r3
2352; P8LE: xxspltw v2, {{[vsf0-9]+}}, 1
2353; P8LE: blr
2354}
2355
2356; Function Attrs: norecurse nounwind readonly
2357define <4 x i32> @spltMemValui(i32* nocapture readonly %ptr) {
2358entry:
2359 %0 = load i32, i32* %ptr, align 4
2360 %splat.splatinsert = insertelement <4 x i32> undef, i32 %0, i32 0
2361 %splat.splat = shufflevector <4 x i32> %splat.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
2362 ret <4 x i32> %splat.splat
2363; P9BE-LABEL: spltMemValui
2364; P9LE-LABEL: spltMemValui
2365; P8BE-LABEL: spltMemValui
2366; P8LE-LABEL: spltMemValui
2367; P9BE: lxvwsx v2, 0, r3
2368; P9BE: blr
2369; P9LE: lxvwsx v2, 0, r3
2370; P9LE: blr
2371; P8BE: lxsiwax {{[vsf0-9]+}}, 0, r3
2372; P8BE: xxspltw v2, {{[vsf0-9]+}}, 1
2373; P8BE: blr
2374; P8LE: lxsiwax {{[vsf0-9]+}}, 0, r3
2375; P8LE: xxspltw v2, {{[vsf0-9]+}}, 1
2376; P8LE: blr
2377}
2378
2379; Function Attrs: norecurse nounwind readnone
2380define <4 x i32> @spltCnstConvftoui() {
2381entry:
2382 ret <4 x i32> <i32 4, i32 4, i32 4, i32 4>
2383; P9BE-LABEL: spltCnstConvftoui
2384; P9LE-LABEL: spltCnstConvftoui
2385; P8BE-LABEL: spltCnstConvftoui
2386; P8LE-LABEL: spltCnstConvftoui
2387; P9BE: vspltisw v2, 4
2388; P9BE: blr
2389; P9LE: vspltisw v2, 4
2390; P9LE: blr
2391; P8BE: vspltisw v2, 4
2392; P8BE: blr
2393; P8LE: vspltisw v2, 4
2394; P8LE: blr
2395}
2396
2397; Function Attrs: norecurse nounwind readnone
2398define <4 x i32> @fromRegsConvftoui(float %a, float %b, float %c, float %d) {
2399entry:
2400 %conv = fptoui float %a to i32
2401 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
2402 %conv1 = fptoui float %b to i32
2403 %vecinit2 = insertelement <4 x i32> %vecinit, i32 %conv1, i32 1
2404 %conv3 = fptoui float %c to i32
2405 %vecinit4 = insertelement <4 x i32> %vecinit2, i32 %conv3, i32 2
2406 %conv5 = fptoui float %d to i32
2407 %vecinit6 = insertelement <4 x i32> %vecinit4, i32 %conv5, i32 3
2408 ret <4 x i32> %vecinit6
2409; P9BE-LABEL: fromRegsConvftoui
2410; P9LE-LABEL: fromRegsConvftoui
2411; P8BE-LABEL: fromRegsConvftoui
2412; P8LE-LABEL: fromRegsConvftoui
2413; P9BE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs1, vs3
2414; P9BE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs2, vs4
2415; P9BE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
2416; P9BE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
2417; P9BE: vmrgew v2, [[REG3]], [[REG4]]
2418; P9BE: xvcvspuxws v2, v2
2419; P9LE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs3, vs1
2420; P9LE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs4, vs2
2421; P9LE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
2422; P9LE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
2423; P9LE: vmrgew v2, [[REG4]], [[REG3]]
2424; P9LE: xvcvspuxws v2, v2
2425; P8BE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs1, vs3
2426; P8BE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs2, vs4
2427; P8BE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
2428; P8BE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
2429; P8BE: vmrgew v2, [[REG3]], [[REG4]]
2430; P8BE: xvcvspuxws v2, v2
2431; P8LE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs3, vs1
2432; P8LE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs4, vs2
2433; P8LE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
2434; P8LE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
2435; P8LE: vmrgew v2, [[REG4]], [[REG3]]
2436; P8LE: xvcvspuxws v2, v2
2437}
2438
2439; Function Attrs: norecurse nounwind readnone
2440define <4 x i32> @fromDiffConstsConvftoui() {
2441entry:
2442 ret <4 x i32> <i32 24, i32 234, i32 988, i32 422>
2443; P9BE-LABEL: fromDiffConstsConvftoui
2444; P9LE-LABEL: fromDiffConstsConvftoui
2445; P8BE-LABEL: fromDiffConstsConvftoui
2446; P8LE-LABEL: fromDiffConstsConvftoui
2447; P9BE: lxvx
2448; P9BE: blr
2449; P9LE: lxvx
2450; P9LE: blr
2451; P8BE: lxvw4x
2452; P8BE: blr
2453; P8LE: lxvd2x
2454; P8LE: xxswapd
2455; P8LE: blr
2456}
2457
2458; Function Attrs: norecurse nounwind readonly
2459define <4 x i32> @fromDiffMemConsAConvftoui(float* nocapture readonly %ptr) {
2460entry:
2461 %0 = bitcast float* %ptr to <4 x float>*
2462 %1 = load <4 x float>, <4 x float>* %0, align 4
2463 %2 = fptoui <4 x float> %1 to <4 x i32>
2464 ret <4 x i32> %2
2465; P9BE-LABEL: fromDiffMemConsAConvftoui
2466; P9LE-LABEL: fromDiffMemConsAConvftoui
2467; P8BE-LABEL: fromDiffMemConsAConvftoui
2468; P8LE-LABEL: fromDiffMemConsAConvftoui
2469; P9BE: lxvx [[REG1:[vs0-9]+]], 0, r3
2470; P9BE: xvcvspuxws v2, [[REG1]]
2471; P9BE: blr
2472; P9LE: lxvx [[REG1:[vs0-9]+]], 0, r3
2473; P9LE: xvcvspuxws v2, [[REG1]]
2474; P9LE: blr
2475; P8BE: lxvw4x [[REG1:[vs0-9]+]], 0, r3
2476; P8BE: xvcvspuxws v2, [[REG1]]
2477; P8BE: blr
2478; P8LE: lxvd2x [[REG1:[vs0-9]+]], 0, r3
2479; P8LE: xxswapd v2, [[REG1]]
2480; P8LE: xvcvspuxws v2, v2
2481; P8LE: blr
2482}
2483
2484; Function Attrs: norecurse nounwind readonly
2485define <4 x i32> @fromDiffMemConsDConvftoui(float* nocapture readonly %ptr) {
2486entry:
2487 %arrayidx = getelementptr inbounds float, float* %ptr, i64 3
2488 %0 = load float, float* %arrayidx, align 4
2489 %conv = fptoui float %0 to i32
2490 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
2491 %arrayidx1 = getelementptr inbounds float, float* %ptr, i64 2
2492 %1 = load float, float* %arrayidx1, align 4
2493 %conv2 = fptoui float %1 to i32
2494 %vecinit3 = insertelement <4 x i32> %vecinit, i32 %conv2, i32 1
2495 %arrayidx4 = getelementptr inbounds float, float* %ptr, i64 1
2496 %2 = load float, float* %arrayidx4, align 4
2497 %conv5 = fptoui float %2 to i32
2498 %vecinit6 = insertelement <4 x i32> %vecinit3, i32 %conv5, i32 2
2499 %3 = load float, float* %ptr, align 4
2500 %conv8 = fptoui float %3 to i32
2501 %vecinit9 = insertelement <4 x i32> %vecinit6, i32 %conv8, i32 3
2502 ret <4 x i32> %vecinit9
2503; P9BE-LABEL: fromDiffMemConsDConvftoui
2504; P9LE-LABEL: fromDiffMemConsDConvftoui
2505; P8BE-LABEL: fromDiffMemConsDConvftoui
2506; P8LE-LABEL: fromDiffMemConsDConvftoui
2507; P9BE: lxvx
2508; P9BE: lxvx
2509; P9BE: vperm
2510; P9BE: xvcvspuxws
2511; P9BE: blr
2512; P9LE: lxvx
2513; P9LE: lxvx
2514; P9LE: vperm
2515; P9LE: xvcvspuxws
2516; P9LE: blr
2517; P8BE: lxvw4x
2518; P8BE: lxvw4x
2519; P8BE: vperm
2520; P8BE: xvcvspuxws
2521; P8BE: blr
2522; P8LE: lxvd2x
2523; P8LE-DAG: lxvd2x
2524; P8LE-DAG: xxswapd
2525; P8LE: xxswapd
2526; P8LE: vperm
2527; P8LE: xvcvspuxws
2528; P8LE: blr
2529}
2530
2531; Function Attrs: norecurse nounwind readonly
2532define <4 x i32> @fromDiffMemVarAConvftoui(float* nocapture readonly %arr, i32 signext %elem) {
2533entry:
2534 %idxprom = sext i32 %elem to i64
2535 %arrayidx = getelementptr inbounds float, float* %arr, i64 %idxprom
2536 %0 = load float, float* %arrayidx, align 4
2537 %conv = fptoui float %0 to i32
2538 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
2539 %add = add nsw i32 %elem, 1
2540 %idxprom1 = sext i32 %add to i64
2541 %arrayidx2 = getelementptr inbounds float, float* %arr, i64 %idxprom1
2542 %1 = load float, float* %arrayidx2, align 4
2543 %conv3 = fptoui float %1 to i32
2544 %vecinit4 = insertelement <4 x i32> %vecinit, i32 %conv3, i32 1
2545 %add5 = add nsw i32 %elem, 2
2546 %idxprom6 = sext i32 %add5 to i64
2547 %arrayidx7 = getelementptr inbounds float, float* %arr, i64 %idxprom6
2548 %2 = load float, float* %arrayidx7, align 4
2549 %conv8 = fptoui float %2 to i32
2550 %vecinit9 = insertelement <4 x i32> %vecinit4, i32 %conv8, i32 2
2551 %add10 = add nsw i32 %elem, 3
2552 %idxprom11 = sext i32 %add10 to i64
2553 %arrayidx12 = getelementptr inbounds float, float* %arr, i64 %idxprom11
2554 %3 = load float, float* %arrayidx12, align 4
2555 %conv13 = fptoui float %3 to i32
2556 %vecinit14 = insertelement <4 x i32> %vecinit9, i32 %conv13, i32 3
2557 ret <4 x i32> %vecinit14
2558; P9BE-LABEL: fromDiffMemVarAConvftoui
2559; P9LE-LABEL: fromDiffMemVarAConvftoui
2560; P8BE-LABEL: fromDiffMemVarAConvftoui
2561; P8LE-LABEL: fromDiffMemVarAConvftoui
2562; FIXME: implement finding consecutive loads with pre-inc
2563; P9BE: lfsux
2564; P9LE: lfsux
2565; P8BE: lfsux
2566; P8LE: lfsux
2567}
2568
2569; Function Attrs: norecurse nounwind readonly
2570define <4 x i32> @fromDiffMemVarDConvftoui(float* nocapture readonly %arr, i32 signext %elem) {
2571entry:
2572 %idxprom = sext i32 %elem to i64
2573 %arrayidx = getelementptr inbounds float, float* %arr, i64 %idxprom
2574 %0 = load float, float* %arrayidx, align 4
2575 %conv = fptoui float %0 to i32
2576 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
2577 %sub = add nsw i32 %elem, -1
2578 %idxprom1 = sext i32 %sub to i64
2579 %arrayidx2 = getelementptr inbounds float, float* %arr, i64 %idxprom1
2580 %1 = load float, float* %arrayidx2, align 4
2581 %conv3 = fptoui float %1 to i32
2582 %vecinit4 = insertelement <4 x i32> %vecinit, i32 %conv3, i32 1
2583 %sub5 = add nsw i32 %elem, -2
2584 %idxprom6 = sext i32 %sub5 to i64
2585 %arrayidx7 = getelementptr inbounds float, float* %arr, i64 %idxprom6
2586 %2 = load float, float* %arrayidx7, align 4
2587 %conv8 = fptoui float %2 to i32
2588 %vecinit9 = insertelement <4 x i32> %vecinit4, i32 %conv8, i32 2
2589 %sub10 = add nsw i32 %elem, -3
2590 %idxprom11 = sext i32 %sub10 to i64
2591 %arrayidx12 = getelementptr inbounds float, float* %arr, i64 %idxprom11
2592 %3 = load float, float* %arrayidx12, align 4
2593 %conv13 = fptoui float %3 to i32
2594 %vecinit14 = insertelement <4 x i32> %vecinit9, i32 %conv13, i32 3
2595 ret <4 x i32> %vecinit14
2596; P9BE-LABEL: fromDiffMemVarDConvftoui
2597; P9LE-LABEL: fromDiffMemVarDConvftoui
2598; P8BE-LABEL: fromDiffMemVarDConvftoui
2599; P8LE-LABEL: fromDiffMemVarDConvftoui
2600; FIXME: implement finding consecutive loads with pre-inc
2601; P9BE: lfsux
2602; P9LE: lfsux
2603; P8BE: lfsux
2604; P8LE: lfsux
2605}
2606
2607; Function Attrs: norecurse nounwind readnone
2608define <4 x i32> @spltRegValConvftoui(float %val) {
2609entry:
2610 %conv = fptoui float %val to i32
2611 %splat.splatinsert = insertelement <4 x i32> undef, i32 %conv, i32 0
2612 %splat.splat = shufflevector <4 x i32> %splat.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
2613 ret <4 x i32> %splat.splat
2614; P9BE-LABEL: spltRegValConvftoui
2615; P9LE-LABEL: spltRegValConvftoui
2616; P8BE-LABEL: spltRegValConvftoui
2617; P8LE-LABEL: spltRegValConvftoui
2618; P9BE: xscvdpuxws f[[REG1:[0-9]+]], f1
2619; P9BE: xxspltw v2, vs[[REG1]], 1
2620; P9BE: blr
2621; P9LE: xscvdpuxws f[[REG1:[0-9]+]], f1
2622; P9LE: xxspltw v2, vs[[REG1]], 1
2623; P9LE: blr
2624; P8BE: xscvdpuxws f[[REG1:[0-9]+]], f1
2625; P8BE: xxspltw v2, vs[[REG1]], 1
2626; P8BE: blr
2627; P8LE: xscvdpuxws f[[REG1:[0-9]+]], f1
2628; P8LE: xxspltw v2, vs[[REG1]], 1
2629; P8LE: blr
2630}
2631
2632; Function Attrs: norecurse nounwind readonly
2633define <4 x i32> @spltMemValConvftoui(float* nocapture readonly %ptr) {
2634entry:
2635 %0 = load float, float* %ptr, align 4
2636 %conv = fptoui float %0 to i32
2637 %splat.splatinsert = insertelement <4 x i32> undef, i32 %conv, i32 0
2638 %splat.splat = shufflevector <4 x i32> %splat.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
2639 ret <4 x i32> %splat.splat
2640; P9BE-LABEL: spltMemValConvftoui
2641; P9LE-LABEL: spltMemValConvftoui
2642; P8BE-LABEL: spltMemValConvftoui
2643; P8LE-LABEL: spltMemValConvftoui
2644; P9BE: lxvwsx [[REG1:[vs0-9]+]], 0, r3
2645; P9BE: xvcvspuxws v2, [[REG1]]
2646; P9LE: [[REG1:[vs0-9]+]], 0, r3
2647; P9LE: xvcvspuxws v2, [[REG1]]
2648; P8BE: lxsspx [[REG1:f[0-9]+]], 0, r3
2649; P8BE: xscvdpuxws f[[REG2:[0-9]+]], [[REG1]]
2650; P8BE: xxspltw v2, vs[[REG2]], 1
2651; P8LE: lxsspx [[REG1:f[0-9]+]], 0, r3
2652; P8LE: xscvdpuxws f[[REG2:[vs0-9]+]], [[REG1]]
2653; P8LE: xxspltw v2, vs[[REG2]], 1
2654}
2655
2656; Function Attrs: norecurse nounwind readnone
2657define <4 x i32> @spltCnstConvdtoui() {
2658entry:
2659 ret <4 x i32> <i32 4, i32 4, i32 4, i32 4>
2660; P9BE-LABEL: spltCnstConvdtoui
2661; P9LE-LABEL: spltCnstConvdtoui
2662; P8BE-LABEL: spltCnstConvdtoui
2663; P8LE-LABEL: spltCnstConvdtoui
2664; P9BE: vspltisw v2, 4
2665; P9BE: blr
2666; P9LE: vspltisw v2, 4
2667; P9LE: blr
2668; P8BE: vspltisw v2, 4
2669; P8BE: blr
2670; P8LE: vspltisw v2, 4
2671; P8LE: blr
2672}
2673
2674; Function Attrs: norecurse nounwind readnone
2675define <4 x i32> @fromRegsConvdtoui(double %a, double %b, double %c, double %d) {
2676entry:
2677 %conv = fptoui double %a to i32
2678 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
2679 %conv1 = fptoui double %b to i32
2680 %vecinit2 = insertelement <4 x i32> %vecinit, i32 %conv1, i32 1
2681 %conv3 = fptoui double %c to i32
2682 %vecinit4 = insertelement <4 x i32> %vecinit2, i32 %conv3, i32 2
2683 %conv5 = fptoui double %d to i32
2684 %vecinit6 = insertelement <4 x i32> %vecinit4, i32 %conv5, i32 3
2685 ret <4 x i32> %vecinit6
2686; P9BE-LABEL: fromRegsConvdtoui
2687; P9LE-LABEL: fromRegsConvdtoui
2688; P8BE-LABEL: fromRegsConvdtoui
2689; P8LE-LABEL: fromRegsConvdtoui
2690; P9BE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs1, vs3
2691; P9BE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs2, vs4
2692; P9BE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
2693; P9BE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
2694; P9BE: vmrgew v2, [[REG3]], [[REG4]]
2695; P9BE: xvcvspuxws v2, v2
2696; P9LE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs3, vs1
2697; P9LE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs4, vs2
2698; P9LE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
2699; P9LE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
2700; P9LE: vmrgew v2, [[REG4]], [[REG3]]
2701; P9LE: xvcvspuxws v2, v2
2702; P8BE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs1, vs3
2703; P8BE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs2, vs4
2704; P8BE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
2705; P8BE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
2706; P8BE: vmrgew v2, [[REG3]], [[REG4]]
2707; P8BE: xvcvspuxws v2, v2
2708; P8LE-DAG: xxmrghd {{[vs]+}}[[REG1:[0-9]+]], vs3, vs1
2709; P8LE-DAG: xxmrghd {{[vs]+}}[[REG2:[0-9]+]], vs4, vs2
2710; P8LE-DAG: xvcvdpsp [[REG3:v[0-9]+]], {{[vs]+}}[[REG1]]
2711; P8LE-DAG: xvcvdpsp [[REG4:v[0-9]+]], {{[vs]+}}[[REG2]]
2712; P8LE: vmrgew v2, [[REG4]], [[REG3]]
2713; P8LE: xvcvspuxws v2, v2
2714}
2715
2716; Function Attrs: norecurse nounwind readnone
2717define <4 x i32> @fromDiffConstsConvdtoui() {
2718entry:
2719 ret <4 x i32> <i32 24, i32 234, i32 988, i32 422>
2720; P9BE-LABEL: fromDiffConstsConvdtoui
2721; P9LE-LABEL: fromDiffConstsConvdtoui
2722; P8BE-LABEL: fromDiffConstsConvdtoui
2723; P8LE-LABEL: fromDiffConstsConvdtoui
2724; P9BE: lxvx
2725; P9BE: blr
2726; P9LE: lxvx
2727; P9LE: blr
2728; P8BE: lxvw4x
2729; P8BE: blr
2730; P8LE: lxvd2x
2731; P8LE: xxswapd
2732; P8LE: blr
2733}
2734
2735; Function Attrs: norecurse nounwind readonly
2736define <4 x i32> @fromDiffMemConsAConvdtoui(double* nocapture readonly %ptr) {
2737entry:
2738 %0 = bitcast double* %ptr to <2 x double>*
2739 %1 = load <2 x double>, <2 x double>* %0, align 8
2740 %2 = fptoui <2 x double> %1 to <2 x i32>
2741 %arrayidx4 = getelementptr inbounds double, double* %ptr, i64 2
2742 %3 = bitcast double* %arrayidx4 to <2 x double>*
2743 %4 = load <2 x double>, <2 x double>* %3, align 8
2744 %5 = fptoui <2 x double> %4 to <2 x i32>
2745 %vecinit9 = shufflevector <2 x i32> %2, <2 x i32> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
2746 ret <4 x i32> %vecinit9
2747; P9BE-LABEL: fromDiffMemConsAConvdtoui
2748; P9LE-LABEL: fromDiffMemConsAConvdtoui
2749; P8BE-LABEL: fromDiffMemConsAConvdtoui
2750; P8LE-LABEL: fromDiffMemConsAConvdtoui
2751; P9BE: lxvx [[REG1:[vs0-9]+]], 0, r3
2752; P9BE: lxvx [[REG2:[vs0-9]+]], r3, r4
2753; P9BE-DAG: xxmrgld [[REG3:[vs0-9]+]], [[REG1]], [[REG2]]
2754; P9BE-DAG: xxmrghd [[REG4:[vs0-9]+]], [[REG1]], [[REG2]]
2755; P9BE-DAG: xvcvdpsp [[REG5:[vs0-9]+]], [[REG3]]
2756; P9BE-DAG: xvcvdpsp [[REG6:[vs0-9]+]], [[REG4]]
2757; P9BE: vmrgew v2, [[REG6]], [[REG5]]
2758; P9BE: xvcvspuxws v2, v2
2759; P9LE: lxvx [[REG1:[vs0-9]+]], 0, r3
2760; P9LE: lxvx [[REG2:[vs0-9]+]], r3, r4
2761; P9LE-DAG: xxmrgld [[REG3:[vs0-9]+]], [[REG2]], [[REG1]]
2762; P9LE-DAG: xxmrghd [[REG4:[vs0-9]+]], [[REG2]], [[REG1]]
2763; P9LE-DAG: xvcvdpsp [[REG5:[vs0-9]+]], [[REG3]]
2764; P9LE-DAG: xvcvdpsp [[REG6:[vs0-9]+]], [[REG4]]
2765; P9LE: vmrgew v2, [[REG6]], [[REG5]]
2766; P9LE: xvcvspuxws v2, v2
2767; P8BE: lxvd2x [[REG1:[vs0-9]+]], 0, r3
2768; P8BE: lxvd2x [[REG2:[vs0-9]+]], r3, r4
2769; P8BE-DAG: xxmrgld [[REG3:[vs0-9]+]], [[REG1]], [[REG2]]
2770; P8BE-DAG: xxmrghd [[REG4:[vs0-9]+]], [[REG1]], [[REG2]]
2771; P8BE-DAG: xvcvdpsp [[REG5:[vs0-9]+]], [[REG3]]
2772; P8BE-DAG: xvcvdpsp [[REG6:[vs0-9]+]], [[REG4]]
2773; P8BE: vmrgew v2, [[REG6]], [[REG5]]
2774; P8BE: xvcvspuxws v2, v2
2775; P8LE: lxvd2x [[REG1:[vs0-9]+]], 0, r3
2776; P8LE: lxvd2x [[REG2:[vs0-9]+]], r3, r4
2777; P8LE-DAG: xxswapd [[REG3:[vs0-9]+]], [[REG1]]
2778; P8LE-DAG: xxswapd [[REG4:[vs0-9]+]], [[REG2]]
2779; P8LE-DAG: xxmrgld [[REG5:[vs0-9]+]], [[REG4]], [[REG3]]
2780; P8LE-DAG: xxmrghd [[REG6:[vs0-9]+]], [[REG4]], [[REG3]]
2781; P8LE-DAG: xvcvdpsp [[REG7:[vs0-9]+]], [[REG5]]
2782; P8LE-DAG: xvcvdpsp [[REG8:[vs0-9]+]], [[REG6]]
2783; P8LE: vmrgew v2, [[REG8]], [[REG7]]
2784; P8LE: xvcvspuxws v2, v2
2785}
2786
2787; Function Attrs: norecurse nounwind readonly
2788define <4 x i32> @fromDiffMemConsDConvdtoui(double* nocapture readonly %ptr) {
2789entry:
2790 %arrayidx = getelementptr inbounds double, double* %ptr, i64 3
2791 %0 = load double, double* %arrayidx, align 8
2792 %conv = fptoui double %0 to i32
2793 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
2794 %arrayidx1 = getelementptr inbounds double, double* %ptr, i64 2
2795 %1 = load double, double* %arrayidx1, align 8
2796 %conv2 = fptoui double %1 to i32
2797 %vecinit3 = insertelement <4 x i32> %vecinit, i32 %conv2, i32 1
2798 %arrayidx4 = getelementptr inbounds double, double* %ptr, i64 1
2799 %2 = load double, double* %arrayidx4, align 8
2800 %conv5 = fptoui double %2 to i32
2801 %vecinit6 = insertelement <4 x i32> %vecinit3, i32 %conv5, i32 2
2802 %3 = load double, double* %ptr, align 8
2803 %conv8 = fptoui double %3 to i32
2804 %vecinit9 = insertelement <4 x i32> %vecinit6, i32 %conv8, i32 3
2805 ret <4 x i32> %vecinit9
2806; P9BE-LABEL: fromDiffMemConsDConvdtoui
2807; P9LE-LABEL: fromDiffMemConsDConvdtoui
2808; P8BE-LABEL: fromDiffMemConsDConvdtoui
2809; P8LE-LABEL: fromDiffMemConsDConvdtoui
2810; P9BE: lfd
2811; P9BE: lfd
2812; P9BE: lfd
2813; P9BE: lfd
2814; P9BE: xxmrghd
2815; P9BE: xxmrghd
2816; P9BE: xvcvdpsp
2817; P9BE: xvcvdpsp
2818; P9BE: vmrgew
2819; P9BE: xvcvspuxws v2
2820; P9LE: lfd
2821; P9LE: lfd
2822; P9LE: lfd
2823; P9LE: lfd
2824; P9LE: xxmrghd
2825; P9LE: xxmrghd
2826; P9LE: xvcvdpsp
2827; P9LE: xvcvdpsp
2828; P9LE: vmrgew
2829; P9LE: xvcvspuxws v2
2830; P8BE: lxsdx
2831; P8BE: lxsdx
2832; P8BE: lxsdx
2833; P8BE: lxsdx
2834; P8BE: xxmrghd
2835; P8BE: xxmrghd
2836; P8BE: xvcvdpsp
2837; P8BE: xvcvdpsp
2838; P8BE: vmrgew
2839; P8BE: xvcvspuxws v2
2840; P8LE: lxsdx
2841; P8LE: lxsdx
2842; P8LE: lxsdx
2843; P8LE: lxsdx
2844; P8LE: xxmrghd
2845; P8LE: xxmrghd
2846; P8LE: xvcvdpsp
2847; P8LE: xvcvdpsp
2848; P8LE: vmrgew
2849; P8LE: xvcvspuxws v2
2850}
2851
2852; Function Attrs: norecurse nounwind readonly
2853define <4 x i32> @fromDiffMemVarAConvdtoui(double* nocapture readonly %arr, i32 signext %elem) {
2854entry:
2855 %idxprom = sext i32 %elem to i64
2856 %arrayidx = getelementptr inbounds double, double* %arr, i64 %idxprom
2857 %0 = load double, double* %arrayidx, align 8
2858 %conv = fptoui double %0 to i32
2859 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
2860 %add = add nsw i32 %elem, 1
2861 %idxprom1 = sext i32 %add to i64
2862 %arrayidx2 = getelementptr inbounds double, double* %arr, i64 %idxprom1
2863 %1 = load double, double* %arrayidx2, align 8
2864 %conv3 = fptoui double %1 to i32
2865 %vecinit4 = insertelement <4 x i32> %vecinit, i32 %conv3, i32 1
2866 %add5 = add nsw i32 %elem, 2
2867 %idxprom6 = sext i32 %add5 to i64
2868 %arrayidx7 = getelementptr inbounds double, double* %arr, i64 %idxprom6
2869 %2 = load double, double* %arrayidx7, align 8
2870 %conv8 = fptoui double %2 to i32
2871 %vecinit9 = insertelement <4 x i32> %vecinit4, i32 %conv8, i32 2
2872 %add10 = add nsw i32 %elem, 3
2873 %idxprom11 = sext i32 %add10 to i64
2874 %arrayidx12 = getelementptr inbounds double, double* %arr, i64 %idxprom11
2875 %3 = load double, double* %arrayidx12, align 8
2876 %conv13 = fptoui double %3 to i32
2877 %vecinit14 = insertelement <4 x i32> %vecinit9, i32 %conv13, i32 3
2878 ret <4 x i32> %vecinit14
2879; P9BE-LABEL: fromDiffMemVarAConvdtoui
2880; P9LE-LABEL: fromDiffMemVarAConvdtoui
2881; P8BE-LABEL: fromDiffMemVarAConvdtoui
2882; P8LE-LABEL: fromDiffMemVarAConvdtoui
2883; P9BE: lfdux
2884; P9BE: lfd
2885; P9BE: lfd
2886; P9BE: lfd
2887; P9BE: xxmrghd
2888; P9BE: xxmrghd
2889; P9BE: xvcvdpsp
2890; P9BE: xvcvdpsp
2891; P9BE: vmrgew
2892; P9BE: xvcvspuxws v2
2893; P9LE: lfdux
2894; P9LE: lfd
2895; P9LE: lfd
2896; P9LE: lfd
2897; P9LE: xxmrghd
2898; P9LE: xxmrghd
2899; P9LE: xvcvdpsp
2900; P9LE: xvcvdpsp
2901; P9LE: vmrgew
2902; P9LE: xvcvspuxws v2
2903; P8BE: lfdux
2904; P8BE: lxsdx
2905; P8BE: lxsdx
2906; P8BE: lxsdx
2907; P8BE: xxmrghd
2908; P8BE: xxmrghd
2909; P8BE: xvcvdpsp
2910; P8BE: xvcvdpsp
2911; P8BE: vmrgew
2912; P8BE: xvcvspuxws v2
2913; P8LE: lfdux
2914; P8LE: lxsdx
2915; P8LE: lxsdx
2916; P8LE: lxsdx
2917; P8LE: xxmrghd
2918; P8LE: xxmrghd
2919; P8LE: xvcvdpsp
2920; P8LE: xvcvdpsp
2921; P8LE: vmrgew
2922; P8LE: xvcvspuxws v2
2923}
2924
2925; Function Attrs: norecurse nounwind readonly
2926define <4 x i32> @fromDiffMemVarDConvdtoui(double* nocapture readonly %arr, i32 signext %elem) {
2927entry:
2928 %idxprom = sext i32 %elem to i64
2929 %arrayidx = getelementptr inbounds double, double* %arr, i64 %idxprom
2930 %0 = load double, double* %arrayidx, align 8
2931 %conv = fptoui double %0 to i32
2932 %vecinit = insertelement <4 x i32> undef, i32 %conv, i32 0
2933 %sub = add nsw i32 %elem, -1
2934 %idxprom1 = sext i32 %sub to i64
2935 %arrayidx2 = getelementptr inbounds double, double* %arr, i64 %idxprom1
2936 %1 = load double, double* %arrayidx2, align 8
2937 %conv3 = fptoui double %1 to i32
2938 %vecinit4 = insertelement <4 x i32> %vecinit, i32 %conv3, i32 1
2939 %sub5 = add nsw i32 %elem, -2
2940 %idxprom6 = sext i32 %sub5 to i64
2941 %arrayidx7 = getelementptr inbounds double, double* %arr, i64 %idxprom6
2942 %2 = load double, double* %arrayidx7, align 8
2943 %conv8 = fptoui double %2 to i32
2944 %vecinit9 = insertelement <4 x i32> %vecinit4, i32 %conv8, i32 2
2945 %sub10 = add nsw i32 %elem, -3
2946 %idxprom11 = sext i32 %sub10 to i64
2947 %arrayidx12 = getelementptr inbounds double, double* %arr, i64 %idxprom11
2948 %3 = load double, double* %arrayidx12, align 8
2949 %conv13 = fptoui double %3 to i32
2950 %vecinit14 = insertelement <4 x i32> %vecinit9, i32 %conv13, i32 3
2951 ret <4 x i32> %vecinit14
2952; P9BE-LABEL: fromDiffMemVarDConvdtoui
2953; P9LE-LABEL: fromDiffMemVarDConvdtoui
2954; P8BE-LABEL: fromDiffMemVarDConvdtoui
2955; P8LE-LABEL: fromDiffMemVarDConvdtoui
2956; P9BE: lfdux
2957; P9BE: lfd
2958; P9BE: lfd
2959; P9BE: lfd
2960; P9BE: xxmrghd
2961; P9BE: xxmrghd
2962; P9BE: xvcvdpsp
2963; P9BE: xvcvdpsp
2964; P9BE: vmrgew
2965; P9BE: xvcvspuxws v2
2966; P9LE: lfdux
2967; P9LE: lfd
2968; P9LE: lfd
2969; P9LE: lfd
2970; P9LE: xxmrghd
2971; P9LE: xxmrghd
2972; P9LE: xvcvdpsp
2973; P9LE: xvcvdpsp
2974; P9LE: vmrgew
2975; P9LE: xvcvspuxws v2
2976; P8BE: lfdux
2977; P8BE: lxsdx
2978; P8BE: lxsdx
2979; P8BE: lxsdx
2980; P8BE: xxmrghd
2981; P8BE: xxmrghd
2982; P8BE: xvcvdpsp
2983; P8BE: xvcvdpsp
2984; P8BE: vmrgew
2985; P8BE: xvcvspuxws v2
2986; P8LE: lfdux
2987; P8LE: lxsdx
2988; P8LE: lxsdx
2989; P8LE: lxsdx
2990; P8LE: xxmrghd
2991; P8LE: xxmrghd
2992; P8LE: xvcvdpsp
2993; P8LE: xvcvdpsp
2994; P8LE: vmrgew
2995; P8LE: xvcvspuxws v2
2996}
2997
2998; Function Attrs: norecurse nounwind readnone
2999define <4 x i32> @spltRegValConvdtoui(double %val) {
3000entry:
3001 %conv = fptoui double %val to i32
3002 %splat.splatinsert = insertelement <4 x i32> undef, i32 %conv, i32 0
3003 %splat.splat = shufflevector <4 x i32> %splat.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
3004 ret <4 x i32> %splat.splat
3005; P9BE-LABEL: spltRegValConvdtoui
3006; P9LE-LABEL: spltRegValConvdtoui
3007; P8BE-LABEL: spltRegValConvdtoui
3008; P8LE-LABEL: spltRegValConvdtoui
3009; P9BE: xscvdpuxws
3010; P9BE: xxspltw
3011; P9BE: blr
3012; P9LE: xscvdpuxws
3013; P9LE: xxspltw
3014; P9LE: blr
3015; P8BE: xscvdpuxws
3016; P8BE: xxspltw
3017; P8BE: blr
3018; P8LE: xscvdpuxws
3019; P8LE: xxspltw
3020; P8LE: blr
3021}
3022
3023; Function Attrs: norecurse nounwind readonly
3024define <4 x i32> @spltMemValConvdtoui(double* nocapture readonly %ptr) {
3025entry:
3026 %0 = load double, double* %ptr, align 8
3027 %conv = fptoui double %0 to i32
3028 %splat.splatinsert = insertelement <4 x i32> undef, i32 %conv, i32 0
3029 %splat.splat = shufflevector <4 x i32> %splat.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
3030 ret <4 x i32> %splat.splat
3031; P9BE-LABEL: spltMemValConvdtoui
3032; P9LE-LABEL: spltMemValConvdtoui
3033; P8BE-LABEL: spltMemValConvdtoui
3034; P8LE-LABEL: spltMemValConvdtoui
3035; P9BE: lfd
3036; P9BE: xscvdpuxws
3037; P9BE: xxspltw
3038; P9BE: blr
3039; P9LE: lfd
3040; P9LE: xscvdpuxws
3041; P9LE: xxspltw
3042; P9LE: blr
3043; P8BE: lxsdx
3044; P8BE: xscvdpuxws
3045; P8BE: xxspltw
3046; P8BE: blr
3047; P8LE: lxsdx
3048; P8LE: xscvdpuxws
3049; P8LE: xxspltw
3050; P8LE: blr
3051}
3052; Function Attrs: norecurse nounwind readnone
3053define <2 x i64> @allZeroll() {
3054entry:
3055 ret <2 x i64> zeroinitializer
3056; P9BE-LABEL: allZeroll
3057; P9LE-LABEL: allZeroll
3058; P8BE-LABEL: allZeroll
3059; P8LE-LABEL: allZeroll
3060; P9BE: xxlxor v2, v2, v2
3061; P9BE: blr
3062; P9LE: xxlxor v2, v2, v2
3063; P9LE: blr
3064; P8BE: xxlxor v2, v2, v2
3065; P8BE: blr
3066; P8LE: xxlxor v2, v2, v2
3067; P8LE: blr
3068}
3069
3070; Function Attrs: norecurse nounwind readnone
3071define <2 x i64> @allOnell() {
3072entry:
3073 ret <2 x i64> <i64 -1, i64 -1>
3074; P9BE-LABEL: allOnell
3075; P9LE-LABEL: allOnell
3076; P8BE-LABEL: allOnell
3077; P8LE-LABEL: allOnell
3078; P9BE: xxspltib v2, 255
3079; P9BE: blr
3080; P9LE: xxspltib v2, 255
3081; P9LE: blr
3082; P8BE: vspltisb v2, -1
3083; P8BE: blr
3084; P8LE: vspltisb v2, -1
3085; P8LE: blr
3086}
3087
3088; Function Attrs: norecurse nounwind readnone
3089define <2 x i64> @spltConst1ll() {
3090entry:
3091 ret <2 x i64> <i64 1, i64 1>
3092; P9BE-LABEL: spltConst1ll
3093; P9LE-LABEL: spltConst1ll
3094; P8BE-LABEL: spltConst1ll
3095; P8LE-LABEL: spltConst1ll
3096; P9BE: lxvx
3097; P9BE: blr
3098; P9LE: lxvx
3099; P9LE: blr
3100; P8BE: lxvd2x
3101; P8BE: blr
3102; P8LE: lxvd2x
3103; P8LE: blr
3104}
3105
3106; Function Attrs: norecurse nounwind readnone
3107define <2 x i64> @spltConst16kll() {
3108entry:
3109 ret <2 x i64> <i64 32767, i64 32767>
3110; P9BE-LABEL: spltConst16kll
3111; P9LE-LABEL: spltConst16kll
3112; P8BE-LABEL: spltConst16kll
3113; P8LE-LABEL: spltConst16kll
3114; P9BE: lxvx
3115; P9BE: blr
3116; P9LE: lxvx
3117; P9LE: blr
3118; P8BE: lxvd2x
3119; P8BE: blr
3120; P8LE: lxvd2x
3121; P8LE: blr
3122}
3123
3124; Function Attrs: norecurse nounwind readnone
3125define <2 x i64> @spltConst32kll() {
3126entry:
3127 ret <2 x i64> <i64 65535, i64 65535>
3128; P9BE-LABEL: spltConst32kll
3129; P9LE-LABEL: spltConst32kll
3130; P8BE-LABEL: spltConst32kll
3131; P8LE-LABEL: spltConst32kll
3132; P9BE: lxvx
3133; P9BE: blr
3134; P9LE: lxvx
3135; P9LE: blr
3136; P8BE: lxvd2x
3137; P8BE: blr
3138; P8LE: lxvd2x
3139; P8LE: blr
3140}
3141
3142; Function Attrs: norecurse nounwind readnone
3143define <2 x i64> @fromRegsll(i64 %a, i64 %b) {
3144entry:
3145 %vecinit = insertelement <2 x i64> undef, i64 %a, i32 0
3146 %vecinit1 = insertelement <2 x i64> %vecinit, i64 %b, i32 1
3147 ret <2 x i64> %vecinit1
3148; P9BE-LABEL: fromRegsll
3149; P9LE-LABEL: fromRegsll
3150; P8BE-LABEL: fromRegsll
3151; P8LE-LABEL: fromRegsll
3152; P9BE: mtvsrdd v2, r3, r4
3153; P9BE: blr
3154; P9LE: mtvsrdd v2, r4, r3
3155; P9LE: blr
3156; P8BE-DAG: mtvsrd {{[vsf0-9]+}}, r3
3157; P8BE-DAG: mtvsrd {{[vsf0-9]+}}, r4
3158; P8BE: xxmrghd v2
3159; P8BE: blr
3160; P8LE-DAG: mtvsrd {{[vsf0-9]+}}, r3
3161; P8LE-DAG: mtvsrd {{[vsf0-9]+}}, r4
3162; P8LE: xxmrghd v2
3163; P8LE: blr
3164}
3165
3166; Function Attrs: norecurse nounwind readnone
3167define <2 x i64> @fromDiffConstsll() {
3168entry:
3169 ret <2 x i64> <i64 242, i64 -113>
3170; P9BE-LABEL: fromDiffConstsll
3171; P9LE-LABEL: fromDiffConstsll
3172; P8BE-LABEL: fromDiffConstsll
3173; P8LE-LABEL: fromDiffConstsll
3174; P9BE: lxvx
3175; P9BE: blr
3176; P9LE: lxvx
3177; P9LE: blr
3178; P8BE: lxvd2x
3179; P8BE: blr
3180; P8LE: lxvd2x
3181; P8LE: blr
3182}
3183
3184; Function Attrs: norecurse nounwind readonly
3185define <2 x i64> @fromDiffMemConsAll(i64* nocapture readonly %arr) {
3186entry:
3187 %0 = load i64, i64* %arr, align 8
3188 %vecinit = insertelement <2 x i64> undef, i64 %0, i32 0
3189 %arrayidx1 = getelementptr inbounds i64, i64* %arr, i64 1
3190 %1 = load i64, i64* %arrayidx1, align 8
3191 %vecinit2 = insertelement <2 x i64> %vecinit, i64 %1, i32 1
3192 ret <2 x i64> %vecinit2
3193; P9BE-LABEL: fromDiffMemConsAll
3194; P9LE-LABEL: fromDiffMemConsAll
3195; P8BE-LABEL: fromDiffMemConsAll
3196; P8LE-LABEL: fromDiffMemConsAll
3197; P9BE: lxvx v2
3198; P9BE: blr
3199; P9LE: lxvx v2
3200; P9LE: blr
3201; P8BE: lxvd2x v2
3202; P8BE: blr
3203; P8LE: lxvd2x
3204; P8LE: xxswapd v2
3205; P8LE: blr
3206}
3207
3208; Function Attrs: norecurse nounwind readonly
3209define <2 x i64> @fromDiffMemConsDll(i64* nocapture readonly %arr) {
3210entry:
3211 %arrayidx = getelementptr inbounds i64, i64* %arr, i64 3
3212 %0 = load i64, i64* %arrayidx, align 8
3213 %vecinit = insertelement <2 x i64> undef, i64 %0, i32 0
3214 %arrayidx1 = getelementptr inbounds i64, i64* %arr, i64 2
3215 %1 = load i64, i64* %arrayidx1, align 8
3216 %vecinit2 = insertelement <2 x i64> %vecinit, i64 %1, i32 1
3217 ret <2 x i64> %vecinit2
3218; P9BE-LABEL: fromDiffMemConsDll
3219; P9LE-LABEL: fromDiffMemConsDll
3220; P8BE-LABEL: fromDiffMemConsDll
3221; P8LE-LABEL: fromDiffMemConsDll
3222; P9BE: lxvx v2
3223; P9BE: blr
3224; P9LE: lxvx
3225; P9LE: xxswapd v2
3226; P9LE: blr
3227; P8BE: lxvd2x
3228; P8BE: xxswapd v2
3229; P8BE-NEXT: blr
3230; P8LE: lxvd2x v2
3231; P8LE-NEXT: blr
3232}
3233
3234; Function Attrs: norecurse nounwind readonly
3235define <2 x i64> @fromDiffMemVarAll(i64* nocapture readonly %arr, i32 signext %elem) {
3236entry:
3237 %idxprom = sext i32 %elem to i64
3238 %arrayidx = getelementptr inbounds i64, i64* %arr, i64 %idxprom
3239 %0 = load i64, i64* %arrayidx, align 8
3240 %vecinit = insertelement <2 x i64> undef, i64 %0, i32 0
3241 %add = add nsw i32 %elem, 1
3242 %idxprom1 = sext i32 %add to i64
3243 %arrayidx2 = getelementptr inbounds i64, i64* %arr, i64 %idxprom1
3244 %1 = load i64, i64* %arrayidx2, align 8
3245 %vecinit3 = insertelement <2 x i64> %vecinit, i64 %1, i32 1
3246 ret <2 x i64> %vecinit3
3247; P9BE-LABEL: fromDiffMemVarAll
3248; P9LE-LABEL: fromDiffMemVarAll
3249; P8BE-LABEL: fromDiffMemVarAll
3250; P8LE-LABEL: fromDiffMemVarAll
3251; P9BE: sldi
3252; P9BE: lxvx v2
3253; P9BE-NEXT: blr
3254; P9LE: sldi
3255; P9LE: lxvx v2
3256; P9LE-NEXT: blr
3257; P8BE: sldi
3258; P8BE: lxvd2x v2
3259; P8BE-NEXT: blr
3260; P8LE: sldi
3261; P8LE: lxvd2x
3262; P8LE: xxswapd v2
3263; P8LE-NEXT: blr
3264}
3265
3266; Function Attrs: norecurse nounwind readonly
3267define <2 x i64> @fromDiffMemVarDll(i64* nocapture readonly %arr, i32 signext %elem) {
3268entry:
3269 %idxprom = sext i32 %elem to i64
3270 %arrayidx = getelementptr inbounds i64, i64* %arr, i64 %idxprom
3271 %0 = load i64, i64* %arrayidx, align 8
3272 %vecinit = insertelement <2 x i64> undef, i64 %0, i32 0
3273 %sub = add nsw i32 %elem, -1
3274 %idxprom1 = sext i32 %sub to i64
3275 %arrayidx2 = getelementptr inbounds i64, i64* %arr, i64 %idxprom1
3276 %1 = load i64, i64* %arrayidx2, align 8
3277 %vecinit3 = insertelement <2 x i64> %vecinit, i64 %1, i32 1
3278 ret <2 x i64> %vecinit3
3279; P9BE-LABEL: fromDiffMemVarDll
3280; P9LE-LABEL: fromDiffMemVarDll
3281; P8BE-LABEL: fromDiffMemVarDll
3282; P8LE-LABEL: fromDiffMemVarDll
3283; P9BE: sldi
3284; P9BE: lxvx
3285; P9BE: xxswapd v2
3286; P9BE-NEXT: blr
3287; P9LE: sldi
3288; P9LE: lxvx
3289; P9LE: xxswapd v2
3290; P9LE-NEXT: blr
3291; P8BE: sldi
3292; P8BE: lxvd2x
3293; P8BE: xxswapd v2
3294; P8BE-NEXT: blr
3295; P8LE: sldi
3296; P8LE: lxvd2x v2
3297; P8LE-NEXT: blr
3298}
3299
3300; Function Attrs: norecurse nounwind readonly
3301define <2 x i64> @fromRandMemConsll(i64* nocapture readonly %arr) {
3302entry:
3303 %arrayidx = getelementptr inbounds i64, i64* %arr, i64 4
3304 %0 = load i64, i64* %arrayidx, align 8
3305 %vecinit = insertelement <2 x i64> undef, i64 %0, i32 0
3306 %arrayidx1 = getelementptr inbounds i64, i64* %arr, i64 18
3307 %1 = load i64, i64* %arrayidx1, align 8
3308 %vecinit2 = insertelement <2 x i64> %vecinit, i64 %1, i32 1
3309 ret <2 x i64> %vecinit2
3310; P9BE-LABEL: fromRandMemConsll
3311; P9LE-LABEL: fromRandMemConsll
3312; P8BE-LABEL: fromRandMemConsll
3313; P8LE-LABEL: fromRandMemConsll
3314; P9BE: ld
3315; P9BE: ld
3316; P9BE: mtvsrdd v2
3317; P9BE-NEXT: blr
3318; P9LE: ld
3319; P9LE: ld
3320; P9LE: mtvsrdd v2
3321; P9LE-NEXT: blr
3322; P8BE: ld
3323; P8BE: ld
3324; P8BE-DAG: mtvsrd
3325; P8BE-DAG: mtvsrd
3326; P8BE: xxmrghd v2
3327; P8BE-NEXT: blr
3328; P8LE: ld
3329; P8LE: ld
3330; P8LE-DAG: mtvsrd
3331; P8LE-DAG: mtvsrd
3332; P8LE: xxmrghd v2
3333; P8LE-NEXT: blr
3334}
3335
3336; Function Attrs: norecurse nounwind readonly
3337define <2 x i64> @fromRandMemVarll(i64* nocapture readonly %arr, i32 signext %elem) {
3338entry:
3339 %add = add nsw i32 %elem, 4
3340 %idxprom = sext i32 %add to i64
3341 %arrayidx = getelementptr inbounds i64, i64* %arr, i64 %idxprom
3342 %0 = load i64, i64* %arrayidx, align 8
3343 %vecinit = insertelement <2 x i64> undef, i64 %0, i32 0
3344 %add1 = add nsw i32 %elem, 1
3345 %idxprom2 = sext i32 %add1 to i64
3346 %arrayidx3 = getelementptr inbounds i64, i64* %arr, i64 %idxprom2
3347 %1 = load i64, i64* %arrayidx3, align 8
3348 %vecinit4 = insertelement <2 x i64> %vecinit, i64 %1, i32 1
3349 ret <2 x i64> %vecinit4
3350; P9BE-LABEL: fromRandMemVarll
3351; P9LE-LABEL: fromRandMemVarll
3352; P8BE-LABEL: fromRandMemVarll
3353; P8LE-LABEL: fromRandMemVarll
3354; P9BE: sldi
3355; P9BE: ld
3356; P9BE: ld
3357; P9BE: mtvsrdd v2
3358; P9BE-NEXT: blr
3359; P9LE: sldi
3360; P9LE: ld
3361; P9LE: ld
3362; P9LE: mtvsrdd v2
3363; P9LE-NEXT: blr
3364; P8BE: sldi
3365; P8BE: ld
3366; P8BE: ld
3367; P8BE: mtvsrd
3368; P8BE: mtvsrd
3369; P8BE: xxmrghd v2
3370; P8BE-NEXT: blr
3371; P8LE: sldi
3372; P8LE: ld
3373; P8LE: ld
3374; P8LE: mtvsrd
3375; P8LE: mtvsrd
3376; P8LE: xxmrghd v2
3377; P8LE-NEXT: blr
3378}
3379
3380; Function Attrs: norecurse nounwind readnone
3381define <2 x i64> @spltRegValll(i64 %val) {
3382entry:
3383 %splat.splatinsert = insertelement <2 x i64> undef, i64 %val, i32 0
3384 %splat.splat = shufflevector <2 x i64> %splat.splatinsert, <2 x i64> undef, <2 x i32> zeroinitializer
3385 ret <2 x i64> %splat.splat
3386; P9BE-LABEL: spltRegValll
3387; P9LE-LABEL: spltRegValll
3388; P8BE-LABEL: spltRegValll
3389; P8LE-LABEL: spltRegValll
3390; P9BE: mtvsrdd v2, r3, r3
3391; P9BE-NEXT: blr
3392; P9LE: mtvsrdd v2, r3, r3
3393; P9LE-NEXT: blr
3394; P8BE: mtvsrd {{[vsf]+}}[[REG1:[0-9]+]], r3
3395; P8BE: xxspltd v2, {{[vsf]+}}[[REG1]], 0
3396; P8BE-NEXT: blr
3397; P8LE: mtvsrd {{[vsf]+}}[[REG1:[0-9]+]], r3
3398; P8LE: xxspltd v2, {{[vsf]+}}[[REG1]], 0
3399; P8LE-NEXT: blr
3400}
3401
3402; Function Attrs: norecurse nounwind readonly
3403define <2 x i64> @spltMemValll(i64* nocapture readonly %ptr) {
3404entry:
3405 %0 = load i64, i64* %ptr, align 8
3406 %splat.splatinsert = insertelement <2 x i64> undef, i64 %0, i32 0
3407 %splat.splat = shufflevector <2 x i64> %splat.splatinsert, <2 x i64> undef, <2 x i32> zeroinitializer
3408 ret <2 x i64> %splat.splat
3409; P9BE-LABEL: spltMemValll
3410; P9LE-LABEL: spltMemValll
3411; P8BE-LABEL: spltMemValll
3412; P8LE-LABEL: spltMemValll
3413; P9BE: lxvdsx v2
3414; P9BE-NEXT: blr
3415; P9LE: lxvdsx v2
3416; P9LE-NEXT: blr
3417; P8BE: lxvdsx v2
3418; P8BE-NEXT: blr
3419; P8LE: lxvdsx v2
3420; P8LE-NEXT: blr
3421}
3422
3423; Function Attrs: norecurse nounwind readnone
3424define <2 x i64> @spltCnstConvftoll() {
3425entry:
3426 ret <2 x i64> <i64 4, i64 4>
3427; P9BE-LABEL: spltCnstConvftoll
3428; P9LE-LABEL: spltCnstConvftoll
3429; P8BE-LABEL: spltCnstConvftoll
3430; P8LE-LABEL: spltCnstConvftoll
3431; P9BE: lxvx
3432; P9BE: blr
3433; P9LE: lxvx
3434; P9LE: blr
3435; P8BE: lxvd2x
3436; P8BE: blr
3437; P8LE: lxvd2x
3438; P8LE: blr
3439}
3440
3441; Function Attrs: norecurse nounwind readnone
3442define <2 x i64> @fromRegsConvftoll(float %a, float %b) {
3443entry:
3444 %conv = fptosi float %a to i64
3445 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
3446 %conv1 = fptosi float %b to i64
3447 %vecinit2 = insertelement <2 x i64> %vecinit, i64 %conv1, i32 1
3448 ret <2 x i64> %vecinit2
3449; P9BE-LABEL: fromRegsConvftoll
3450; P9LE-LABEL: fromRegsConvftoll
3451; P8BE-LABEL: fromRegsConvftoll
3452; P8LE-LABEL: fromRegsConvftoll
3453; P9BE: xxmrghd
3454; P9BE: xvcvdpsxds v2
3455; P9BE-NEXT: blr
3456; P9LE: xxmrghd
3457; P9LE: xvcvdpsxds v2
3458; P9LE-NEXT: blr
3459; P8BE: xxmrghd
3460; P8BE: xvcvdpsxds v2
3461; P8BE-NEXT: blr
3462; P8LE: xxmrghd
3463; P8LE: xvcvdpsxds v2
3464; P8LE-NEXT: blr
3465}
3466
3467; Function Attrs: norecurse nounwind readnone
3468define <2 x i64> @fromDiffConstsConvftoll() {
3469entry:
3470 ret <2 x i64> <i64 24, i64 234>
3471; P9BE-LABEL: fromDiffConstsConvftoll
3472; P9LE-LABEL: fromDiffConstsConvftoll
3473; P8BE-LABEL: fromDiffConstsConvftoll
3474; P8LE-LABEL: fromDiffConstsConvftoll
3475; P9BE: lxvx v2
3476; P9BE: blr
3477; P9LE: lxvx v2
3478; P9LE: blr
3479; P8BE: lxvd2x v2
3480; P8BE: blr
3481; P8LE: lxvd2x
3482; P8LE: xxswapd v2
3483; P8LE: blr
3484}
3485
3486; Function Attrs: norecurse nounwind readonly
3487define <2 x i64> @fromDiffMemConsAConvftoll(float* nocapture readonly %ptr) {
3488entry:
3489 %0 = load float, float* %ptr, align 4
3490 %conv = fptosi float %0 to i64
3491 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
3492 %arrayidx1 = getelementptr inbounds float, float* %ptr, i64 1
3493 %1 = load float, float* %arrayidx1, align 4
3494 %conv2 = fptosi float %1 to i64
3495 %vecinit3 = insertelement <2 x i64> %vecinit, i64 %conv2, i32 1
3496 ret <2 x i64> %vecinit3
3497; P9BE-LABEL: fromDiffMemConsAConvftoll
3498; P9LE-LABEL: fromDiffMemConsAConvftoll
3499; P8BE-LABEL: fromDiffMemConsAConvftoll
3500; P8LE-LABEL: fromDiffMemConsAConvftoll
3501; P9BE: lfs
3502; P9BE: lfs
3503; P9BE: xxmrghd
3504; P9BE-NEXT: xvcvdpsxds v2
3505; P9BE-NEXT: blr
3506; P9LE: lfs
3507; P9LE: lfs
3508; P9LE: xxmrghd
3509; P9LE-NEXT: xvcvdpsxds v2
3510; P9LE-NEXT: blr
3511; P8BE: lxsspx
3512; P8BE: lxsspx
3513; P8BE: xxmrghd
3514; P8BE-NEXT: xvcvdpsxds v2
3515; P8BE-NEXT: blr
3516; P8LE: lxsspx
3517; P8LE: lxsspx
3518; P8LE: xxmrghd
3519; P8LE-NEXT: xvcvdpsxds v2
3520; P8LE-NEXT: blr
3521}
3522
3523; Function Attrs: norecurse nounwind readonly
3524define <2 x i64> @fromDiffMemConsDConvftoll(float* nocapture readonly %ptr) {
3525entry:
3526 %arrayidx = getelementptr inbounds float, float* %ptr, i64 3
3527 %0 = load float, float* %arrayidx, align 4
3528 %conv = fptosi float %0 to i64
3529 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
3530 %arrayidx1 = getelementptr inbounds float, float* %ptr, i64 2
3531 %1 = load float, float* %arrayidx1, align 4
3532 %conv2 = fptosi float %1 to i64
3533 %vecinit3 = insertelement <2 x i64> %vecinit, i64 %conv2, i32 1
3534 ret <2 x i64> %vecinit3
3535; P9BE-LABEL: fromDiffMemConsDConvftoll
3536; P9LE-LABEL: fromDiffMemConsDConvftoll
3537; P8BE-LABEL: fromDiffMemConsDConvftoll
3538; P8LE-LABEL: fromDiffMemConsDConvftoll
3539; P9BE: lfs
3540; P9BE: lfs
3541; P9BE: xxmrghd
3542; P9BE-NEXT: xvcvdpsxds v2
3543; P9BE-NEXT: blr
3544; P9LE: lfs
3545; P9LE: lfs
3546; P9LE: xxmrghd
3547; P9LE-NEXT: xvcvdpsxds v2
3548; P9LE-NEXT: blr
3549; P8BE: lxsspx
3550; P8BE: lxsspx
3551; P8BE: xxmrghd
3552; P8BE-NEXT: xvcvdpsxds v2
3553; P8BE-NEXT: blr
3554; P8LE: lxsspx
3555; P8LE: lxsspx
3556; P8LE: xxmrghd
3557; P8LE-NEXT: xvcvdpsxds v2
3558; P8LE-NEXT: blr
3559}
3560
3561; Function Attrs: norecurse nounwind readonly
3562define <2 x i64> @fromDiffMemVarAConvftoll(float* nocapture readonly %arr, i32 signext %elem) {
3563entry:
3564 %idxprom = sext i32 %elem to i64
3565 %arrayidx = getelementptr inbounds float, float* %arr, i64 %idxprom
3566 %0 = load float, float* %arrayidx, align 4
3567 %conv = fptosi float %0 to i64
3568 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
3569 %add = add nsw i32 %elem, 1
3570 %idxprom1 = sext i32 %add to i64
3571 %arrayidx2 = getelementptr inbounds float, float* %arr, i64 %idxprom1
3572 %1 = load float, float* %arrayidx2, align 4
3573 %conv3 = fptosi float %1 to i64
3574 %vecinit4 = insertelement <2 x i64> %vecinit, i64 %conv3, i32 1
3575 ret <2 x i64> %vecinit4
3576; P9BE-LABEL: fromDiffMemVarAConvftoll
3577; P9LE-LABEL: fromDiffMemVarAConvftoll
3578; P8BE-LABEL: fromDiffMemVarAConvftoll
3579; P8LE-LABEL: fromDiffMemVarAConvftoll
3580; P9BE: sldi
3581; P9BE: lfsux
3582; P9BE: lfs
3583; P9BE: xxmrghd
3584; P9BE-NEXT: xvcvdpsxds v2
3585; P9BE-NEXT: blr
3586; P9LE: sldi
3587; P9LE: lfsux
3588; P9LE: lfs
3589; P9LE: xxmrghd
3590; P9LE-NEXT: xvcvdpsxds v2
3591; P9LE-NEXT: blr
3592; P8BE: sldi
3593; P8BE: lfsux
3594; P8BE: lxsspx
3595; P8BE: xxmrghd
3596; P8BE-NEXT: xvcvdpsxds v2
3597; P8BE-NEXT: blr
3598; P8LE: sldi
3599; P8LE: lfsux
3600; P8LE: lxsspx
3601; P8LE: xxmrghd
3602; P8LE-NEXT: xvcvdpsxds v2
3603; P8LE-NEXT: blr
3604}
3605
3606; Function Attrs: norecurse nounwind readonly
3607define <2 x i64> @fromDiffMemVarDConvftoll(float* nocapture readonly %arr, i32 signext %elem) {
3608entry:
3609 %idxprom = sext i32 %elem to i64
3610 %arrayidx = getelementptr inbounds float, float* %arr, i64 %idxprom
3611 %0 = load float, float* %arrayidx, align 4
3612 %conv = fptosi float %0 to i64
3613 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
3614 %sub = add nsw i32 %elem, -1
3615 %idxprom1 = sext i32 %sub to i64
3616 %arrayidx2 = getelementptr inbounds float, float* %arr, i64 %idxprom1
3617 %1 = load float, float* %arrayidx2, align 4
3618 %conv3 = fptosi float %1 to i64
3619 %vecinit4 = insertelement <2 x i64> %vecinit, i64 %conv3, i32 1
3620 ret <2 x i64> %vecinit4
3621; P9BE-LABEL: fromDiffMemVarDConvftoll
3622; P9LE-LABEL: fromDiffMemVarDConvftoll
3623; P8BE-LABEL: fromDiffMemVarDConvftoll
3624; P8LE-LABEL: fromDiffMemVarDConvftoll
3625; P9BE: sldi
3626; P9BE: lfsux
3627; P9BE: lfs
3628; P9BE: xxmrghd
3629; P9BE-NEXT: xvcvdpsxds v2
3630; P9BE-NEXT: blr
3631; P9LE: sldi
3632; P9LE: lfsux
3633; P9LE: lfs
3634; P9LE: xxmrghd
3635; P9LE-NEXT: xvcvdpsxds v2
3636; P9LE-NEXT: blr
3637; P8BE: sldi
3638; P8BE: lfsux
3639; P8BE: lxsspx
3640; P8BE: xxmrghd
3641; P8BE-NEXT: xvcvdpsxds v2
3642; P8BE-NEXT: blr
3643; P8LE: sldi
3644; P8LE: lfsux
3645; P8LE: lxsspx
3646; P8LE: xxmrghd
3647; P8LE-NEXT: xvcvdpsxds v2
3648; P8LE-NEXT: blr
3649}
3650
3651; Function Attrs: norecurse nounwind readnone
3652define <2 x i64> @spltRegValConvftoll(float %val) {
3653entry:
3654 %conv = fptosi float %val to i64
3655 %splat.splatinsert = insertelement <2 x i64> undef, i64 %conv, i32 0
3656 %splat.splat = shufflevector <2 x i64> %splat.splatinsert, <2 x i64> undef, <2 x i32> zeroinitializer
3657 ret <2 x i64> %splat.splat
3658; P9BE-LABEL: spltRegValConvftoll
3659; P9LE-LABEL: spltRegValConvftoll
3660; P8BE-LABEL: spltRegValConvftoll
3661; P8LE-LABEL: spltRegValConvftoll
3662; P9BE: xscvdpsxds
3663; P9BE-NEXT: xxspltd v2
3664; P9BE-NEXT: blr
3665; P9LE: xscvdpsxds
3666; P9LE-NEXT: xxspltd v2
3667; P9LE-NEXT: blr
3668; P8BE: xscvdpsxds
3669; P8BE-NEXT: xxspltd v2
3670; P8BE-NEXT: blr
3671; P8LE: xscvdpsxds
3672; P8LE-NEXT: xxspltd v2
3673; P8LE-NEXT: blr
3674}
3675
3676; Function Attrs: norecurse nounwind readonly
3677define <2 x i64> @spltMemValConvftoll(float* nocapture readonly %ptr) {
3678entry:
3679 %0 = load float, float* %ptr, align 4
3680 %conv = fptosi float %0 to i64
3681 %splat.splatinsert = insertelement <2 x i64> undef, i64 %conv, i32 0
3682 %splat.splat = shufflevector <2 x i64> %splat.splatinsert, <2 x i64> undef, <2 x i32> zeroinitializer
3683 ret <2 x i64> %splat.splat
3684; P9BE-LABEL: spltMemValConvftoll
3685; P9LE-LABEL: spltMemValConvftoll
3686; P8BE-LABEL: spltMemValConvftoll
3687; P8LE-LABEL: spltMemValConvftoll
3688; P9BE: lfs
3689; P9BE-NEXT: xscvdpsxds
3690; P9BE-NEXT: xxspltd v2
3691; P9BE-NEXT: blr
3692; P9LE: lfs
3693; P9LE-NEXT: xscvdpsxds
3694; P9LE-NEXT: xxspltd v2
3695; P9LE-NEXT: blr
3696; P8BE: lxsspx
3697; P8BE-NEXT: xscvdpsxds
3698; P8BE-NEXT: xxspltd v2
3699; P8BE-NEXT: blr
3700; P8LE: lxsspx
3701; P8LE-NEXT: xscvdpsxds
3702; P8LE-NEXT: xxspltd v2
3703; P8LE-NEXT: blr
3704}
3705
3706; Function Attrs: norecurse nounwind readnone
3707define <2 x i64> @spltCnstConvdtoll() {
3708entry:
3709 ret <2 x i64> <i64 4, i64 4>
3710; P9BE-LABEL: spltCnstConvdtoll
3711; P9LE-LABEL: spltCnstConvdtoll
3712; P8BE-LABEL: spltCnstConvdtoll
3713; P8LE-LABEL: spltCnstConvdtoll
3714; P9BE: lxvx
3715; P9BE: blr
3716; P9LE: lxvx
3717; P9LE: blr
3718; P8BE: lxvd2x
3719; P8BE: blr
3720; P8LE: lxvd2x
3721; P8LE: blr
3722}
3723
3724; Function Attrs: norecurse nounwind readnone
3725define <2 x i64> @fromRegsConvdtoll(double %a, double %b) {
3726entry:
3727 %conv = fptosi double %a to i64
3728 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
3729 %conv1 = fptosi double %b to i64
3730 %vecinit2 = insertelement <2 x i64> %vecinit, i64 %conv1, i32 1
3731 ret <2 x i64> %vecinit2
3732; P9BE-LABEL: fromRegsConvdtoll
3733; P9LE-LABEL: fromRegsConvdtoll
3734; P8BE-LABEL: fromRegsConvdtoll
3735; P8LE-LABEL: fromRegsConvdtoll
3736; P9BE: xxmrghd
3737; P9BE-NEXT: xvcvdpsxds
3738; P9BE-NEXT: blr
3739; P9LE: xxmrghd
3740; P9LE-NEXT: xvcvdpsxds
3741; P9LE-NEXT: blr
3742; P8BE: xxmrghd
3743; P8BE-NEXT: xvcvdpsxds
3744; P8BE-NEXT: blr
3745; P8LE: xxmrghd
3746; P8LE-NEXT: xvcvdpsxds
3747; P8LE-NEXT: blr
3748}
3749
3750; Function Attrs: norecurse nounwind readnone
3751define <2 x i64> @fromDiffConstsConvdtoll() {
3752entry:
3753 ret <2 x i64> <i64 24, i64 234>
3754; P9BE-LABEL: fromDiffConstsConvdtoll
3755; P9LE-LABEL: fromDiffConstsConvdtoll
3756; P8BE-LABEL: fromDiffConstsConvdtoll
3757; P8LE-LABEL: fromDiffConstsConvdtoll
3758; P9BE: lxvx
3759; P9BE: blr
3760; P9LE: lxvx
3761; P9LE: blr
3762; P8BE: lxvd2x
3763; P8BE: blr
3764; P8LE: lxvd2x
3765; P8LE: blr
3766}
3767
3768; Function Attrs: norecurse nounwind readonly
3769define <2 x i64> @fromDiffMemConsAConvdtoll(double* nocapture readonly %ptr) {
3770entry:
3771 %0 = bitcast double* %ptr to <2 x double>*
3772 %1 = load <2 x double>, <2 x double>* %0, align 8
3773 %2 = fptosi <2 x double> %1 to <2 x i64>
3774 ret <2 x i64> %2
3775; P9BE-LABEL: fromDiffMemConsAConvdtoll
3776; P9LE-LABEL: fromDiffMemConsAConvdtoll
3777; P8BE-LABEL: fromDiffMemConsAConvdtoll
3778; P8LE-LABEL: fromDiffMemConsAConvdtoll
3779; P9BE: lxvx
3780; P9BE-NEXT: xvcvdpsxds v2
3781; P9BE-NEXT: blr
3782; P9LE: lxvx
3783; P9LE-NEXT: xvcvdpsxds v2
3784; P9LE-NEXT: blr
3785; P8BE: lxvd2x
3786; P8BE-NEXT: xvcvdpsxds v2
3787; P8BE-NEXT: blr
3788; P8LE: lxvd2x
3789; P8LE: xxswapd
3790; P8LE-NEXT: xvcvdpsxds v2
3791; P8LE-NEXT: blr
3792}
3793
3794; Function Attrs: norecurse nounwind readonly
3795define <2 x i64> @fromDiffMemConsDConvdtoll(double* nocapture readonly %ptr) {
3796entry:
3797 %arrayidx = getelementptr inbounds double, double* %ptr, i64 3
3798 %0 = load double, double* %arrayidx, align 8
3799 %conv = fptosi double %0 to i64
3800 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
3801 %arrayidx1 = getelementptr inbounds double, double* %ptr, i64 2
3802 %1 = load double, double* %arrayidx1, align 8
3803 %conv2 = fptosi double %1 to i64
3804 %vecinit3 = insertelement <2 x i64> %vecinit, i64 %conv2, i32 1
3805 ret <2 x i64> %vecinit3
3806; P9BE-LABEL: fromDiffMemConsDConvdtoll
3807; P9LE-LABEL: fromDiffMemConsDConvdtoll
3808; P8BE-LABEL: fromDiffMemConsDConvdtoll
3809; P8LE-LABEL: fromDiffMemConsDConvdtoll
3810; P9BE: lxvx
3811; P9BE-NEXT: xxswapd
3812; P9BE-NEXT: xvcvdpsxds v2
3813; P9BE-NEXT: blr
3814; P9LE: lxvx
3815; P9LE-NEXT: xxswapd
3816; P9LE-NEXT: xvcvdpsxds v2
3817; P9LE-NEXT: blr
3818; P8BE: lxvd2x
3819; P8BE-NEXT: xxswapd
3820; P8BE-NEXT: xvcvdpsxds v2
3821; P8BE-NEXT: blr
3822; P8LE: lxvd2x
3823; P8LE-NEXT: xvcvdpsxds v2
3824; P8LE-NEXT: blr
3825}
3826
3827; Function Attrs: norecurse nounwind readonly
3828define <2 x i64> @fromDiffMemVarAConvdtoll(double* nocapture readonly %arr, i32 signext %elem) {
3829entry:
3830 %idxprom = sext i32 %elem to i64
3831 %arrayidx = getelementptr inbounds double, double* %arr, i64 %idxprom
3832 %0 = load double, double* %arrayidx, align 8
3833 %conv = fptosi double %0 to i64
3834 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
3835 %add = add nsw i32 %elem, 1
3836 %idxprom1 = sext i32 %add to i64
3837 %arrayidx2 = getelementptr inbounds double, double* %arr, i64 %idxprom1
3838 %1 = load double, double* %arrayidx2, align 8
3839 %conv3 = fptosi double %1 to i64
3840 %vecinit4 = insertelement <2 x i64> %vecinit, i64 %conv3, i32 1
3841 ret <2 x i64> %vecinit4
3842; P9BE-LABEL: fromDiffMemVarAConvdtoll
3843; P9LE-LABEL: fromDiffMemVarAConvdtoll
3844; P8BE-LABEL: fromDiffMemVarAConvdtoll
3845; P8LE-LABEL: fromDiffMemVarAConvdtoll
3846; P9BE: sldi
3847; P9BE: lxvx
3848; P9BE-NEXT: xvcvdpsxds v2
3849; P9BE-NEXT: blr
3850; P9LE: sldi
3851; P9LE: lxvx
3852; P9LE-NEXT: xvcvdpsxds v2
3853; P9LE-NEXT: blr
3854; P8BE: sldi
3855; P8BE: lxvd2x
3856; P8BE-NEXT: xvcvdpsxds v2
3857; P8BE-NEXT: blr
3858; P8LE: sldi
3859; P8LE: lxvd2x
3860; P8LE-NEXT: xxswapd
3861; P8LE-NEXT: xvcvdpsxds v2
3862; P8LE-NEXT: blr
3863}
3864
3865; Function Attrs: norecurse nounwind readonly
3866define <2 x i64> @fromDiffMemVarDConvdtoll(double* nocapture readonly %arr, i32 signext %elem) {
3867entry:
3868 %idxprom = sext i32 %elem to i64
3869 %arrayidx = getelementptr inbounds double, double* %arr, i64 %idxprom
3870 %0 = load double, double* %arrayidx, align 8
3871 %conv = fptosi double %0 to i64
3872 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
3873 %sub = add nsw i32 %elem, -1
3874 %idxprom1 = sext i32 %sub to i64
3875 %arrayidx2 = getelementptr inbounds double, double* %arr, i64 %idxprom1
3876 %1 = load double, double* %arrayidx2, align 8
3877 %conv3 = fptosi double %1 to i64
3878 %vecinit4 = insertelement <2 x i64> %vecinit, i64 %conv3, i32 1
3879 ret <2 x i64> %vecinit4
3880; P9BE-LABEL: fromDiffMemVarDConvdtoll
3881; P9LE-LABEL: fromDiffMemVarDConvdtoll
3882; P8BE-LABEL: fromDiffMemVarDConvdtoll
3883; P8LE-LABEL: fromDiffMemVarDConvdtoll
3884; P9BE: sldi
3885; P9BE: lxvx
3886; P9BE-NEXT: xxswapd
3887; P9BE-NEXT: xvcvdpsxds v2
3888; P9BE-NEXT: blr
3889; P9LE: sldi
3890; P9LE: lxvx
3891; P9LE-NEXT: xxswapd
3892; P9LE-NEXT: xvcvdpsxds v2
3893; P9LE-NEXT: blr
3894; P8BE: sldi
3895; P8BE: lxvd2x
3896; P8BE-NEXT: xxswapd
3897; P8BE-NEXT: xvcvdpsxds v2
3898; P8BE-NEXT: blr
3899; P8LE: sldi
3900; P8LE: lxvd2x
3901; P8LE-NEXT: xvcvdpsxds v2
3902; P8LE-NEXT: blr
3903}
3904
3905; Function Attrs: norecurse nounwind readnone
3906define <2 x i64> @spltRegValConvdtoll(double %val) {
3907entry:
3908 %conv = fptosi double %val to i64
3909 %splat.splatinsert = insertelement <2 x i64> undef, i64 %conv, i32 0
3910 %splat.splat = shufflevector <2 x i64> %splat.splatinsert, <2 x i64> undef, <2 x i32> zeroinitializer
3911 ret <2 x i64> %splat.splat
3912; P9BE-LABEL: spltRegValConvdtoll
3913; P9LE-LABEL: spltRegValConvdtoll
3914; P8BE-LABEL: spltRegValConvdtoll
3915; P8LE-LABEL: spltRegValConvdtoll
3916; P9BE: xscvdpsxds
3917; P9BE-NEXT: xxspltd v2
3918; P9BE-NEXT: blr
3919; P9LE: xscvdpsxds
3920; P9LE-NEXT: xxspltd v2
3921; P9LE-NEXT: blr
3922; P8BE: xscvdpsxds
3923; P8BE-NEXT: xxspltd v2
3924; P8BE-NEXT: blr
3925; P8LE: xscvdpsxds
3926; P8LE-NEXT: xxspltd v2
3927; P8LE-NEXT: blr
3928}
3929
3930; Function Attrs: norecurse nounwind readonly
3931define <2 x i64> @spltMemValConvdtoll(double* nocapture readonly %ptr) {
3932entry:
3933 %0 = load double, double* %ptr, align 8
3934 %conv = fptosi double %0 to i64
3935 %splat.splatinsert = insertelement <2 x i64> undef, i64 %conv, i32 0
3936 %splat.splat = shufflevector <2 x i64> %splat.splatinsert, <2 x i64> undef, <2 x i32> zeroinitializer
3937 ret <2 x i64> %splat.splat
3938; P9BE-LABEL: spltMemValConvdtoll
3939; P9LE-LABEL: spltMemValConvdtoll
3940; P8BE-LABEL: spltMemValConvdtoll
3941; P8LE-LABEL: spltMemValConvdtoll
3942; P9BE: lxvdsx
3943; P9BE-NEXT: xvcvdpsxds
3944; P9BE-NEXT: blr
3945; P9LE: lxvdsx
3946; P9LE-NEXT: xvcvdpsxds
3947; P9LE-NEXT: blr
3948; P8BE: lxvdsx
3949; P8BE-NEXT: xvcvdpsxds
3950; P8BE-NEXT: blr
3951; P8LE: lxvdsx
3952; P8LE-NEXT: xvcvdpsxds
3953; P8LE-NEXT: blr
3954}
3955
3956; Function Attrs: norecurse nounwind readnone
3957define <2 x i64> @allZeroull() {
3958entry:
3959 ret <2 x i64> zeroinitializer
3960; P9BE-LABEL: allZeroull
3961; P9LE-LABEL: allZeroull
3962; P8BE-LABEL: allZeroull
3963; P8LE-LABEL: allZeroull
3964; P9BE: xxlxor v2, v2, v2
3965; P9BE: blr
3966; P9LE: xxlxor v2, v2, v2
3967; P9LE: blr
3968; P8BE: xxlxor v2, v2, v2
3969; P8BE: blr
3970; P8LE: xxlxor v2, v2, v2
3971; P8LE: blr
3972}
3973
3974; Function Attrs: norecurse nounwind readnone
3975define <2 x i64> @allOneull() {
3976entry:
3977 ret <2 x i64> <i64 -1, i64 -1>
3978; P9BE-LABEL: allOneull
3979; P9LE-LABEL: allOneull
3980; P8BE-LABEL: allOneull
3981; P8LE-LABEL: allOneull
3982; P9BE: xxspltib v2, 255
3983; P9BE: blr
3984; P9LE: xxspltib v2, 255
3985; P9LE: blr
3986; P8BE: vspltisb v2, -1
3987; P8BE: blr
3988; P8LE: vspltisb v2, -1
3989; P8LE: blr
3990}
3991
3992; Function Attrs: norecurse nounwind readnone
3993define <2 x i64> @spltConst1ull() {
3994entry:
3995 ret <2 x i64> <i64 1, i64 1>
3996; P9BE-LABEL: spltConst1ull
3997; P9LE-LABEL: spltConst1ull
3998; P8BE-LABEL: spltConst1ull
3999; P8LE-LABEL: spltConst1ull
4000; P9BE: lxvx
4001; P9BE: blr
4002; P9LE: lxvx
4003; P9LE: blr
4004; P8BE: lxvd2x
4005; P8BE: blr
4006; P8LE: lxvd2x
4007; P8LE: blr
4008}
4009
4010; Function Attrs: norecurse nounwind readnone
4011define <2 x i64> @spltConst16kull() {
4012entry:
4013 ret <2 x i64> <i64 32767, i64 32767>
4014; P9BE-LABEL: spltConst16kull
4015; P9LE-LABEL: spltConst16kull
4016; P8BE-LABEL: spltConst16kull
4017; P8LE-LABEL: spltConst16kull
4018; P9BE: lxvx
4019; P9BE: blr
4020; P9LE: lxvx
4021; P9LE: blr
4022; P8BE: lxvd2x
4023; P8BE: blr
4024; P8LE: lxvd2x
4025; P8LE: blr
4026}
4027
4028; Function Attrs: norecurse nounwind readnone
4029define <2 x i64> @spltConst32kull() {
4030entry:
4031 ret <2 x i64> <i64 65535, i64 65535>
4032; P9BE-LABEL: spltConst32kull
4033; P9LE-LABEL: spltConst32kull
4034; P8BE-LABEL: spltConst32kull
4035; P8LE-LABEL: spltConst32kull
4036; P9BE: lxvx
4037; P9BE: blr
4038; P9LE: lxvx
4039; P9LE: blr
4040; P8BE: lxvd2x
4041; P8BE: blr
4042; P8LE: lxvd2x
4043; P8LE: blr
4044}
4045
4046; Function Attrs: norecurse nounwind readnone
4047define <2 x i64> @fromRegsull(i64 %a, i64 %b) {
4048entry:
4049 %vecinit = insertelement <2 x i64> undef, i64 %a, i32 0
4050 %vecinit1 = insertelement <2 x i64> %vecinit, i64 %b, i32 1
4051 ret <2 x i64> %vecinit1
4052; P9BE-LABEL: fromRegsull
4053; P9LE-LABEL: fromRegsull
4054; P8BE-LABEL: fromRegsull
4055; P8LE-LABEL: fromRegsull
4056; P9BE: mtvsrdd v2, r3, r4
4057; P9BE: blr
4058; P9LE: mtvsrdd v2, r4, r3
4059; P9LE: blr
4060; P8BE-DAG: mtvsrd {{[vsf0-9]+}}, r3
4061; P8BE-DAG: mtvsrd {{[vsf0-9]+}}, r4
4062; P8BE: xxmrghd v2
4063; P8BE: blr
4064; P8LE-DAG: mtvsrd {{[vsf0-9]+}}, r3
4065; P8LE-DAG: mtvsrd {{[vsf0-9]+}}, r4
4066; P8LE: xxmrghd v2
4067; P8LE: blr
4068}
4069
4070; Function Attrs: norecurse nounwind readnone
4071define <2 x i64> @fromDiffConstsull() {
4072entry:
4073 ret <2 x i64> <i64 242, i64 -113>
4074; P9BE-LABEL: fromDiffConstsull
4075; P9LE-LABEL: fromDiffConstsull
4076; P8BE-LABEL: fromDiffConstsull
4077; P8LE-LABEL: fromDiffConstsull
4078; P9BE: lxvx
4079; P9BE: blr
4080; P9LE: lxvx
4081; P9LE: blr
4082; P8BE: lxvd2x
4083; P8BE: blr
4084; P8LE: lxvd2x
4085; P8LE: blr
4086}
4087
4088; Function Attrs: norecurse nounwind readonly
4089define <2 x i64> @fromDiffMemConsAull(i64* nocapture readonly %arr) {
4090entry:
4091 %0 = load i64, i64* %arr, align 8
4092 %vecinit = insertelement <2 x i64> undef, i64 %0, i32 0
4093 %arrayidx1 = getelementptr inbounds i64, i64* %arr, i64 1
4094 %1 = load i64, i64* %arrayidx1, align 8
4095 %vecinit2 = insertelement <2 x i64> %vecinit, i64 %1, i32 1
4096 ret <2 x i64> %vecinit2
4097; P9BE-LABEL: fromDiffMemConsAull
4098; P9LE-LABEL: fromDiffMemConsAull
4099; P8BE-LABEL: fromDiffMemConsAull
4100; P8LE-LABEL: fromDiffMemConsAull
4101; P9BE: lxvx v2
4102; P9BE: blr
4103; P9LE: lxvx v2
4104; P9LE: blr
4105; P8BE: lxvd2x v2
4106; P8BE: blr
4107; P8LE: lxvd2x
4108; P8LE: xxswapd v2
4109; P8LE: blr
4110}
4111
4112; Function Attrs: norecurse nounwind readonly
4113define <2 x i64> @fromDiffMemConsDull(i64* nocapture readonly %arr) {
4114entry:
4115 %arrayidx = getelementptr inbounds i64, i64* %arr, i64 3
4116 %0 = load i64, i64* %arrayidx, align 8
4117 %vecinit = insertelement <2 x i64> undef, i64 %0, i32 0
4118 %arrayidx1 = getelementptr inbounds i64, i64* %arr, i64 2
4119 %1 = load i64, i64* %arrayidx1, align 8
4120 %vecinit2 = insertelement <2 x i64> %vecinit, i64 %1, i32 1
4121 ret <2 x i64> %vecinit2
4122; P9BE-LABEL: fromDiffMemConsDull
4123; P9LE-LABEL: fromDiffMemConsDull
4124; P8BE-LABEL: fromDiffMemConsDull
4125; P8LE-LABEL: fromDiffMemConsDull
4126; P9BE: lxvx v2
4127; P9BE: blr
4128; P9LE: lxvx
4129; P9LE: xxswapd v2
4130; P9LE: blr
4131; P8BE: lxvd2x
4132; P8BE: xxswapd v2
4133; P8BE-NEXT: blr
4134; P8LE: lxvd2x v2
4135; P8LE-NEXT: blr
4136}
4137
4138; Function Attrs: norecurse nounwind readonly
4139define <2 x i64> @fromDiffMemVarAull(i64* nocapture readonly %arr, i32 signext %elem) {
4140entry:
4141 %idxprom = sext i32 %elem to i64
4142 %arrayidx = getelementptr inbounds i64, i64* %arr, i64 %idxprom
4143 %0 = load i64, i64* %arrayidx, align 8
4144 %vecinit = insertelement <2 x i64> undef, i64 %0, i32 0
4145 %add = add nsw i32 %elem, 1
4146 %idxprom1 = sext i32 %add to i64
4147 %arrayidx2 = getelementptr inbounds i64, i64* %arr, i64 %idxprom1
4148 %1 = load i64, i64* %arrayidx2, align 8
4149 %vecinit3 = insertelement <2 x i64> %vecinit, i64 %1, i32 1
4150 ret <2 x i64> %vecinit3
4151; P9BE-LABEL: fromDiffMemVarAull
4152; P9LE-LABEL: fromDiffMemVarAull
4153; P8BE-LABEL: fromDiffMemVarAull
4154; P8LE-LABEL: fromDiffMemVarAull
4155; P9BE: sldi
4156; P9BE: lxvx v2
4157; P9BE-NEXT: blr
4158; P9LE: sldi
4159; P9LE: lxvx v2
4160; P9LE-NEXT: blr
4161; P8BE: sldi
4162; P8BE: lxvd2x v2
4163; P8BE-NEXT: blr
4164; P8LE: sldi
4165; P8LE: lxvd2x
4166; P8LE: xxswapd v2
4167; P8LE-NEXT: blr
4168}
4169
4170; Function Attrs: norecurse nounwind readonly
4171define <2 x i64> @fromDiffMemVarDull(i64* nocapture readonly %arr, i32 signext %elem) {
4172entry:
4173 %idxprom = sext i32 %elem to i64
4174 %arrayidx = getelementptr inbounds i64, i64* %arr, i64 %idxprom
4175 %0 = load i64, i64* %arrayidx, align 8
4176 %vecinit = insertelement <2 x i64> undef, i64 %0, i32 0
4177 %sub = add nsw i32 %elem, -1
4178 %idxprom1 = sext i32 %sub to i64
4179 %arrayidx2 = getelementptr inbounds i64, i64* %arr, i64 %idxprom1
4180 %1 = load i64, i64* %arrayidx2, align 8
4181 %vecinit3 = insertelement <2 x i64> %vecinit, i64 %1, i32 1
4182 ret <2 x i64> %vecinit3
4183; P9BE-LABEL: fromDiffMemVarDull
4184; P9LE-LABEL: fromDiffMemVarDull
4185; P8BE-LABEL: fromDiffMemVarDull
4186; P8LE-LABEL: fromDiffMemVarDull
4187; P9BE: sldi
4188; P9BE: lxvx
4189; P9BE: xxswapd v2
4190; P9BE-NEXT: blr
4191; P9LE: sldi
4192; P9LE: lxvx
4193; P9LE: xxswapd v2
4194; P9LE-NEXT: blr
4195; P8BE: sldi
4196; P8BE: lxvd2x
4197; P8BE: xxswapd v2
4198; P8BE-NEXT: blr
4199; P8LE: sldi
4200; P8LE: lxvd2x v2
4201; P8LE-NEXT: blr
4202}
4203
4204; Function Attrs: norecurse nounwind readonly
4205define <2 x i64> @fromRandMemConsull(i64* nocapture readonly %arr) {
4206entry:
4207 %arrayidx = getelementptr inbounds i64, i64* %arr, i64 4
4208 %0 = load i64, i64* %arrayidx, align 8
4209 %vecinit = insertelement <2 x i64> undef, i64 %0, i32 0
4210 %arrayidx1 = getelementptr inbounds i64, i64* %arr, i64 18
4211 %1 = load i64, i64* %arrayidx1, align 8
4212 %vecinit2 = insertelement <2 x i64> %vecinit, i64 %1, i32 1
4213 ret <2 x i64> %vecinit2
4214; P9BE-LABEL: fromRandMemConsull
4215; P9LE-LABEL: fromRandMemConsull
4216; P8BE-LABEL: fromRandMemConsull
4217; P8LE-LABEL: fromRandMemConsull
4218; P9BE: ld
4219; P9BE: ld
4220; P9BE: mtvsrdd v2
4221; P9BE-NEXT: blr
4222; P9LE: ld
4223; P9LE: ld
4224; P9LE: mtvsrdd v2
4225; P9LE-NEXT: blr
4226; P8BE: ld
4227; P8BE: ld
4228; P8BE-DAG: mtvsrd
4229; P8BE-DAG: mtvsrd
4230; P8BE: xxmrghd v2
4231; P8BE-NEXT: blr
4232; P8LE: ld
4233; P8LE: ld
4234; P8LE-DAG: mtvsrd
4235; P8LE-DAG: mtvsrd
4236; P8LE: xxmrghd v2
4237; P8LE-NEXT: blr
4238}
4239
4240; Function Attrs: norecurse nounwind readonly
4241define <2 x i64> @fromRandMemVarull(i64* nocapture readonly %arr, i32 signext %elem) {
4242entry:
4243 %add = add nsw i32 %elem, 4
4244 %idxprom = sext i32 %add to i64
4245 %arrayidx = getelementptr inbounds i64, i64* %arr, i64 %idxprom
4246 %0 = load i64, i64* %arrayidx, align 8
4247 %vecinit = insertelement <2 x i64> undef, i64 %0, i32 0
4248 %add1 = add nsw i32 %elem, 1
4249 %idxprom2 = sext i32 %add1 to i64
4250 %arrayidx3 = getelementptr inbounds i64, i64* %arr, i64 %idxprom2
4251 %1 = load i64, i64* %arrayidx3, align 8
4252 %vecinit4 = insertelement <2 x i64> %vecinit, i64 %1, i32 1
4253 ret <2 x i64> %vecinit4
4254; P9BE-LABEL: fromRandMemVarull
4255; P9LE-LABEL: fromRandMemVarull
4256; P8BE-LABEL: fromRandMemVarull
4257; P8LE-LABEL: fromRandMemVarull
4258; P9BE: sldi
4259; P9BE: ld
4260; P9BE: ld
4261; P9BE: mtvsrdd v2
4262; P9BE-NEXT: blr
4263; P9LE: sldi
4264; P9LE: ld
4265; P9LE: ld
4266; P9LE: mtvsrdd v2
4267; P9LE-NEXT: blr
4268; P8BE: sldi
4269; P8BE: ld
4270; P8BE: ld
4271; P8BE: mtvsrd
4272; P8BE: mtvsrd
4273; P8BE: xxmrghd v2
4274; P8BE-NEXT: blr
4275; P8LE: sldi
4276; P8LE: ld
4277; P8LE: ld
4278; P8LE: mtvsrd
4279; P8LE: mtvsrd
4280; P8LE: xxmrghd v2
4281; P8LE-NEXT: blr
4282}
4283
4284; Function Attrs: norecurse nounwind readnone
4285define <2 x i64> @spltRegValull(i64 %val) {
4286entry:
4287 %splat.splatinsert = insertelement <2 x i64> undef, i64 %val, i32 0
4288 %splat.splat = shufflevector <2 x i64> %splat.splatinsert, <2 x i64> undef, <2 x i32> zeroinitializer
4289 ret <2 x i64> %splat.splat
4290; P9BE-LABEL: spltRegValull
4291; P9LE-LABEL: spltRegValull
4292; P8BE-LABEL: spltRegValull
4293; P8LE-LABEL: spltRegValull
4294; P9BE: mtvsrdd v2, r3, r3
4295; P9BE-NEXT: blr
4296; P9LE: mtvsrdd v2, r3, r3
4297; P9LE-NEXT: blr
4298; P8BE: mtvsrd {{[vsf]+}}[[REG1:[0-9]+]], r3
4299; P8BE: xxspltd v2, {{[vsf]+}}[[REG1]], 0
4300; P8BE-NEXT: blr
4301; P8LE: mtvsrd {{[vsf]+}}[[REG1:[0-9]+]], r3
4302; P8LE: xxspltd v2, {{[vsf]+}}[[REG1]], 0
4303; P8LE-NEXT: blr
4304}
4305
4306; Function Attrs: norecurse nounwind readonly
4307define <2 x i64> @spltMemValull(i64* nocapture readonly %ptr) {
4308entry:
4309 %0 = load i64, i64* %ptr, align 8
4310 %splat.splatinsert = insertelement <2 x i64> undef, i64 %0, i32 0
4311 %splat.splat = shufflevector <2 x i64> %splat.splatinsert, <2 x i64> undef, <2 x i32> zeroinitializer
4312 ret <2 x i64> %splat.splat
4313; P9BE-LABEL: spltMemValull
4314; P9LE-LABEL: spltMemValull
4315; P8BE-LABEL: spltMemValull
4316; P8LE-LABEL: spltMemValull
4317; P9BE: lxvdsx v2
4318; P9BE-NEXT: blr
4319; P9LE: lxvdsx v2
4320; P9LE-NEXT: blr
4321; P8BE: lxvdsx v2
4322; P8BE-NEXT: blr
4323; P8LE: lxvdsx v2
4324; P8LE-NEXT: blr
4325}
4326
4327; Function Attrs: norecurse nounwind readnone
4328define <2 x i64> @spltCnstConvftoull() {
4329entry:
4330 ret <2 x i64> <i64 4, i64 4>
4331; P9BE-LABEL: spltCnstConvftoull
4332; P9LE-LABEL: spltCnstConvftoull
4333; P8BE-LABEL: spltCnstConvftoull
4334; P8LE-LABEL: spltCnstConvftoull
4335; P9BE: lxvx
4336; P9BE: blr
4337; P9LE: lxvx
4338; P9LE: blr
4339; P8BE: lxvd2x
4340; P8BE: blr
4341; P8LE: lxvd2x
4342; P8LE: blr
4343}
4344
4345; Function Attrs: norecurse nounwind readnone
4346define <2 x i64> @fromRegsConvftoull(float %a, float %b) {
4347entry:
4348 %conv = fptoui float %a to i64
4349 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
4350 %conv1 = fptoui float %b to i64
4351 %vecinit2 = insertelement <2 x i64> %vecinit, i64 %conv1, i32 1
4352 ret <2 x i64> %vecinit2
4353; P9BE-LABEL: fromRegsConvftoull
4354; P9LE-LABEL: fromRegsConvftoull
4355; P8BE-LABEL: fromRegsConvftoull
4356; P8LE-LABEL: fromRegsConvftoull
4357; P9BE: xxmrghd
4358; P9BE: xvcvdpuxds v2
4359; P9BE-NEXT: blr
4360; P9LE: xxmrghd
4361; P9LE: xvcvdpuxds v2
4362; P9LE-NEXT: blr
4363; P8BE: xxmrghd
4364; P8BE: xvcvdpuxds v2
4365; P8BE-NEXT: blr
4366; P8LE: xxmrghd
4367; P8LE: xvcvdpuxds v2
4368; P8LE-NEXT: blr
4369}
4370
4371; Function Attrs: norecurse nounwind readnone
4372define <2 x i64> @fromDiffConstsConvftoull() {
4373entry:
4374 ret <2 x i64> <i64 24, i64 234>
4375; P9BE-LABEL: fromDiffConstsConvftoull
4376; P9LE-LABEL: fromDiffConstsConvftoull
4377; P8BE-LABEL: fromDiffConstsConvftoull
4378; P8LE-LABEL: fromDiffConstsConvftoull
4379; P9BE: lxvx v2
4380; P9BE: blr
4381; P9LE: lxvx v2
4382; P9LE: blr
4383; P8BE: lxvd2x v2
4384; P8BE: blr
4385; P8LE: lxvd2x
4386; P8LE: xxswapd v2
4387; P8LE: blr
4388}
4389
4390; Function Attrs: norecurse nounwind readonly
4391define <2 x i64> @fromDiffMemConsAConvftoull(float* nocapture readonly %ptr) {
4392entry:
4393 %0 = load float, float* %ptr, align 4
4394 %conv = fptoui float %0 to i64
4395 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
4396 %arrayidx1 = getelementptr inbounds float, float* %ptr, i64 1
4397 %1 = load float, float* %arrayidx1, align 4
4398 %conv2 = fptoui float %1 to i64
4399 %vecinit3 = insertelement <2 x i64> %vecinit, i64 %conv2, i32 1
4400 ret <2 x i64> %vecinit3
4401; P9BE-LABEL: fromDiffMemConsAConvftoull
4402; P9LE-LABEL: fromDiffMemConsAConvftoull
4403; P8BE-LABEL: fromDiffMemConsAConvftoull
4404; P8LE-LABEL: fromDiffMemConsAConvftoull
4405; P9BE: lfs
4406; P9BE: lfs
4407; P9BE: xxmrghd
4408; P9BE-NEXT: xvcvdpuxds v2
4409; P9BE-NEXT: blr
4410; P9LE: lfs
4411; P9LE: lfs
4412; P9LE: xxmrghd
4413; P9LE-NEXT: xvcvdpuxds v2
4414; P9LE-NEXT: blr
4415; P8BE: lxsspx
4416; P8BE: lxsspx
4417; P8BE: xxmrghd
4418; P8BE-NEXT: xvcvdpuxds v2
4419; P8BE-NEXT: blr
4420; P8LE: lxsspx
4421; P8LE: lxsspx
4422; P8LE: xxmrghd
4423; P8LE-NEXT: xvcvdpuxds v2
4424; P8LE-NEXT: blr
4425}
4426
4427; Function Attrs: norecurse nounwind readonly
4428define <2 x i64> @fromDiffMemConsDConvftoull(float* nocapture readonly %ptr) {
4429entry:
4430 %arrayidx = getelementptr inbounds float, float* %ptr, i64 3
4431 %0 = load float, float* %arrayidx, align 4
4432 %conv = fptoui float %0 to i64
4433 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
4434 %arrayidx1 = getelementptr inbounds float, float* %ptr, i64 2
4435 %1 = load float, float* %arrayidx1, align 4
4436 %conv2 = fptoui float %1 to i64
4437 %vecinit3 = insertelement <2 x i64> %vecinit, i64 %conv2, i32 1
4438 ret <2 x i64> %vecinit3
4439; P9BE-LABEL: fromDiffMemConsDConvftoull
4440; P9LE-LABEL: fromDiffMemConsDConvftoull
4441; P8BE-LABEL: fromDiffMemConsDConvftoull
4442; P8LE-LABEL: fromDiffMemConsDConvftoull
4443; P9BE: lfs
4444; P9BE: lfs
4445; P9BE: xxmrghd
4446; P9BE-NEXT: xvcvdpuxds v2
4447; P9BE-NEXT: blr
4448; P9LE: lfs
4449; P9LE: lfs
4450; P9LE: xxmrghd
4451; P9LE-NEXT: xvcvdpuxds v2
4452; P9LE-NEXT: blr
4453; P8BE: lxsspx
4454; P8BE: lxsspx
4455; P8BE: xxmrghd
4456; P8BE-NEXT: xvcvdpuxds v2
4457; P8BE-NEXT: blr
4458; P8LE: lxsspx
4459; P8LE: lxsspx
4460; P8LE: xxmrghd
4461; P8LE-NEXT: xvcvdpuxds v2
4462; P8LE-NEXT: blr
4463}
4464
4465; Function Attrs: norecurse nounwind readonly
4466define <2 x i64> @fromDiffMemVarAConvftoull(float* nocapture readonly %arr, i32 signext %elem) {
4467entry:
4468 %idxprom = sext i32 %elem to i64
4469 %arrayidx = getelementptr inbounds float, float* %arr, i64 %idxprom
4470 %0 = load float, float* %arrayidx, align 4
4471 %conv = fptoui float %0 to i64
4472 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
4473 %add = add nsw i32 %elem, 1
4474 %idxprom1 = sext i32 %add to i64
4475 %arrayidx2 = getelementptr inbounds float, float* %arr, i64 %idxprom1
4476 %1 = load float, float* %arrayidx2, align 4
4477 %conv3 = fptoui float %1 to i64
4478 %vecinit4 = insertelement <2 x i64> %vecinit, i64 %conv3, i32 1
4479 ret <2 x i64> %vecinit4
4480; P9BE-LABEL: fromDiffMemVarAConvftoull
4481; P9LE-LABEL: fromDiffMemVarAConvftoull
4482; P8BE-LABEL: fromDiffMemVarAConvftoull
4483; P8LE-LABEL: fromDiffMemVarAConvftoull
4484; P9BE: sldi
4485; P9BE: lfsux
4486; P9BE: lfs
4487; P9BE: xxmrghd
4488; P9BE-NEXT: xvcvdpuxds v2
4489; P9BE-NEXT: blr
4490; P9LE: sldi
4491; P9LE: lfsux
4492; P9LE: lfs
4493; P9LE: xxmrghd
4494; P9LE-NEXT: xvcvdpuxds v2
4495; P9LE-NEXT: blr
4496; P8BE: sldi
4497; P8BE: lfsux
4498; P8BE: lxsspx
4499; P8BE: xxmrghd
4500; P8BE-NEXT: xvcvdpuxds v2
4501; P8BE-NEXT: blr
4502; P8LE: sldi
4503; P8LE: lfsux
4504; P8LE: lxsspx
4505; P8LE: xxmrghd
4506; P8LE-NEXT: xvcvdpuxds v2
4507; P8LE-NEXT: blr
4508}
4509
4510; Function Attrs: norecurse nounwind readonly
4511define <2 x i64> @fromDiffMemVarDConvftoull(float* nocapture readonly %arr, i32 signext %elem) {
4512entry:
4513 %idxprom = sext i32 %elem to i64
4514 %arrayidx = getelementptr inbounds float, float* %arr, i64 %idxprom
4515 %0 = load float, float* %arrayidx, align 4
4516 %conv = fptoui float %0 to i64
4517 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
4518 %sub = add nsw i32 %elem, -1
4519 %idxprom1 = sext i32 %sub to i64
4520 %arrayidx2 = getelementptr inbounds float, float* %arr, i64 %idxprom1
4521 %1 = load float, float* %arrayidx2, align 4
4522 %conv3 = fptoui float %1 to i64
4523 %vecinit4 = insertelement <2 x i64> %vecinit, i64 %conv3, i32 1
4524 ret <2 x i64> %vecinit4
4525; P9BE-LABEL: fromDiffMemVarDConvftoull
4526; P9LE-LABEL: fromDiffMemVarDConvftoull
4527; P8BE-LABEL: fromDiffMemVarDConvftoull
4528; P8LE-LABEL: fromDiffMemVarDConvftoull
4529; P9BE: sldi
4530; P9BE: lfsux
4531; P9BE: lfs
4532; P9BE: xxmrghd
4533; P9BE-NEXT: xvcvdpuxds v2
4534; P9BE-NEXT: blr
4535; P9LE: sldi
4536; P9LE: lfsux
4537; P9LE: lfs
4538; P9LE: xxmrghd
4539; P9LE-NEXT: xvcvdpuxds v2
4540; P9LE-NEXT: blr
4541; P8BE: sldi
4542; P8BE: lfsux
4543; P8BE: lxsspx
4544; P8BE: xxmrghd
4545; P8BE-NEXT: xvcvdpuxds v2
4546; P8BE-NEXT: blr
4547; P8LE: sldi
4548; P8LE: lfsux
4549; P8LE: lxsspx
4550; P8LE: xxmrghd
4551; P8LE-NEXT: xvcvdpuxds v2
4552; P8LE-NEXT: blr
4553}
4554
4555; Function Attrs: norecurse nounwind readnone
4556define <2 x i64> @spltRegValConvftoull(float %val) {
4557entry:
4558 %conv = fptoui float %val to i64
4559 %splat.splatinsert = insertelement <2 x i64> undef, i64 %conv, i32 0
4560 %splat.splat = shufflevector <2 x i64> %splat.splatinsert, <2 x i64> undef, <2 x i32> zeroinitializer
4561 ret <2 x i64> %splat.splat
4562; P9BE-LABEL: spltRegValConvftoull
4563; P9LE-LABEL: spltRegValConvftoull
4564; P8BE-LABEL: spltRegValConvftoull
4565; P8LE-LABEL: spltRegValConvftoull
4566; P9BE: xscvdpuxds
4567; P9BE-NEXT: xxspltd v2
4568; P9BE-NEXT: blr
4569; P9LE: xscvdpuxds
4570; P9LE-NEXT: xxspltd v2
4571; P9LE-NEXT: blr
4572; P8BE: xscvdpuxds
4573; P8BE-NEXT: xxspltd v2
4574; P8BE-NEXT: blr
4575; P8LE: xscvdpuxds
4576; P8LE-NEXT: xxspltd v2
4577; P8LE-NEXT: blr
4578}
4579
4580; Function Attrs: norecurse nounwind readonly
4581define <2 x i64> @spltMemValConvftoull(float* nocapture readonly %ptr) {
4582entry:
4583 %0 = load float, float* %ptr, align 4
4584 %conv = fptoui float %0 to i64
4585 %splat.splatinsert = insertelement <2 x i64> undef, i64 %conv, i32 0
4586 %splat.splat = shufflevector <2 x i64> %splat.splatinsert, <2 x i64> undef, <2 x i32> zeroinitializer
4587 ret <2 x i64> %splat.splat
4588; P9BE-LABEL: spltMemValConvftoull
4589; P9LE-LABEL: spltMemValConvftoull
4590; P8BE-LABEL: spltMemValConvftoull
4591; P8LE-LABEL: spltMemValConvftoull
4592; P9BE: lfs
4593; P9BE-NEXT: xscvdpuxds
4594; P9BE-NEXT: xxspltd v2
4595; P9BE-NEXT: blr
4596; P9LE: lfs
4597; P9LE-NEXT: xscvdpuxds
4598; P9LE-NEXT: xxspltd v2
4599; P9LE-NEXT: blr
4600; P8BE: lxsspx
4601; P8BE-NEXT: xscvdpuxds
4602; P8BE-NEXT: xxspltd v2
4603; P8BE-NEXT: blr
4604; P8LE: lxsspx
4605; P8LE-NEXT: xscvdpuxds
4606; P8LE-NEXT: xxspltd v2
4607; P8LE-NEXT: blr
4608}
4609
4610; Function Attrs: norecurse nounwind readnone
4611define <2 x i64> @spltCnstConvdtoull() {
4612entry:
4613 ret <2 x i64> <i64 4, i64 4>
4614; P9BE-LABEL: spltCnstConvdtoull
4615; P9LE-LABEL: spltCnstConvdtoull
4616; P8BE-LABEL: spltCnstConvdtoull
4617; P8LE-LABEL: spltCnstConvdtoull
4618; P9BE: lxvx
4619; P9BE: blr
4620; P9LE: lxvx
4621; P9LE: blr
4622; P8BE: lxvd2x
4623; P8BE: blr
4624; P8LE: lxvd2x
4625; P8LE: blr
4626}
4627
4628; Function Attrs: norecurse nounwind readnone
4629define <2 x i64> @fromRegsConvdtoull(double %a, double %b) {
4630entry:
4631 %conv = fptoui double %a to i64
4632 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
4633 %conv1 = fptoui double %b to i64
4634 %vecinit2 = insertelement <2 x i64> %vecinit, i64 %conv1, i32 1
4635 ret <2 x i64> %vecinit2
4636; P9BE-LABEL: fromRegsConvdtoull
4637; P9LE-LABEL: fromRegsConvdtoull
4638; P8BE-LABEL: fromRegsConvdtoull
4639; P8LE-LABEL: fromRegsConvdtoull
4640; P9BE: xxmrghd
4641; P9BE-NEXT: xvcvdpuxds
4642; P9BE-NEXT: blr
4643; P9LE: xxmrghd
4644; P9LE-NEXT: xvcvdpuxds
4645; P9LE-NEXT: blr
4646; P8BE: xxmrghd
4647; P8BE-NEXT: xvcvdpuxds
4648; P8BE-NEXT: blr
4649; P8LE: xxmrghd
4650; P8LE-NEXT: xvcvdpuxds
4651; P8LE-NEXT: blr
4652}
4653
4654; Function Attrs: norecurse nounwind readnone
4655define <2 x i64> @fromDiffConstsConvdtoull() {
4656entry:
4657 ret <2 x i64> <i64 24, i64 234>
4658; P9BE-LABEL: fromDiffConstsConvdtoull
4659; P9LE-LABEL: fromDiffConstsConvdtoull
4660; P8BE-LABEL: fromDiffConstsConvdtoull
4661; P8LE-LABEL: fromDiffConstsConvdtoull
4662; P9BE: lxvx
4663; P9BE: blr
4664; P9LE: lxvx
4665; P9LE: blr
4666; P8BE: lxvd2x
4667; P8BE: blr
4668; P8LE: lxvd2x
4669; P8LE: blr
4670}
4671
4672; Function Attrs: norecurse nounwind readonly
4673define <2 x i64> @fromDiffMemConsAConvdtoull(double* nocapture readonly %ptr) {
4674entry:
4675 %0 = bitcast double* %ptr to <2 x double>*
4676 %1 = load <2 x double>, <2 x double>* %0, align 8
4677 %2 = fptoui <2 x double> %1 to <2 x i64>
4678 ret <2 x i64> %2
4679; P9BE-LABEL: fromDiffMemConsAConvdtoull
4680; P9LE-LABEL: fromDiffMemConsAConvdtoull
4681; P8BE-LABEL: fromDiffMemConsAConvdtoull
4682; P8LE-LABEL: fromDiffMemConsAConvdtoull
4683; P9BE: lxvx
4684; P9BE-NEXT: xvcvdpuxds v2
4685; P9BE-NEXT: blr
4686; P9LE: lxvx
4687; P9LE-NEXT: xvcvdpuxds v2
4688; P9LE-NEXT: blr
4689; P8BE: lxvd2x
4690; P8BE-NEXT: xvcvdpuxds v2
4691; P8BE-NEXT: blr
4692; P8LE: lxvd2x
4693; P8LE: xxswapd
4694; P8LE-NEXT: xvcvdpuxds v2
4695; P8LE-NEXT: blr
4696}
4697
4698; Function Attrs: norecurse nounwind readonly
4699define <2 x i64> @fromDiffMemConsDConvdtoull(double* nocapture readonly %ptr) {
4700entry:
4701 %arrayidx = getelementptr inbounds double, double* %ptr, i64 3
4702 %0 = load double, double* %arrayidx, align 8
4703 %conv = fptoui double %0 to i64
4704 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
4705 %arrayidx1 = getelementptr inbounds double, double* %ptr, i64 2
4706 %1 = load double, double* %arrayidx1, align 8
4707 %conv2 = fptoui double %1 to i64
4708 %vecinit3 = insertelement <2 x i64> %vecinit, i64 %conv2, i32 1
4709 ret <2 x i64> %vecinit3
4710; P9BE-LABEL: fromDiffMemConsDConvdtoull
4711; P9LE-LABEL: fromDiffMemConsDConvdtoull
4712; P8BE-LABEL: fromDiffMemConsDConvdtoull
4713; P8LE-LABEL: fromDiffMemConsDConvdtoull
4714; P9BE: lxvx
4715; P9BE-NEXT: xxswapd
4716; P9BE-NEXT: xvcvdpuxds v2
4717; P9BE-NEXT: blr
4718; P9LE: lxvx
4719; P9LE-NEXT: xxswapd
4720; P9LE-NEXT: xvcvdpuxds v2
4721; P9LE-NEXT: blr
4722; P8BE: lxvd2x
4723; P8BE-NEXT: xxswapd
4724; P8BE-NEXT: xvcvdpuxds v2
4725; P8BE-NEXT: blr
4726; P8LE: lxvd2x
4727; P8LE-NEXT: xvcvdpuxds v2
4728; P8LE-NEXT: blr
4729}
4730
4731; Function Attrs: norecurse nounwind readonly
4732define <2 x i64> @fromDiffMemVarAConvdtoull(double* nocapture readonly %arr, i32 signext %elem) {
4733entry:
4734 %idxprom = sext i32 %elem to i64
4735 %arrayidx = getelementptr inbounds double, double* %arr, i64 %idxprom
4736 %0 = load double, double* %arrayidx, align 8
4737 %conv = fptoui double %0 to i64
4738 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
4739 %add = add nsw i32 %elem, 1
4740 %idxprom1 = sext i32 %add to i64
4741 %arrayidx2 = getelementptr inbounds double, double* %arr, i64 %idxprom1
4742 %1 = load double, double* %arrayidx2, align 8
4743 %conv3 = fptoui double %1 to i64
4744 %vecinit4 = insertelement <2 x i64> %vecinit, i64 %conv3, i32 1
4745 ret <2 x i64> %vecinit4
4746; P9BE-LABEL: fromDiffMemVarAConvdtoull
4747; P9LE-LABEL: fromDiffMemVarAConvdtoull
4748; P8BE-LABEL: fromDiffMemVarAConvdtoull
4749; P8LE-LABEL: fromDiffMemVarAConvdtoull
4750; P9BE: sldi
4751; P9BE: lxvx
4752; P9BE-NEXT: xvcvdpuxds v2
4753; P9BE-NEXT: blr
4754; P9LE: sldi
4755; P9LE: lxvx
4756; P9LE-NEXT: xvcvdpuxds v2
4757; P9LE-NEXT: blr
4758; P8BE: sldi
4759; P8BE: lxvd2x
4760; P8BE-NEXT: xvcvdpuxds v2
4761; P8BE-NEXT: blr
4762; P8LE: sldi
4763; P8LE: lxvd2x
4764; P8LE-NEXT: xxswapd
4765; P8LE-NEXT: xvcvdpuxds v2
4766; P8LE-NEXT: blr
4767}
4768
4769; Function Attrs: norecurse nounwind readonly
4770define <2 x i64> @fromDiffMemVarDConvdtoull(double* nocapture readonly %arr, i32 signext %elem) {
4771entry:
4772 %idxprom = sext i32 %elem to i64
4773 %arrayidx = getelementptr inbounds double, double* %arr, i64 %idxprom
4774 %0 = load double, double* %arrayidx, align 8
4775 %conv = fptoui double %0 to i64
4776 %vecinit = insertelement <2 x i64> undef, i64 %conv, i32 0
4777 %sub = add nsw i32 %elem, -1
4778 %idxprom1 = sext i32 %sub to i64
4779 %arrayidx2 = getelementptr inbounds double, double* %arr, i64 %idxprom1
4780 %1 = load double, double* %arrayidx2, align 8
4781 %conv3 = fptoui double %1 to i64
4782 %vecinit4 = insertelement <2 x i64> %vecinit, i64 %conv3, i32 1
4783 ret <2 x i64> %vecinit4
4784; P9BE-LABEL: fromDiffMemVarDConvdtoull
4785; P9LE-LABEL: fromDiffMemVarDConvdtoull
4786; P8BE-LABEL: fromDiffMemVarDConvdtoull
4787; P8LE-LABEL: fromDiffMemVarDConvdtoull
4788; P9BE: sldi
4789; P9BE: lxvx
4790; P9BE-NEXT: xxswapd
4791; P9BE-NEXT: xvcvdpuxds v2
4792; P9BE-NEXT: blr
4793; P9LE: sldi
4794; P9LE: lxvx
4795; P9LE-NEXT: xxswapd
4796; P9LE-NEXT: xvcvdpuxds v2
4797; P9LE-NEXT: blr
4798; P8BE: sldi
4799; P8BE: lxvd2x
4800; P8BE-NEXT: xxswapd
4801; P8BE-NEXT: xvcvdpuxds v2
4802; P8BE-NEXT: blr
4803; P8LE: sldi
4804; P8LE: lxvd2x
4805; P8LE-NEXT: xvcvdpuxds v2
4806; P8LE-NEXT: blr
4807}
4808
4809; Function Attrs: norecurse nounwind readnone
4810define <2 x i64> @spltRegValConvdtoull(double %val) {
4811entry:
4812 %conv = fptoui double %val to i64
4813 %splat.splatinsert = insertelement <2 x i64> undef, i64 %conv, i32 0
4814 %splat.splat = shufflevector <2 x i64> %splat.splatinsert, <2 x i64> undef, <2 x i32> zeroinitializer
4815 ret <2 x i64> %splat.splat
4816; P9BE-LABEL: spltRegValConvdtoull
4817; P9LE-LABEL: spltRegValConvdtoull
4818; P8BE-LABEL: spltRegValConvdtoull
4819; P8LE-LABEL: spltRegValConvdtoull
4820; P9BE: xscvdpuxds
4821; P9BE-NEXT: xxspltd v2
4822; P9BE-NEXT: blr
4823; P9LE: xscvdpuxds
4824; P9LE-NEXT: xxspltd v2
4825; P9LE-NEXT: blr
4826; P8BE: xscvdpuxds
4827; P8BE-NEXT: xxspltd v2
4828; P8BE-NEXT: blr
4829; P8LE: xscvdpuxds
4830; P8LE-NEXT: xxspltd v2
4831; P8LE-NEXT: blr
4832}
4833
4834; Function Attrs: norecurse nounwind readonly
4835define <2 x i64> @spltMemValConvdtoull(double* nocapture readonly %ptr) {
4836entry:
4837 %0 = load double, double* %ptr, align 8
4838 %conv = fptoui double %0 to i64
4839 %splat.splatinsert = insertelement <2 x i64> undef, i64 %conv, i32 0
4840 %splat.splat = shufflevector <2 x i64> %splat.splatinsert, <2 x i64> undef, <2 x i32> zeroinitializer
4841 ret <2 x i64> %splat.splat
4842; P9BE-LABEL: spltMemValConvdtoull
4843; P9LE-LABEL: spltMemValConvdtoull
4844; P8BE-LABEL: spltMemValConvdtoull
4845; P8LE-LABEL: spltMemValConvdtoull
4846; P9BE: lxvdsx
4847; P9BE-NEXT: xvcvdpuxds
4848; P9BE-NEXT: blr
4849; P9LE: lxvdsx
4850; P9LE-NEXT: xvcvdpuxds
4851; P9LE-NEXT: blr
4852; P8BE: lxvdsx
4853; P8BE-NEXT: xvcvdpuxds
4854; P8BE-NEXT: blr
4855; P8LE: lxvdsx
4856; P8LE-NEXT: xvcvdpuxds
4857; P8LE-NEXT: blr
4858}