xref: /llvm-project/llvm/test/CodeGen/X86/umul_fix_sat.ll (revision 2e3de997ab7cd8728c484bc39e24fecbb97dfae8) 
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=x86_64-linux | FileCheck %s --check-prefix=X64
3; RUN: llc < %s -mtriple=i686 -mattr=cmov | FileCheck %s --check-prefix=X86
4
5declare  i4  @llvm.umul.fix.sat.i4   (i4,  i4, i32)
6declare  i32 @llvm.umul.fix.sat.i32  (i32, i32, i32)
7declare  i64 @llvm.umul.fix.sat.i64  (i64, i64, i32)
8declare  <4 x i32> @llvm.umul.fix.sat.v4i32(<4 x i32>, <4 x i32>, i32)
9
10define i32 @func(i32 %x, i32 %y) nounwind {
11; X64-LABEL: func:
12; X64:       # %bb.0:
13; X64-NEXT:    movl %esi, %eax
14; X64-NEXT:    movl %edi, %ecx
15; X64-NEXT:    imulq %rax, %rcx
16; X64-NEXT:    movq %rcx, %rax
17; X64-NEXT:    shrq $32, %rax
18; X64-NEXT:    shrdl $2, %eax, %ecx
19; X64-NEXT:    cmpl $4, %eax
20; X64-NEXT:    movl $-1, %eax
21; X64-NEXT:    cmovbl %ecx, %eax
22; X64-NEXT:    retq
23;
24; X86-LABEL: func:
25; X86:       # %bb.0:
26; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
27; X86-NEXT:    mull {{[0-9]+}}(%esp)
28; X86-NEXT:    shrdl $2, %edx, %eax
29; X86-NEXT:    cmpl $4, %edx
30; X86-NEXT:    movl $-1, %ecx
31; X86-NEXT:    cmovael %ecx, %eax
32; X86-NEXT:    retl
33  %tmp = call i32 @llvm.umul.fix.sat.i32(i32 %x, i32 %y, i32 2)
34  ret i32 %tmp
35}
36
37define i64 @func2(i64 %x, i64 %y) nounwind {
38; X64-LABEL: func2:
39; X64:       # %bb.0:
40; X64-NEXT:    movq %rdi, %rax
41; X64-NEXT:    mulq %rsi
42; X64-NEXT:    shrdq $2, %rdx, %rax
43; X64-NEXT:    cmpq $4, %rdx
44; X64-NEXT:    movq $-1, %rcx
45; X64-NEXT:    cmovaeq %rcx, %rax
46; X64-NEXT:    retq
47;
48; X86-LABEL: func2:
49; X86:       # %bb.0:
50; X86-NEXT:    pushl %ebp
51; X86-NEXT:    pushl %ebx
52; X86-NEXT:    pushl %edi
53; X86-NEXT:    pushl %esi
54; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
55; X86-NEXT:    movl {{[0-9]+}}(%esp), %edi
56; X86-NEXT:    movl %ecx, %eax
57; X86-NEXT:    mull %edi
58; X86-NEXT:    movl %edx, %esi
59; X86-NEXT:    movl %eax, %ebx
60; X86-NEXT:    movl %ecx, %eax
61; X86-NEXT:    mull {{[0-9]+}}(%esp)
62; X86-NEXT:    movl %eax, %ecx
63; X86-NEXT:    addl %edx, %ebx
64; X86-NEXT:    adcl $0, %esi
65; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
66; X86-NEXT:    mull %edi
67; X86-NEXT:    movl %edx, %edi
68; X86-NEXT:    movl %eax, %ebp
69; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
70; X86-NEXT:    mull {{[0-9]+}}(%esp)
71; X86-NEXT:    addl %ebx, %eax
72; X86-NEXT:    adcl %esi, %edx
73; X86-NEXT:    adcl $0, %edi
74; X86-NEXT:    addl %ebp, %edx
75; X86-NEXT:    adcl $0, %edi
76; X86-NEXT:    shrdl $2, %eax, %ecx
77; X86-NEXT:    shrdl $2, %edx, %eax
78; X86-NEXT:    shrl $2, %edx
79; X86-NEXT:    orl %edi, %edx
80; X86-NEXT:    movl $-1, %edx
81; X86-NEXT:    cmovnel %edx, %ecx
82; X86-NEXT:    cmovel %eax, %edx
83; X86-NEXT:    movl %ecx, %eax
84; X86-NEXT:    popl %esi
85; X86-NEXT:    popl %edi
86; X86-NEXT:    popl %ebx
87; X86-NEXT:    popl %ebp
88; X86-NEXT:    retl
89  %tmp = call i64 @llvm.umul.fix.sat.i64(i64 %x, i64 %y, i32 2)
90  ret i64 %tmp
91}
92
93define i4 @func3(i4 %x, i4 %y) nounwind {
94; X64-LABEL: func3:
95; X64:       # %bb.0:
96; X64-NEXT:    andl $15, %esi
97; X64-NEXT:    shlb $4, %dil
98; X64-NEXT:    movzbl %dil, %eax
99; X64-NEXT:    imull %esi, %eax
100; X64-NEXT:    movl %eax, %ecx
101; X64-NEXT:    shrb $2, %cl
102; X64-NEXT:    shrl $8, %eax
103; X64-NEXT:    movl %eax, %edx
104; X64-NEXT:    shlb $6, %dl
105; X64-NEXT:    orb %cl, %dl
106; X64-NEXT:    movzbl %dl, %ecx
107; X64-NEXT:    cmpb $4, %al
108; X64-NEXT:    movl $255, %eax
109; X64-NEXT:    cmovbl %ecx, %eax
110; X64-NEXT:    shrb $4, %al
111; X64-NEXT:    # kill: def $al killed $al killed $eax
112; X64-NEXT:    retq
113;
114; X86-LABEL: func3:
115; X86:       # %bb.0:
116; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
117; X86-NEXT:    andb $15, %al
118; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
119; X86-NEXT:    movzbl %al, %edx
120; X86-NEXT:    shlb $4, %cl
121; X86-NEXT:    movzbl %cl, %eax
122; X86-NEXT:    imull %edx, %eax
123; X86-NEXT:    movb %ah, %cl
124; X86-NEXT:    shlb $6, %cl
125; X86-NEXT:    shrb $2, %al
126; X86-NEXT:    orb %cl, %al
127; X86-NEXT:    movzbl %al, %ecx
128; X86-NEXT:    cmpb $4, %ah
129; X86-NEXT:    movl $255, %eax
130; X86-NEXT:    cmovbl %ecx, %eax
131; X86-NEXT:    shrb $4, %al
132; X86-NEXT:    # kill: def $al killed $al killed $eax
133; X86-NEXT:    retl
134  %tmp = call i4 @llvm.umul.fix.sat.i4(i4 %x, i4 %y, i32 2)
135  ret i4 %tmp
136}
137
138define <4 x i32> @vec(<4 x i32> %x, <4 x i32> %y) nounwind {
139; X64-LABEL: vec:
140; X64:       # %bb.0:
141; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[3,3,3,3]
142; X64-NEXT:    movd %xmm2, %eax
143; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[3,3,3,3]
144; X64-NEXT:    movd %xmm2, %ecx
145; X64-NEXT:    imulq %rax, %rcx
146; X64-NEXT:    movq %rcx, %rax
147; X64-NEXT:    shrq $32, %rax
148; X64-NEXT:    shrdl $2, %eax, %ecx
149; X64-NEXT:    cmpl $4, %eax
150; X64-NEXT:    movl $-1, %eax
151; X64-NEXT:    cmovael %eax, %ecx
152; X64-NEXT:    movd %ecx, %xmm2
153; X64-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[2,3,2,3]
154; X64-NEXT:    movd %xmm3, %ecx
155; X64-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,3,2,3]
156; X64-NEXT:    movd %xmm3, %edx
157; X64-NEXT:    imulq %rcx, %rdx
158; X64-NEXT:    movq %rdx, %rcx
159; X64-NEXT:    shrq $32, %rcx
160; X64-NEXT:    shrdl $2, %ecx, %edx
161; X64-NEXT:    cmpl $4, %ecx
162; X64-NEXT:    cmovael %eax, %edx
163; X64-NEXT:    movd %edx, %xmm3
164; X64-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
165; X64-NEXT:    movd %xmm1, %ecx
166; X64-NEXT:    movd %xmm0, %edx
167; X64-NEXT:    imulq %rcx, %rdx
168; X64-NEXT:    movq %rdx, %rcx
169; X64-NEXT:    shrq $32, %rcx
170; X64-NEXT:    shrdl $2, %ecx, %edx
171; X64-NEXT:    cmpl $4, %ecx
172; X64-NEXT:    cmovael %eax, %edx
173; X64-NEXT:    movd %edx, %xmm2
174; X64-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[1,1,1,1]
175; X64-NEXT:    movd %xmm1, %ecx
176; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
177; X64-NEXT:    movd %xmm0, %edx
178; X64-NEXT:    imulq %rcx, %rdx
179; X64-NEXT:    movq %rdx, %rcx
180; X64-NEXT:    shrq $32, %rcx
181; X64-NEXT:    shrdl $2, %ecx, %edx
182; X64-NEXT:    cmpl $4, %ecx
183; X64-NEXT:    cmovael %eax, %edx
184; X64-NEXT:    movd %edx, %xmm0
185; X64-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
186; X64-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
187; X64-NEXT:    movdqa %xmm2, %xmm0
188; X64-NEXT:    retq
189;
190; X86-LABEL: vec:
191; X86:       # %bb.0:
192; X86-NEXT:    pushl %ebp
193; X86-NEXT:    pushl %ebx
194; X86-NEXT:    pushl %edi
195; X86-NEXT:    pushl %esi
196; X86-NEXT:    movl {{[0-9]+}}(%esp), %edi
197; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebp
198; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebx
199; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
200; X86-NEXT:    mull {{[0-9]+}}(%esp)
201; X86-NEXT:    movl %eax, %esi
202; X86-NEXT:    shrdl $2, %edx, %esi
203; X86-NEXT:    cmpl $4, %edx
204; X86-NEXT:    movl $-1, %ecx
205; X86-NEXT:    cmovael %ecx, %esi
206; X86-NEXT:    movl %ebx, %eax
207; X86-NEXT:    mull {{[0-9]+}}(%esp)
208; X86-NEXT:    movl %eax, %ebx
209; X86-NEXT:    shrdl $2, %edx, %ebx
210; X86-NEXT:    cmpl $4, %edx
211; X86-NEXT:    cmovael %ecx, %ebx
212; X86-NEXT:    movl %ebp, %eax
213; X86-NEXT:    mull {{[0-9]+}}(%esp)
214; X86-NEXT:    movl %eax, %ebp
215; X86-NEXT:    shrdl $2, %edx, %ebp
216; X86-NEXT:    cmpl $4, %edx
217; X86-NEXT:    cmovael %ecx, %ebp
218; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
219; X86-NEXT:    mull {{[0-9]+}}(%esp)
220; X86-NEXT:    shrdl $2, %edx, %eax
221; X86-NEXT:    cmpl $4, %edx
222; X86-NEXT:    cmovael %ecx, %eax
223; X86-NEXT:    movl %eax, 12(%edi)
224; X86-NEXT:    movl %ebp, 8(%edi)
225; X86-NEXT:    movl %ebx, 4(%edi)
226; X86-NEXT:    movl %esi, (%edi)
227; X86-NEXT:    movl %edi, %eax
228; X86-NEXT:    popl %esi
229; X86-NEXT:    popl %edi
230; X86-NEXT:    popl %ebx
231; X86-NEXT:    popl %ebp
232; X86-NEXT:    retl $4
233  %tmp = call <4 x i32> @llvm.umul.fix.sat.v4i32(<4 x i32> %x, <4 x i32> %y, i32 2)
234  ret <4 x i32> %tmp
235}
236
237; These result in regular integer multiplication
238define i32 @func4(i32 %x, i32 %y) nounwind {
239; X64-LABEL: func4:
240; X64:       # %bb.0:
241; X64-NEXT:    movl %edi, %eax
242; X64-NEXT:    mull %esi
243; X64-NEXT:    movl $-1, %ecx
244; X64-NEXT:    cmovol %ecx, %eax
245; X64-NEXT:    retq
246;
247; X86-LABEL: func4:
248; X86:       # %bb.0:
249; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
250; X86-NEXT:    mull {{[0-9]+}}(%esp)
251; X86-NEXT:    movl $-1, %ecx
252; X86-NEXT:    cmovol %ecx, %eax
253; X86-NEXT:    retl
254  %tmp = call i32 @llvm.umul.fix.sat.i32(i32 %x, i32 %y, i32 0)
255  ret i32 %tmp
256}
257
258define i64 @func5(i64 %x, i64 %y) {
259; X64-LABEL: func5:
260; X64:       # %bb.0:
261; X64-NEXT:    movq %rdi, %rax
262; X64-NEXT:    mulq %rsi
263; X64-NEXT:    movq $-1, %rcx
264; X64-NEXT:    cmovoq %rcx, %rax
265; X64-NEXT:    retq
266;
267; X86-LABEL: func5:
268; X86:       # %bb.0:
269; X86-NEXT:    pushl %ebp
270; X86-NEXT:    .cfi_def_cfa_offset 8
271; X86-NEXT:    pushl %ebx
272; X86-NEXT:    .cfi_def_cfa_offset 12
273; X86-NEXT:    pushl %edi
274; X86-NEXT:    .cfi_def_cfa_offset 16
275; X86-NEXT:    pushl %esi
276; X86-NEXT:    .cfi_def_cfa_offset 20
277; X86-NEXT:    .cfi_offset %esi, -20
278; X86-NEXT:    .cfi_offset %edi, -16
279; X86-NEXT:    .cfi_offset %ebx, -12
280; X86-NEXT:    .cfi_offset %ebp, -8
281; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebp
282; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
283; X86-NEXT:    movl {{[0-9]+}}(%esp), %esi
284; X86-NEXT:    testl %esi, %esi
285; X86-NEXT:    setne %dl
286; X86-NEXT:    testl %eax, %eax
287; X86-NEXT:    setne %cl
288; X86-NEXT:    andb %dl, %cl
289; X86-NEXT:    mull {{[0-9]+}}(%esp)
290; X86-NEXT:    movl %eax, %edi
291; X86-NEXT:    seto %bl
292; X86-NEXT:    movl %esi, %eax
293; X86-NEXT:    mull %ebp
294; X86-NEXT:    seto %ch
295; X86-NEXT:    orb %bl, %ch
296; X86-NEXT:    orb %cl, %ch
297; X86-NEXT:    leal (%edi,%eax), %esi
298; X86-NEXT:    movl %ebp, %eax
299; X86-NEXT:    mull {{[0-9]+}}(%esp)
300; X86-NEXT:    addl %esi, %edx
301; X86-NEXT:    setb %cl
302; X86-NEXT:    orb %ch, %cl
303; X86-NEXT:    movl $-1, %ecx
304; X86-NEXT:    cmovnel %ecx, %eax
305; X86-NEXT:    cmovnel %ecx, %edx
306; X86-NEXT:    popl %esi
307; X86-NEXT:    .cfi_def_cfa_offset 16
308; X86-NEXT:    popl %edi
309; X86-NEXT:    .cfi_def_cfa_offset 12
310; X86-NEXT:    popl %ebx
311; X86-NEXT:    .cfi_def_cfa_offset 8
312; X86-NEXT:    popl %ebp
313; X86-NEXT:    .cfi_def_cfa_offset 4
314; X86-NEXT:    retl
315  %tmp = call i64 @llvm.umul.fix.sat.i64(i64 %x, i64 %y, i32 0)
316  ret i64 %tmp
317}
318
319define i4 @func6(i4 %x, i4 %y) nounwind {
320; X64-LABEL: func6:
321; X64:       # %bb.0:
322; X64-NEXT:    movl %edi, %eax
323; X64-NEXT:    andb $15, %sil
324; X64-NEXT:    shlb $4, %al
325; X64-NEXT:    # kill: def $al killed $al killed $eax
326; X64-NEXT:    mulb %sil
327; X64-NEXT:    movzbl %al, %ecx
328; X64-NEXT:    movl $255, %eax
329; X64-NEXT:    cmovnol %ecx, %eax
330; X64-NEXT:    shrb $4, %al
331; X64-NEXT:    # kill: def $al killed $al killed $eax
332; X64-NEXT:    retq
333;
334; X86-LABEL: func6:
335; X86:       # %bb.0:
336; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx
337; X86-NEXT:    andb $15, %cl
338; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
339; X86-NEXT:    shlb $4, %al
340; X86-NEXT:    mulb %cl
341; X86-NEXT:    movzbl %al, %ecx
342; X86-NEXT:    movl $255, %eax
343; X86-NEXT:    cmovnol %ecx, %eax
344; X86-NEXT:    shrb $4, %al
345; X86-NEXT:    # kill: def $al killed $al killed $eax
346; X86-NEXT:    retl
347  %tmp = call i4 @llvm.umul.fix.sat.i4(i4 %x, i4 %y, i32 0)
348  ret i4 %tmp
349}
350
351define <4 x i32> @vec2(<4 x i32> %x, <4 x i32> %y) nounwind {
352; X64-LABEL: vec2:
353; X64:       # %bb.0:
354; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm0[3,3,3,3]
355; X64-NEXT:    movd %xmm2, %eax
356; X64-NEXT:    pshufd {{.*#+}} xmm2 = xmm1[3,3,3,3]
357; X64-NEXT:    movd %xmm2, %ecx
358; X64-NEXT:    mull %ecx
359; X64-NEXT:    movl $-1, %ecx
360; X64-NEXT:    cmovol %ecx, %eax
361; X64-NEXT:    movd %eax, %xmm2
362; X64-NEXT:    pshufd {{.*#+}} xmm3 = xmm0[2,3,2,3]
363; X64-NEXT:    movd %xmm3, %eax
364; X64-NEXT:    pshufd {{.*#+}} xmm3 = xmm1[2,3,2,3]
365; X64-NEXT:    movd %xmm3, %edx
366; X64-NEXT:    mull %edx
367; X64-NEXT:    cmovol %ecx, %eax
368; X64-NEXT:    movd %eax, %xmm3
369; X64-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
370; X64-NEXT:    movd %xmm0, %eax
371; X64-NEXT:    movd %xmm1, %edx
372; X64-NEXT:    mull %edx
373; X64-NEXT:    cmovol %ecx, %eax
374; X64-NEXT:    movd %eax, %xmm2
375; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,1,1,1]
376; X64-NEXT:    movd %xmm0, %eax
377; X64-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,1,1,1]
378; X64-NEXT:    movd %xmm0, %edx
379; X64-NEXT:    mull %edx
380; X64-NEXT:    cmovol %ecx, %eax
381; X64-NEXT:    movd %eax, %xmm0
382; X64-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm0[0],xmm2[1],xmm0[1]
383; X64-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
384; X64-NEXT:    movdqa %xmm2, %xmm0
385; X64-NEXT:    retq
386;
387; X86-LABEL: vec2:
388; X86:       # %bb.0:
389; X86-NEXT:    pushl %ebp
390; X86-NEXT:    pushl %ebx
391; X86-NEXT:    pushl %edi
392; X86-NEXT:    pushl %esi
393; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
394; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebp
395; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebx
396; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
397; X86-NEXT:    mull {{[0-9]+}}(%esp)
398; X86-NEXT:    movl %eax, %esi
399; X86-NEXT:    movl $-1, %edi
400; X86-NEXT:    cmovol %edi, %esi
401; X86-NEXT:    movl %ebx, %eax
402; X86-NEXT:    mull {{[0-9]+}}(%esp)
403; X86-NEXT:    movl %eax, %ebx
404; X86-NEXT:    cmovol %edi, %ebx
405; X86-NEXT:    movl %ebp, %eax
406; X86-NEXT:    mull {{[0-9]+}}(%esp)
407; X86-NEXT:    movl %eax, %ebp
408; X86-NEXT:    cmovol %edi, %ebp
409; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
410; X86-NEXT:    mull {{[0-9]+}}(%esp)
411; X86-NEXT:    cmovol %edi, %eax
412; X86-NEXT:    movl %eax, 12(%ecx)
413; X86-NEXT:    movl %ebp, 8(%ecx)
414; X86-NEXT:    movl %ebx, 4(%ecx)
415; X86-NEXT:    movl %esi, (%ecx)
416; X86-NEXT:    movl %ecx, %eax
417; X86-NEXT:    popl %esi
418; X86-NEXT:    popl %edi
419; X86-NEXT:    popl %ebx
420; X86-NEXT:    popl %ebp
421; X86-NEXT:    retl $4
422  %tmp = call <4 x i32> @llvm.umul.fix.sat.v4i32(<4 x i32> %x, <4 x i32> %y, i32 0)
423  ret <4 x i32> %tmp
424}
425
426define i64 @func7(i64 %x, i64 %y) nounwind {
427; X64-LABEL: func7:
428; X64:       # %bb.0:
429; X64-NEXT:    movq %rdi, %rax
430; X64-NEXT:    mulq %rsi
431; X64-NEXT:    shrdq $32, %rdx, %rax
432; X64-NEXT:    movl $4294967295, %ecx # imm = 0xFFFFFFFF
433; X64-NEXT:    cmpq %rcx, %rdx
434; X64-NEXT:    movq $-1, %rcx
435; X64-NEXT:    cmovaq %rcx, %rax
436; X64-NEXT:    retq
437;
438; X86-LABEL: func7:
439; X86:       # %bb.0:
440; X86-NEXT:    pushl %ebp
441; X86-NEXT:    pushl %ebx
442; X86-NEXT:    pushl %edi
443; X86-NEXT:    pushl %esi
444; X86-NEXT:    movl {{[0-9]+}}(%esp), %edi
445; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebx
446; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebp
447; X86-NEXT:    movl %edi, %eax
448; X86-NEXT:    mull %ebp
449; X86-NEXT:    movl %edx, %ecx
450; X86-NEXT:    movl %eax, %esi
451; X86-NEXT:    movl %edi, %eax
452; X86-NEXT:    mull %ebx
453; X86-NEXT:    addl %edx, %esi
454; X86-NEXT:    adcl $0, %ecx
455; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
456; X86-NEXT:    mull %ebp
457; X86-NEXT:    movl %edx, %edi
458; X86-NEXT:    movl %eax, %ebp
459; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
460; X86-NEXT:    mull %ebx
461; X86-NEXT:    addl %esi, %eax
462; X86-NEXT:    adcl %ecx, %edx
463; X86-NEXT:    adcl $0, %edi
464; X86-NEXT:    addl %ebp, %edx
465; X86-NEXT:    adcl $0, %edi
466; X86-NEXT:    xorl %ecx, %ecx
467; X86-NEXT:    negl %edi
468; X86-NEXT:    sbbl %ecx, %ecx
469; X86-NEXT:    orl %ecx, %eax
470; X86-NEXT:    orl %ecx, %edx
471; X86-NEXT:    popl %esi
472; X86-NEXT:    popl %edi
473; X86-NEXT:    popl %ebx
474; X86-NEXT:    popl %ebp
475; X86-NEXT:    retl
476  %tmp = call i64 @llvm.umul.fix.sat.i64(i64 %x, i64 %y, i32 32)
477  ret i64 %tmp
478}
479
480define i64 @func8(i64 %x, i64 %y) nounwind {
481; X64-LABEL: func8:
482; X64:       # %bb.0:
483; X64-NEXT:    movq %rdi, %rax
484; X64-NEXT:    mulq %rsi
485; X64-NEXT:    shrdq $63, %rdx, %rax
486; X64-NEXT:    movabsq $9223372036854775807, %rcx # imm = 0x7FFFFFFFFFFFFFFF
487; X64-NEXT:    cmpq %rcx, %rdx
488; X64-NEXT:    movq $-1, %rcx
489; X64-NEXT:    cmovaq %rcx, %rax
490; X64-NEXT:    retq
491;
492; X86-LABEL: func8:
493; X86:       # %bb.0:
494; X86-NEXT:    pushl %ebp
495; X86-NEXT:    pushl %ebx
496; X86-NEXT:    pushl %edi
497; X86-NEXT:    pushl %esi
498; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
499; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebx
500; X86-NEXT:    movl {{[0-9]+}}(%esp), %ebp
501; X86-NEXT:    movl %ecx, %eax
502; X86-NEXT:    mull %ebp
503; X86-NEXT:    movl %edx, %esi
504; X86-NEXT:    movl %eax, %edi
505; X86-NEXT:    movl %ecx, %eax
506; X86-NEXT:    mull %ebx
507; X86-NEXT:    addl %edx, %edi
508; X86-NEXT:    adcl $0, %esi
509; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
510; X86-NEXT:    mull %ebp
511; X86-NEXT:    movl %edx, %ecx
512; X86-NEXT:    movl %eax, %ebp
513; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
514; X86-NEXT:    mull %ebx
515; X86-NEXT:    addl %edi, %eax
516; X86-NEXT:    adcl %esi, %edx
517; X86-NEXT:    adcl $0, %ecx
518; X86-NEXT:    addl %ebp, %edx
519; X86-NEXT:    adcl $0, %ecx
520; X86-NEXT:    shldl $1, %edx, %ecx
521; X86-NEXT:    shrdl $31, %edx, %eax
522; X86-NEXT:    testl $-2147483648, %edx # imm = 0x80000000
523; X86-NEXT:    movl $-1, %edx
524; X86-NEXT:    cmovnel %edx, %eax
525; X86-NEXT:    cmovnel %edx, %ecx
526; X86-NEXT:    movl %ecx, %edx
527; X86-NEXT:    popl %esi
528; X86-NEXT:    popl %edi
529; X86-NEXT:    popl %ebx
530; X86-NEXT:    popl %ebp
531; X86-NEXT:    retl
532  %tmp = call i64 @llvm.umul.fix.sat.i64(i64 %x, i64 %y, i32 63)
533  ret i64 %tmp
534}
535