1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=i386-apple-darwin -mattr=+sse2 | FileCheck %s --check-prefix=X86
3 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+sse2 | FileCheck %s --check-prefix=X64
5 define i32 @t(ptr %ref_frame_ptr, i32 %ref_frame_stride, i32 %idxX, i32 %idxY) nounwind {
7 ; X86: ## %bb.0: ## %entry
8 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
9 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx
10 ; X86-NEXT: imull {{[0-9]+}}(%esp), %ecx
11 ; X86-NEXT: addl {{[0-9]+}}(%esp), %ecx
12 ; X86-NEXT: movl (%eax,%ecx), %eax
16 ; X64: ## %bb.0: ## %entry
17 ; X64-NEXT: imull %ecx, %esi
18 ; X64-NEXT: addl %edx, %esi
19 ; X64-NEXT: movslq %esi, %rax
20 ; X64-NEXT: movl (%rdi,%rax), %eax
23 %tmp7 = mul i32 %idxY, %ref_frame_stride ; <i32> [#uses=2]
24 %tmp9 = add i32 %tmp7, %idxX ; <i32> [#uses=1]
25 %tmp11 = getelementptr i8, ptr %ref_frame_ptr, i32 %tmp9 ; <ptr> [#uses=1]
26 %tmp13 = load i32, ptr %tmp11, align 4 ; <i32> [#uses=1]
27 %tmp18 = add i32 %idxX, 4 ; <i32> [#uses=1]
28 %tmp20.sum = add i32 %tmp18, %tmp7 ; <i32> [#uses=1]
29 %tmp21 = getelementptr i8, ptr %ref_frame_ptr, i32 %tmp20.sum ; <ptr> [#uses=1]
30 %tmp23 = load i16, ptr %tmp21, align 2 ; <i16> [#uses=1]
31 %tmp2425 = zext i16 %tmp23 to i64 ; <i64> [#uses=1]
32 %tmp26 = shl i64 %tmp2425, 32 ; <i64> [#uses=1]
33 %tmp2728 = zext i32 %tmp13 to i64 ; <i64> [#uses=1]
34 %tmp29 = or i64 %tmp26, %tmp2728 ; <i64> [#uses=1]
35 %tmp3454 = bitcast i64 %tmp29 to double ; <double> [#uses=1]
36 %tmp35 = insertelement <2 x double> undef, double %tmp3454, i32 0 ; <<2 x double>> [#uses=1]
37 %tmp36 = insertelement <2 x double> %tmp35, double 0.000000e+00, i32 1 ; <<2 x double>> [#uses=1]
38 %tmp42 = bitcast <2 x double> %tmp36 to <8 x i16> ; <<8 x i16>> [#uses=1]
39 %tmp43 = shufflevector <8 x i16> %tmp42, <8 x i16> undef, <8 x i32> < i32 0, i32 1, i32 1, i32 2, i32 4, i32 5, i32 6, i32 7 > ; <<8 x i16>> [#uses=1]
40 %tmp47 = bitcast <8 x i16> %tmp43 to <4 x i32> ; <<4 x i32>> [#uses=1]
41 %tmp48 = extractelement <4 x i32> %tmp47, i32 0 ; <i32> [#uses=1]
45 ; Test CSE for SDAG nodes with multiple results (UMUL_LOHI).
46 define i96 @square_high(i96 %x) nounwind {
47 ; X86-LABEL: square_high:
48 ; X86: ## %bb.0: ## %entry
49 ; X86-NEXT: pushl %ebp
50 ; X86-NEXT: pushl %ebx
51 ; X86-NEXT: pushl %edi
52 ; X86-NEXT: pushl %esi
53 ; X86-NEXT: pushl %eax
54 ; X86-NEXT: movl {{[0-9]+}}(%esp), %esi
55 ; X86-NEXT: movl {{[0-9]+}}(%esp), %edi
56 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ebx
57 ; X86-NEXT: movl %edi, %eax
59 ; X86-NEXT: movl %edx, %ecx
60 ; X86-NEXT: movl %ebx, %eax
62 ; X86-NEXT: addl %eax, %ecx
63 ; X86-NEXT: movl %edx, %ebp
64 ; X86-NEXT: adcl $0, %ebp
65 ; X86-NEXT: addl %eax, %ecx
66 ; X86-NEXT: adcl %edx, %ebp
68 ; X86-NEXT: movzbl %al, %ecx
69 ; X86-NEXT: movl %ebx, %eax
71 ; X86-NEXT: movl %eax, %ebx
72 ; X86-NEXT: addl %ebp, %ebx
73 ; X86-NEXT: adcl %edx, %ecx
74 ; X86-NEXT: movl %esi, %eax
76 ; X86-NEXT: movl %edx, (%esp) ## 4-byte Spill
77 ; X86-NEXT: movl %eax, %ebp
78 ; X86-NEXT: movl %esi, %eax
79 ; X86-NEXT: mull {{[0-9]+}}(%esp)
80 ; X86-NEXT: movl %edx, %esi
81 ; X86-NEXT: movl %eax, %edi
82 ; X86-NEXT: addl (%esp), %edi ## 4-byte Folded Reload
83 ; X86-NEXT: adcl $0, %esi
84 ; X86-NEXT: addl %ebp, %ebx
85 ; X86-NEXT: adcl %edi, %ecx
86 ; X86-NEXT: movl %esi, %eax
87 ; X86-NEXT: adcl $0, %eax
89 ; X86-NEXT: addl %ebp, %ebx
90 ; X86-NEXT: adcl %ecx, %edi
91 ; X86-NEXT: movzbl %dl, %ecx
92 ; X86-NEXT: adcl %eax, %esi
93 ; X86-NEXT: adcl $0, %ecx
94 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
96 ; X86-NEXT: addl %eax, %esi
97 ; X86-NEXT: adcl %edx, %ecx
98 ; X86-NEXT: movl %edi, %eax
99 ; X86-NEXT: movl %esi, %edx
100 ; X86-NEXT: addl $4, %esp
101 ; X86-NEXT: popl %esi
102 ; X86-NEXT: popl %edi
103 ; X86-NEXT: popl %ebx
104 ; X86-NEXT: popl %ebp
107 ; X64-LABEL: square_high:
108 ; X64: ## %bb.0: ## %entry
109 ; X64-NEXT: movl %esi, %ecx
110 ; X64-NEXT: movq %rcx, %rax
111 ; X64-NEXT: mulq %rdi
112 ; X64-NEXT: movq %rdx, %rsi
113 ; X64-NEXT: movq %rax, %r8
114 ; X64-NEXT: movq %rdi, %rax
115 ; X64-NEXT: mulq %rdi
116 ; X64-NEXT: addq %r8, %rdx
117 ; X64-NEXT: movq %rsi, %rax
118 ; X64-NEXT: adcq $0, %rax
119 ; X64-NEXT: addq %rdx, %r8
120 ; X64-NEXT: adcq %rsi, %rax
121 ; X64-NEXT: imulq %rcx, %rcx
122 ; X64-NEXT: addq %rax, %rcx
123 ; X64-NEXT: shrdq $32, %rcx, %r8
124 ; X64-NEXT: shrq $32, %rcx
125 ; X64-NEXT: movq %r8, %rax
126 ; X64-NEXT: movq %rcx, %rdx
129 %conv = zext i96 %x to i192
130 %mul = mul nuw i192 %conv, %conv
131 %shr = lshr i192 %mul, 96
132 %conv2 = trunc i192 %shr to i96