| blob | 76341cb6afe078334874abce544e480e024d9faa |
1 ; RUN: opt %loadPolly -polly-opt-isl -polly-pattern-matching-based-opts=true \
2 ; RUN: -polly-target-throughput-vector-fma=1 \
3 ; RUN: -polly-target-latency-vector-fma=8 \
4 ; RUN: -analyze -polly-ast -polly-target-1st-cache-level-size=0 \
5 ; RUN: -polly-target-vector-register-bitwidth=256 \
6 ; RUN: < %s 2>&1 | FileCheck %s
8 ; RUN: opt %loadPolly -polly-opt-isl -polly-pattern-matching-based-opts=true \
9 ; RUN: -polly-target-throughput-vector-fma=1 \
10 ; RUN: -polly-target-latency-vector-fma=8 \
11 ; RUN: -analyze -polly-ast -polly-target-1st-cache-level-associativity=8 \
12 ; RUN: -polly-target-2nd-cache-level-associativity=8 \
13 ; RUN: -polly-target-1st-cache-level-size=32768 \
14 ; RUN: -polly-target-vector-register-bitwidth=256 \
15 ; RUN: -polly-target-2nd-cache-level-size=262144 < %s 2>&1 \
16 ; RUN: | FileCheck %s --check-prefix=EXTRACTION-OF-MACRO-KERNEL
17 ;
18 ; /* C := alpha*A*B + beta*C */
19 ; for (i = 0; i < _PB_NI; i++)
20 ; for (j = 0; j < _PB_NJ; j++)
21 ; {
22 ; C[i][j] *= beta;
23 ; for (k = 0; k < _PB_NK; ++k)
24 ; C[i][j] += alpha * A[i][k] * B[k][j];
25 ; }
26 ;
27 ; CHECK: {
28 ; CHECK-NEXT: // 1st level tiling - Tiles
29 ; CHECK-NEXT: for (int c0 = 0; c0 <= 32; c0 += 1)
30 ; CHECK-NEXT: for (int c1 = 0; c1 <= 32; c1 += 1) {
31 ; CHECK-NEXT: // 1st level tiling - Points
32 ; CHECK-NEXT: for (int c2 = 0; c2 <= 31; c2 += 1)
33 ; CHECK-NEXT: for (int c3 = 0; c3 <= 31; c3 += 1)
34 ; CHECK-NEXT: Stmt_bb9(32 * c0 + c2, 32 * c1 + c3);
35 ; CHECK-NEXT: }
36 ; CHECK-NEXT: // Inter iteration alias-free
37 ; CHECK-NEXT: // Register tiling - Tiles
38 ; CHECK-NEXT: for (int c0 = 0; c0 <= 131; c0 += 1)
39 ; CHECK-NEXT: for (int c1 = 0; c1 <= 263; c1 += 1)
40 ; CHECK-NEXT: for (int c2 = 0; c2 <= 1023; c2 += 1) {
41 ; CHECK-NEXT: // Register tiling - Points
42 ; CHECK-NEXT: {
43 ; CHECK-NEXT: Stmt_Copy_0(4 * c1, 8 * c0, c2);
44 ; CHECK-NEXT: Stmt_Copy_0(4 * c1, 8 * c0 + 1, c2);
45 ; CHECK-NEXT: Stmt_Copy_0(4 * c1, 8 * c0 + 2, c2);
46 ; CHECK-NEXT: Stmt_Copy_0(4 * c1, 8 * c0 + 3, c2);
47 ; CHECK-NEXT: Stmt_Copy_0(4 * c1, 8 * c0 + 4, c2);
48 ; CHECK-NEXT: Stmt_Copy_0(4 * c1, 8 * c0 + 5, c2);
49 ; CHECK-NEXT: Stmt_Copy_0(4 * c1, 8 * c0 + 6, c2);
50 ; CHECK-NEXT: Stmt_Copy_0(4 * c1, 8 * c0 + 7, c2);
51 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 1, 8 * c0, c2);
52 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 1, 8 * c0 + 1, c2);
53 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 1, 8 * c0 + 2, c2);
54 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 1, 8 * c0 + 3, c2);
55 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 1, 8 * c0 + 4, c2);
56 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 1, 8 * c0 + 5, c2);
57 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 1, 8 * c0 + 6, c2);
58 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 1, 8 * c0 + 7, c2);
59 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 2, 8 * c0, c2);
60 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 2, 8 * c0 + 1, c2);
61 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 2, 8 * c0 + 2, c2);
62 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 2, 8 * c0 + 3, c2);
63 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 2, 8 * c0 + 4, c2);
64 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 2, 8 * c0 + 5, c2);
65 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 2, 8 * c0 + 6, c2);
66 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 2, 8 * c0 + 7, c2);
67 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 3, 8 * c0, c2);
68 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 3, 8 * c0 + 1, c2);
69 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 3, 8 * c0 + 2, c2);
70 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 3, 8 * c0 + 3, c2);
71 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 3, 8 * c0 + 4, c2);
72 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 3, 8 * c0 + 5, c2);
73 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 3, 8 * c0 + 6, c2);
74 ; CHECK-NEXT: Stmt_Copy_0(4 * c1 + 3, 8 * c0 + 7, c2);
75 ; CHECK-NEXT: }
76 ; CHECK-NEXT: }
77 ; CHECK-NEXT: }
78 ;
79 ; EXTRACTION-OF-MACRO-KERNEL: {
80 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: // 1st level tiling - Tiles
81 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: for (int c0 = 0; c0 <= 32; c0 += 1)
82 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: for (int c1 = 0; c1 <= 32; c1 += 1) {
83 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: // 1st level tiling - Points
84 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: for (int c2 = 0; c2 <= 31; c2 += 1)
85 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: for (int c3 = 0; c3 <= 31; c3 += 1)
86 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_bb9(32 * c0 + c2, 32 * c1 + c3);
87 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: }
88 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: // Inter iteration alias-free
89 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: // 1st level tiling - Tiles
90 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: for (int c1 = 0; c1 <= 3; c1 += 1) {
91 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: for (int c3 = 0; c3 <= 1055; c3 += 1)
92 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: for (int c4 = 256 * c1; c4 <= 256 * c1 + 255; c4 += 1)
93 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: CopyStmt_0(0, c3, c4);
94 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: for (int c2 = 0; c2 <= 10; c2 += 1) {
95 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: for (int c3 = 96 * c2; c3 <= 96 * c2 + 95; c3 += 1)
96 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: for (int c5 = 256 * c1; c5 <= 256 * c1 + 255; c5 += 1)
97 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: CopyStmt_1(c3, 0, c5);
98 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: // 1st level tiling - Points
99 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: // Register tiling - Tiles
100 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: for (int c3 = 0; c3 <= 131; c3 += 1)
101 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: for (int c4 = 0; c4 <= 23; c4 += 1)
102 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: for (int c5 = 0; c5 <= 255; c5 += 1) {
103 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: // Loop Vectorizer Disabled
104 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: // Register tiling - Points
105 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: {
106 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4, 8 * c3, 256 * c1 + c5);
107 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4, 8 * c3 + 1, 256 * c1 + c5);
108 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4, 8 * c3 + 2, 256 * c1 + c5);
109 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4, 8 * c3 + 3, 256 * c1 + c5);
110 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4, 8 * c3 + 4, 256 * c1 + c5);
111 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4, 8 * c3 + 5, 256 * c1 + c5);
112 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4, 8 * c3 + 6, 256 * c1 + c5);
113 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4, 8 * c3 + 7, 256 * c1 + c5);
114 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 1, 8 * c3, 256 * c1 + c5);
115 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 1, 8 * c3 + 1, 256 * c1 + c5);
116 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 1, 8 * c3 + 2, 256 * c1 + c5);
117 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 1, 8 * c3 + 3, 256 * c1 + c5);
118 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 1, 8 * c3 + 4, 256 * c1 + c5);
119 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 1, 8 * c3 + 5, 256 * c1 + c5);
120 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 1, 8 * c3 + 6, 256 * c1 + c5);
121 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 1, 8 * c3 + 7, 256 * c1 + c5);
122 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 2, 8 * c3, 256 * c1 + c5);
123 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 2, 8 * c3 + 1, 256 * c1 + c5);
124 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 2, 8 * c3 + 2, 256 * c1 + c5);
125 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 2, 8 * c3 + 3, 256 * c1 + c5);
126 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 2, 8 * c3 + 4, 256 * c1 + c5);
127 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 2, 8 * c3 + 5, 256 * c1 + c5);
128 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 2, 8 * c3 + 6, 256 * c1 + c5);
129 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 2, 8 * c3 + 7, 256 * c1 + c5);
130 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 3, 8 * c3, 256 * c1 + c5);
131 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 3, 8 * c3 + 1, 256 * c1 + c5);
132 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 3, 8 * c3 + 2, 256 * c1 + c5);
133 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 3, 8 * c3 + 3, 256 * c1 + c5);
134 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 3, 8 * c3 + 4, 256 * c1 + c5);
135 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 3, 8 * c3 + 5, 256 * c1 + c5);
136 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 3, 8 * c3 + 6, 256 * c1 + c5);
137 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: Stmt_Copy_0(96 * c2 + 4 * c4 + 3, 8 * c3 + 7, 256 * c1 + c5);
138 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: }
139 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: }
140 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: }
141 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: }
142 ; EXTRACTION-OF-MACRO-KERNEL-NEXT: }
143 ;
144 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
145 target triple = "x86_64-unknown-unknown"
147 define internal void @kernel_gemm(i32 %arg, i32 %arg1, i32 %arg2, double %arg3, double %arg4, [1056 x double]* %arg5, [1024 x double]* %arg6, [1056 x double]* %arg7) #0 {
148 bb:
149 br label %bb8
151 bb8: ; preds = %bb29, %bb
152 %tmp = phi i64 [ 0, %bb ], [ %tmp30, %bb29 ]
153 br label %bb9
155 bb9: ; preds = %bb26, %bb8
156 %tmp10 = phi i64 [ 0, %bb8 ], [ %tmp27, %bb26 ]
157 %tmp11 = getelementptr inbounds [1056 x double], [1056 x double]* %arg5, i64 %tmp, i64 %tmp10
158 %tmp12 = load double, double* %tmp11, align 8
159 %tmp13 = fmul double %tmp12, %arg4
160 store double %tmp13, double* %tmp11, align 8
161 br label %Copy_0
163 Copy_0: ; preds = %Copy_0, %bb9
164 %tmp15 = phi i64 [ 0, %bb9 ], [ %tmp24, %Copy_0 ]
165 %tmp16 = getelementptr inbounds [1024 x double], [1024 x double]* %arg6, i64 %tmp, i64 %tmp15
166 %tmp17 = load double, double* %tmp16, align 8
167 %tmp18 = fmul double %tmp17, %arg3
168 %tmp19 = getelementptr inbounds [1056 x double], [1056 x double]* %arg7, i64 %tmp15, i64 %tmp10
169 %tmp20 = load double, double* %tmp19, align 8
170 %tmp21 = fmul double %tmp18, %tmp20
171 %tmp22 = load double, double* %tmp11, align 8
172 %tmp23 = fadd double %tmp22, %tmp21
173 store double %tmp23, double* %tmp11, align 8
174 %tmp24 = add nuw nsw i64 %tmp15, 1
175 %tmp25 = icmp ne i64 %tmp24, 1024
176 br i1 %tmp25, label %Copy_0, label %bb26
178 bb26: ; preds = %Copy_0
179 %tmp27 = add nuw nsw i64 %tmp10, 1
180 %tmp28 = icmp ne i64 %tmp27, 1056
181 br i1 %tmp28, label %bb9, label %bb29
183 bb29: ; preds = %bb26
184 %tmp30 = add nuw nsw i64 %tmp, 1
185 %tmp31 = icmp ne i64 %tmp30, 1056
186 br i1 %tmp31, label %bb8, label %bb32
188 bb32: ; preds = %bb29
189 ret void
190 }