maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/polly/test/ScopInfo/invariant-loads-leave-read-...

127 lines
5.2 KiB
LLVM
Raw Normal View History

Remove read-only statements from the SCoP We do not need to model read-only statements in the SCoP as they will not cause any side effects that are visible to the outside anyway. Removing them should safe us time and might even simplify the ASTs we generate. Differential Revision: http://reviews.llvm.org/D14272 llvm-svn: 251948
2015-11-04 00:54:49 +08:00
; RUN: opt %loadPolly -polly-scops -analyze < %s | FileCheck %s
; RUN: opt %loadPolly -polly-codegen -analyze < %s
Prepare unit tests for update to ISL 0.16 ISL 0.16 will change how sets are printed which breaks 117 unit tests that text-compare printed sets. This patch re-formats most of these unit tests using a script and small manual editing on top of that. When actually updating ISL, most work is done by just re-running the script to adapt to the changed output. Some tests that compare IR and tests with single CHECK-lines that can be easily updated manually are not included here. The re-format script will also be committed afterwards. The per-test formatter invocation command lines options will not be added in the near future because it is ad hoc and would overwrite the manual edits. Ideally it also shouldn't be required anymore because ISL's set printing has become more stable in 0.16. Differential Revision: http://reviews.llvm.org/D16095 llvm-svn: 257851
2016-01-15 08:48:42 +08:00
; CHECK: Statements {
; CHECK-NEXT: Stmt_top_split
; CHECK-NEXT: Domain :=
; CHECK-NEXT: [p_0, p_1, p_2] -> { Stmt_top_split[] };
; CHECK-NEXT: Schedule :=
; CHECK-NEXT: [p_0, p_1, p_2] -> { Stmt_top_split[] -> [0, 0, 0, 0] };
; CHECK-NEXT: MustWriteAccess := [Reduction Type: NONE] [Scalar: 1]
Follow uses to create value MemoryAccesses The previously implemented approach is to follow value definitions and create write accesses ("push defs") while searching for uses. This requires the same relatively validity- and requirement conditions to be replicated at multiple locations (PHI instructions, other instructions, uses by PHIs). We replace this by iterating over the uses in a SCoP ("pull in requirements"), and add writes only when at least one read has been added. It turns out to be simpler code because each use is only iterated over once and writes are added for the first access that reads it. We need another iteration to identify escaping values (uses not in the SCoP), which also makes the difference between such accesses more obvious. As a side-effect, the order of scalar MemoryAccess can change. Differential Revision: http://reviews.llvm.org/D15706 llvm-svn: 259987
2016-02-06 17:19:40 +08:00
; CHECK-NEXT: [p_0, p_1, p_2] -> { Stmt_top_split[] -> MemRef_25[] };
Prepare unit tests for update to ISL 0.16 ISL 0.16 will change how sets are printed which breaks 117 unit tests that text-compare printed sets. This patch re-formats most of these unit tests using a script and small manual editing on top of that. When actually updating ISL, most work is done by just re-running the script to adapt to the changed output. Some tests that compare IR and tests with single CHECK-lines that can be easily updated manually are not included here. The re-format script will also be committed afterwards. The per-test formatter invocation command lines options will not be added in the near future because it is ad hoc and would overwrite the manual edits. Ideally it also shouldn't be required anymore because ISL's set printing has become more stable in 0.16. Differential Revision: http://reviews.llvm.org/D16095 llvm-svn: 257851
2016-01-15 08:48:42 +08:00
; CHECK-NEXT: Stmt_L_4
; CHECK-NEXT: Domain :=
Update to ISL 0.16.1 llvm-svn: 257898
2016-01-15 23:54:45 +08:00
; CHECK-NEXT: [p_0, p_1, p_2] -> { Stmt_L_4[i0, i1, i2] : 0 <= i0 < p_0 and 0 <= i1 < p_0 and 0 <= i2 < p_0 };
Prepare unit tests for update to ISL 0.16 ISL 0.16 will change how sets are printed which breaks 117 unit tests that text-compare printed sets. This patch re-formats most of these unit tests using a script and small manual editing on top of that. When actually updating ISL, most work is done by just re-running the script to adapt to the changed output. Some tests that compare IR and tests with single CHECK-lines that can be easily updated manually are not included here. The re-format script will also be committed afterwards. The per-test formatter invocation command lines options will not be added in the near future because it is ad hoc and would overwrite the manual edits. Ideally it also shouldn't be required anymore because ISL's set printing has become more stable in 0.16. Differential Revision: http://reviews.llvm.org/D16095 llvm-svn: 257851
2016-01-15 08:48:42 +08:00
; CHECK-NEXT: Schedule :=
; CHECK-NEXT: [p_0, p_1, p_2] -> { Stmt_L_4[i0, i1, i2] -> [1, i0, i1, i2] };
; CHECK-NEXT: ReadAccess := [Reduction Type: NONE] [Scalar: 0]
; CHECK-NEXT: [p_0, p_1, p_2] -> { Stmt_L_4[i0, i1, i2] -> MemRef_19[i1, i0] };
; CHECK-NEXT: ReadAccess := [Reduction Type: NONE] [Scalar: 0]
; CHECK-NEXT: [p_0, p_1, p_2] -> { Stmt_L_4[i0, i1, i2] -> MemRef_5[i2, i0] };
; CHECK-NEXT: ReadAccess := [Reduction Type: NONE] [Scalar: 0]
; CHECK-NEXT: [p_0, p_1, p_2] -> { Stmt_L_4[i0, i1, i2] -> MemRef_12[i2, i1] };
; CHECK-NEXT: MustWriteAccess := [Reduction Type: NONE] [Scalar: 0]
; CHECK-NEXT: [p_0, p_1, p_2] -> { Stmt_L_4[i0, i1, i2] -> MemRef_19[i1, i0] };
Follow uses to create value MemoryAccesses The previously implemented approach is to follow value definitions and create write accesses ("push defs") while searching for uses. This requires the same relatively validity- and requirement conditions to be replicated at multiple locations (PHI instructions, other instructions, uses by PHIs). We replace this by iterating over the uses in a SCoP ("pull in requirements"), and add writes only when at least one read has been added. It turns out to be simpler code because each use is only iterated over once and writes are added for the first access that reads it. We need another iteration to identify escaping values (uses not in the SCoP), which also makes the difference between such accesses more obvious. As a side-effect, the order of scalar MemoryAccess can change. Differential Revision: http://reviews.llvm.org/D15706 llvm-svn: 259987
2016-02-06 17:19:40 +08:00
; CHECK-NEXT: ReadAccess := [Reduction Type: NONE] [Scalar: 1]
; CHECK-NEXT: [p_0, p_1, p_2] -> { Stmt_L_4[i0, i1, i2] -> MemRef_25[] };
Prepare unit tests for update to ISL 0.16 ISL 0.16 will change how sets are printed which breaks 117 unit tests that text-compare printed sets. This patch re-formats most of these unit tests using a script and small manual editing on top of that. When actually updating ISL, most work is done by just re-running the script to adapt to the changed output. Some tests that compare IR and tests with single CHECK-lines that can be easily updated manually are not included here. The re-format script will also be committed afterwards. The per-test formatter invocation command lines options will not be added in the near future because it is ad hoc and would overwrite the manual edits. Ideally it also shouldn't be required anymore because ISL's set printing has become more stable in 0.16. Differential Revision: http://reviews.llvm.org/D16095 llvm-svn: 257851
2016-01-15 08:48:42 +08:00
; CHECK-NEXT: }
Remove read-only statements from the SCoP We do not need to model read-only statements in the SCoP as they will not cause any side effects that are visible to the outside anyway. Removing them should safe us time and might even simplify the ASTs we generate. Differential Revision: http://reviews.llvm.org/D14272 llvm-svn: 251948
2015-11-04 00:54:49 +08:00
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
%jl_value_t = type { %jl_value_t* }
define %jl_value_t* @julia_gemm_22583(%jl_value_t*, %jl_value_t**, i32) {
top:
br label %top.split
top.split: ; preds = %top
%3 = load %jl_value_t*, %jl_value_t** %1, align 8
%4 = bitcast %jl_value_t* %3 to double**
%5 = load double*, double** %4, align 8
%6 = getelementptr inbounds %jl_value_t, %jl_value_t* %3, i64 3, i32 0
%7 = bitcast %jl_value_t** %6 to i64*
%8 = load i64, i64* %7, align 8
%9 = getelementptr %jl_value_t*, %jl_value_t** %1, i64 1
%10 = load %jl_value_t*, %jl_value_t** %9, align 8
%11 = bitcast %jl_value_t* %10 to double**
%12 = load double*, double** %11, align 8
%13 = getelementptr inbounds %jl_value_t, %jl_value_t* %10, i64 3, i32 0
%14 = bitcast %jl_value_t** %13 to i64*
%15 = load i64, i64* %14, align 8
%16 = getelementptr %jl_value_t*, %jl_value_t** %1, i64 2
%17 = load %jl_value_t*, %jl_value_t** %16, align 8
%18 = bitcast %jl_value_t* %17 to double**
%19 = load double*, double** %18, align 8
%20 = getelementptr inbounds %jl_value_t, %jl_value_t* %17, i64 3, i32 0
%21 = bitcast %jl_value_t** %20 to i64*
%22 = load i64, i64* %21, align 8
%23 = icmp sgt i64 %8, 0
%24 = select i1 %23, i64 %8, i64 0
%25 = add i64 %24, 1
%26 = icmp eq i64 %24, 0
br i1 %26, label %L.11, label %L.preheader
L.preheader: ; preds = %top.split
br label %L
L: ; preds = %L.preheader, %L.9
%"#s5.0" = phi i64 [ %27, %L.9 ], [ 1, %L.preheader ]
%27 = add i64 %"#s5.0", 1
br i1 %26, label %L.9, label %L.2.preheader
L.2.preheader: ; preds = %L
br label %L.2
L.2: ; preds = %L.2.preheader, %L.7
%"#s4.0" = phi i64 [ %28, %L.7 ], [ 1, %L.2.preheader ]
%28 = add i64 %"#s4.0", 1
br i1 %26, label %L.7, label %L.4.preheader
L.4.preheader: ; preds = %L.2
br label %L.4
L.4: ; preds = %L.4.preheader, %L.4
%"#s3.0" = phi i64 [ %29, %L.4 ], [ 1, %L.4.preheader ]
%29 = add i64 %"#s3.0", 1
%30 = add i64 %"#s5.0", -1
%31 = add i64 %"#s4.0", -1
%32 = mul i64 %31, %22
%33 = add i64 %32, %30
%34 = getelementptr double, double* %19, i64 %33
%35 = load double, double* %34, align 8
%36 = add i64 %"#s3.0", -1
%37 = mul i64 %36, %8
%38 = add i64 %37, %30
%39 = getelementptr double, double* %5, i64 %38
%40 = load double, double* %39, align 8
%41 = mul i64 %36, %15
%42 = add i64 %41, %31
%43 = getelementptr double, double* %12, i64 %42
%44 = load double, double* %43, align 8
%45 = fmul double %40, %44
%46 = fadd double %35, %45
store double %46, double* %34, align 8
%47 = icmp eq i64 %29, %25
br i1 %47, label %L.7.loopexit, label %L.4
L.7.loopexit: ; preds = %L.4
br label %L.7
L.7: ; preds = %L.7.loopexit, %L.2
%48 = icmp eq i64 %28, %25
br i1 %48, label %L.9.loopexit, label %L.2
L.9.loopexit: ; preds = %L.7
br label %L.9
L.9: ; preds = %L.9.loopexit, %L
%49 = icmp eq i64 %27, %25
br i1 %49, label %L.11.loopexit, label %L
L.11.loopexit: ; preds = %L.9
br label %L.11
L.11: ; preds = %L.11.loopexit, %top.split
ret %jl_value_t* inttoptr (i64 140220477440016 to %jl_value_t*)
}