Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Sanitized statement splitting #92

Merged
merged 1 commit into from
May 24, 2021
Merged

Sanitized statement splitting #92

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
578 changes: 578 additions & 0 deletions example/polybench/split-and-merge/2mm/2mm.heuristic.polymer.par.log
View file
Open in desktop

Large diffs are not rendered by default.

117 changes: 80 additions & 37 deletions example/polybench/split-and-merge/2mm/2mm.heuristic.polymer.par.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,14 @@
#map7 = affine_map<(d0, d1) -> (2200, d0 * 32 - d1 * 32 + 32)>
#map8 = affine_map<(d0) -> (0, (d0 * 32 - 1799) ceildiv 32)>
#map9 = affine_map<(d0, d1) -> (1800, d0 * 32 - d1 * 32 + 32)>
#map10 = affine_map<()[s0] -> ((s0 - 1) floordiv 32 + 1)>
#map11 = affine_map<(d0)[s0] -> (s0, d0 * 32 + 32)>
#map12 = affine_map<()[s0, s1] -> ((s0 + s1 - 2) floordiv 32 + 1)>
#map13 = affine_map<(d0)[s0] -> (0, (d0 * 32 - s0 + 1) ceildiv 32)>
#map14 = affine_map<(d0)[s0] -> ((s0 - 1) floordiv 32 + 1, d0 + 1)>
#map15 = affine_map<(d0, d1)[s0] -> (s0, d0 * 32 - d1 * 32 + 32)>
#set0 = affine_set<()[s0] : (s0 - 1 >= 0)>
#set1 = affine_set<()[s0, s1] : (s0 - 1 >= 0, s1 - 1 >= 0)>
module attributes {llvm.data_layout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128", llvm.target_triple = "x86_64-unknown-linux-gnu"} {
llvm.mlir.global internal constant @str7("==END DUMP_ARRAYS==\0A\00")
llvm.mlir.global internal constant @str6("\0Aend dump: %s\0A\00")
Expand Down Expand Up @@ -234,44 +242,79 @@ module attributes {llvm.data_layout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i6
func private @polybench_timer_start()
func private @polybench_timer_stop()
func private @polybench_timer_print()
func private @S0(%arg0: memref<?x1800xf64>, %arg1: index, %arg2: index) attributes {scop.stmt} {
func @kernel_2mm(%arg0: i32, %arg1: i32, %arg2: i32, %arg3: i32, %arg4: f64, %arg5: f64, %arg6: memref<?x1800xf64>, %arg7: memref<?x2200xf64>, %arg8: memref<?x1800xf64>, %arg9: memref<?x2400xf64>, %arg10: memref<?x2400xf64>, %arg11: memref<?xf64>, %arg12: memref<?xf64>) {
%cst = constant 0.000000e+00 : f64
affine.store %cst, %arg0[symbol(%arg1), symbol(%arg2)] : memref<?x1800xf64>
return
}
func private @S1(%arg0: memref<?xf64>, %arg1: index, %arg2: memref<?x1800xf64>, %arg3: index, %arg4: f64, %arg5: memref<?x2200xf64>, %arg6: index) attributes {scop.stmt} {
%0 = affine.load %arg5[symbol(%arg6), symbol(%arg1)] : memref<?x2200xf64>
%1 = mulf %arg4, %0 : f64
%2 = affine.load %arg2[symbol(%arg1), symbol(%arg3)] : memref<?x1800xf64>
%3 = mulf %1, %2 {scop.splittable = 0 : index} : f64
affine.store %3, %arg0[symbol(%arg1)] : memref<?xf64>
return
}
func private @S2(%arg0: memref<?x1800xf64>, %arg1: index, %arg2: index, %arg3: memref<?xf64>, %arg4: index) attributes {scop.stmt} {
%0 = affine.load %arg0[symbol(%arg1), symbol(%arg2)] : memref<?x1800xf64>
%1 = affine.load %arg3[symbol(%arg4)] : memref<?xf64>
%2 = addf %0, %1 : f64
affine.store %2, %arg0[symbol(%arg1), symbol(%arg2)] : memref<?x1800xf64>
return
}
func private @S3(%arg0: memref<?x2400xf64>, %arg1: index, %arg2: index, %arg3: f64) attributes {scop.stmt} {
%0 = affine.load %arg0[symbol(%arg1), symbol(%arg2)] : memref<?x2400xf64>
%1 = mulf %0, %arg3 : f64
affine.store %1, %arg0[symbol(%arg1), symbol(%arg2)] : memref<?x2400xf64>
return
}
func private @S4(%arg0: memref<?xf64>, %arg1: index, %arg2: memref<?x2400xf64>, %arg3: index, %arg4: memref<?x1800xf64>, %arg5: index) attributes {scop.stmt} {
%0 = affine.load %arg4[symbol(%arg5), symbol(%arg1)] : memref<?x1800xf64>
%1 = affine.load %arg2[symbol(%arg1), symbol(%arg3)] : memref<?x2400xf64>
%2 = mulf %0, %1 {scop.splittable = 1 : index} : f64
affine.store %2, %arg0[symbol(%arg1)] : memref<?xf64>
return
}
func private @S5(%arg0: memref<?x2400xf64>, %arg1: index, %arg2: index, %arg3: memref<?xf64>, %arg4: index) attributes {scop.stmt} {
%0 = affine.load %arg0[symbol(%arg1), symbol(%arg2)] : memref<?x2400xf64>
%1 = affine.load %arg3[symbol(%arg4)] : memref<?xf64>
%2 = addf %0, %1 : f64
affine.store %2, %arg0[symbol(%arg1), symbol(%arg2)] : memref<?x2400xf64>
%0 = index_cast %arg0 : i32 to index
%1 = index_cast %arg1 : i32 to index
%2 = index_cast %arg2 : i32 to index
%3 = index_cast %arg3 : i32 to index
affine.if #set0()[%0] {
affine.if #set0()[%3] {
affine.for %arg13 = 0 to #map10()[%0] {
affine.for %arg14 = 0 to #map10()[%3] {
affine.for %arg15 = #map0(%arg13) to min #map11(%arg13)[%0] {
affine.for %arg16 = #map0(%arg14) to min #map11(%arg14)[%3] {
%4 = affine.load %arg10[%arg15, %arg16] : memref<?x2400xf64>
%5 = mulf %4, %arg5 : f64
affine.store %5, %arg10[%arg15, %arg16] : memref<?x2400xf64>
}
}
}
} {scop.parallelizable}
}
affine.if #set0()[%1] {
affine.for %arg13 = 0 to #map10()[%0] {
affine.for %arg14 = 0 to #map10()[%1] {
affine.for %arg15 = #map0(%arg13) to min #map11(%arg13)[%0] {
affine.for %arg16 = #map0(%arg14) to min #map11(%arg14)[%1] {
affine.store %cst, %arg6[%arg15, %arg16] : memref<?x1800xf64>
}
}
}
} {scop.parallelizable}
}
affine.if #set1()[%1, %2] {
affine.for %arg13 = 0 to #map12()[%0, %2] {
affine.for %arg14 = max #map13(%arg13)[%2] to min #map14(%arg13)[%0] {
affine.for %arg15 = #map6(%arg13, %arg14) to min #map15(%arg13, %arg14)[%2] {
affine.for %arg16 = #map0(%arg14) to min #map11(%arg14)[%0] {
affine.for %arg17 = 0 to %1 {
%4 = affine.load %arg7[%arg16, %arg15] : memref<?x2200xf64>
%5 = mulf %arg4, %4 : f64
%6 = affine.load %arg8[%arg15, %arg17] : memref<?x1800xf64>
%7 = mulf %5, %6 {scop.splittable = 0 : index} : f64
affine.store %7, %arg11[%arg15] : memref<?xf64>
%8 = affine.load %arg6[%arg16, %arg17] : memref<?x1800xf64>
%9 = affine.load %arg11[%arg15] : memref<?xf64>
%10 = addf %8, %9 : f64
affine.store %10, %arg6[%arg16, %arg17] : memref<?x1800xf64>
}
}
}
} {scop.parallelizable}
}
}
affine.if #set1()[%1, %3] {
affine.for %arg13 = 0 to #map12()[%0, %1] {
affine.for %arg14 = max #map13(%arg13)[%1] to min #map14(%arg13)[%0] {
affine.for %arg15 = #map6(%arg13, %arg14) to min #map15(%arg13, %arg14)[%1] {
affine.for %arg16 = #map0(%arg14) to min #map11(%arg14)[%0] {
affine.for %arg17 = 0 to %3 {
%4 = affine.load %arg6[%arg16, %arg15] : memref<?x1800xf64>
%5 = affine.load %arg9[%arg15, %arg17] : memref<?x2400xf64>
%6 = mulf %4, %5 {scop.splittable = 1 : index} : f64
affine.store %6, %arg11[%arg15] : memref<?xf64>
%7 = affine.load %arg10[%arg16, %arg17] : memref<?x2400xf64>
%8 = affine.load %arg11[%arg15] : memref<?xf64>
%9 = addf %7, %8 : f64
affine.store %9, %arg10[%arg16, %arg17] : memref<?x2400xf64>
}
}
}
} {scop.parallelizable}
}
}
}
return
}
func private @print_array(%arg0: i32, %arg1: i32, %arg2: memref<?x2400xf64>) {
Expand Down
Loading