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

[Analysis] Use verify-diagnostics for print-based tests (NFC) #5970

Merged
merged 2 commits into from
Feb 20, 2025
Merged

[Analysis] Use verify-diagnostics for print-based tests (NFC) #5970

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
345057b
[Analysis] Use `verify-diagnostics` for print-based tests (NFC)
Mogball Feb 20, 2025
bfd5d5c
try disabling threading
Mogball Feb 20, 2025
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
113 changes: 51 additions & 62 deletions test/Analysis/test-alias.mlir
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
// RUN: triton-opt %s --mlir-disable-threading -test-print-alias -split-input-file 2>&1 | FileCheck %s
// RUN: triton-opt %s -mlir-disable-threading -test-print-alias -verify-diagnostics -o /dev/null

#AL = #ttg.blocked<{sizePerThread = [1, 4], threadsPerWarp = [4, 8], warpsPerCTA = [4, 1], order = [1, 0]}>
#BL = #ttg.blocked<{sizePerThread = [1, 4], threadsPerWarp = [1, 32], warpsPerCTA = [4, 1], order = [1, 0]}>
Expand All @@ -11,8 +11,6 @@

module attributes {"ttg.num-warps" = 4 : i32, "ttg.target" = "cuda:80"} {

// CHECK-LABEL: matmul_loop
// CHECK-NOT: ->
// There shouldn't be any aliasing with the dot op encoding.
tt.func @matmul_loop(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f16>, %B : !tt.ptr<f16>) {
%a_ptr_init = tt.splat %A : !tt.ptr<f16> -> tensor<128x32x!tt.ptr<f16>, #AL>
Expand All @@ -38,47 +36,42 @@ tt.func @matmul_loop(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f16>,
tt.return
}

// CHECK-LABEL: alloc
tt.func @alloc(%A : !tt.ptr<f16>) {
// CHECK: %0 -> %0
// expected-remark @below {{%0 -> %0}}
%cst2 = ttg.local_alloc : () -> !ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory, mutable>
tt.return
}

// CHECK-LABEL: alloc_init
tt.func @alloc_init(%A : !tt.ptr<f16>) {
%cst0 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #AL>
// CHECK: %0 -> %0
// expected-remark @below {{%0 -> %0}}
%cst1 = ttg.local_alloc %cst0 : (tensor<16x16xf16, #AL>) -> !ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory>
tt.return
}

// CHECK-LABEL: trans
tt.func @trans(%A : !tt.ptr<f16>) {
// CHECK: %0 -> %0
// expected-remark @below {{%0 -> %0}}
%tensor = ttg.local_alloc : () -> !ttg.memdesc<16x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK: %1 -> %0
// expected-remark @below {{%1 -> %0}}
%b = ttg.memdesc_trans %tensor {order=array<i32: 1,0>} : !ttg.memdesc<16x32xf16, #A_SHARED, #ttg.shared_memory, mutable> -> !ttg.memdesc<32x16xf16, #A_SHARED_T, #ttg.shared_memory, mutable>
tt.return
}

// CHECK-LABEL: subview
tt.func @subview(%A : !ttg.memdesc<1x16x16xf16, #A_SHARED, #ttg.shared_memory>) {
%index = arith.constant 0 : i32
// CHECK: %0 -> %0
// expected-remark @below {{%0 -> %0}}
%a = ttg.local_alloc : () -> !ttg.memdesc<1x16x16xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK-NEXT: %1 -> %0
// expected-remark @below {{%1 -> %0}}
%cst1 = ttg.memdesc_subview %a[%index, %index, %index] : !ttg.memdesc<1x16x16xf16, #A_SHARED, #ttg.shared_memory, mutable> -> !ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory, mutable>
tt.return
}

// CHECK-LABEL: if_alias
tt.func @if_alias(%i1 : i1) {
// CHECK: %0 -> %0
// expected-remark @below {{%0 -> %0}}
%a = ttg.local_alloc : () -> !ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK: %1 -> %1
// expected-remark @below {{%1 -> %1}}
%b = ttg.local_alloc : () -> !ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK-NEXT: %2 -> %0,%1
// expected-remark @below {{%2 -> %0,%1}}
%cst2 = scf.if %i1 -> !ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory, mutable> {
scf.yield %a : !ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory, mutable>
} else {
Expand All @@ -87,46 +80,44 @@ tt.func @if_alias(%i1 : i1) {
tt.return
}

// CHECK-LABEL: for
tt.func @for(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f16>, %B : !tt.ptr<f16>) {
// CHECK: %0 -> %0
// expected-remark @below {{%0 -> %0}}
%a = ttg.local_alloc : () -> !ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK: %1 -> %1
// expected-remark @below {{%1 -> %1}}
%b = ttg.local_alloc : () -> !ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK: %2 -> %2
// expected-remark @below {{%2 -> %2}}
%c = ttg.local_alloc : () -> !ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK-NEXT: %arg6 -> %0
// CHECK-NEXT: %arg7 -> %1
// CHECK-NEXT: %arg8 -> %2
// CHECK-NEXT: %3#0 -> %0,%1
// CHECK-NEXT: %3#1 -> %0,%1
// CHECK-NEXT: %3#2 -> %0,%1,%2
// expected-remark @below {{%arg6 -> %0}}
// expected-remark @below {{%arg7 -> %1}}
// expected-remark @below {{%arg8 -> %2}}
// expected-remark @below {{%3#0 -> %0,%1}}
// expected-remark @below {{%3#1 -> %0,%1}}
// expected-remark @below {{%3#2 -> %0,%1,%2}}
%a_shared, %b_shared, %c_shared = scf.for %iv = %lb to %ub step %step iter_args(%a_shared = %a, %b_shared = %b, %c_shared = %c) ->
(!ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory, mutable>, !ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory, mutable>, !ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory, mutable>) {
scf.yield %b_shared, %a_shared, %a_shared : !ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory, mutable>, !ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory, mutable>, !ttg.memdesc<16x16xf16, #A_SHARED, #ttg.shared_memory, mutable>
}
tt.return
}

// CHECK-LABEL: for_if
tt.func @for_if(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f16>, %B : !tt.ptr<f16>, %i1 : i1) {
// CHECK: %0 -> %0
// expected-remark @below {{%0 -> %0}}
%a_shared_init = ttg.local_alloc : () -> !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK-NEXT: %1 -> %1
// expected-remark @below {{%1 -> %1}}
%b_shared_init = ttg.local_alloc : () -> !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK-NEXT: %2 -> %2
// expected-remark @below {{%2 -> %2}}
%c_shared_init = ttg.local_alloc : () -> !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK-NEXT: %arg7 -> %0
// CHECK-NEXT: %arg8 -> %1
// CHECK-NEXT: %arg9 -> %2
// CHECK-NEXT: %3#0 -> %0,%1
// CHECK-NEXT: %3#1 -> %0,%1
// CHECK-NEXT: %3#2 -> %0,%1,%2
// expected-remark @below {{%arg7 -> %0}}
// expected-remark @below {{%arg8 -> %1}}
// expected-remark @below {{%arg9 -> %2}}
// expected-remark @below {{%3#0 -> %0,%1}}
// expected-remark @below {{%3#1 -> %0,%1}}
// expected-remark @below {{%3#2 -> %0,%1,%2}}
%a_shared, %b_shared, %c_shared = scf.for %iv = %lb to %ub step %step iter_args(%a_shared = %a_shared_init, %b_shared = %b_shared_init, %c_shared = %c_shared_init) ->
(!ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>, !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>, !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>) {
scf.if %i1 {
%index = arith.constant 8 : i32
// CHECK-NEXT: %4 -> %0,%1
// expected-remark @below {{%4 -> %0,%1}}
%cst0 = ttg.memdesc_subview %a_shared[%index, %index] : !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable> -> !ttg.memdesc<32xf16, #A_SHARED, #ttg.shared_memory, mutable>
scf.yield
}
Expand All @@ -135,32 +126,31 @@ tt.func @for_if(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f16>, %B :
tt.return
}

// CHECK-LABEL: for_for_if
tt.func @for_for_if(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f16>, %B : !tt.ptr<f16>, %i1 : i1) {
// CHECK: %0 -> %0
// expected-remark @below {{%0 -> %0}}
%a_shared_init = ttg.local_alloc : () -> !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK-NEXT: %1 -> %1
// expected-remark @below {{%1 -> %1}}
%b_shared_init = ttg.local_alloc : () -> !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK-NEXT: %2 -> %2
// expected-remark @below {{%2 -> %2}}
%c_shared_init = ttg.local_alloc : () -> !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK-NEXT: %arg7 -> %0
// CHECK-NEXT: %arg8 -> %1
// CHECK-NEXT: %arg9 -> %2
// CHECK-NEXT: %3#0 -> %0
// CHECK-NEXT: %3#1 -> %1
// CHECK-NEXT: %3#2 -> %2,%6,%6
// expected-remark @below {{%arg7 -> %0}}
// expected-remark @below {{%arg8 -> %1}}
// expected-remark @below {{%arg9 -> %2}}
// expected-remark @below {{%3#0 -> %0}}
// expected-remark @below {{%3#1 -> %1}}
// expected-remark @below {{%3#2 -> %2,%6,%6}}
%a_shared, %b_shared, %c_shared = scf.for %iv = %lb to %ub step %step iter_args(%a_shared = %a_shared_init, %b_shared = %b_shared_init, %c_shared = %c_shared_init) ->
(!ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>, !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>, !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>) {
// CHECK-NEXT: %arg11 -> %2,%6,%6
// CHECK-NEXT: %4 -> %2,%6,%6
// expected-remark @below {{%arg11 -> %2,%6,%6}}
// expected-remark @below {{%4 -> %2,%6,%6}}
%c_shared_next = scf.for %jv = %lb to %ub step %step iter_args(%c_shared_next = %c_shared) -> (!ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>) {
// CHECK-NEXT: %5 -> %6,%6
// expected-remark @below {{%5 -> %6,%6}}
%c_shared_next_next = scf.if %i1 -> !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable> {
// CHECK-NEXT: %6 -> %6
// expected-remark @below {{%6 -> %6}}
%cst0 = ttg.local_alloc : () -> !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
scf.yield %cst0 : !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
} else {
// CHECK-NEXT: %6 -> %6
// expected-remark @below {{%6 -> %6}}
%cst0 = ttg.local_alloc : () -> !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
scf.yield %cst0 : !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
}
Expand All @@ -171,32 +161,31 @@ tt.func @for_for_if(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f16>,
tt.return
}

// CHECK-LABEL: cf_for
tt.func @cf_for(%arg0: index, %arg1: index, %arg2: index, %arg3: !tt.ptr<f16>, %arg4: !tt.ptr<f16>) {
%idx = arith.constant 0 : i32
// CHECK: %0 -> %0
// expected-remark @below {{%0 -> %0}}
%cst = ttg.local_alloc : () -> !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK-NEXT: %1 -> %1
// expected-remark @below {{%1 -> %1}}
%cst_0 = ttg.local_alloc : () -> !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK-NEXT: %2 -> %0
// expected-remark @below {{%2 -> %0}}
%0 = ttg.memdesc_subview %cst[%idx, %idx] : !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable> -> !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
gpu.barrier
// CHECK-NEXT: %3 -> %3
// expected-remark @below {{%3 -> %3}}
%cst_1 = ttg.local_alloc : () -> !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
// CHECK-NEXT: %5 -> %0,%1,%3
// CHECK-NEXT: %6 -> %0,%1,%3
// CHECK-NEXT: %7 -> %0,%1,%3
cf.br ^bb1(%arg0, %cst, %cst_0, %cst_1 : index, !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>, !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>, !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>)
^bb1(%1: index, %2: !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>, %3: !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>, %4: !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>): // 2 preds: ^bb0, ^bb2
%5 = arith.cmpi slt, %1, %arg1 : index
// expected-remark @below {{%5 -> %0,%1,%3}}
// expected-remark @below {{%6 -> %0,%1,%3}}
// expected-remark @below {{%7 -> %0,%1,%3}}
cf.cond_br %5, ^bb2, ^bb3
^bb2: // pred: ^bb1
gpu.barrier
%8 = arith.addi %1, %arg2 : index
cf.br ^bb1(%8, %4, %2, %3 : index, !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>, !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>, !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>)
^bb3: // pred: ^bb1
gpu.barrier
// CHECK-NEXT: %10 -> %0
// expected-remark @below {{%10 -> %0}}
%9 = ttg.memdesc_subview %0[%idx, %idx] : !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable> -> !ttg.memdesc<128x32xf16, #A_SHARED, #ttg.shared_memory, mutable>
tt.return
}
Expand Down
Loading
Loading