gaspersic/triton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[Triton-MLIR] tt.dot operands now must have DotOperand layout; also added prefetch pass prototype (#712)

Browse Source 
Co-authored-by: Jokeren <kerenzhou@openai.com>
Co-authored-by: Phil Tillet <phil@openai.com>
Co-authored-by: Superjomn <yanchunwei@outlook.com>
This commit is contained in:
Da Yan
2022-11-10 13:57:27 +08:00
committed by GitHub
parent 8832e32683
commit 4946167241
29 changed files with 1227 additions and 507 deletions
Download Patch File Download Diff File Expand all files Collapse all files

188
test/Analysis/test-allocation.mlir
View File

@@ -3,9 +3,11 @@
#AL = #triton_gpu.blocked<{sizePerThread = [1, 4], threadsPerWarp = [4, 8], warpsPerCTA = [4, 1], order = [1, 0]}>
#sliceAd0 = #triton_gpu.slice<{dim = 0, parent = #AL}>
#BL = #triton_gpu.blocked<{sizePerThread = [1, 4], threadsPerWarp = [1, 32], warpsPerCTA = [4, 1], order = [1, 0]}>
#A = #triton_gpu.shared<{vec = 2, perPhase = 2, maxPhase = 4, order = [1, 0]}>
#B = #triton_gpu.shared<{vec = 2, perPhase = 2, maxPhase = 4, order = [1, 0]}>
#A_SHARED = #triton_gpu.shared<{vec = 2, perPhase = 2, maxPhase = 4, order = [1, 0]}>
#B_SHARED = #triton_gpu.shared<{vec = 2, perPhase = 2, maxPhase = 4, order = [1, 0]}>
#C = #triton_gpu.mma<{version = 2, warpsPerCTA = [4, 1]}>
#A_DOT = #triton_gpu.dot_op<{opIdx = 0, parent = #C}>
#B_DOT = #triton_gpu.dot_op<{opIdx = 1, parent = #C}>
// CHECK-LABEL: matmul_loop
func @matmul_loop(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f16>, %B : !tt.ptr<f16>) {
@@ -23,20 +25,20 @@ func @matmul_loop(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f16>, %B
scf.for %iv = %lb to %ub step %step iter_args(%a_ptr = %a_ptr_init, %b_ptr = %b_ptr_init, %prev_c = %c_init) -> (tensor<128x32x!tt.ptr<f16>, #AL>, tensor<32x128x!tt.ptr<f16>, #BL>, tensor<128x128xf32, #C>) {
%a_ = tt.load %a_ptr, %a_mask, %a_other {cache = 1 : i32, evict = 1 : i32, isVolatile = false} : tensor<128x32xf16, #AL>
// CHECK: offset = 0, size = 8192
%a = triton_gpu.convert_layout %a_ : (tensor<128x32xf16, #AL>) -> tensor<128x32xf16, #A>
// CHECK: offset = 0, size = 4608
%a = triton_gpu.convert_layout %a_ : (tensor<128x32xf16, #AL>) -> tensor<128x32xf16, #A_DOT>
%b_ = tt.load %b_ptr, %b_mask, %b_other {cache = 1 : i32, evict = 1 : i32, isVolatile = false} : tensor<32x128xf16, #BL>
// CHECK-NEXT: offset = 8192, size = 8192
%b = triton_gpu.convert_layout %b_ : (tensor<32x128xf16, #BL>) -> tensor<32x128xf16, #B>
// CHECK-NEXT: offset = 0, size = 4224
%b = triton_gpu.convert_layout %b_ : (tensor<32x128xf16, #BL>) -> tensor<32x128xf16, #B_DOT>
%c = tt.dot %a, %b, %prev_c {allowTF32 = true, transA = false, transB = false} : tensor<128x32xf16, #A> * tensor<32x128xf16, #B> -> tensor<128x128xf32, #C>
%c = tt.dot %a, %b, %prev_c {allowTF32 = true, transA = false, transB = false} : tensor<128x32xf16, #A_DOT> * tensor<32x128xf16, #B_DOT> -> tensor<128x128xf32, #C>
%next_a_ptr = tt.addptr %a_ptr, %a_off : tensor<128x32x!tt.ptr<f16>, #AL>
%next_b_ptr = tt.addptr %b_ptr, %b_off : tensor<32x128x!tt.ptr<f16>, #BL>
scf.yield %next_a_ptr, %next_b_ptr, %c : tensor<128x32x!tt.ptr<f16>, #AL>, tensor<32x128x!tt.ptr<f16>, #BL>, tensor<128x128xf32, #C>
}
return
// CHECK-NEXT: size = 16384
// CHECK-NEXT: size = 4608
}
// Shared memory is available after a tensor's liveness range ends
@@ -51,21 +53,21 @@ func @reusable(%A : !tt.ptr<f16>) {
%a_ptr = tt.broadcast %A : (!tt.ptr<f16>) -> tensor<128x32x!tt.ptr<f16>, #AL>
%b_ptr = tt.broadcast %A : (!tt.ptr<f16>) -> tensor<32x128x!tt.ptr<f16>, #AL>
%a1_ = tt.load %a_ptr, %cst1, %cst2 {cache = 1 : i32, evict = 1 : i32, isVolatile = false} : tensor<128x32xf16, #AL>
// CHECK-NEXT: offset = 0, size = 8192
%a1 = triton_gpu.convert_layout %a1_ : (tensor<128x32xf16, #AL>) -> tensor<128x32xf16, #A>
// CHECK-NEXT: offset = 0, size = 4608
%a1 = triton_gpu.convert_layout %a1_ : (tensor<128x32xf16, #AL>) -> tensor<128x32xf16, #A_DOT>
%a2_ = tt.load %b_ptr, %cst3, %cst4 {cache = 1 : i32, evict = 1 : i32, isVolatile = false} : tensor<32x128xf16, #AL>
// CHECK-NEXT: offset = 8192, size = 8192
%a2 = triton_gpu.convert_layout %a2_ : (tensor<32x128xf16, #AL>) -> tensor<32x128xf16, #A>
// CHECK-NEXT: offset = 0, size = 1152
%a2 = triton_gpu.convert_layout %a2_ : (tensor<32x128xf16, #AL>) -> tensor<32x128xf16, #B_DOT>
%a3_ = tt.load %a_ptr, %cst1, %cst2 {cache = 1 : i32, evict = 1 : i32, isVolatile = false} : tensor<128x32xf16, #AL>
// CHECK-NEXT: offset = 16384, size = 8192
%a3 = triton_gpu.convert_layout %a3_ : (tensor<128x32xf16, #AL>) -> tensor<128x32xf16, #A>
%c = tt.dot %a1, %a2, %c_init {allowTF32 = true, transA = false, transB = false} : tensor<128x32xf16, #A> * tensor<32x128xf16, #B> -> tensor<128x128xf32, #C>
// CHECK-NEXT: offset = 0, size = 4608
%a3 = triton_gpu.convert_layout %a3_ : (tensor<128x32xf16, #AL>) -> tensor<128x32xf16, #A_DOT>
%c = tt.dot %a1, %a2, %c_init {allowTF32 = true, transA = false, transB = false} : tensor<128x32xf16, #A_DOT> * tensor<32x128xf16, #B_DOT> -> tensor<128x128xf32, #C>
%a4_ = tt.load %b_ptr, %cst3, %cst4 {cache = 1 : i32, evict = 1 : i32, isVolatile = false} : tensor<32x128xf16, #AL>
// CHECK-NEXT: offset = 0, size = 8192
%a4 = triton_gpu.convert_layout %a4_ : (tensor<32x128xf16, #AL>) -> tensor<32x128xf16, #A>
%c1 = tt.dot %a3, %a4, %c {allowTF32 = true, transA = false, transB = false} : tensor<128x32xf16, #A> * tensor<32x128xf16, #B> -> tensor<128x128xf32, #C>
// CHECK-NEXT: offset = 0, size = 1152
%a4 = triton_gpu.convert_layout %a4_ : (tensor<32x128xf16, #AL>) -> tensor<32x128xf16, #B_DOT>
%c1 = tt.dot %a3, %a4, %c {allowTF32 = true, transA = false, transB = false} : tensor<128x32xf16, #A_DOT> * tensor<32x128xf16, #B_DOT> -> tensor<128x128xf32, #C>
return
// CHECK-NEXT: size = 24576
// CHECK-NEXT: size = 4608
}
// A tensor's shared memory offset is larger than it needs to accommodate further tensors
@@ -75,33 +77,33 @@ func @reusable(%A : !tt.ptr<f16>) {
// CHECK-LABEL: preallocate
func @preallocate(%A : !tt.ptr<f16>) {
// CHECK: offset = 0, size = 512
%cst0 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst0 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 1024, size = 512
%cst1 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst1 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 1536, size = 512
%cst2 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst2 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 2048, size = 1024
%a = tt.cat %cst0, %cst1 {axis = 0} : (tensor<16x16xf16, #A>, tensor<16x16xf16, #A>) -> tensor<32x16xf16, #A>
%a = tt.cat %cst0, %cst1 {axis = 0} : (tensor<16x16xf16, #A_SHARED>, tensor<16x16xf16, #A_SHARED>) -> tensor<32x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 3072, size = 1024
%b = tt.cat %cst0, %cst2 {axis = 0} : (tensor<16x16xf16, #A>, tensor<16x16xf16, #A>) -> tensor<32x16xf16, #A>
%b = tt.cat %cst0, %cst2 {axis = 0} : (tensor<16x16xf16, #A_SHARED>, tensor<16x16xf16, #A_SHARED>) -> tensor<32x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 0, size = 1024
%c = tt.cat %cst1, %cst2 {axis = 0} : (tensor<16x16xf16, #A>, tensor<16x16xf16, #A>) -> tensor<32x16xf16, #A>
%c = tt.cat %cst1, %cst2 {axis = 0} : (tensor<16x16xf16, #A_SHARED>, tensor<16x16xf16, #A_SHARED>) -> tensor<32x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 1024, size = 1024
%cst4 = arith.constant dense<0.000000e+00> : tensor<32x16xf16, #A>
%cst4 = arith.constant dense<0.000000e+00> : tensor<32x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 6144, size = 2048
%e = tt.cat %a, %cst4 {axis = 0} : (tensor<32x16xf16, #A>, tensor<32x16xf16, #A>) -> tensor<64x16xf16, #A>
%e = tt.cat %a, %cst4 {axis = 0} : (tensor<32x16xf16, #A_SHARED>, tensor<32x16xf16, #A_SHARED>) -> tensor<64x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 8192, size = 2048
%d = tt.cat %b, %cst4 {axis = 0} : (tensor<32x16xf16, #A>, tensor<32x16xf16, #A>) -> tensor<64x16xf16, #A>
%d = tt.cat %b, %cst4 {axis = 0} : (tensor<32x16xf16, #A_SHARED>, tensor<32x16xf16, #A_SHARED>) -> tensor<64x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 10240, size = 2048
%f = tt.cat %c, %cst4 {axis = 0} : (tensor<32x16xf16, #A>, tensor<32x16xf16, #A>) -> tensor<64x16xf16, #A>
%f = tt.cat %c, %cst4 {axis = 0} : (tensor<32x16xf16, #A_SHARED>, tensor<32x16xf16, #A_SHARED>) -> tensor<64x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 0, size = 2048
%cst5 = arith.constant dense<0.000000e+00> : tensor<64x16xf16, #A>
%cst5 = arith.constant dense<0.000000e+00> : tensor<64x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 2048, size = 4096
%g = tt.cat %e, %cst5 {axis = 0} : (tensor<64x16xf16, #A>, tensor<64x16xf16, #A>) -> tensor<128x16xf16, #A>
%g = tt.cat %e, %cst5 {axis = 0} : (tensor<64x16xf16, #A_SHARED>, tensor<64x16xf16, #A_SHARED>) -> tensor<128x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 2048, size = 4096
%h = tt.cat %d, %cst5 {axis = 0} : (tensor<64x16xf16, #A>, tensor<64x16xf16, #A>) -> tensor<128x16xf16, #A>
%h = tt.cat %d, %cst5 {axis = 0} : (tensor<64x16xf16, #A_SHARED>, tensor<64x16xf16, #A_SHARED>) -> tensor<128x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 2048, size = 4096
%i = tt.cat %f, %cst5 {axis = 0} : (tensor<64x16xf16, #A>, tensor<64x16xf16, #A>) -> tensor<128x16xf16, #A>
%i = tt.cat %f, %cst5 {axis = 0} : (tensor<64x16xf16, #A_SHARED>, tensor<64x16xf16, #A_SHARED>) -> tensor<128x16xf16, #A_SHARED>
return
// CHECK-NEXT: size = 12288
}
@@ -110,13 +112,13 @@ func @preallocate(%A : !tt.ptr<f16>) {
// CHECK-LABEL: unused
func @unused(%A : !tt.ptr<f16>) {
// CHECK: offset = 0, size = 1024
%cst0 = arith.constant dense<0.000000e+00> : tensor<32x16xf16, #A>
%cst0 = arith.constant dense<0.000000e+00> : tensor<32x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 0, size = 512
%cst1 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst1 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 512, size = 512
%cst2 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst2 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 1024, size = 1024
%a = tt.cat %cst1, %cst2 {axis = 0} : (tensor<16x16xf16, #A>, tensor<16x16xf16, #A>) -> tensor<32x16xf16, #A>
%a = tt.cat %cst1, %cst2 {axis = 0} : (tensor<16x16xf16, #A_SHARED>, tensor<16x16xf16, #A_SHARED>) -> tensor<32x16xf16, #A_SHARED>
return
// CHECK: size = 2048
}
@@ -125,27 +127,27 @@ func @unused(%A : !tt.ptr<f16>) {
// CHECK-LABEL: longlive
func @longlive(%A : !tt.ptr<f16>) {
// CHECK: offset = 0, size = 512
%cst0 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst0 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 512, size = 512
%cst1 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst1 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 1024, size = 512
%cst2 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst2 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 1536, size = 1024
%a = tt.cat %cst1, %cst2 {axis = 0} : (tensor<16x16xf16, #A>, tensor<16x16xf16, #A>) -> tensor<32x16xf16, #A>
%a = tt.cat %cst1, %cst2 {axis = 0} : (tensor<16x16xf16, #A_SHARED>, tensor<16x16xf16, #A_SHARED>) -> tensor<32x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 512, size = 512
%cst3 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst3 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 1024, size = 512
%cst4 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst4 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 1536, size = 1024
%b = tt.cat %cst3, %cst4 {axis = 0} : (tensor<16x16xf16, #A>, tensor<16x16xf16, #A>) -> tensor<32x16xf16, #A>
%b = tt.cat %cst3, %cst4 {axis = 0} : (tensor<16x16xf16, #A_SHARED>, tensor<16x16xf16, #A_SHARED>) -> tensor<32x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 1536, size = 512
%cst5 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst5 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 1536, size = 512
%cst6 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst6 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 1536, size = 1024
%c = tt.cat %cst3, %cst4 {axis = 0} : (tensor<16x16xf16, #A>, tensor<16x16xf16, #A>) -> tensor<32x16xf16, #A>
%c = tt.cat %cst3, %cst4 {axis = 0} : (tensor<16x16xf16, #A_SHARED>, tensor<16x16xf16, #A_SHARED>) -> tensor<32x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 512, size = 1024
%d = tt.cat %cst0, %cst0 {axis = 0} : (tensor<16x16xf16, #A>, tensor<16x16xf16, #A>) -> tensor<32x16xf16, #A>
%d = tt.cat %cst0, %cst0 {axis = 0} : (tensor<16x16xf16, #A_SHARED>, tensor<16x16xf16, #A_SHARED>) -> tensor<32x16xf16, #A_SHARED>
return
// CHECK-NEXT: size = 2560
}
@@ -153,10 +155,10 @@ func @longlive(%A : !tt.ptr<f16>) {
// CHECK-LABEL: alloc
func @alloc(%A : !tt.ptr<f16>) {
// CHECK: offset = 0, size = 512
%cst0 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst0 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
%cst1 = arith.constant dense<0.000000e+00> : tensor<16x32xf16, #AL>
// CHECK-NEXT: offset = 0, size = 512
%cst2 = triton_gpu.alloc_tensor : tensor<16x16xf16, #A>
%cst2 = triton_gpu.alloc_tensor : tensor<16x16xf16, #A_SHARED>
return
// CHECK-NEXT: size = 512
}
@@ -176,9 +178,9 @@ func @insert_slice_async(%A : !tt.ptr<f16>, %i1 : i1) {
%mask = tt.splat %i1 : (i1) -> tensor<16x16xi1, #AL>
%other = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #AL>
// CHECK: offset = 0, size = 512
%tensor = arith.constant dense<0.000000e+00> : tensor<1x16x16xf16, #A>
%tensor = arith.constant dense<0.000000e+00> : tensor<1x16x16xf16, #A_SHARED>
%index = arith.constant 0 : i32
%a = triton_gpu.insert_slice_async %a_ptr, %tensor, %index, %mask, %other {axis = 0 : i32, cache = 1 : i32, evict = 1 : i32, isVolatile = false} : tensor<16x16x!tt.ptr<f16>, #AL> -> tensor<1x16x16xf16, #A>
%a = triton_gpu.insert_slice_async %a_ptr, %tensor, %index, %mask, %other {axis = 0 : i32, cache = 1 : i32, evict = 1 : i32, isVolatile = false} : tensor<16x16x!tt.ptr<f16>, #AL> -> tensor<1x16x16xf16, #A_SHARED>
return
// CHECK-NEXT: size = 512
}
@@ -186,9 +188,9 @@ func @insert_slice_async(%A : !tt.ptr<f16>, %i1 : i1) {
// CHECK-LABEL: extract_slice
func @extract_slice(%A : !tt.ptr<f16>) {
// CHECK: offset = 0, size = 512
%cst0 = arith.constant dense<0.000000e+00> : tensor<1x16x16xf16, #A>
%cst0 = arith.constant dense<0.000000e+00> : tensor<1x16x16xf16, #A_SHARED>
%index = arith.constant 0 : index
%cst1 = tensor.extract_slice %cst0[%index, 0, 0][1, 16, 16][1,1,1] : tensor<1x16x16xf16, #A> to tensor<16x16xf16, #A>
%cst1 = tensor.extract_slice %cst0[%index, 0, 0][1, 16, 16][1,1,1] : tensor<1x16x16xf16, #A_SHARED> to tensor<16x16xf16, #A_SHARED>
return
// CHECK-NEXT: size = 512
}
@@ -198,21 +200,21 @@ func @extract_slice(%A : !tt.ptr<f16>) {
// CHECK-LABEL: if
func @if(%i1 : i1) {
// CHECK: offset = 0, size = 512
%cst0 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst0 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 512, size = 512
%cst1 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst1 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
scf.if %i1 {
// CHECK-NEXT: offset = 1024, size = 1024
%a = tt.cat %cst0, %cst1 {axis = 0} : (tensor<16x16xf16, #A>, tensor<16x16xf16, #A>) -> tensor<32x16xf16, #A>
%a = tt.cat %cst0, %cst1 {axis = 0} : (tensor<16x16xf16, #A_SHARED>, tensor<16x16xf16, #A_SHARED>) -> tensor<32x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 1024, size = 1024
%b = tt.cat %cst0, %cst1 {axis = 0} : (tensor<16x16xf16, #A>, tensor<16x16xf16, #A>) -> tensor<32x16xf16, #A>
%b = tt.cat %cst0, %cst1 {axis = 0} : (tensor<16x16xf16, #A_SHARED>, tensor<16x16xf16, #A_SHARED>) -> tensor<32x16xf16, #A_SHARED>
}
// CHECK-NEXT: offset = 0, size = 512
%cst2 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst2 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 512, size = 512
%cst3 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst3 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 1024, size = 1024
%a = tt.cat %cst2, %cst3 {axis = 0} : (tensor<16x16xf16, #A>, tensor<16x16xf16, #A>) -> tensor<32x16xf16, #A>
%a = tt.cat %cst2, %cst3 {axis = 0} : (tensor<16x16xf16, #A_SHARED>, tensor<16x16xf16, #A_SHARED>) -> tensor<32x16xf16, #A_SHARED>
return
// CHECK-NEXT: size = 2048
}
@@ -222,24 +224,24 @@ func @if(%i1 : i1) {
// CHECK-LABEL: if_else
func @if_else(%i1 : i1) {
// CHECK: offset = 0, size = 512
%cst0 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst0 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 512, size = 512
%cst1 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst1 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
scf.if %i1 {
// CHECK-NEXT: offset = 1024, size = 1024
%a = tt.cat %cst0, %cst1 {axis = 0} : (tensor<16x16xf16, #A>, tensor<16x16xf16, #A>) -> tensor<32x16xf16, #A>
%a = tt.cat %cst0, %cst1 {axis = 0} : (tensor<16x16xf16, #A_SHARED>, tensor<16x16xf16, #A_SHARED>) -> tensor<32x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 1024, size = 1024
%b = tt.cat %cst0, %cst1 {axis = 0} : (tensor<16x16xf16, #A>, tensor<16x16xf16, #A>) -> tensor<32x16xf16, #A>
%b = tt.cat %cst0, %cst1 {axis = 0} : (tensor<16x16xf16, #A_SHARED>, tensor<16x16xf16, #A_SHARED>) -> tensor<32x16xf16, #A_SHARED>
} else {
// CHECK-NEXT: offset = 1024, size = 512
%cst2 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst2 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 1536, size = 512
%cst3 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A>
%cst3 = arith.constant dense<0.000000e+00> : tensor<16x16xf16, #A_SHARED>
// CHECK-NEXT: offset = 2048, size = 1024
%a = tt.cat %cst2, %cst3 {axis = 0} : (tensor<16x16xf16, #A>, tensor<16x16xf16, #A>) -> tensor<32x16xf16, #A>
%a = tt.cat %cst2, %cst3 {axis = 0} : (tensor<16x16xf16, #A_SHARED>, tensor<16x16xf16, #A_SHARED>) -> tensor<32x16xf16, #A_SHARED>
}
// CHECK-NEXT: offset = 1024, size = 1024
%a = tt.cat %cst0, %cst1 {axis = 0} : (tensor<16x16xf16, #A>, tensor<16x16xf16, #A>) -> tensor<32x16xf16, #A>
%a = tt.cat %cst0, %cst1 {axis = 0} : (tensor<16x16xf16, #A_SHARED>, tensor<16x16xf16, #A_SHARED>) -> tensor<32x16xf16, #A_SHARED>
return
// CHECK-NEXT: size = 3072
}
@@ -249,13 +251,13 @@ func @if_else(%i1 : i1) {
// CHECK-LABEL: for
func @for(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f16>, %B : !tt.ptr<f16>) {
// CHECK: offset = 0, size = 8192
%a_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A>
%a_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A_SHARED>
// CHECK-NEXT: offset = 8192, size = 8192
%b_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A>
%b_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A_SHARED>
// CHECK-NEXT: offset = 16384, size = 8192
%c_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A>
%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) -> (tensor<128x32xf16, #A>, tensor<128x32xf16, #A>, tensor<128x32xf16, #A>) {
scf.yield %b_shared, %a_shared, %a_shared : tensor<128x32xf16, #A>, tensor<128x32xf16, #A>, tensor<128x32xf16, #A>
%c_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A_SHARED>
%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) -> (tensor<128x32xf16, #A_SHARED>, tensor<128x32xf16, #A_SHARED>, tensor<128x32xf16, #A_SHARED>) {
scf.yield %b_shared, %a_shared, %a_shared : tensor<128x32xf16, #A_SHARED>, tensor<128x32xf16, #A_SHARED>, tensor<128x32xf16, #A_SHARED>
}
return
// CHECK-NEXT: size = 24576
@@ -264,18 +266,18 @@ func @for(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f16>, %B : !tt.p
// CHECK-LABEL: for_if_slice
func @for_if_slice(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f16>, %B : !tt.ptr<f16>, %i1 : i1) {
// CHECK: offset = 0, size = 8192
%a_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A>
%a_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A_SHARED>
// CHECK-NEXT: offset = 8192, size = 8192
%b_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A>
%b_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A_SHARED>
// CHECK-NEXT: offset = 16384, size = 8192
%c_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A>
%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) -> (tensor<128x32xf16, #A>, tensor<128x32xf16, #A>, tensor<128x32xf16, #A>) {
%c_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A_SHARED>
%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) -> (tensor<128x32xf16, #A_SHARED>, tensor<128x32xf16, #A_SHARED>, tensor<128x32xf16, #A_SHARED>) {
scf.if %i1 {
%index = arith.constant 8 : index
%cst0 = tensor.extract_slice %a_shared[%index, 0][1, 32][1, 1] : tensor<128x32xf16, #A> to tensor<32xf16, #A>
%cst0 = tensor.extract_slice %a_shared[%index, 0][1, 32][1, 1] : tensor<128x32xf16, #A_SHARED> to tensor<32xf16, #A_SHARED>
scf.yield
}
scf.yield %b_shared, %a_shared, %a_shared : tensor<128x32xf16, #A>, tensor<128x32xf16, #A>, tensor<128x32xf16, #A>
scf.yield %b_shared, %a_shared, %a_shared : tensor<128x32xf16, #A_SHARED>, tensor<128x32xf16, #A_SHARED>, tensor<128x32xf16, #A_SHARED>
}
return
// CHECK-NEXT: size = 24576
@@ -286,28 +288,28 @@ func @for_if_slice(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f16>, %
// CHECK-LABEL: for_if_for
func @for_if_for(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f16>, %B : !tt.ptr<f16>, %i1 : i1) {
// CHECK: offset = 0, size = 8192
%a_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A>
%a_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A_SHARED>
// CHECK-NEXT: offset = 8192, size = 8192
%b_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A>
%b_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A_SHARED>
// CHECK-NEXT: offset = 16384, size = 8192
%c_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A>
%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) -> (tensor<128x32xf16, #A>, tensor<128x32xf16, #A>, tensor<128x32xf16, #A>) {
%c_shared_next = scf.for %jv = %lb to %ub step %step iter_args(%c_shared_next = %c_shared) -> (tensor<128x32xf16, #A>) {
%c_shared_next_next = scf.if %i1 -> tensor<128x32xf16, #A> {
%c_shared_init = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A_SHARED>
%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) -> (tensor<128x32xf16, #A_SHARED>, tensor<128x32xf16, #A_SHARED>, tensor<128x32xf16, #A_SHARED>) {
%c_shared_next = scf.for %jv = %lb to %ub step %step iter_args(%c_shared_next = %c_shared) -> (tensor<128x32xf16, #A_SHARED>) {
%c_shared_next_next = scf.if %i1 -> tensor<128x32xf16, #A_SHARED> {
// CHECK-NEXT: offset = 24576, size = 8192
%cst0 = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A>
scf.yield %cst0 : tensor<128x32xf16, #A>
%cst0 = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A_SHARED>
scf.yield %cst0 : tensor<128x32xf16, #A_SHARED>
} else {
// CHECK-NEXT: offset = 32768, size = 8192
%cst1 = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A>
scf.yield %cst1 : tensor<128x32xf16, #A>
%cst1 = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A_SHARED>
scf.yield %cst1 : tensor<128x32xf16, #A_SHARED>
}
scf.yield %c_shared_next_next : tensor<128x32xf16, #A>
scf.yield %c_shared_next_next : tensor<128x32xf16, #A_SHARED>
}
scf.yield %a_shared, %b_shared, %c_shared_next : tensor<128x32xf16, #A>, tensor<128x32xf16, #A>, tensor<128x32xf16, #A>
scf.yield %a_shared, %b_shared, %c_shared_next : tensor<128x32xf16, #A_SHARED>, tensor<128x32xf16, #A_SHARED>, tensor<128x32xf16, #A_SHARED>
}
// CHECK-NEXT: offset = 0, size = 8192
%cst2 = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A>
%cst2 = arith.constant dense<0.00e+00> : tensor<128x32xf16, #A_SHARED>
return
// CHECK-NEXT: size = 40960
}