[TritonIR] make other optional and remove isOtherUnspecified (#67)
[Triton] make other optional and remove isOtherUnspecified
This commit is contained in:
@@ -33,9 +33,9 @@ func @matmul_loop(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f16>, %B
|
||||
%b_off = arith.constant dense<4> : tensor<32x128xi32, #BL>
|
||||
|
||||
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, isOtherUnspecified = false, isVolatile = false} : tensor<128x32xf16, #AL>
|
||||
%a_ = tt.load %a_ptr, %a_mask, %a_other {cache = 1 : i32, evict = 1 : i32, isVolatile = false} : tensor<128x32xf16, #AL>
|
||||
%a = triton_gpu.convert_layout %a_ : (tensor<128x32xf16, #AL>) -> tensor<128x32xf16, #A>
|
||||
%b_ = tt.load %b_ptr, %b_mask, %b_other {cache = 1 : i32, evict = 1 : i32, isOtherUnspecified = false, isVolatile = false} : tensor<32x128xf16, #BL>
|
||||
%b_ = tt.load %b_ptr, %b_mask, %b_other {cache = 1 : i32, evict = 1 : i32, isVolatile = false} : tensor<32x128xf16, #BL>
|
||||
%b = triton_gpu.convert_layout %b_ : (tensor<32x128xf16, #BL>) -> tensor<32x128xf16, #B>
|
||||
|
||||
%c = tt.dot %a, %b, %prev_c {allowTF32 = true} : tensor<128x32xf16, #A> * tensor<32x128xf16, #B> -> tensor<128x128xf32, #C>
|
||||
@@ -75,9 +75,9 @@ func @matmul_loop_nested(%lb : index, %ub : index, %step : index, %A : !tt.ptr<f
|
||||
%b_off = arith.constant dense<4> : tensor<32x128xi32, #BL>
|
||||
|
||||
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, isOtherUnspecified = false, isVolatile = false} : tensor<128x32xf16, #AL>
|
||||
%a_ = tt.load %a_ptr, %a_mask, %a_other {cache = 1 : i32, evict = 1 : i32, isVolatile = false} : tensor<128x32xf16, #AL>
|
||||
%a = triton_gpu.convert_layout %a_ : (tensor<128x32xf16, #AL>) -> tensor<128x32xf16, #A>
|
||||
%b_ = tt.load %b_ptr, %b_mask, %b_other {cache = 1 : i32, evict = 1 : i32, isOtherUnspecified = false, isVolatile = false} : tensor<32x128xf16, #BL>
|
||||
%b_ = tt.load %b_ptr, %b_mask, %b_other {cache = 1 : i32, evict = 1 : i32, isVolatile = false} : tensor<32x128xf16, #BL>
|
||||
%b = triton_gpu.convert_layout %b_ : (tensor<32x128xf16, #BL>) -> tensor<32x128xf16, #B>
|
||||
|
||||
%c = tt.dot %a, %b, %prev_c {allowTF32 = true} : tensor<128x32xf16, #A> * tensor<32x128xf16, #B> -> tensor<128x128xf32, #C>
|
||||
@@ -106,7 +106,7 @@ func @matmul_loop_single_pipeline(%lb : index, %ub : index, %step : index, %A :
|
||||
%a_mask = arith.constant dense<true> : tensor<128x32xi1, #AL>
|
||||
%a_other = arith.constant dense<0.00e+00> : tensor<128x32xf16, #AL>
|
||||
|
||||
%a_ = tt.load %a_ptr_init, %a_mask, %a_other {cache = 1 : i32, evict = 1 : i32, isOtherUnspecified = false, isVolatile = false} : tensor<128x32xf16, #AL>
|
||||
%a_ = tt.load %a_ptr_init, %a_mask, %a_other {cache = 1 : i32, evict = 1 : i32, isVolatile = false} : tensor<128x32xf16, #AL>
|
||||
%a = triton_gpu.convert_layout %a_ : (tensor<128x32xf16, #AL>) -> tensor<128x32xf16, #A>
|
||||
|
||||
%b_mask = arith.constant dense<true> : tensor<32x128xi1, #BL>
|
||||
@@ -116,7 +116,7 @@ func @matmul_loop_single_pipeline(%lb : index, %ub : index, %step : index, %A :
|
||||
%b_off = arith.constant dense<4> : tensor<32x128xi32, #BL>
|
||||
|
||||
scf.for %iv = %lb to %ub step %step iter_args(%b_ptr = %b_ptr_init, %prev_c = %c_init) -> (tensor<32x128x!tt.ptr<f16>, #BL>, tensor<128x128xf32, #C>) {
|
||||
%b_ = tt.load %b_ptr, %b_mask, %b_other {cache = 1 : i32, evict = 1 : i32, isOtherUnspecified = false, isVolatile = false} : tensor<32x128xf16, #BL>
|
||||
%b_ = tt.load %b_ptr, %b_mask, %b_other {cache = 1 : i32, evict = 1 : i32, isVolatile = false} : tensor<32x128xf16, #BL>
|
||||
%b = triton_gpu.convert_layout %b_ : (tensor<32x128xf16, #BL>) -> tensor<32x128xf16, #B>
|
||||
%c = tt.dot %a, %b, %prev_c {allowTF32 = true} : tensor<128x32xf16, #A> * tensor<32x128xf16, #B> -> tensor<128x128xf32, #C>
|
||||
%next_b_ptr = tt.getelementptr %b_ptr, %b_off : tensor<32x128x!tt.ptr<f16>, #BL>
|
||||
|
Reference in New Issue
Block a user