Skip to content

[flang][cuda] Add more interfaces for __ldca, __ldcs, __ldlu and __ldcv #130218

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

Merged
merged 2 commits into from
Mar 7, 2025
Merged

[flang][cuda] Add more interfaces for __ldca, __ldcs, __ldlu and __ldcv #130218

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
c954784
[flang][cuda] Add more interfaces for __ldca, __ldcs, __ldlu and __ldcv
clementval Mar 7, 2025
2dc746b
Add more tests
clementval Mar 7, 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
100 changes: 100 additions & 0 deletions flang/module/cudadevice.f90
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1118,6 +1118,31 @@ attributes(device) integer function match_any_syncjd(mask, val)
!dir$ ignore_tkr (d) x
complex(8), intent(in) :: x
end function
attributes(device) pure function __ldca_i4x4(x) result(y)
!dir$ ignore_tkr (d) x
integer(4), dimension(4), intent(in) :: x
integer(4), dimension(4) :: y
end function
attributes(device) pure function __ldca_i8x2(x) result(y)
!dir$ ignore_tkr (d) x
integer(8), dimension(2), intent(in) :: x
integer(8), dimension(2) :: y
end function
attributes(device) pure function __ldca_r2x2(x) result(y)
!dir$ ignore_tkr (d) x
real(2), dimension(2), intent(in) :: x
real(2), dimension(2) :: y
end function
attributes(device) pure function __ldca_r4x4(x) result(y)
!dir$ ignore_tkr (d) x
real(4), dimension(4), intent(in) :: x
real(4), dimension(4) :: y
end function
attributes(device) pure function __ldca_r8x2(x) result(y)
!dir$ ignore_tkr (d) x
real(8), dimension(2), intent(in) :: x
real(8), dimension(2) :: y
end function
end interface

! LDCS
Expand Down Expand Up @@ -1158,6 +1183,31 @@ attributes(device) integer function match_any_syncjd(mask, val)
!dir$ ignore_tkr (d) x
complex(8), intent(in) :: x
end function
attributes(device) pure function __ldcs_i4x4(x) result(y)
!dir$ ignore_tkr (d) x
integer(4), dimension(4), intent(in) :: x
integer(4), dimension(4) :: y
end function
attributes(device) pure function __ldcs_i8x2(x) result(y)
!dir$ ignore_tkr (d) x
integer(8), dimension(2), intent(in) :: x
integer(8), dimension(2) :: y
end function
attributes(device) pure function __ldcs_r2x2(x) result(y)
!dir$ ignore_tkr (d) x
real(2), dimension(2), intent(in) :: x
real(2), dimension(2) :: y
end function
attributes(device) pure function __ldcs_r4x4(x) result(y)
!dir$ ignore_tkr (d) x
real(4), dimension(4), intent(in) :: x
real(4), dimension(4) :: y
end function
attributes(device) pure function __ldcs_r8x2(x) result(y)
!dir$ ignore_tkr (d) x
real(8), dimension(2), intent(in) :: x
real(8), dimension(2) :: y
end function
end interface

! LDLU
Expand Down Expand Up @@ -1198,6 +1248,31 @@ attributes(device) integer function match_any_syncjd(mask, val)
!dir$ ignore_tkr (d) x
complex(8), intent(in) :: x
end function
attributes(device) pure function __ldlu_i4x4(x) result(y)
!dir$ ignore_tkr (d) x
integer(4), dimension(4), intent(in) :: x
integer(4), dimension(4) :: y
end function
attributes(device) pure function __ldlu_i8x2(x) result(y)
!dir$ ignore_tkr (d) x
integer(8), dimension(2), intent(in) :: x
integer(8), dimension(2) :: y
end function
attributes(device) pure function __ldlu_r2x2(x) result(y)
!dir$ ignore_tkr (d) x
real(2), dimension(2), intent(in) :: x
real(2), dimension(2) :: y
end function
attributes(device) pure function __ldlu_r4x4(x) result(y)
!dir$ ignore_tkr (d) x
real(4), dimension(4), intent(in) :: x
real(4), dimension(4) :: y
end function
attributes(device) pure function __ldlu_r8x2(x) result(y)
!dir$ ignore_tkr (d) x
real(8), dimension(2), intent(in) :: x
real(8), dimension(2) :: y
end function
end interface

! LDCV
Expand Down Expand Up @@ -1238,6 +1313,31 @@ attributes(device) integer function match_any_syncjd(mask, val)
!dir$ ignore_tkr (d) x
complex(8), intent(in) :: x
end function
attributes(device) pure function __ldcv_i4x4(x) result(y)
!dir$ ignore_tkr (d) x
integer(4), dimension(4), intent(in) :: x
integer(4), dimension(4) :: y
end function
attributes(device) pure function __ldcv_i8x2(x) result(y)
!dir$ ignore_tkr (d) x
integer(8), dimension(2), intent(in) :: x
integer(8), dimension(2) :: y
end function
attributes(device) pure function __ldcv_r2x2(x) result(y)
!dir$ ignore_tkr (d) x
real(2), dimension(2), intent(in) :: x
real(2), dimension(2) :: y
end function
attributes(device) pure function __ldcv_r4x4(x) result(y)
!dir$ ignore_tkr (d) x
real(4), dimension(4), intent(in) :: x
real(4), dimension(4) :: y
end function
attributes(device) pure function __ldcv_r8x2(x) result(y)
!dir$ ignore_tkr (d) x
real(8), dimension(2), intent(in) :: x
real(8), dimension(2) :: y
end function
end interface

! STWB
Expand Down
80 changes: 80 additions & 0 deletions flang/test/Lower/CUDA/cuda-device-proc.cuf
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -198,3 +198,83 @@ end subroutine
! CHECK: %[[CAST:.*]] = builtin.unrealized_conversion_cast %{{.*}}#1 : !fir.ref<f64> to !llvm.ptr
! CHECK: %[[ATOMIC:.*]] = llvm.cmpxchg %[[CAST]], %[[BCAST1]], %[[BCAST2]] acq_rel monotonic : !llvm.ptr, i64
! CHECK: %[[RES:.*]] = llvm.extractvalue %[[ATOMIC]][1] : !llvm.struct<(i64, i1)>

attributes(global) subroutine __ldXXi4(b)
integer, device :: b(*)
integer, device :: x(4)
x(1:4) = __ldca(b(i:j))
x = __ldcg(b(i:j))
x = __ldcs(b(i:j))
x(1:4) = __ldlu(b(i:j))
x(1:4) = __ldcv(b(i:j))
end

! CHECK-LABEL: func.func @_QP__ldxxi4
! CHECK: __ldca_i4x4
! CHECK: __ldcg_i4x4
! CHECK: __ldcs_i4x4
! CHECK: __ldlu_i4x4

attributes(global) subroutine __ldXXi8(b)
integer(8), device :: b(*)
integer(8), device :: x(2)
x(1:2) = __ldca(b(i:j))
x = __ldcg(b(i:j))
x = __ldcs(b(i:j))
x(1:2) = __ldlu(b(i:j))
x(1:2) = __ldcv(b(i:j))
end

! CHECK-LABEL: func.func @_QP__ldxxi8
! CHECK: __ldca_i8x2
! CHECK: __ldcg_i8x2
! CHECK: __ldcs_i8x2
! CHECK: __ldlu_i8x2

attributes(global) subroutine __ldXXr4(b)
real, device :: b(*)
real, device :: x(4)
x(1:4) = __ldca(b(i:j))
x = __ldcg(b(i:j))
x = __ldcs(b(i:j))
x(1:4) = __ldlu(b(i:j))
x(1:4) = __ldcv(b(i:j))
end

! CHECK-LABEL: func.func @_QP__ldxxr4
! CHECK: __ldca_r4x4
! CHECK: __ldcg_r4x4
! CHECK: __ldcs_r4x4
! CHECK: __ldlu_r4x4

attributes(global) subroutine __ldXXr2(b)
real(2), device :: b(*)
real(2), device :: x(2)
x(1:2) = __ldca(b(i:j))
x = __ldcg(b(i:j))
x = __ldcs(b(i:j))
x(1:2) = __ldlu(b(i:j))
x(1:2) = __ldcv(b(i:j))
end

! CHECK-LABEL: func.func @_QP__ldxxr2
! CHECK: __ldca_r2x2
! CHECK: __ldcg_r2x2
! CHECK: __ldcs_r2x2
! CHECK: __ldlu_r2x2

attributes(global) subroutine __ldXXr8(b)
real(8), device :: b(*)
real(8), device :: x(2)
x(1:2) = __ldca(b(i:j))
x = __ldcg(b(i:j))
x = __ldcs(b(i:j))
x(1:2) = __ldlu(b(i:j))
x(1:2) = __ldcv(b(i:j))
end

! CHECK-LABEL: func.func @_QP__ldxxr8
! CHECK: __ldca_r8x2
! CHECK: __ldcg_r8x2
! CHECK: __ldcs_r8x2
! CHECK: __ldlu_r8x2