This is an archive of the discontinued LLVM Phabricator instance.

Differential D95387

[flang] Fix errors in ISO_FORTRAN_ENV module for REAL128
ClosedPublic

Authored by klausler on Jan 25 2021, 12:28 PM.

Details

Reviewers
PeteSteinfeld
Commits
rG03f1c9fa85d4: [flang] Fix errors in ISO_FORTRAN_ENV module for REAL128
Summary

Some parameters of 128-bit IEEE floating-point numbers were
incorrect, leading to a failure to define REAL128 to 16.

Diff Detail

Event Timeline

klausler created this revision.Jan 25 2021, 12:28 PM
Herald added a subscriber: jdoerfert. · View Herald TranscriptJan 25 2021, 12:28 PM
klausler requested review of this revision.Jan 25 2021, 12:28 PM
Herald added a project: Restricted Project. · View Herald TranscriptJan 25 2021, 12:28 PM
Herald added a subscriber: llvm-commits. · View Herald Transcript
Harbormaster completed remote builds in B86615: Diff 319089.Jan 25 2021, 1:10 PM
PeteSteinfeld accepted this revision.Jan 25 2021, 1:23 PM

All builds, tests, and looks good.

This revision is now accepted and ready to land.Jan 25 2021, 1:23 PM
Closed by commit rG03f1c9fa85d4: [flang] Fix errors in ISO_FORTRAN_ENV module for REAL128 (authored by klausler). · Explain WhyJan 25 2021, 1:41 PM
This revision was automatically updated to reflect the committed changes.
klausler added a commit: rG03f1c9fa85d4: [flang] Fix errors in ISO_FORTRAN_ENV module for REAL128.

Revision Contents

PathSize
flang/
module/
4 lines

Diff 319089

flang/module/iso_fortran_env.f90

Context not available.
selectedReal64 = selected_real_kind(15, 307), & ! IEEE double selectedReal64 = selected_real_kind(15, 307), & ! IEEE double
selectedReal80 = selected_real_kind(18, 4931), & ! 80x87 extended selectedReal80 = selected_real_kind(18, 4931), & ! 80x87 extended
selectedReal64x2 = selected_real_kind(31, 307), & ! "double-double" selectedReal64x2 = selected_real_kind(31, 307), & ! "double-double"
selectedReal128 = selected_real_kind(33, 9863), & ! IEEE quad selectedReal128 = selected_real_kind(33, 4931), & ! IEEE quad
safeReal16 = merge(selectedReal16, selected_real_kind(0,0), & safeReal16 = merge(selectedReal16, selected_real_kind(0,0), &
selectedReal16 >= 0), & selectedReal16 >= 0), &
safeBfloat16 = merge(selectedBfloat16, selected_real_kind(0,0), & safeBfloat16 = merge(selectedBfloat16, selected_real_kind(0,0), &
Context not available.
real64x2 = merge(selectedReal64x2, merge(-2, -1, selectedReal64x2 >= 0), & real64x2 = merge(selectedReal64x2, merge(-2, -1, selectedReal64x2 >= 0), &
digits(real(0,kind=safeReal64x2)) == 106), & digits(real(0,kind=safeReal64x2)) == 106), &
real128 = merge(selectedReal128, merge(-2, -1, selectedReal128 >= 0), & real128 = merge(selectedReal128, merge(-2, -1, selectedReal128 >= 0), &
digits(real(0,kind=safeReal128)) == 112) digits(real(0,kind=safeReal128)) == 113)
integer, parameter :: real_kinds(*) = [ & integer, parameter :: real_kinds(*) = [ &
[(real16, integer :: j=1, count([real16 >= 0]))], & [(real16, integer :: j=1, count([real16 >= 0]))], &
Context not available.