- cmucl-cvs - mailman3.common-lisp.net

[Git][cmucl/cmucl][issue-425-correctly-rounded-math-functions] TEMPORARY: Use clang in Config.x86_linux
by Raymond Toy (＠rtoy) 09 Aug '25

09 Aug '25

Raymond Toy pushed to branch issue-425-correctly-rounded-math-functions at cmucl / cmucl Commits: 44d0795c by Raymond Toy at 2025-08-09T06:50:37-07:00 TEMPORARY: Use clang in Config.x86_linux See if clang on the CI machines can build with core-math. If so, we need to modify .gitlab-ci.yml so that we can use Config.x86_linux_clang to build instead of the default. - - - - - 1 changed file: - src/lisp/Config.x86_linux Changes: ===================================== src/lisp/Config.x86_linux ===================================== @@ -1,6 +1,9 @@ # -*- Mode: makefile -*- include Config.x86_common +# TEMP FIX! +CC = clang + # We want errno support because we used to have that. CORE_MATH = -DCORE_MATH_SUPPORT_ERRNO View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/commit/44d0795cfada2045adb9d39… -- View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/commit/44d0795cfada2045adb9d39… You're receiving this email because of your account on gitlab.common-lisp.net.

1 0

[Git][cmucl/cmucl][issue-425-correctly-rounded-math-functions] 6 commits: The float-modes type is a 30 bit unsigned byte
by Raymond Toy (＠rtoy) 09 Aug '25

09 Aug '25

Raymond Toy pushed to branch issue-425-correctly-rounded-math-functions at cmucl / cmucl Commits: 70542bd9 by Raymond Toy at 2025-08-07T17:10:40-07:00 The float-modes type is a 30 bit unsigned byte Since we use a single 32-bit word to set the x87 FP modes, we need at least 30 bits because the top 2 bits are reserved so they shouldn't be set. This is because the control word is in the top 16 bits. Previously, the float-modes type was an (unsigned-byte 24), which is too small now. - - - - - 5edb2d90 by Raymond Toy at 2025-08-08T06:49:06-07:00 Fix #426: Define float-modes type correctly The definition of `float-modes` is `(unsigned-byte 24)`. This is too small if we are trying to set the precision control field for x87 via `(setf (x86::x87-floating-point-modes) ...)`. The field needs to be at least 30 bits to that we can assign values for the x87 control and status words. Thus, set the type to `(unsigned-byte 32)`. - - - - - 387ed142 by Raymond Toy at 2025-08-08T07:11:42-07:00 Add test for (setf (x86::x87-floating-point-modes)) - - - - - 4d1f287c by Raymond Toy at 2025-08-08T18:29:40-07:00 Merge branch 'issue-426-fix-float-modes-type-def' into issue-425-correctly-rounded-math-functions - - - - - 84ee222f by Raymond Toy at 2025-08-08T19:59:15-07:00 lisp.a needs to have the core math routines too. - - - - - f47f8f07 by Raymond Toy at 2025-08-08T20:00:17-07:00 Set the x87 precision to double when calling out to C When we call out to C, we save the x87 FPU control word. Then update the FPU control to have the precision set to double (53-bit) precision. This is needed for the core-math routines that depend on the x87 precision control to be double. When the function returns, we restore the original x87 control word. - - - - - 4 changed files: - src/compiler/x86/float-sse2.lisp - src/compiler/x86/sse2-c-call.lisp - src/lisp/GNUmakefile - tests/float.lisp Changes: ===================================== src/compiler/x86/float-sse2.lisp ===================================== @@ -1265,7 +1265,7 @@ ;;;; Float mode hackery: -(deftype float-modes () '(unsigned-byte 24)) +(deftype float-modes () '(unsigned-byte 30)) ;; For the record, here is the format of the MXCSR register. ;; ===================================== src/compiler/x86/sse2-c-call.lisp ===================================== @@ -36,12 +36,23 @@ :from :eval :to :result) edx) (:temporary (:sc single-stack) temp-single) (:temporary (:sc double-stack) temp-double) + (:temporary (:sc unsigned-stack) save-fpu-cw) + (:temporary (:sc unsigned-stack) fpu-cw) (:node-var node) (:vop-var vop) (:save-p t) - (:ignore args ecx edx) + (:ignore args ecx #+nil edx) (:guard (backend-featurep :sse2)) - (:generator 0 + (:generator 0 + ;; Save the x87 FPU control word. Then modify it to set the + ;; precision bits to double (2). + (inst fnstcw save-fpu-cw) + (move edx save-fpu-cw) + (inst and edx (dpb 0 (byte 2 8) #xffff)) ; Zap the precision control bits + (inst or edx (dpb 2 (byte 2 8) 0)) ; Set precision control to double + (move fpu-cw edx) + (inst fldcw fpu-cw) ; New CW + (cond ((policy node (> space speed)) (move eax function) (inst call (make-fixup (extern-alien-name "call_into_c") :foreign))) @@ -63,7 +74,9 @@ (inst movss xmm0-tn (ea-for-sf-stack temp-single))) (double-reg (inst fstpd (ea-for-df-stack temp-double)) - (inst movsd xmm0-tn (ea-for-df-stack temp-double))))))) + (inst movsd xmm0-tn (ea-for-df-stack temp-double))))) + ;; Restore the x87 FPU control settings + (inst fldcw save-fpu-cw))) (define-vop (alloc-number-stack-space) (:info amount) ===================================== src/lisp/GNUmakefile ===================================== @@ -83,7 +83,7 @@ lisp: ${OBJS} ${CORE64_OBJS} version.o # Create a library out of all the object files so we can build an # executable. However, we need to remove exec-init.o from the library lisp.a: version.o ${OBJS} ${CORE64_OBJS} ${EXEC_FINAL_OBJ} - $(AR) crs lisp.a ${OBJS} version.o + $(AR) crs lisp.a ${OBJS} ${CORE64_OBJS} version.o ifneq (${EXEC_FINAL_OBJ},) $(AR) d lisp.a exec-init.o endif ===================================== tests/float.lisp ===================================== @@ -333,3 +333,13 @@ (declare (ignore c)) (values (invoke-restart 'lisp::largest-float))))) (read-from-string string)))))) + +#+x86 +(define-test x87-set-floating-point-modes + (:tag :issues) + (let ((new-mode (setf (x86::x87-floating-point-modes) + (dpb 2 (byte 2 24) + (x86::x87-floating-point-modes))))) + (assert-true (typep new-mode 'x86::float-modes)) + (assert-equal new-mode (setf (x86::x87-floating-point-modes) new-mode)))) + View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/compare/a38612dca7b478767b47a6… -- View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/compare/a38612dca7b478767b47a6… You're receiving this email because of your account on gitlab.common-lisp.net.

1 0

[Git][cmucl/cmucl][issue-426-fix-float-modes-type-def] Add test for (setf (x86::x87-floating-point-modes))
by Raymond Toy (＠rtoy) 08 Aug '25

08 Aug '25

Raymond Toy pushed to branch issue-426-fix-float-modes-type-def at cmucl / cmucl Commits: 387ed142 by Raymond Toy at 2025-08-08T07:11:42-07:00 Add test for (setf (x86::x87-floating-point-modes)) - - - - - 1 changed file: - tests/float.lisp Changes: ===================================== tests/float.lisp ===================================== @@ -333,3 +333,13 @@ (declare (ignore c)) (values (invoke-restart 'lisp::largest-float))))) (read-from-string string)))))) + +#+x86 +(define-test x87-set-floating-point-modes + (:tag :issues) + (let ((new-mode (setf (x86::x87-floating-point-modes) + (dpb 2 (byte 2 24) + (x86::x87-floating-point-modes))))) + (assert-true (typep new-mode 'x86::float-modes)) + (assert-equal new-mode (setf (x86::x87-floating-point-modes) new-mode)))) + View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/commit/387ed142e6e279b10bf9ee2… -- View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/commit/387ed142e6e279b10bf9ee2… You're receiving this email because of your account on gitlab.common-lisp.net.

1 0

[Git][cmucl/cmucl] Pushed new branch issue-426-fix-float-modes-type-def
by Raymond Toy (＠rtoy) 08 Aug '25

08 Aug '25

Raymond Toy pushed new branch issue-426-fix-float-modes-type-def at cmucl / cmucl -- View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/tree/issue-426-fix-float-modes… You're receiving this email because of your account on gitlab.common-lisp.net.

1 0

[Git][cmucl/cmucl] Pushed new branch issue-425-correctly-rounded-math-functions
by Raymond Toy (＠rtoy) 06 Aug '25

06 Aug '25

Raymond Toy pushed new branch issue-425-correctly-rounded-math-functions at cmucl / cmucl -- View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/tree/issue-425-correctly-round… You're receiving this email because of your account on gitlab.common-lisp.net.

1 0

[Git][cmucl/cmucl] Pushed new branch issue-424-fix-hypot
by Raymond Toy (＠rtoy) 05 Aug '25

05 Aug '25

Raymond Toy pushed new branch issue-424-fix-hypot at cmucl / cmucl -- View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/tree/issue-424-fix-hypot You're receiving this email because of your account on gitlab.common-lisp.net.

1 0

[Git][cmucl/cmucl][master] 2 commits: Fix #423: Use new binaries for CI
by Raymond Toy (＠rtoy) 13 Jul '25

13 Jul '25

Raymond Toy pushed to branch master at cmucl / cmucl Commits: 2c94b54e by Raymond Toy at 2025-07-12T18:18:16-07:00 Fix #423: Use new binaries for CI - - - - - 9a8fd3c4 by Raymond Toy at 2025-07-12T18:18:16-07:00 Merge branch 'issue-423-update-ci-for-snapshot' into 'master' Fix #423: Use new binaries for CI Closes #423 See merge request cmucl/cmucl!310 - - - - - 1 changed file: - .gitlab-ci.yml Changes: ===================================== .gitlab-ci.yml ===================================== @@ -1,7 +1,10 @@ variables: - download_url: "https://common-lisp.net/project/cmucl/downloads/snapshots/2024/08" - version: "2024-08-x86" - bootstrap: "-B boot-2024-08" + year: "2025" + month: "07" + download_url: "https://common-lisp.net/project/cmucl/downloads/snapshots/$year/$month" + version: "$year-$month-x86" + tar_ext: "xz" + bootstrap: "" # Default install configuration to download the cmucl tarballs to use # for building. @@ -15,10 +18,10 @@ variables: - ls -F /usr/local/bin - type -all gitlab-runner # Download binaries. (Do we really need the extras tarball?) - - $CURL -o cmucl-$version-$osname.tar.bz2 $download_url/cmucl-$version-$osname.tar.bz2 - - $CURL -o cmucl-$version-$osname.extra.tar.bz2 $download_url/cmucl-$version-$osname.extra.tar.bz2 + - $CURL -o cmucl-$version-$osname.tar.$tar_ext $download_url/cmucl-$version-$osname.tar.$tar_ext + - $CURL -o cmucl-$version-$osname.extra.tar.$tar_ext $download_url/cmucl-$version-$osname.extra.tar.$tar_ext - mkdir snapshot - - (cd snapshot; tar xjf ../cmucl-$version-$osname.tar.bz2; tar xjf ../cmucl-$version-$osname.extra.tar.bz2) + - (cd snapshot; tar xf ../cmucl-$version-$osname.tar.$tar_ext; tar xf ../cmucl-$version-$osname.extra.tar.$tar_ext) # Default build configuration to be added to each build stage for each # OS. This assumes we don't need anything special between OSes, and View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/compare/38b2787c5c70a2b9af4006… -- View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/compare/38b2787c5c70a2b9af4006… You're receiving this email because of your account on gitlab.common-lisp.net.

1 0

[Git][cmucl/cmucl][issue-423-update-ci-for-snapshot] Add some new variables so we have the correct values everywhere
by Raymond Toy (＠rtoy) 13 Jul '25

13 Jul '25

Raymond Toy pushed to branch issue-423-update-ci-for-snapshot at cmucl / cmucl Commits: 179cb29a by Raymond Toy at 2025-07-12T17:15:46-07:00 Add some new variables so we have the correct values everywhere Add variables "year" and "month" so we have the correct year and month of the snapshot we want. (We'll have to do something different when we use a release.) Use these variables in the download_url and the version. Finally, add variable "tar_ext" so we use the correct tar extension everywhere too. - - - - - 1 changed file: - .gitlab-ci.yml Changes: ===================================== .gitlab-ci.yml ===================================== @@ -1,6 +1,9 @@ variables: - download_url: "https://common-lisp.net/project/cmucl/downloads/snapshots/2024/08" - version: "2025-07-x86" + year: "2025" + month: "07" + download_url: "https://common-lisp.net/project/cmucl/downloads/snapshots/$year/$month" + version: "$year-$month-x86" + tar_ext: "xz" bootstrap: "" # Default install configuration to download the cmucl tarballs to use @@ -15,10 +18,10 @@ variables: - ls -F /usr/local/bin - type -all gitlab-runner # Download binaries. (Do we really need the extras tarball?) - - $CURL -o cmucl-$version-$osname.tar.bz2 $download_url/cmucl-$version-$osname.tar.bz2 - - $CURL -o cmucl-$version-$osname.extra.tar.bz2 $download_url/cmucl-$version-$osname.extra.tar.bz2 + - $CURL -o cmucl-$version-$osname.tar.$tar_ext $download_url/cmucl-$version-$osname.tar.$tar_ext + - $CURL -o cmucl-$version-$osname.extra.tar.$tar_ext $download_url/cmucl-$version-$osname.extra.tar.$tar_ext - mkdir snapshot - - (cd snapshot; tar xf ../cmucl-$version-$osname.tar.xz; tar xf ../cmucl-$version-$osname.extra.tar.xz) + - (cd snapshot; tar xf ../cmucl-$version-$osname.tar.$tar_ext; tar xf ../cmucl-$version-$osname.extra.tar.$tar_ext) # Default build configuration to be added to each build stage for each # OS. This assumes we don't need anything special between OSes, and View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/commit/179cb29ac3dead48c60d3a4… -- View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/commit/179cb29ac3dead48c60d3a4… You're receiving this email because of your account on gitlab.common-lisp.net.

1 0

[Git][cmucl/cmucl][issue-423-update-ci-for-snapshot] Use correct tar options and tarball extension
by Raymond Toy (＠rtoy) 13 Jul '25

13 Jul '25

Raymond Toy pushed to branch issue-423-update-ci-for-snapshot at cmucl / cmucl Commits: f720c990 by Raymond Toy at 2025-07-12T17:11:15-07:00 Use correct tar options and tarball extension Oops. The latest snapshots have an "xz" extension so we need to use the correct extension and the correct options for tar. - - - - - 1 changed file: - .gitlab-ci.yml Changes: ===================================== .gitlab-ci.yml ===================================== @@ -18,7 +18,7 @@ variables: - $CURL -o cmucl-$version-$osname.tar.bz2 $download_url/cmucl-$version-$osname.tar.bz2 - $CURL -o cmucl-$version-$osname.extra.tar.bz2 $download_url/cmucl-$version-$osname.extra.tar.bz2 - mkdir snapshot - - (cd snapshot; tar xjf ../cmucl-$version-$osname.tar.bz2; tar xjf ../cmucl-$version-$osname.extra.tar.bz2) + - (cd snapshot; tar xf ../cmucl-$version-$osname.tar.xz; tar xf ../cmucl-$version-$osname.extra.tar.xz) # Default build configuration to be added to each build stage for each # OS. This assumes we don't need anything special between OSes, and View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/commit/f720c990e71166c04d083bb… -- View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/commit/f720c990e71166c04d083bb… You're receiving this email because of your account on gitlab.common-lisp.net.

1 0

[Git][cmucl/cmucl] Pushed new branch issue-423-update-ci-for-snapshot
by Raymond Toy (＠rtoy) 12 Jul '25

12 Jul '25

Raymond Toy pushed new branch issue-423-update-ci-for-snapshot at cmucl / cmucl -- View it on GitLab: https://gitlab.common-lisp.net/cmucl/cmucl/-/tree/issue-423-update-ci-for-s… You're receiving this email because of your account on gitlab.common-lisp.net.

1 0