Re: [llvm-branch-commits] [cfe-branch] r214778 - Merging r214777:
On Mon, Aug 4, 2014 at 2:11 PM, Hans Wennborg h...@hanshq.net wrote: Author: hans Date: Mon Aug 4 16:11:53 2014 New Revision: 214778 URL: http://llvm.org/viewvc/llvm-project?rev=214778view=rev Log: Merging r214777: Excluding the /Zp option, because that was added after the 3.5 branch. r214777 | hans | 2014-08-04 14:07:58 -0700 (Mon, 04 Aug 2014) | 1 line UsersManual: update clang-cl options Modified: cfe/branches/release_35/ (props changed) cfe/branches/release_35/docs/UsersManual.rst cfe/branches/release_35/test/Analysis/MismatchedDeallocator-checker-test.mm (props changed) cfe/branches/release_35/test/Analysis/NewDelete-checker-test.cpp (props changed) cfe/branches/release_35/test/SemaCXX/warn-unreachable.cpp (props changed) Weird.. I have no idea why it committed props changes to files in test/, that was certainly not my intention. Since it's just metadata, maybe I can treat it as benign, or do you think I should try to revert that part of the commit? ___ llvm-branch-commits mailing list llvm-branch-commits@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [compiler-rt-branch] r216035 - Merging r215295:
Author: rengolin Date: Tue Aug 19 18:30:32 2014 New Revision: 216035 URL: http://llvm.org/viewvc/llvm-project?rev=216035view=rev Log: Merging r215295: r215295 | compnerd | 2014-08-09 21:17:37 +0100 (Sat, 09 Aug 2014) | 10 lines builtins: correct __umodsi3, __udivsi3 on ARM When building the builtins for a modern CPU (idiv support), __umodsi3 was completely incorrect as it would behave as __udivmosi3, which takes a tertiary parameter which is a pointer. __udivsi3 was also incorrect, returning the remainder in r1. Although this would not result in any crash or invalid behaviour as r1 is a caller saved register in AAPCS, this is unnecessary. Simply perform the division ignoring the remainder. Modified: compiler-rt/branches/release_35/lib/builtins/arm/udivsi3.S compiler-rt/branches/release_35/lib/builtins/arm/umodsi3.S Modified: compiler-rt/branches/release_35/lib/builtins/arm/udivsi3.S URL: http://llvm.org/viewvc/llvm-project/compiler-rt/branches/release_35/lib/builtins/arm/udivsi3.S?rev=216035r1=216034r2=216035view=diff == --- compiler-rt/branches/release_35/lib/builtins/arm/udivsi3.S (original) +++ compiler-rt/branches/release_35/lib/builtins/arm/udivsi3.S Tue Aug 19 18:30:32 2014 @@ -35,9 +35,7 @@ DEFINE_COMPILERRT_FUNCTION(__udivsi3) #if __ARM_ARCH_EXT_IDIV__ tst r1, r1 beq LOCAL_LABEL(divby0) - mov r3, r0 - udivr0, r3, r1 - mls r1, r0, r1, r3 + udivr0, r0, r1 bx lr #else cmp r1, #1 Modified: compiler-rt/branches/release_35/lib/builtins/arm/umodsi3.S URL: http://llvm.org/viewvc/llvm-project/compiler-rt/branches/release_35/lib/builtins/arm/umodsi3.S?rev=216035r1=216034r2=216035view=diff == --- compiler-rt/branches/release_35/lib/builtins/arm/umodsi3.S (original) +++ compiler-rt/branches/release_35/lib/builtins/arm/umodsi3.S Tue Aug 19 18:30:32 2014 @@ -33,10 +33,8 @@ DEFINE_COMPILERRT_FUNCTION(__umodsi3) #if __ARM_ARCH_EXT_IDIV__ tst r1, r1 beq LOCAL_LABEL(divby0) - mov r3, r0 - udivr0, r3, r1 - mls r1, r0, r1, r3 - str r1, [r2] + udivr2, r0, r1 + mls r0, r2, r1, r0 bx lr #else cmp r1, #1 ___ llvm-branch-commits mailing list llvm-branch-commits@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [cfe-branch] r214823 - Merging r214801:
Author: wschmidt Date: Mon Aug 4 18:48:10 2014 New Revision: 214823 URL: http://llvm.org/viewvc/llvm-project?rev=214823view=rev Log: Merging r214801: r214801 | wschmidt | 2014-08-04 18:21:26 -0500 (Mon, 04 Aug 2014) | 19 lines [PPC64LE] Fix wrong IR for vec_sld and vec_vsldoi My original LE implementation of the vsldoi instruction, with its altivec.h interfaces vec_sld and vec_vsldoi, produces incorrect shufflevector operations in the LLVM IR. Correct code is generated because the back end handles the incorrect shufflevector in a consistent manner. This patch and a companion patch for LLVM correct this problem by removing the fixup from altivec.h and the corresponding fixup from the PowerPC back end. Several test cases are also modified to reflect the now-correct LLVM IR. The vec_sums and vec_vsumsws interfaces in altivec.h are also fixed, because they used vec_perm calls intended to be recognized as vsldoi instructions. These vec_perm calls are now replaced with code that more clearly shows the intent of the transformation. Modified: cfe/branches/release_35/lib/Headers/altivec.h cfe/branches/release_35/test/CodeGen/builtins-ppc-altivec.c Modified: cfe/branches/release_35/lib/Headers/altivec.h URL: http://llvm.org/viewvc/llvm-project/cfe/branches/release_35/lib/Headers/altivec.h?rev=214823r1=214822r2=214823view=diff == --- cfe/branches/release_35/lib/Headers/altivec.h (original) +++ cfe/branches/release_35/lib/Headers/altivec.h Mon Aug 4 18:48:10 2014 @@ -5224,113 +5224,65 @@ vec_vslw(vector unsigned int __a, vector static vector signed char __ATTRS_o_ai vec_sld(vector signed char __a, vector signed char __b, unsigned char __c) { -#ifdef __LITTLE_ENDIAN__ - return vec_perm(__a, __b, (vector unsigned char) -(__c, __c-1, __c-2, __c-3, __c-4, __c-5, __c-6, __c-7, - __c-8, __c-9, __c-10, __c-11, __c-12, __c-13, __c-14, __c-15)); -#else return vec_perm(__a, __b, (vector unsigned char) (__c, __c+1, __c+2, __c+3, __c+4, __c+5, __c+6, __c+7, __c+8, __c+9, __c+10, __c+11, __c+12, __c+13, __c+14, __c+15)); -#endif } static vector unsigned char __ATTRS_o_ai vec_sld(vector unsigned char __a, vector unsigned char __b, unsigned char __c) { -#ifdef __LITTLE_ENDIAN__ - return vec_perm(__a, __b, (vector unsigned char) -(__c, __c-1, __c-2, __c-3, __c-4, __c-5, __c-6, __c-7, - __c-8, __c-9, __c-10, __c-11, __c-12, __c-13, __c-14, __c-15)); -#else return vec_perm(__a, __b, (vector unsigned char) (__c, __c+1, __c+2, __c+3, __c+4, __c+5, __c+6, __c+7, __c+8, __c+9, __c+10, __c+11, __c+12, __c+13, __c+14, __c+15)); -#endif } static vector short __ATTRS_o_ai vec_sld(vector short __a, vector short __b, unsigned char __c) { -#ifdef __LITTLE_ENDIAN__ - return vec_perm(__a, __b, (vector unsigned char) -(__c, __c-1, __c-2, __c-3, __c-4, __c-5, __c-6, __c-7, - __c-8, __c-9, __c-10, __c-11, __c-12, __c-13, __c-14, __c-15)); -#else return vec_perm(__a, __b, (vector unsigned char) (__c, __c+1, __c+2, __c+3, __c+4, __c+5, __c+6, __c+7, __c+8, __c+9, __c+10, __c+11, __c+12, __c+13, __c+14, __c+15)); -#endif } static vector unsigned short __ATTRS_o_ai vec_sld(vector unsigned short __a, vector unsigned short __b, unsigned char __c) { -#ifdef __LITTLE_ENDIAN__ - return vec_perm(__a, __b, (vector unsigned char) -(__c, __c-1, __c-2, __c-3, __c-4, __c-5, __c-6, __c-7, - __c-8, __c-9, __c-10, __c-11, __c-12, __c-13, __c-14, __c-15)); -#else return vec_perm(__a, __b, (vector unsigned char) (__c, __c+1, __c+2, __c+3, __c+4, __c+5, __c+6, __c+7, __c+8, __c+9, __c+10, __c+11, __c+12, __c+13, __c+14, __c+15)); -#endif } static vector pixel __ATTRS_o_ai vec_sld(vector pixel __a, vector pixel __b, unsigned char __c) { -#ifdef __LITTLE_ENDIAN__ - return vec_perm(__a, __b, (vector unsigned char) -(__c, __c-1, __c-2, __c-3, __c-4, __c-5, __c-6, __c-7, - __c-8, __c-9, __c-10, __c-11, __c-12, __c-13, __c-14, __c-15)); -#else return vec_perm(__a, __b, (vector unsigned char) (__c, __c+1, __c+2, __c+3, __c+4, __c+5, __c+6, __c+7, __c+8, __c+9, __c+10, __c+11, __c+12, __c+13, __c+14, __c+15)); -#endif } static vector int __ATTRS_o_ai vec_sld(vector int __a, vector int __b, unsigned char __c) { -#ifdef __LITTLE_ENDIAN__ - return vec_perm(__a, __b, (vector unsigned char) -(__c, __c-1, __c-2, __c-3, __c-4, __c-5, __c-6, __c-7, - __c-8, __c-9, __c-10, __c-11, __c-12, __c-13, __c-14, __c-15)); -#else return vec_perm(__a, __b, (vector unsigned char) (__c, __c+1, __c+2, __c+3, __c+4, __c+5, __c+6, __c+7, __c+8, __c+9, __c+10, __c+11, __c+12, __c+13, __c+14, __c+15)); -#endif } static vector unsigned int
Re: [llvm-branch-commits] [compiler-rt-branch] r216035 - Merging r215295:
As I mentioned in a previous email, it’s far too late to be patching the release branch, especially without telling me first. Please revert this. -bw On Aug 19, 2014, at 4:30 PM, Renato Golin renato.go...@linaro.org wrote: Author: rengolin Date: Tue Aug 19 18:30:32 2014 New Revision: 216035 URL: http://llvm.org/viewvc/llvm-project?rev=216035view=rev Log: Merging r215295: r215295 | compnerd | 2014-08-09 21:17:37 +0100 (Sat, 09 Aug 2014) | 10 lines builtins: correct __umodsi3, __udivsi3 on ARM When building the builtins for a modern CPU (idiv support), __umodsi3 was completely incorrect as it would behave as __udivmosi3, which takes a tertiary parameter which is a pointer. __udivsi3 was also incorrect, returning the remainder in r1. Although this would not result in any crash or invalid behaviour as r1 is a caller saved register in AAPCS, this is unnecessary. Simply perform the division ignoring the remainder. Modified: compiler-rt/branches/release_35/lib/builtins/arm/udivsi3.S compiler-rt/branches/release_35/lib/builtins/arm/umodsi3.S Modified: compiler-rt/branches/release_35/lib/builtins/arm/udivsi3.S URL: http://llvm.org/viewvc/llvm-project/compiler-rt/branches/release_35/lib/builtins/arm/udivsi3.S?rev=216035r1=216034r2=216035view=diff == --- compiler-rt/branches/release_35/lib/builtins/arm/udivsi3.S (original) +++ compiler-rt/branches/release_35/lib/builtins/arm/udivsi3.S Tue Aug 19 18:30:32 2014 @@ -35,9 +35,7 @@ DEFINE_COMPILERRT_FUNCTION(__udivsi3) #if __ARM_ARCH_EXT_IDIV__ tst r1, r1 beq LOCAL_LABEL(divby0) - mov r3, r0 - udivr0, r3, r1 - mls r1, r0, r1, r3 + udivr0, r0, r1 bx lr #else cmp r1, #1 Modified: compiler-rt/branches/release_35/lib/builtins/arm/umodsi3.S URL: http://llvm.org/viewvc/llvm-project/compiler-rt/branches/release_35/lib/builtins/arm/umodsi3.S?rev=216035r1=216034r2=216035view=diff == --- compiler-rt/branches/release_35/lib/builtins/arm/umodsi3.S (original) +++ compiler-rt/branches/release_35/lib/builtins/arm/umodsi3.S Tue Aug 19 18:30:32 2014 @@ -33,10 +33,8 @@ DEFINE_COMPILERRT_FUNCTION(__umodsi3) #if __ARM_ARCH_EXT_IDIV__ tst r1, r1 beq LOCAL_LABEL(divby0) - mov r3, r0 - udivr0, r3, r1 - mls r1, r0, r1, r3 - str r1, [r2] + udivr2, r0, r1 + mls r0, r2, r1, r0 bx lr #else cmp r1, #1 ___ llvm-branch-commits mailing list llvm-branch-commits@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/llvm-branch-commits ___ llvm-branch-commits mailing list llvm-branch-commits@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/llvm-branch-commits
Re: [llvm-branch-commits] [cfe-branch] r214823 - Merging r214801:
Okay. Thanks for this. I didn’t realize that these were old emails. Sorry for freaking out. -bw On Aug 26, 2014, at 4:36 PM, Bill Schmidt wschm...@linux.vnet.ibm.com wrote: Hi Bill, I forgot to CC you when I just responded on-list, but it might sit in a moderator queue like all these others did, so here is what I wrote: Hi Bill, These were committed back on August 4th. I followed your instructions to use the merge tool, but I didn't have write-authority to the branch-commits list so all the notifications sat in a queue waiting for verification from a moderator. I have to assume that this finally happened... So anyway, nothing untoward here. These changes were from several weeks ago. Thanks, Bill On Tue, 2014-08-26 at 15:31 -0700, Bill Wendling wrote: Um…What the hell is this? It’s far too late to be applying things to the branch, especially without my approval -bw On Aug 4, 2014, at 4:48 PM, Bill Schmidt wschm...@linux.vnet.ibm.com wrote: Author: wschmidt Date: Mon Aug 4 18:48:10 2014 New Revision: 214823 URL: http://llvm.org/viewvc/llvm-project?rev=214823view=rev Log: Merging r214801: r214801 | wschmidt | 2014-08-04 18:21:26 -0500 (Mon, 04 Aug 2014) | 19 lines [PPC64LE] Fix wrong IR for vec_sld and vec_vsldoi My original LE implementation of the vsldoi instruction, with its altivec.h interfaces vec_sld and vec_vsldoi, produces incorrect shufflevector operations in the LLVM IR. Correct code is generated because the back end handles the incorrect shufflevector in a consistent manner. This patch and a companion patch for LLVM correct this problem by removing the fixup from altivec.h and the corresponding fixup from the PowerPC back end. Several test cases are also modified to reflect the now-correct LLVM IR. The vec_sums and vec_vsumsws interfaces in altivec.h are also fixed, because they used vec_perm calls intended to be recognized as vsldoi instructions. These vec_perm calls are now replaced with code that more clearly shows the intent of the transformation. Modified: cfe/branches/release_35/lib/Headers/altivec.h cfe/branches/release_35/test/CodeGen/builtins-ppc-altivec.c Modified: cfe/branches/release_35/lib/Headers/altivec.h URL: http://llvm.org/viewvc/llvm-project/cfe/branches/release_35/lib/Headers/altivec.h?rev=214823r1=214822r2=214823view=diff == --- cfe/branches/release_35/lib/Headers/altivec.h (original) +++ cfe/branches/release_35/lib/Headers/altivec.h Mon Aug 4 18:48:10 2014 @@ -5224,113 +5224,65 @@ vec_vslw(vector unsigned int __a, vector static vector signed char __ATTRS_o_ai vec_sld(vector signed char __a, vector signed char __b, unsigned char __c) { -#ifdef __LITTLE_ENDIAN__ - return vec_perm(__a, __b, (vector unsigned char) -(__c, __c-1, __c-2, __c-3, __c-4, __c-5, __c-6, __c-7, - __c-8, __c-9, __c-10, __c-11, __c-12, __c-13, __c-14, __c-15)); -#else return vec_perm(__a, __b, (vector unsigned char) (__c, __c+1, __c+2, __c+3, __c+4, __c+5, __c+6, __c+7, __c+8, __c+9, __c+10, __c+11, __c+12, __c+13, __c+14, __c+15)); -#endif } static vector unsigned char __ATTRS_o_ai vec_sld(vector unsigned char __a, vector unsigned char __b, unsigned char __c) { -#ifdef __LITTLE_ENDIAN__ - return vec_perm(__a, __b, (vector unsigned char) -(__c, __c-1, __c-2, __c-3, __c-4, __c-5, __c-6, __c-7, - __c-8, __c-9, __c-10, __c-11, __c-12, __c-13, __c-14, __c-15)); -#else return vec_perm(__a, __b, (vector unsigned char) (__c, __c+1, __c+2, __c+3, __c+4, __c+5, __c+6, __c+7, __c+8, __c+9, __c+10, __c+11, __c+12, __c+13, __c+14, __c+15)); -#endif } static vector short __ATTRS_o_ai vec_sld(vector short __a, vector short __b, unsigned char __c) { -#ifdef __LITTLE_ENDIAN__ - return vec_perm(__a, __b, (vector unsigned char) -(__c, __c-1, __c-2, __c-3, __c-4, __c-5, __c-6, __c-7, - __c-8, __c-9, __c-10, __c-11, __c-12, __c-13, __c-14, __c-15)); -#else return vec_perm(__a, __b, (vector unsigned char) (__c, __c+1, __c+2, __c+3, __c+4, __c+5, __c+6, __c+7, __c+8, __c+9, __c+10, __c+11, __c+12, __c+13, __c+14, __c+15)); -#endif } static vector unsigned short __ATTRS_o_ai vec_sld(vector unsigned short __a, vector unsigned short __b, unsigned char __c) { -#ifdef __LITTLE_ENDIAN__ - return vec_perm(__a, __b, (vector unsigned char) -(__c, __c-1, __c-2, __c-3, __c-4, __c-5, __c-6, __c-7, - __c-8, __c-9, __c-10, __c-11, __c-12, __c-13, __c-14, __c-15)); -#else return vec_perm(__a, __b, (vector unsigned char) (__c, __c+1, __c+2, __c+3, __c+4, __c+5, __c+6, __c+7, __c+8, __c+9, __c+10, __c+11,