[llvm-dev] LLVM 10 is miscompiling this code on x86_64 with an invalid shift count for `shrd`
Riyaz Puthiyapurayil via llvm-dev
llvm-dev at lists.llvm.org
Tue Jan 26 12:09:03 PST 2021
Thanks Craig!
From: Craig Topper <craig.topper at gmail.com>
Sent: Tuesday, January 26, 2021 12:06 PM
To: Riyaz Puthiyapurayil <riyaz at synopsys.com>
Cc: Roman Lebedev <lebedev.ri at gmail.com>; llvm-dev at lists.llvm.org
Subject: Re: [llvm-dev] LLVM 10 is miscompiling this code on x86_64 with an invalid shift count for `shrd`
Which was flagged in post-commit review about a year later. https://reviews.llvm.org/D58475<https://urldefense.com/v3/__https:/reviews.llvm.org/D58475__;!!A4F2R9G_pg!LZwRdXlbAwQyD5-XlUqg2SEhRa_HhVybtA8bwSpe8ds1VQXwsCwI5QR-jHCTCqL9_U6F8ks7vCRy$>
~Craig
On Tue, Jan 26, 2021 at 12:04 PM Craig Topper <craig.topper at gmail.com<mailto:craig.topper at gmail.com>> wrote:
There's a typo in this line of X86InstrCompiler.td in llvm 10. It should be shiftMask64 not shiftMask32.
// (shift x (and y, 63)) ==> (shift x, y)
def : Pat<(frag GR64:$src1, GR64:$src2, (shiftMask32 CL)),
(!cast<Instruction>(name # "64rrCL") GR64:$src1, GR64:$src2)>;
~Craig
On Tue, Jan 26, 2021 at 11:57 AM Riyaz Puthiyapurayil via llvm-dev <llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>> wrote:
Curiously, for the following test, with -fsanitize=undefined -O3, clang-10 reports no UB errors and produces the correct run-time result. But without -fsanitize=undefined, it produces wrong output. Clang (trunk) produces correct results either way. I think there is definitely a bug that was fixed recently (in the last year or so). Could it be this one:
https://reviews.llvm.org/D75748?id=<https://urldefense.com/v3/__https:/reviews.llvm.org/D75748?id=__;!!A4F2R9G_pg!LZwRdXlbAwQyD5-XlUqg2SEhRa_HhVybtA8bwSpe8ds1VQXwsCwI5QR-jHCTCqL9_U6F8nTqMJUS$>
Roman, you were a reviewer for this.
-----Original Message-----
From: Riyaz Puthiyapurayil
Sent: Tuesday, January 26, 2021 11:48 AM
To: 'Roman Lebedev' <lebedev.ri at gmail.com<mailto:lebedev.ri at gmail.com>>
Cc: llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>
Subject: RE: [llvm-dev] LLVM 10 is miscompiling this code on x86_64 with an invalid shift count for `shrd`
My bad! There was a typo that caused the UBSAN error. Here is the corrected test. This time UBSAN reports no errors. But clang-10 still produces unexpected result whereas trunk does not.
#include <stdio.h>
#include <stdint.h>
#include <string.h>
__attribute__ ((noinline)) uint64_t foo4(__int128_t* arr1) {
unsigned chunk = 34 >> 5;
unsigned bitIndex = 34 - chunk * 32;
unsigned* arr = (unsigned *)arr1 + chunk;
__int128_t v1 = *arr1;
v1 = v1 >> bitIndex;
return (uint64_t) v1;
}
__attribute__ ((noinline)) uint64_t foo3(__int128_t* arr1, unsigned index) {
unsigned chunk = index >> 5;
unsigned bitIndex = index - chunk * 32;
unsigned *arr = (unsigned *)arr1 + chunk;
__int128_t v1 = *arr1;
v1 = v1 >> bitIndex;
return (uint64_t) v1;
}
int main() {
unsigned arr[4];
arr[0]= 0x6f7cfd6f;
arr[1]= 0xd96c9806;
arr[2]= 0x89420144;
arr[3]= 0x8a20f548;
__int128_t arr1;
memcpy(&arr1, arr, sizeof(arr1));
printf("foo3 %lx\n",foo3(&arr1,34) & 0x3ffffffff);
printf("foo4 %lx\n",foo4(&arr1) & 0x3ffffffff); }
-----Original Message-----
From: Roman Lebedev <lebedev.ri at gmail.com<mailto:lebedev.ri at gmail.com>>
Sent: Tuesday, January 26, 2021 11:31 AM
To: Riyaz Puthiyapurayil <riyaz at synopsys.com<mailto:riyaz at synopsys.com>>
Cc: llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>
Subject: Re: [llvm-dev] LLVM 10 is miscompiling this code on x86_64 with an invalid shift count for `shrd`
https://urldefense.com/v3/__https://godbolt.org/z/5bcsj9__;!!A4F2R9G_pg!P3wQMqAfQM_jzlQy2F-1LD2x1ekeAKGQckArJiV5UsQIR2NhtjHbH5ivGPqtfl8bZLzgtjq5KIG7$<https://urldefense.com/v3/__https:/godbolt.org/z/5bcsj9__;!!A4F2R9G_pg!P3wQMqAfQM_jzlQy2F-1LD2x1ekeAKGQckArJiV5UsQIR2NhtjHbH5ivGPqtfl8bZLzgtjq5KIG7$>
example.c:43:37: runtime error: subtraction of unsigned offset from
0x000000000022 overflowed to 0x7fffd01217c8
SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior example.c:43:37 in
example.c:19:37: runtime error: applying non-zero offset
140576899311424 to null pointer
SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior example.c:19:37 in
foo3 19bdf3f5b
foo4 19bdf3f5b
On Tue, Jan 26, 2021 at 10:29 PM Riyaz Puthiyapurayil <Riyaz.Puthiyapurayil at synopsys.com<mailto:Riyaz.Puthiyapurayil at synopsys.com>> wrote:
>
> Apologies, outlook messed up the formatting in the previous email. Here is it again.
>
>
>
> Do you see any UB in the following modified test? Here also, I see a difference between foo3 and foo4. Again, trunk and gcc match. But not Clang-10.
>
>
>
> #include <stdio.h>
>
> #include <stdint.h>
>
> #include <string.h>
>
>
>
> __attribute__ ((noinline)) uint64_t foo4(__int128_t* arr1)
>
> {
>
> unsigned chunk = 34 >> 5;
>
> unsigned bitIndex = 34 - chunk * 32;
>
>
>
> unsigned* arr = (unsigned *)arr + chunk;
>
> __int128_t v1 = *arr1;
>
> v1 = v1 >> bitIndex;
>
>
>
> return (uint64_t) v1;
>
> }
>
>
>
> __attribute__ ((noinline)) uint64_t foo3(__int128_t* arr1, unsigned
> index)
>
> {
>
> unsigned chunk = index >> 5;
>
> unsigned bitIndex = index - chunk * 32;
>
>
>
> unsigned *arr = (unsigned *)arr + chunk;
>
> __int128_t v1 = *arr1;
>
> v1 = v1 >> bitIndex;
>
>
>
> return (uint64_t) v1;
>
> }
>
>
>
> int main() {
>
>
>
> unsigned arr[4];
>
>
>
> arr[0]= 0x6f7cfd6f;
>
> arr[1]= 0xd96c9806;
>
> arr[2]= 0x89420144;
>
> arr[3]= 0x8a20f548;
>
>
>
> __int128_t arr1;
>
>
>
> memcpy(&arr1, arr, sizeof(arr1));
>
>
>
> printf("foo3 %lx\n",foo3(&arr1,34) & 0x3ffffffff);
>
> printf("foo4 %lx\n",foo4(&arr1) & 0x3ffffffff);
>
> }
>
>
>
>
>
>
>
> From: Riyaz Puthiyapurayil
> Sent: Tuesday, January 26, 2021 11:25 AM
> To: 'Roman Lebedev' <lebedev.ri at gmail.com<mailto:lebedev.ri at gmail.com>>
> Cc: llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>
> Subject: RE: [llvm-dev] LLVM 10 is miscompiling this code on x86_64
> with an invalid shift count for `shrd`
>
>
>
> Do you see any UB in the following modified test? Here also, I see a
> difference between foo3 and foo4. Again, trunk and gcc match. But not
> Clang-10
>
>
>
> #include <stdio.h>
>
> #include <stdint.h>
>
> #include <string.h>
>
>
>
> __attribute__ ((noinline)) uint64_t foo4(__int128_t* arr1)
>
> {
>
> unsigned chunk = 34 >> 5;
>
> unsigned bitIndex = 34 - chunk * 32;
>
>
>
> unsigned* arr = (unsigned *)arr + chunk;
>
> __int128_t v1 = *arr1;
>
> v1 = v1 >> bitIndex;
>
>
>
> return (uint64_t) v1;
>
> }
>
>
>
> __attribute__ ((noinline)) uint64_t foo3(__int128_t* arr1, unsigned
> index)
>
> {
>
> unsigned chunk = index >> 5;
>
> unsigned bitIndex = index - chunk * 32;
>
>
>
> unsigned *arr = (unsigned *)arr + chunk;
>
> __int128_t v1 = *arr1;
>
> v1 = v1 >> bitIndex;
>
>
>
> return (uint64_t) v1;
>
> }
>
>
>
> int main() {
>
>
>
> unsigned arr[4];
>
>
>
> arr[0]= 0x6f7cfd6f;
>
> arr[1]= 0xd96c9806;
>
> arr[2]= 0x89420144;
>
> arr[3]= 0x8a20f548;
>
>
>
> __int128_t arr1;
>
>
>
> memcpy(&arr1, arr, sizeof(arr1));
>
>
>
> printf("foo3 %lx\n",foo3(&arr1,34) & 0x3ffffffff); // prints
> 1365b2601 (clang-10) 19bdf3f5b (gcc, clang-12)
>
> printf("foo4 %lx\n",foo4(&arr1) & 0x3ffffffff); // prints 19bdf3f5b
>
> }
>
>
>
> -----Original Message-----
> From: Riyaz Puthiyapurayil
> Sent: Tuesday, January 26, 2021 11:10 AM
> To: 'Roman Lebedev' <lebedev.ri at gmail.com<mailto:lebedev.ri at gmail.com>>
> Subject: RE: [llvm-dev] LLVM 10 is miscompiling this code on x86_64
> with an invalid shift count for `shrd`
>
>
>
> You may be right about that the cast from unsigned* to __int128_t* being UB. But I think that has nothing to do with the bug. That code can be fixed to use a __int128_t array and change all the casts from __int128_t* to unsigned* (instead of the other way around) and we would still have the issue.
>
>
>
> -----Original Message-----
>
> From: Roman Lebedev <lebedev.ri at gmail.com<mailto:lebedev.ri at gmail.com>>
>
> Sent: Tuesday, January 26, 2021 10:34 AM
>
> To: Riyaz Puthiyapurayil <riyaz at synopsys.com<mailto:riyaz at synopsys.com>>
>
> Cc: llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>
>
> Subject: Re: [llvm-dev] LLVM 10 is miscompiling this code on x86_64
> with an invalid shift count for `shrd`
>
>
>
> On Tue, Jan 26, 2021 at 9:28 PM Riyaz Puthiyapurayil via llvm-dev <llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>> wrote:
>
> >
>
> > It appears that LLVM 10 is miscompiling the following test on x86_64 but the trunk version seems to work fine. Before I spend any time debugging this, it would be helpful if someone can point me to a bug fix if there was one.
>
> >
>
> >
>
> >
>
> > #include <stdio.h>
>
> >
>
> > #include <stdint.h>
>
> >
>
> >
>
> >
>
> > __attribute__ ((noinline)) uint64_t foo4(unsigned* arr)
>
> >
>
> > {
>
> >
>
> > unsigned chunk = 34 >> 5;
>
> >
>
> > unsigned bitIndex = 34 - chunk * 32;
>
> >
>
> >
>
> >
>
> > arr = arr + chunk;
>
> >
>
> > __int128_t* arr1 = (__int128_t*) arr;
>
> I'm pretty sure this is UB, because the alignment of __int128_t is bigger than that of unsigned.
>
>
>
> >
>
> > __int128_t v1 = *arr1;
>
> >
>
> > v1 = v1 >> bitIndex;
>
> >
>
> >
>
> >
>
> > return (uint64_t) v1;
>
> >
>
> > }
>
> >
>
> >
>
> >
>
> > __attribute__ ((noinline)) uint64_t foo3(unsigned* arr, unsigned
>
> > index)
>
> >
>
> > {
>
> >
>
> > unsigned chunk = index >> 5;
>
> >
>
> > unsigned bitIndex = index - chunk * 32;
>
> >
>
> >
>
> >
>
> > arr = arr + chunk;
>
> >
>
> > __int128_t* arr1 = (__int128_t*) arr;
>
> >
>
> > __int128_t v1 = *arr1;
>
> >
>
> > v1 = v1 >> bitIndex;
>
> >
>
> >
>
> >
>
> > return (uint64_t) v1;
>
> >
>
> > }
>
> >
>
> >
>
> >
>
> >
>
> >
>
> > int main() {
>
> >
>
> >
>
> >
>
> > unsigned arr[5];
>
> >
>
> >
>
> >
>
> > arr[0]= 0x6f7cfd6f;
>
> >
>
> > arr[1]= 0xd96c9806;
>
> >
>
> > arr[2]= 0x89420144;
>
> >
>
> > arr[3]= 0x8a20f548;
>
> >
>
> >
>
> >
>
> > printf("foo3 %lx\n",foo3(arr,34) & 0x3ffffffff); // prints
>
> > 222508051 instead of 1365b2601
>
> >
>
> > printf("foo4 %lx\n",foo4(arr) & 0x3ffffffff); // prints
> > 1365b2601
>
> > as expected
>
> >
>
> > }
>
> >
>
> >
>
> >
>
> > If you look at the assembly, there is an ‘and cl, 31’ missing in foo3 which is needed to make sure that the shift count is in the range [0...31]. This is missing in the generated code. Note that if the count in CL register is greater than 31, x86_64 manual states that the behavior of shrd is undefined.
>
> >
>
> >
>
> >
>
> > # clang 10
>
> >
>
> > foo3(unsigned int*, unsigned int): # @foo3(unsigned int*, unsigned int)
>
> >
>
> > mov ecx, esi
>
> >
>
> > mov edx, esi
>
> >
>
> > shr edx, 5
>
> >
>
> > mov rax, qword ptr [rdi + 4*rdx]
>
> >
>
> > mov rdx, qword ptr [rdi + 4*rdx + 8]
>
> >
>
> > shrd rax, rdx, cl
>
> >
>
> > ret
>
> >
>
> >
>
> >
>
> > # Trunk
>
> >
>
> > foo3(unsigned int*, unsigned int): # @foo3(unsigned int*, unsigned int)
>
> >
>
> > mov ecx, esi
>
> >
>
> > mov edx, esi
>
> >
>
> > shr edx, 5
>
> >
>
> > mov rax, qword ptr [rdi + 4*rdx]
>
> >
>
> > mov rdx, qword ptr [rdi + 4*rdx + 8]
>
> >
>
> > and cl, 31
>
> >
>
> > shrd rax, rdx, cl
>
> >
>
> > ret
>
> >
>
>
>
> Roman.
>
>
>
> > _______________________________________________
>
> > LLVM Developers mailing list
>
> > llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>
>
> > https://urldefense.com/v3/__https://lists.llvm.org/cgi-bin/mailman/l<https://urldefense.com/v3/__https:/lists.llvm.org/cgi-bin/mailman/l>
> > is
>
> > tinfo/llvm-dev__;!!A4F2R9G_pg!Mq1ZgXCY7kwDMwNwdn4Tcz-KgkoXMlP_rjq7_U
> > 9M
>
> > rh1gbrtqvDoSi6cMuSLZaXIPgtijPW55JJ_8$
_______________________________________________
LLVM Developers mailing list
llvm-dev at lists.llvm.org<mailto:llvm-dev at lists.llvm.org>
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev<https://urldefense.com/v3/__https:/lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev__;!!A4F2R9G_pg!LZwRdXlbAwQyD5-XlUqg2SEhRa_HhVybtA8bwSpe8ds1VQXwsCwI5QR-jHCTCqL9_U6F8kv5Y8Tw$>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20210126/fd630b88/attachment.html>
More information about the llvm-dev
mailing list