Implementation notes: amd64, dali, crypto_sort/int32

Test failure

error 111

Compiler output

api.cpp: In file included from api.cpp:3:
api.cpp: In file included from ./aspas.h:204:
api.cpp: In file included from ./aspas.tcc:31:
api.cpp: In file included from ./sorter.h:120:
api.cpp: In file included from ./sorter_avx2.tcc:27:
api.cpp: ./util/extintrin.h:37:25: warning: unused variable 'max2' [-Wunused-variable]
api.cpp:     __m128i max1, min1, max2, min2;
api.cpp:                         ^
api.cpp: ./util/extintrin.h:37:13: warning: unused variable 'max1' [-Wunused-variable]
api.cpp:     __m128i max1, min1, max2, min2;
api.cpp:             ^
api.cpp: ./util/extintrin.h:53:31: warning: unused variable 'min2' [-Wunused-variable]
api.cpp:     __m128i max1, min1, max2, min2;
api.cpp:                               ^
api.cpp: ./util/extintrin.h:53:19: warning: unused variable 'min1' [-Wunused-variable]
api.cpp:     __m128i max1, min1, max2, min2;
api.cpp:                   ^
api.cpp: In file included from api.cpp:3:
api.cpp: In file included from ./aspas.h:204:
api.cpp: In file included from ./aspas.tcc:32:
api.cpp: In file included from ./merger.h:64:
api.cpp: ./merger.tcc:187:17: warning: & has lower precedence than ==; == will be evaluated first [-Wparentheses]
api.cpp:         if (num & 1 == 1)
api.cpp:                 ^~~~~~~~
api.cpp: ./merger.tcc:187:17: note: place parentheses around the '==' expression to silence this warning
api.cpp: ...

Compiler output

api.cpp: In file included from api.cpp:3:
api.cpp: In file included from ./aspas.h:204:
api.cpp: In file included from ./aspas.tcc:31:
api.cpp: In file included from ./sorter.h:120:
api.cpp: In file included from ./sorter_avx2.tcc:27:
api.cpp: ./util/extintrin.h:37:13: warning: unused variable 'max1' [-Wunused-variable]
api.cpp:     __m128i max1, min1, max2, min2;
api.cpp:             ^
api.cpp: ./util/extintrin.h:37:25: warning: unused variable 'max2' [-Wunused-variable]
api.cpp:     __m128i max1, min1, max2, min2;
api.cpp:                         ^
api.cpp: ./util/extintrin.h:53:19: warning: unused variable 'min1' [-Wunused-variable]
api.cpp:     __m128i max1, min1, max2, min2;
api.cpp:                   ^
api.cpp: ./util/extintrin.h:53:31: warning: unused variable 'min2' [-Wunused-variable]
api.cpp:     __m128i max1, min1, max2, min2;
api.cpp:                               ^
api.cpp: In file included from api.cpp:3:
api.cpp: In file included from ./aspas.h:204:
api.cpp: In file included from ./aspas.tcc:32:
api.cpp: In file included from ./merger.h:64:
api.cpp: ./merger.tcc:187:17: warning: & has lower precedence than ==; == will be evaluated first [-Wparentheses]
api.cpp:         if (num & 1 == 1)
api.cpp:                 ^~~~~~~~
api.cpp: ./merger.tcc:187:17: note: place parentheses around the '==' expression to silence this warning
api.cpp: ...

Compiler output

api.cpp: In file included from sorter_avx2.tcc:27,
api.cpp:                  from sorter.h:120,
api.cpp:                  from aspas.tcc:31,
api.cpp:                  from aspas.h:204,
api.cpp:                  from api.cpp:3:
api.cpp: util/extintrin.h: In function '__m256i util::_my_mm256_min_epi32(__m256i, __m256i)':
api.cpp: util/extintrin.h:37:13: warning: unused variable 'max1' [-Wunused-variable]
api.cpp:    37 |     __m128i max1, min1, max2, min2;
api.cpp:       |             ^~~~
api.cpp: util/extintrin.h:37:25: warning: unused variable 'max2' [-Wunused-variable]
api.cpp:    37 |     __m128i max1, min1, max2, min2;
api.cpp:       |                         ^~~~
api.cpp: util/extintrin.h: In function '__m256i util::_my_mm256_max_epi32(__m256i, __m256i)':
api.cpp: util/extintrin.h:53:19: warning: unused variable 'min1' [-Wunused-variable]
api.cpp:    53 |     __m128i max1, min1, max2, min2;
api.cpp:       |                   ^~~~
api.cpp: util/extintrin.h:53:31: warning: unused variable 'min2' [-Wunused-variable]
api.cpp:    53 |     __m128i max1, min1, max2, min2;
api.cpp:       |                               ^~~~
api.cpp: In file included from sorter.h:120,
api.cpp:                  from aspas.tcc:31,
api.cpp:                  from aspas.h:204,
api.cpp:                  from api.cpp:3:
api.cpp: sorter_avx2.tcc: At global scope:
api.cpp: sorter_avx2.tcc:96:47: warning: ignoring attributes on template argument '__m256' [-Wignored-attributes]
api.cpp: ...

Compiler output

sort.c: sort.c:940:40: error: always_inline function '_mm256_set1_epi32' requires target feature 'avx', but would be inlined into function 'int32_sort' that is compiled without support for 'avx'
sort.c:     for (i = q>>3;i < q>>2;++i) y[i] = _mm256_set1_epi32(0x7fffffff);
sort.c:                                        ^
sort.c: sort.c:940:40: error: AVX vector return of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
sort.c: sort.c:956:22: error: always_inline function '_mm256_loadu_si256' requires target feature 'avx', but would be inlined into function 'int32_sort' that is compiled without support for 'avx'
sort.c:         int32x8 x0 = int32x8_load(&x[i]);
sort.c:                      ^
sort.c: sort.c:9:25: note: expanded from macro 'int32x8_load'
sort.c: #define int32x8_load(z) _mm256_loadu_si256((__m256i *) (z))
sort.c:                         ^
sort.c: sort.c:956:22: error: AVX vector return of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
sort.c: sort.c:9:25: note: expanded from macro 'int32x8_load'
sort.c: #define int32x8_load(z) _mm256_loadu_si256((__m256i *) (z))
sort.c:                         ^
sort.c: sort.c:957:22: error: always_inline function '_mm256_loadu_si256' requires target feature 'avx', but would be inlined into function 'int32_sort' that is compiled without support for 'avx'
sort.c:         int32x8 x1 = int32x8_load(&x[i+q]);
sort.c:                      ^
sort.c: sort.c:9:25: note: expanded from macro 'int32x8_load'
sort.c: #define int32x8_load(z) _mm256_loadu_si256((__m256i *) (z))
sort.c:                         ^
sort.c: sort.c:957:22: error: AVX vector return of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
sort.c: sort.c:9:25: note: expanded from macro 'int32x8_load'
sort.c: #define int32x8_load(z) _mm256_loadu_si256((__m256i *) (z))
sort.c:                         ^
sort.c: sort.c:958:22: error: always_inline function '_mm256_loadu_si256' requires target feature 'avx', but would be inlined into function 'int32_sort' that is compiled without support for 'avx'
sort.c: ...

Compiler output

sort.c: sort.c:133:17: warning: passing 'int32 [n]' to parameter of type 'uint32 *' (aka 'unsigned int *') converts between pointers to integer types with different sign [-Wpointer-sign]
sort.c:   RadixSort11(x,y,n);
sort.c:                 ^
sort.c: sort.c:47:48: note: passing argument to parameter 'sort' here
sort.c: static void RadixSort11(uint32 *array, uint32 *sort, uint32 elements)
sort.c:                                                ^
sort.c: 1 warning generated.

Compiler output

sort.c: sort.c: In function 'crypto_sort_int32_herf_timingleaks':
sort.c: sort.c:133:17: warning: pointer targets in passing argument 2 of 'RadixSort11' differ in signedness [-Wpointer-sign]
sort.c:   133 |   RadixSort11(x,y,n);
sort.c:       |                 ^
sort.c:       |                 |
sort.c:       |                 int32 * {aka int *}
sort.c: sort.c:47:48: note: expected 'uint32 *' {aka 'unsigned int *'} but argument is of type 'int32 *' {aka 'int *'}
sort.c:    47 | static void RadixSort11(uint32 *array, uint32 *sort, uint32 elements)
sort.c:       |                                        ~~~~~~~~^~~~

Compiler output

int32_sort.c: int32_sort.c:330:15: error: always_inline function '_mm256_loadu_si256' requires target feature 'avx', but would be inlined into function 'minmax8' that is compiled without support for 'avx'
int32_sort.c:   __m256i a = _mm256_loadu_si256((__m256i *) x);
int32_sort.c:               ^
int32_sort.c: int32_sort.c:330:15: error: AVX vector return of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
int32_sort.c: int32_sort.c:331:15: error: always_inline function '_mm256_loadu_si256' requires target feature 'avx', but would be inlined into function 'minmax8' that is compiled without support for 'avx'
int32_sort.c:   __m256i b = _mm256_loadu_si256((__m256i *) y);
int32_sort.c:               ^
int32_sort.c: int32_sort.c:331:15: error: AVX vector return of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
int32_sort.c: int32_sort.c:332:37: error: always_inline function '_mm256_min_epi32' requires target feature 'avx2', but would be inlined into function 'minmax8' that is compiled without support for 'avx2'
int32_sort.c:   _mm256_storeu_si256((__m256i *) x,_mm256_min_epi32(a,b));
int32_sort.c:                                     ^
int32_sort.c: int32_sort.c:332:37: error: AVX vector argument of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
int32_sort.c: int32_sort.c:332:3: error: always_inline function '_mm256_storeu_si256' requires target feature 'avx', but would be inlined into function 'minmax8' that is compiled without support for 'avx'
int32_sort.c:   _mm256_storeu_si256((__m256i *) x,_mm256_min_epi32(a,b));
int32_sort.c:   ^
int32_sort.c: int32_sort.c:332:3: error: AVX vector argument of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
int32_sort.c: int32_sort.c:333:37: error: always_inline function '_mm256_max_epi32' requires target feature 'avx2', but would be inlined into function 'minmax8' that is compiled without support for 'avx2'
int32_sort.c:   _mm256_storeu_si256((__m256i *) y,_mm256_max_epi32(a,b));
int32_sort.c:                                     ^
int32_sort.c: int32_sort.c:333:37: error: AVX vector argument of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
int32_sort.c: int32_sort.c:333:3: error: always_inline function '_mm256_storeu_si256' requires target feature 'avx', but would be inlined into function 'minmax8' that is compiled without support for 'avx'
int32_sort.c:   _mm256_storeu_si256((__m256i *) y,_mm256_max_epi32(a,b));
int32_sort.c:   ^
int32_sort.c: int32_sort.c:333:3: error: AVX vector argument of type '__m256i' (vector of 4 'long long' values) without 'avx' enabled changes the ABI
int32_sort.c: int32_sort.c:364:34: error: always_inline function '_mm_min_epi32' requires target feature 'sse4.1', but would be inlined into function 'minmax4' that is compiled without support for 'sse4.1'
int32_sort.c: ...

Passed TIMECOP

TIMECOP iterations: 10