Skip to content

[Clang][X86] Add tests for AVX512 integer comparison intrinsics to be used in constexpr #169452

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 10 commits into
base: main
Choose a base branch
from
Open

[Clang][X86] Add tests for AVX512 integer comparison intrinsics to be used in constexpr #169452

Changes from 2 commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
dccf63b
add: tests for _mm_cmp*_epi8
shri-acha Nov 25, 2025
6d54113
Merge branch 'llvm:main' into main
shri-acha Nov 25, 2025
120d264
add tests for _mm_cmp*_epi16
shri-acha Nov 25, 2025
448d362
add: tests for _mm_cmp*_epu8_mask
shri-acha Nov 25, 2025
ca27b8a
Merge branch 'main' of github.com:shri-acha/llvm-project
shri-acha Nov 25, 2025
e4e4e01
add: tests for _mm_cmp*_epu16_mask
shri-acha Nov 25, 2025
bffb0d8
add: tests for _mm256_cmp*_epi8_mask
shri-acha Nov 28, 2025
e8893d5
Merge branch 'main' into main
RKSimon Nov 28, 2025
cf7039c
add: tests for _mm256_cmpeq_epu8
shri-acha Nov 30, 2025
60764b0
Merge branch 'main' of github.com:shri-acha/llvm-project
shri-acha Nov 30, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
121 changes: 112 additions & 9 deletions clang/test/CodeGen/X86/avx512vlbw-builtins.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -645,20 +645,123 @@ __mmask16 test_mm_cmp_epi8_mask(__m128i __a, __m128i __b) {
return (__mmask16)_mm_cmp_epi8_mask(__a, __b, 0);
}

// cmpeq tests
TEST_CONSTEXPR(_mm_cmpeq_epi8_mask(
((__m128i)(__v16qi){5, 3, 7, 2, 9, 3, 7, 1, 5, 4, 8, 2, 9, 6, 7, 5}),
((__m128i)(__v16qi){5, 2, 7, 3, 9, 4, 6, 1, 5, 3, 8, 1, 9, 5, 7, 5})
) == (__mmask16)0xd595);
((__m128i)(__v16qi){0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}),
((__m128i)(__v16qi){0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0})
) == (__mmask16)0xffff);
Copy link
Collaborator

@RKSimon RKSimon Nov 25, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

avoid 'splat' tests like this - all ones/zero results always make me suspicious :)

Copy link
Author

@shri-acha shri-acha Nov 25, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

:) okay ,will modify the 'splat' cases here onwards.


TEST_CONSTEXPR(_mm_cmpeq_epi8_mask(
((__m128i)(__v16qi){1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}),
((__m128i)(__v16qi){1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1})
) == (__mmask16)0xffff);

TEST_CONSTEXPR(_mm_cmpeq_epi8_mask(
((__m128i)(__v16qi){127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127}),
((__m128i)(__v16qi){-128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128})
Copy link
Collaborator

@RKSimon RKSimon Nov 28, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

you have to use __v16qs / __v32qs / __v64qs to allow you to use signed chars

Copy link
Author

@shri-acha shri-acha Nov 30, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

__vqs in epi variants and __vqi for epu variants then?

Copy link
Collaborator

@RKSimon RKSimon Nov 30, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

vXqu for unsigned char

) == (__mmask16)0x0000);

TEST_CONSTEXPR(_mm_cmpeq_epi8_mask(
((__m128i)(__v16qi){-31, 90, -66, 3, 27, -22, -64, 111, -87, 105, -19, 0, 26, -111, 24, -72}),
((__m128i)(__v16qi){-84, -98, 20, -83, -98, 80, -46, -9, 22, -120, -123, 53, 117, -85, 50, 94})
) == (__mmask16)0x0000);

Copy link
Collaborator

@RKSimon RKSimon Nov 25, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

You need _mm_mask_cmpeq_epi8_mask (et al) test coverage as well

Copy link
Author

@shri-acha shri-acha Nov 25, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

yes, I do have those in mind aswell, i will cover those after all of the _mm_mask_cmp*_epi* are complete.

// cmpneq tests
TEST_CONSTEXPR(_mm_cmpneq_epi8_mask(
((__m128i)(__v16qi){0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}),
((__m128i)(__v16qi){0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0})
) == (__mmask16)0x0000);

TEST_CONSTEXPR(_mm_cmpneq_epi8_mask(
((__m128i)(__v16qi){1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}),
((__m128i)(__v16qi){1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1})
) == (__mmask16)0x0000);

TEST_CONSTEXPR(_mm_cmpneq_epi8_mask(
((__m128i)(__v16qi){127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127}),
((__m128i)(__v16qi){-128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128})
) == (__mmask16)0xffff);

TEST_CONSTEXPR(_mm_cmpneq_epi8_mask(
((__m128i)(__v16qi){-2, 49, -5, -11, 21, -70, 121, -111, 20, 112, -121, 18, -34, -73, 89, 122}),
((__m128i)(__v16qi){14, 36, 2, 3, 118, 88, -126, -21, 104, -125, -1, 39, 99, -12, 35, -126})
) == (__mmask16)0xffff);

// cmplt tests
TEST_CONSTEXPR(_mm_cmplt_epi8_mask(
((__m128i)(__v16qi){0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}),
((__m128i)(__v16qi){0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0})
) == (__mmask16)0x0000);

TEST_CONSTEXPR(_mm_cmplt_epi8_mask(
((__m128i)(__v16qi){1, 5, 3, 7, 2, 8, 4, 6, 9, 5, 3, 11, 2, 6, 15, 8}),
((__m128i)(__v16qi){2, 4, 6, 8, 3, 5, 7, 9, 4, 6, 8, 10, 5, 7, 9, 11})
) == (__mmask16)0xb6dd);
((__m128i)(__v16qi){1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}),
((__m128i)(__v16qi){1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1})
) == (__mmask16)0x0000);

TEST_CONSTEXPR(_mm_cmplt_epi8_mask(
((__m128i)(__v16qi){127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127}),
((__m128i)(__v16qi){-128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128})
) == (__mmask16)0x0000);

TEST_CONSTEXPR(_mm_cmplt_epi8_mask(
((__m128i)(__v16qi){-111, -10, -60, -123, -6, -110, -43, -32, -58, -7, 42, -128, -21, 24, 8, -101}),
((__m128i)(__v16qi){-108, 30, 71, 73, 20, 117, 63, -93, 79, -30, 99, -100, 34, 49, 83, 68})
) == (__mmask16)0xfd7f);

// cmple tests
TEST_CONSTEXPR(_mm_cmple_epi8_mask(
((__m128i)(__v16qi){0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}),
((__m128i)(__v16qi){0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0})
) == (__mmask16)0xffff);
TEST_CONSTEXPR(_mm_cmple_epi8_mask(
((__m128i)(__v16qi){1, 3, 5, 7, 2, 6, 6, 8, 1, 3, 9, 7, 2, 4, 6, 10}),
((__m128i)(__v16qi){2, 3, 4, 7, 3, 4, 5, 8, 2, 3, 4, 7, 3, 4, 5, 8})
) == (__mmask16)0x3b9b);
((__m128i)(__v16qi){1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}),
((__m128i)(__v16qi){1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1})
) == (__mmask16)0xffff);
TEST_CONSTEXPR(_mm_cmple_epi8_mask(
((__m128i)(__v16qi){127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127}),
((__m128i)(__v16qi){-128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128})
) == (__mmask16)0x0000);
TEST_CONSTEXPR(_mm_cmple_epi8_mask(
((__m128i)(__v16qi){122, 6, -22, -94, 78, -35, -43, -16, -69, 124, -2, 24, -117, 8, -17, 118}),
((__m128i)(__v16qi){53, -50, 104, 11, 63, -77, -25, 102, 46, 62, 27, -28, -61, 68, 40, -65})
) == (__mmask16)0x75cc);

// cmpge tests
TEST_CONSTEXPR(_mm_cmpge_epi8_mask(
((__m128i)(__v16qi){0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}),
((__m128i)(__v16qi){0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0})
) == (__mmask16)0xffff);
TEST_CONSTEXPR(_mm_cmpge_epi8_mask(
((__m128i)(__v16qi){1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}),
((__m128i)(__v16qi){1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1})
) == (__mmask16)0xffff);
TEST_CONSTEXPR(_mm_cmpge_epi8_mask(
((__m128i)(__v16qi){127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127}),
((__m128i)(__v16qi){-128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128})
) == (__mmask16)0xffff);
TEST_CONSTEXPR(_mm_cmpge_epi8_mask(
((__m128i)(__v16qi){-11, 0, 97, 123, -48, 105, 26, -118, 62, -86, -94, -32, 14, -4, -50, 72}),
((__m128i)(__v16qi){-33, 49, 22, 31, -4, -81, 6, -22, 40, 127, -1, -106, 6, -64, 12, 8})
) == (__mmask16)0xb96d);

// cmpgt tests
TEST_CONSTEXPR(_mm_cmpgt_epi8_mask(
((__m128i)(__v16qi){0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0}),
((__m128i)(__v16qi){0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0})
) == (__mmask16)0x0000);
TEST_CONSTEXPR(_mm_cmpgt_epi8_mask(
((__m128i)(__v16qi){1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1}),
((__m128i)(__v16qi){1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1})
) == (__mmask16)0x0000);
TEST_CONSTEXPR(_mm_cmpgt_epi8_mask(
((__m128i)(__v16qi){127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127, 127}),
((__m128i)(__v16qi){-128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128})
) == (__mmask16)0xffff);
TEST_CONSTEXPR(_mm_cmpgt_epi8_mask(
((__m128i)(__v16qi){-127, 37, -123, -60, 98, -68, -115, 96, 80, -27, -105, 64, -6, -51, -45, -81}),
((__m128i)(__v16qi){-124, 59, 8, 59, 122, 47, -74, 120, 19, -43, -33, -76, 7, -22, -24, -14})
) == (__mmask16)0x0b00);


__mmask16 test_mm_mask_cmp_epi8_mask(__mmask16 __u, __m128i __a, __m128i __b) {
// CHECK-LABEL: test_mm_mask_cmp_epi8_mask
Expand Down
Loading