Skip to content

ARROW-13944: [C++] Bump xsimd to latest version #11142

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

Closed
cyb70289 wants to merge 1 commit into from
Closed

ARROW-13944: [C++] Bump xsimd to latest version #11142

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
8 changes: 7 additions & 1 deletion cpp/src/arrow/util/bpacking_simd128_generated.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,13 @@ using ::arrow::util::SafeLoad;
template <DispatchLevel level>
struct UnpackBits128 {

using simd_batch = xsimd::batch<uint32_t, 4>;
#ifdef ARROW_HAVE_NEON
using simd_arch = xsimd::neon64;
#else
using simd_arch = xsimd::sse4_2;
#endif

using simd_batch = xsimd::batch<uint32_t, simd_arch>;

inline static const uint32_t* unpack0_32(const uint32_t* in, uint32_t* out) {
memset(out, 0x0, 32 * sizeof(*out));
Expand Down
3 changes: 2 additions & 1 deletion cpp/src/arrow/util/bpacking_simd256_generated.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,8 @@ using ::arrow::util::SafeLoad;
template <DispatchLevel level>
struct UnpackBits256 {

using simd_batch = xsimd::batch<uint32_t, 8>;
using simd_arch = xsimd::avx2;
using simd_batch = xsimd::batch<uint32_t, simd_arch>;

inline static const uint32_t* unpack0_32(const uint32_t* in, uint32_t* out) {
memset(out, 0x0, 32 * sizeof(*out));
Expand Down
3 changes: 2 additions & 1 deletion cpp/src/arrow/util/bpacking_simd512_generated.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,8 @@ using ::arrow::util::SafeLoad;
template <DispatchLevel level>
struct UnpackBits512 {

using simd_batch = xsimd::batch<uint32_t, 16>;
using simd_arch = xsimd::avx512bw;
using simd_batch = xsimd::batch<uint32_t, simd_arch>;

inline static const uint32_t* unpack0_32(const uint32_t* in, uint32_t* out) {
memset(out, 0x0, 32 * sizeof(*out));
Expand Down
16 changes: 15 additions & 1 deletion cpp/src/arrow/util/bpacking_simd_codegen.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -152,6 +152,19 @@ def main(simd_width):

struct_name = f"UnpackBits{simd_width}"

define_simd_arch = {
# ugly format to get aligned output
128: dedent("""\
#ifdef ARROW_HAVE_NEON
using simd_arch = xsimd::neon64;
#else
using simd_arch = xsimd::sse4_2;
#endif
"""),
256: "using simd_arch = xsimd::avx2;",
512: "using simd_arch = xsimd::avx512bw;"
}

# NOTE: templating the UnpackBits struct on the dispatch level avoids
# potential name collisions if there are several UnpackBits generations
# with the same SIMD width on a given architecture.
Expand All @@ -176,7 +189,8 @@ def main(simd_width):
template <DispatchLevel level>
struct {struct_name} {{

using simd_batch = xsimd::batch<uint32_t, {simd_width // 32}>;
{define_simd_arch[simd_width]}
using simd_batch = xsimd::batch<uint32_t, simd_arch>;
"""))

gen = UnpackGenerator(simd_width)
Expand Down
10 changes: 7 additions & 3 deletions cpp/src/arrow/util/utf8.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -248,16 +248,20 @@ inline bool ValidateAsciiSw(const uint8_t* data, int64_t len) {

#if defined(ARROW_HAVE_NEON) || defined(ARROW_HAVE_SSE4_2)
inline bool ValidateAsciiSimd(const uint8_t* data, int64_t len) {
using simd_batch = xsimd::batch<int8_t, 16>;
#ifdef ARROW_HAVE_NEON
using simd_batch = xsimd::batch<int8_t, xsimd::neon64>;
#else
using simd_batch = xsimd::batch<int8_t, xsimd::sse4_2>;
#endif

if (len >= 32) {
const simd_batch zero(static_cast<int8_t>(0));
const uint8_t* data2 = data + 16;
simd_batch or1 = zero, or2 = zero;

while (len >= 32) {
or1 |= simd_batch(reinterpret_cast<const int8_t*>(data), xsimd::unaligned_mode{});
or2 |= simd_batch(reinterpret_cast<const int8_t*>(data2), xsimd::unaligned_mode{});
or1 |= simd_batch::load_unaligned(reinterpret_cast<const int8_t*>(data));
or2 |= simd_batch::load_unaligned(reinterpret_cast<const int8_t*>(data2));
data += 32;
data2 += 32;
len -= 32;
Expand Down
4 changes: 2 additions & 2 deletions cpp/thirdparty/versions.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -82,8 +82,8 @@ ARROW_THRIFT_BUILD_VERSION=0.13.0
ARROW_THRIFT_BUILD_SHA256_CHECKSUM=7ad348b88033af46ce49148097afe354d513c1fca7c607b59c33ebb6064b5179
ARROW_UTF8PROC_BUILD_VERSION=v2.6.1
ARROW_UTF8PROC_BUILD_SHA256_CHECKSUM=4c06a9dc4017e8a2438ef80ee371d45868bda2237a98b26554de7a95406b283b
ARROW_XSIMD_BUILD_VERSION=e9234cd6e6f4428fc260073b2c34ffe86fda1f34
ARROW_XSIMD_BUILD_SHA256_CHECKSUM=1e98bae41abae7f3f6fa4c70ec2dcad008d831876009aa047fb69fd5b24076fd
ARROW_XSIMD_BUILD_VERSION=f212f3c3801924bf218bc39705230a747467edcb
ARROW_XSIMD_BUILD_SHA256_CHECKSUM=8f362dfbc12026e6689563cbe5f1680443d50c2e133110245d270da5e2edc0a9
ARROW_ZLIB_BUILD_VERSION=1.2.11
ARROW_ZLIB_BUILD_SHA256_CHECKSUM=c3e5e9fdd5004dcb542feda5ee4f0ff0744628baf8ed2dd5d66f8ca1197cb1a1
ARROW_ZSTD_BUILD_VERSION=v1.5.0
Expand Down