Skip to content
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

Jump to bottom

[SYCL] Support non-x86 platforms #2333

Merged
merged 5 commits into from
Sep 3, 2020
Merged

[SYCL] Support non-x86 platforms #2333

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
abedcbd
support non-x86 platforms
Aug 12, 2020
e44b026
clang-format fixes
Aug 18, 2020
1c3bb5c
NIT camel case fix
Aug 23, 2020
95c3fb5
remove #warning in #else case
Sep 2, 2020
76b6e28
silence compiler issues on Windows
Sep 3, 2020
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
44 changes: 40 additions & 4 deletions sycl/source/detail/platform_util.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,7 +11,11 @@
#include <detail/platform_util.hpp>

#if defined(SYCL_RT_OS_LINUX)
#include <errno.h>
#include <unistd.h>
#if defined(__x86_64__) || defined(__i386__)
#include <cpuid.h>
#endif
#elif defined(SYCL_RT_OS_WINDOWS)
#include <intrin.h>
#endif
Expand All @@ -20,6 +24,7 @@ __SYCL_INLINE_NAMESPACE(cl) {
namespace sycl {
namespace detail {

#if defined(__x86_64__) || defined(__i386__)
// Used by methods that duplicate OpenCL behaviour in order to get CPU info
static void cpuid(uint32_t *CPUInfo, uint32_t Type, uint32_t SubType = 0) {
#if defined(SYCL_RT_OS_LINUX)
Expand All @@ -28,11 +33,13 @@ static void cpuid(uint32_t *CPUInfo, uint32_t Type, uint32_t SubType = 0) {
__cpuidex(reinterpret_cast<int *>(CPUInfo), Type, SubType);
#endif
}
#endif

uint32_t PlatformUtil::getMaxClockFrequency() {
throw runtime_error(
"max_clock_frequency parameter is not supported for host device",
PI_INVALID_DEVICE);
#if defined(__x86_64__) || defined(__i386__)
uint32_t CPUInfo[4];
string_class Buff(sizeof(CPUInfo) * 3 + 1, 0);
size_t Offset = 0;
Expand Down Expand Up @@ -62,21 +69,43 @@ uint32_t PlatformUtil::getMaxClockFrequency() {
Buff = Buff.substr(Buff.rfind(' '), Buff.length());
Freq *= std::stod(Buff);
return Freq;
#endif
return 0;
}

uint32_t PlatformUtil::getMemCacheLineSize() {
#if defined(__x86_64__) || defined(__i386__)
uint32_t CPUInfo[4];
cpuid(CPUInfo, 0x80000006);
return CPUInfo[2] & 0xff;
#elif defined(SYCL_RT_OS_LINUX) && defined(_SC_LEVEL2_DCACHE_LINESIZE)
long lineSize = sysconf(_SC_LEVEL2_DCACHE_LINESIZE);
if (lineSize > 0) {
return lineSize;
}
#endif
return 8;
}

uint64_t PlatformUtil::getMemCacheSize() {
#if defined(__x86_64__) || defined(__i386__)
uint32_t CPUInfo[4];
cpuid(CPUInfo, 0x80000006);
return static_cast<uint64_t>(CPUInfo[2] >> 16) * 1024;
#elif defined(SYCL_RT_OS_LINUX) && defined(_SC_LEVEL2_DCACHE_SIZE)
long cacheSize = sysconf(_SC_LEVEL2_DCACHE_SIZE);
if (cacheSize > 0) {
return cacheSize;
}
#endif
return static_cast<uint64_t>(16 * 1024);
}

uint32_t PlatformUtil::getNativeVectorWidth(PlatformUtil::TypeIndex TIndex) {

#if defined(__x86_64__) || defined(__i386__)
uint32_t Index = static_cast<uint32_t>(TIndex);

// SSE4.2 has 16 byte (XMM) registers
static constexpr uint32_t VECTOR_WIDTH_SSE42[] = {16, 8, 4, 2, 4, 2, 0};
// AVX supports 32 byte (YMM) registers only for floats and doubles
Expand All @@ -86,8 +115,6 @@ uint32_t PlatformUtil::getNativeVectorWidth(PlatformUtil::TypeIndex TIndex) {
// AVX512 has 64 byte (ZMM) registers
static constexpr uint32_t VECTOR_WIDTH_AVX512[] = {64, 32, 16, 8, 16, 8, 0};

uint32_t Index = static_cast<uint32_t>(TIndex);

#if defined(SYCL_RT_OS_LINUX)
if (__builtin_cpu_supports("avx512f"))
return VECTOR_WIDTH_AVX512[Index];
Expand Down Expand Up @@ -119,14 +146,23 @@ uint32_t PlatformUtil::getNativeVectorWidth(PlatformUtil::TypeIndex TIndex) {
#endif

return VECTOR_WIDTH_SSE42[Index];

#elif defined(__ARM_NEON)
uint32_t Index = static_cast<uint32_t>(TIndex);

// NEON has 16 byte registers
static constexpr uint32_t VECTOR_WIDTH_NEON[] = {16, 8, 4, 2, 4, 2, 0};
return VECTOR_WIDTH_NEON[Index];

#endif
return 0;
}

void PlatformUtil::prefetch(const char *Ptr, size_t NumBytes) {
if (!Ptr)
return;

// The current implementation assumes 64-byte x86 cache lines.
const size_t CacheLineSize = 64;
const size_t CacheLineSize = PlatformUtil::getMemCacheLineSize();
const size_t CacheLineMask = ~(CacheLineSize - 1);
const char *PtrEnd = Ptr + NumBytes;

Expand Down