aboutsummaryrefslogtreecommitdiff
path: root/libc
diff options
context:
space:
mode:
Diffstat (limited to 'libc')
-rw-r--r--libc/src/__support/FPUtil/CMakeLists.txt2
-rw-r--r--libc/src/__support/FPUtil/bfloat16.h25
-rw-r--r--libc/src/__support/FPUtil/rounding_mode.h67
-rw-r--r--libc/src/__support/GPU/allocator.cpp138
-rw-r--r--libc/src/__support/RPC/rpc_server.h6
5 files changed, 152 insertions, 86 deletions
diff --git a/libc/src/__support/FPUtil/CMakeLists.txt b/libc/src/__support/FPUtil/CMakeLists.txt
index f157d90..3024ac6 100644
--- a/libc/src/__support/FPUtil/CMakeLists.txt
+++ b/libc/src/__support/FPUtil/CMakeLists.txt
@@ -16,6 +16,7 @@ add_header_library(
rounding_mode.h
DEPENDS
libc.hdr.fenv_macros
+ libc.src.__support.CPP.type_traits
libc.src.__support.macros.attributes
libc.src.__support.macros.properties.architectures
libc.src.__support.macros.sanitizer
@@ -274,6 +275,7 @@ add_header_library(
bfloat16.h
DEPENDS
.cast
+ .comparison_operations
.dyadic_float
libc.src.__support.CPP.bit
libc.src.__support.CPP.type_traits
diff --git a/libc/src/__support/FPUtil/bfloat16.h b/libc/src/__support/FPUtil/bfloat16.h
index bc0b8b2..84a0dca 100644
--- a/libc/src/__support/FPUtil/bfloat16.h
+++ b/libc/src/__support/FPUtil/bfloat16.h
@@ -12,6 +12,7 @@
#include "src/__support/CPP/bit.h"
#include "src/__support/CPP/type_traits.h"
#include "src/__support/FPUtil/cast.h"
+#include "src/__support/FPUtil/comparison_operations.h"
#include "src/__support/FPUtil/dyadic_float.h"
#include "src/__support/macros/config.h"
#include "src/__support/macros/properties/types.h"
@@ -57,6 +58,30 @@ struct BFloat16 {
uint32_t x_bits = static_cast<uint32_t>(bits) << 16U;
return cpp::bit_cast<float>(x_bits);
}
+
+ LIBC_INLINE bool operator==(BFloat16 other) const {
+ return fputil::equals(*this, other);
+ }
+
+ LIBC_INLINE bool operator!=(BFloat16 other) const {
+ return !fputil::equals(*this, other);
+ }
+
+ LIBC_INLINE bool operator<(BFloat16 other) const {
+ return fputil::less_than(*this, other);
+ }
+
+ LIBC_INLINE bool operator<=(BFloat16 other) const {
+ return fputil::less_than_or_equals(*this, other);
+ }
+
+ LIBC_INLINE bool operator>(BFloat16 other) const {
+ return fputil::greater_than(*this, other);
+ }
+
+ LIBC_INLINE bool operator>=(BFloat16 other) const {
+ return fputil::greater_than_or_equals(*this, other);
+ }
}; // struct BFloat16
} // namespace fputil
diff --git a/libc/src/__support/FPUtil/rounding_mode.h b/libc/src/__support/FPUtil/rounding_mode.h
index bc66d09..4ee0a0b 100644
--- a/libc/src/__support/FPUtil/rounding_mode.h
+++ b/libc/src/__support/FPUtil/rounding_mode.h
@@ -10,6 +10,7 @@
#define LLVM_LIBC_SRC___SUPPORT_FPUTIL_ROUNDING_MODE_H
#include "hdr/fenv_macros.h"
+#include "src/__support/CPP/type_traits.h" // is_constant_evaluated
#include "src/__support/macros/attributes.h" // LIBC_INLINE
#include "src/__support/macros/config.h"
@@ -20,18 +21,26 @@ namespace fputil {
// Using the following observation:
// 1.0f + 2^-25 = 1.0f for FE_TONEAREST, FE_DOWNWARD, FE_TOWARDZERO
// = 0x1.000002f for FE_UPWARD.
-LIBC_INLINE bool fenv_is_round_up() {
- volatile float x = 0x1.0p-25f;
- return (1.0f + x != 1.0f);
+LIBC_INLINE static constexpr bool fenv_is_round_up() {
+ if (cpp::is_constant_evaluated()) {
+ return false;
+ } else {
+ volatile float x = 0x1.0p-25f;
+ return (1.0f + x != 1.0f);
+ }
}
// Quick free-standing test whether fegetround() == FE_DOWNWARD.
// Using the following observation:
// -1.0f - 2^-25 = -1.0f for FE_TONEAREST, FE_UPWARD, FE_TOWARDZERO
// = -0x1.000002f for FE_DOWNWARD.
-LIBC_INLINE bool fenv_is_round_down() {
- volatile float x = 0x1.0p-25f;
- return (-1.0f - x != -1.0f);
+LIBC_INLINE static constexpr bool fenv_is_round_down() {
+ if (cpp::is_constant_evaluated()) {
+ return false;
+ } else {
+ volatile float x = 0x1.0p-25f;
+ return (-1.0f - x != -1.0f);
+ }
}
// Quick free-standing test whether fegetround() == FE_TONEAREST.
@@ -40,10 +49,14 @@ LIBC_INLINE bool fenv_is_round_down() {
// = 0x1.100002p0f for FE_UPWARD,
// 1.5f - 2^-24 = 1.5f for FE_TONEAREST, FE_UPWARD
// = 0x1.0ffffep-1f for FE_DOWNWARD, FE_TOWARDZERO
-LIBC_INLINE bool fenv_is_round_to_nearest() {
- static volatile float x = 0x1.0p-24f;
- float y = x;
- return (1.5f + y == 1.5f - y);
+LIBC_INLINE static constexpr bool fenv_is_round_to_nearest() {
+ if (cpp::is_constant_evaluated()) {
+ return true;
+ } else {
+ volatile float x = 0x1.0p-24f;
+ float y = 1.5f + x;
+ return (y == 1.5f - x);
+ }
}
// Quick free-standing test whether fegetround() == FE_TOWARDZERO.
@@ -56,23 +69,31 @@ LIBC_INLINE bool fenv_is_round_to_nearest() {
// (0x1.000002p0f + 2^-24) + (-1.0f - 2^-24) = 2^-23 for FE_TOWARDZERO
// = 2^-22 for FE_TONEAREST, FE_UPWARD
// = 0 for FE_DOWNWARD
-LIBC_INLINE bool fenv_is_round_to_zero() {
- static volatile float x = 0x1.0p-24f;
- float y = x;
- return ((0x1.000002p0f + y) + (-1.0f - y) == 0x1.0p-23f);
+LIBC_INLINE static constexpr bool fenv_is_round_to_zero() {
+ if (cpp::is_constant_evaluated()) {
+ return false;
+ } else {
+ volatile float x = 0x1.0p-24f;
+ volatile float y = 0x1.000002p0f + x;
+ return (y + (-1.0f - x) == 0x1.0p-23f);
+ }
}
// Quick free standing get rounding mode based on the above observations.
-LIBC_INLINE int quick_get_round() {
- static volatile float x = 0x1.0p-24f;
- float y = x;
- float z = (0x1.000002p0f + y) + (-1.0f - y);
+LIBC_INLINE static constexpr int quick_get_round() {
+ if (cpp::is_constant_evaluated()) {
+ return FE_TONEAREST;
+ } else {
+ volatile float x = 0x1.0p-24f;
+ volatile float y = 0x1.000002p0f + x;
+ float z = y + (-1.0f - x);
- if (z == 0.0f)
- return FE_DOWNWARD;
- if (z == 0x1.0p-23f)
- return FE_TOWARDZERO;
- return (2.0f + y == 2.0f) ? FE_TONEAREST : FE_UPWARD;
+ if (z == 0.0f)
+ return FE_DOWNWARD;
+ if (z == 0x1.0p-23f)
+ return FE_TOWARDZERO;
+ return (2.0f + x == 2.0f) ? FE_TONEAREST : FE_UPWARD;
+ }
}
} // namespace fputil
diff --git a/libc/src/__support/GPU/allocator.cpp b/libc/src/__support/GPU/allocator.cpp
index 7923fbb..866aea7 100644
--- a/libc/src/__support/GPU/allocator.cpp
+++ b/libc/src/__support/GPU/allocator.cpp
@@ -34,13 +34,12 @@ constexpr static uint32_t BITS_IN_WORD = sizeof(uint32_t) * 8;
constexpr static uint32_t MIN_SIZE = 16;
constexpr static uint32_t MIN_ALIGNMENT = MIN_SIZE - 1;
+// The number of times to attempt claiming an in-progress slab allocation.
+constexpr static uint32_t MAX_TRIES = 128;
+
// A sentinel used to indicate an invalid but non-null pointer value.
constexpr static uint64_t SENTINEL = cpp::numeric_limits<uint64_t>::max();
-// The number of times we will try starting on a single index before skipping
-// past it.
-constexpr static uint32_t MAX_TRIES = 512;
-
static_assert(!(ARRAY_SIZE & (ARRAY_SIZE - 1)), "Must be a power of two");
namespace impl {
@@ -92,20 +91,10 @@ static inline uint32_t xorshift32(uint32_t &state) {
return state * 0x9e3779bb;
}
-// Final stage of murmurhash used to get a unique index for the global array
-static inline uint32_t hash(uint32_t x) {
- x ^= x >> 16;
- x *= 0x85ebca6b;
- x ^= x >> 13;
- x *= 0xc2b2ae35;
- x ^= x >> 16;
- return x;
-}
-
// Rounds the input value to the closest permitted chunk size. Here we accept
// the sum of the closest three powers of two. For a 2MiB slab size this is 48
// different chunk sizes. This gives us average internal fragmentation of 87.5%.
-static inline uint32_t get_chunk_size(uint32_t x) {
+static inline constexpr uint32_t get_chunk_size(uint32_t x) {
uint32_t y = x < MIN_SIZE ? MIN_SIZE : x;
uint32_t pow2 = BITS_IN_WORD - cpp::countl_zero(y - 1);
@@ -123,6 +112,16 @@ static inline uint32_t get_chunk_size(uint32_t x) {
return (s3 + MIN_ALIGNMENT) & ~MIN_ALIGNMENT;
}
+// Converts a chunk size into an index suitable for a statically sized array.
+static inline constexpr uint32_t get_chunk_id(uint32_t x) {
+ if (x <= MIN_SIZE)
+ return 0;
+ uint32_t y = x >> 4;
+ if (x < MIN_SIZE << 2)
+ return cpp::popcount(y);
+ return cpp::popcount(y) + 3 * (BITS_IN_WORD - cpp::countl_zero(y)) - 7;
+}
+
// Rounds to the nearest power of two.
template <uint32_t N, typename T>
static inline constexpr T round_up(const T x) {
@@ -143,6 +142,12 @@ static inline constexpr bool is_pow2(uint64_t x) {
return x && (x & (x - 1)) == 0;
}
+// Where this chunk size should start looking in the global array.
+static inline constexpr uint32_t start_index(uint32_t chunk_index) {
+ return (ARRAY_SIZE * impl::get_chunk_id(chunk_index)) /
+ impl::get_chunk_id(SLAB_SIZE / 2);
+}
+
} // namespace impl
/// A slab allocator used to hand out identically sized slabs of memory.
@@ -251,12 +256,18 @@ struct Slab {
// The uniform mask represents which lanes contain a uniform target pointer.
// We attempt to place these next to each other.
void *result = nullptr;
+ uint32_t after = ~0u;
+ uint32_t old_index = 0;
for (uint64_t mask = lane_mask; mask;
mask = gpu::ballot(lane_mask, !result)) {
if (result)
continue;
- uint32_t start = gpu::broadcast_value(lane_mask, impl::xorshift32(state));
+ // We try using any known empty bits from the previous attempt first.
+ uint32_t start = gpu::shuffle(mask, cpp::countr_zero(uniform & mask),
+ ~after ? (old_index & ~(BITS_IN_WORD - 1)) +
+ cpp::countr_zero(~after)
+ : impl::xorshift32(state));
uint32_t id = impl::lane_count(uniform & mask);
uint32_t index = (start + id) % usable_bits(chunk_size);
@@ -266,8 +277,9 @@ struct Slab {
// Get the mask of bits destined for the same slot and coalesce it.
uint64_t match = uniform & gpu::match_any(mask, slot);
uint32_t length = cpp::popcount(match);
- uint32_t bitmask = static_cast<uint32_t>((uint64_t(1) << length) - 1)
- << bit;
+ uint32_t bitmask = gpu::shuffle(
+ mask, cpp::countr_zero(match),
+ static_cast<uint32_t>((uint64_t(1) << length) - 1) << bit);
uint32_t before = 0;
if (gpu::get_lane_id() == static_cast<uint32_t>(cpp::countr_zero(match)))
@@ -278,6 +290,9 @@ struct Slab {
result = ptr_from_index(index, chunk_size);
else
sleep_briefly();
+
+ after = before | bitmask;
+ old_index = index;
}
cpp::atomic_thread_fence(cpp::MemoryOrder::ACQUIRE);
@@ -451,66 +466,65 @@ public:
// The global array used to search for a valid slab to allocate from.
static GuardPtr slots[ARRAY_SIZE] = {};
+// Keep a cache of the last successful slot for each chunk size. Initialize it
+// to an even spread of the total size. Must be updated if the chunking scheme
+// changes.
+#define S(X) (impl::start_index(X))
+static cpp::Atomic<uint32_t> indices[] = {
+ S(16), S(32), S(48), S(64), S(96), S(112), S(128),
+ S(192), S(224), S(256), S(384), S(448), S(512), S(768),
+ S(896), S(1024), S(1536), S(1792), S(2048), S(3072), S(3584),
+ S(4096), S(6144), S(7168), S(8192), S(12288), S(14336), S(16384),
+ S(24576), S(28672), S(32768), S(49152), S(57344), S(65536), S(98304),
+ S(114688), S(131072), S(196608), S(229376), S(262144), S(393216), S(458752),
+ S(524288), S(786432), S(917504), S(1048576)};
+#undef S
+
// Tries to find a slab in the table that can support the given chunk size.
static Slab *find_slab(uint32_t chunk_size) {
- // We start at a hashed value to spread out different chunk sizes.
- uint32_t start = impl::hash(chunk_size);
- uint64_t lane_mask = gpu::get_lane_mask();
- uint64_t uniform = gpu::match_any(lane_mask, chunk_size);
-
- Slab *result = nullptr;
- uint32_t nudge = 0;
- for (uint64_t mask = lane_mask; mask;
- mask = gpu::ballot(lane_mask, !result), ++nudge) {
- uint32_t index = cpp::numeric_limits<uint32_t>::max();
- for (uint32_t offset = nudge / MAX_TRIES;
- gpu::ballot(lane_mask, index == cpp::numeric_limits<uint32_t>::max());
- offset += cpp::popcount(uniform & lane_mask)) {
- uint32_t candidate =
- (start + offset + impl::lane_count(uniform & lane_mask)) % ARRAY_SIZE;
- uint64_t available =
- gpu::ballot(lane_mask, slots[candidate].use_count() <
- Slab::available_chunks(chunk_size));
- uint32_t new_index = gpu::shuffle(
- lane_mask, cpp::countr_zero(available & uniform), candidate);
-
- // Each uniform group will use the first empty slot they find.
- if ((index == cpp::numeric_limits<uint32_t>::max() &&
- (available & uniform)))
- index = new_index;
-
- // Guaruntees that this loop will eventuall exit if there is no space.
- if (offset >= ARRAY_SIZE) {
- result = reinterpret_cast<Slab *>(SENTINEL);
- index = 0;
- }
- }
+ // We start at the index of the last successful allocation for this kind.
+ uint32_t chunk_id = impl::get_chunk_id(chunk_size);
+ uint32_t start = indices[chunk_id].load(cpp::MemoryOrder::RELAXED);
+ uint64_t uniform = gpu::match_any(gpu::get_lane_mask(), chunk_size);
+
+ for (uint32_t offset = 0; offset < ARRAY_SIZE; ++offset) {
+ uint32_t index =
+ !offset ? start : (impl::start_index(chunk_size) + offset) % ARRAY_SIZE;
- // Try to claim a slot for the found slot.
- if (!result) {
+ if (slots[index].use_count() < Slab::available_chunks(chunk_size)) {
+ uint64_t lane_mask = gpu::get_lane_mask();
uint64_t reserved = 0;
- Slab *slab = slots[index].try_lock(lane_mask & mask, uniform & mask,
+
+ Slab *slab = slots[index].try_lock(lane_mask, uniform & lane_mask,
reserved, chunk_size, index);
+
+ // If there is a slab allocation in progress we retry a few times.
+ for (uint32_t retries = 0;
+ retries < MAX_TRIES && !slab && reserved != SENTINEL; retries++) {
+ uint64_t lane_mask = gpu::get_lane_mask();
+ slab = slots[index].try_lock(lane_mask, uniform & lane_mask, reserved,
+ chunk_size, index);
+ sleep_briefly();
+ }
+
// If we find a slab with a matching chunk size then we store the result.
// Otherwise, we need to free the claimed lock and continue. In the case
- // of out-of-memory we return a sentinel value.
+ // of out-of-memory we receive a sentinel value and return a failure.
if (slab && reserved <= Slab::available_chunks(chunk_size) &&
slab->get_chunk_size() == chunk_size) {
- result = slab;
+ if (index != start)
+ indices[chunk_id].store(index, cpp::MemoryOrder::RELAXED);
+ return slab;
} else if (slab && (reserved > Slab::available_chunks(chunk_size) ||
slab->get_chunk_size() != chunk_size)) {
- if (slab->get_chunk_size() != chunk_size)
- start = index + 1;
slots[index].unlock(gpu::get_lane_mask(),
gpu::get_lane_mask() & uniform);
- } else if (!slab && reserved == cpp::numeric_limits<uint64_t>::max()) {
- result = reinterpret_cast<Slab *>(SENTINEL);
- } else {
- sleep_briefly();
+ } else if (!slab && reserved == SENTINEL) {
+ return nullptr;
}
}
}
- return result;
+ return nullptr;
}
// Release the lock associated with a given slab.
diff --git a/libc/src/__support/RPC/rpc_server.h b/libc/src/__support/RPC/rpc_server.h
index dc3d803..beea4dd 100644
--- a/libc/src/__support/RPC/rpc_server.h
+++ b/libc/src/__support/RPC/rpc_server.h
@@ -15,13 +15,17 @@
#ifndef LLVM_LIBC_SRC___SUPPORT_RPC_RPC_SERVER_H
#define LLVM_LIBC_SRC___SUPPORT_RPC_RPC_SERVER_H
+#include "src/__support/macros/properties/compiler.h"
+
// Workaround for missing __has_builtin in < GCC 10.
#ifndef __has_builtin
#define __has_builtin(x) 0
#endif
// Workaround for missing __builtin_is_constant_evaluated in < GCC 10.
-#ifndef __builtin_is_constant_evaluated
+// Also this builtin is defined for GCC 9.
+#if !(__has_builtin(__builtin_is_constant_evaluated) || \
+ (defined(LIBC_COMPILER_IS_GCC) && (LIBC_COMPILER_GCC_VER >= 900)))
#define __builtin_is_constant_evaluated(x) 0
#endif
generated by cgit v1.1 at 2025-07-24 02:27:38 +0000