sglang_v0.5.2/pytorch_2.8.0/third_party/XNNPACK/test/f32-raddextexp.cc

118 lines
5.9 KiB
C++

// Copyright 2019 Google LLC
//
// This source code is licensed under the BSD-style license found in the
// LICENSE file in the root directory of this source tree.
#include <algorithm>
#include <cassert>
#include <cmath>
#include <cstddef>
#include <cstdlib>
#include <functional>
#include <random>
#include <vector>
#include <gtest/gtest.h>
#include "replicable_random_device.h"
#include "xnnpack.h"
#include "xnnpack/buffer.h"
#include "xnnpack/common.h"
#include "xnnpack/isa-checks.h"
#include "xnnpack/microfnptr.h"
#include "xnnpack/raddextexp.h"
class RAddExtExpMicrokernelTester {
public:
RAddExtExpMicrokernelTester& elements(size_t elements) {
assert(elements != 0);
this->elements_ = elements;
return *this;
}
size_t elements() const {
return this->elements_;
}
RAddExtExpMicrokernelTester& iterations(size_t iterations) {
this->iterations_ = iterations;
return *this;
}
size_t iterations() const {
return this->iterations_;
}
void Test(xnn_f32_raddextexp_ukernel_fn raddextexp) const {
xnnpack::ReplicableRandomDevice rng;
// Choose such range that expf(x[i]) overflows, but double-precision exp doesn't overflow.
auto f32rng = [&rng]() {
return std::uniform_real_distribution<float>(90.0f, 100.0f)(rng);
};
xnnpack::Buffer<float> x(elements() + XNN_EXTRA_BYTES / sizeof(float));
for (size_t iteration = 0; iteration < iterations(); iteration++) {
std::generate(x.begin(), x.end(), std::ref(f32rng));
// Compute reference results.
double sum_ref = 0.0f;
for (size_t i = 0; i < elements(); i++) {
sum_ref += exp(double(x[i]));
}
// Call optimized micro-kernel.
float sum[2];
raddextexp(elements() * sizeof(float), x.data(), sum);
// Verify results.
ASSERT_NEAR(sum_ref, exp2(double(sum[1])) * double(sum[0]), std::abs(sum_ref) * 1.0e-6)
<< "elements = " << elements() << ", y:value = " << sum[0] << ", y:exponent = " << sum[1];
}
}
private:
size_t elements_{1};
size_t iterations_{15};
};
#define XNN_TEST_RADDEXTEXP_ELEMENT_EQ(ukernel, arch_flags, element_tile, ...) \
TEST(ukernel, element_eq) \
{ \
TEST_REQUIRES_ARCH_FLAGS(arch_flags); \
RAddExtExpMicrokernelTester().elements(element_tile).Test(ukernel); \
}
#define XNN_TEST_RADDEXTEXP_ELEMENT_DIV(ukernel, arch_flags, element_tile, ...) \
TEST(ukernel, element_gt) \
{ \
TEST_REQUIRES_ARCH_FLAGS(arch_flags); \
for (size_t element_size = element_tile * 2; element_size < element_tile * 10; element_size += element_tile) { \
RAddExtExpMicrokernelTester().elements(element_size).Test(ukernel); \
} \
}
#define XNN_TEST_RADDEXTEXP_ELEMENT_LT(ukernel, arch_flags, element_tile, ...) \
TEST(ukernel, element_lt) \
{ \
TEST_REQUIRES_ARCH_FLAGS(arch_flags); \
for (size_t element_size = 1; element_size < element_tile; element_size++) { \
RAddExtExpMicrokernelTester().elements(element_size).Test(ukernel); \
} \
}
#define XNN_TEST_RADDEXTEXP_ELEMENT_GT(ukernel, arch_flags, element_tile, ...) \
TEST(ukernel, element_div) \
{ \
TEST_REQUIRES_ARCH_FLAGS(arch_flags); \
for (size_t element_size = element_tile + 1; element_size < (element_tile == 1 ? 10 : element_tile * 2); \
element_size++) { \
RAddExtExpMicrokernelTester().elements(element_size).Test(ukernel); \
} \
}
#define XNN_UKERNEL_WITH_PARAMS(arch_flags, ukernel, element_tile, datatype, params_type, init_params) \
XNN_TEST_RADDEXTEXP_ELEMENT_EQ(ukernel, arch_flags, element_tile, init_params); \
XNN_TEST_RADDEXTEXP_ELEMENT_DIV(ukernel, arch_flags, element_tile, init_params); \
XNN_TEST_RADDEXTEXP_ELEMENT_LT(ukernel, arch_flags, element_tile, init_params); \
XNN_TEST_RADDEXTEXP_ELEMENT_GT(ukernel, arch_flags, element_tile, init_params);
#include "f32-raddextexp/f32-raddextexp.h"
#undef XNN_UKERNEL_WITH_PARAMS