Added log pow and fixed exp - #349
Merged
morousg merged 3 commits intoAug 19, 2026
Merged
Conversation
Contributor
There was a problem hiding this comment.
Pull request overview
This PR extends the cxp:: constexpr math utilities by adding log/pow support and adjusting the exp implementation, with accompanying unit tests to validate runtime vs constexpr behavior.
Changes:
- Added new
cxp::log/cxp::logfandcxp::pow/cxp::powfAPIs inconstexpr_cmath.h. - Refactored
expinto a templatedcxp::expplus a float wrappercxp::expf, and adjusted numeric-limits handling for CUDA builds. - Expanded utests to include sweeps/edge cases for
expf, and compile-time + runtime tests forlogandpow.
Reviewed changes
Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
include/fused_kernel/core/constexpr_libs/constexpr_cmath.h |
Adds log/pow, refactors exp, and adjusts CUDA numeric_limits plumbing. |
utests/core/constexpr_libs/utest_constexpr_expf_exact.h |
Adds runtime/constexpr sweep checks for cxp::expf vs std::expf. |
utests/core/constexpr_libs/utest_constexpr_cmath.h |
Adds compile-time and runtime validation for new log/pow APIs. |
.github/copilot-instructions.md |
Documents guidance about not declaring CUDA compilation blocked based on ad hoc nvcc checks. |
Suppressed comments (1)
include/fused_kernel/core/constexpr_libs/constexpr_cmath.h:1044
- Same shadowing issue in powf::BaseFunc::exec: the parameter name
basepreventsbase::is_constant_evaluated()from compiling. Rename the parameter and keepcxp::basenamespace accesses unshadowed.
using InstanceType = fk::BinaryType;
FK_HOST_DEVICE_FUSE float exec(const float base, const float exponent) {
if (base::is_constant_evaluated()) {
return pow::BaseFunc::exec(base, exponent);
💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.
Comment on lines
+948
to
+952
| using InstanceType = fk::BinaryType; | ||
| template <std::floating_point ST> | ||
| FK_HOST_DEVICE_FUSE ST exec(const ST base, const ST exponent) { | ||
| if (base::is_constant_evaluated()) { | ||
| return c_pow(base, exponent); |
Contributor
There was a problem hiding this comment.
Pull request overview
Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.
Suppressed comments (3)
include/fused_kernel/core/constexpr_libs/constexpr_cmath.h:954
pow::BaseFunc::execnames its first parameterbase, which shadows thecxp::basenamespace. As written,base::is_constant_evaluated()will not compile (attempts to use::on a function parameter). Rename the parameter or fully-qualify the namespace (e.g.cxp::base::is_constant_evaluated()).
FK_HOST_DEVICE_FUSE ST exec(const ST base, const ST exponent) {
if (base::is_constant_evaluated()) {
return c_pow(base, exponent);
} else {
return std::pow(base, exponent);
include/fused_kernel/core/constexpr_libs/constexpr_cmath.h:907
log::BaseFunc::c_log(double)similarly assumes a normal double when extracting exponent/mantissa. Subnormal doubles (exponent bits == 0, nonzero mantissa) will produce an incorrect result. Consider normalizing (e.g. scale by 2^54, adjust exponent) before computingeandm.
ulonglong bits = bit_cast<ulonglong>(x);
int e = static_cast<int>(bits >> 52) - 1023;
double m = bit_cast<double>((bits & 0x000FFFFFFFFFFFFFull) | 0x3FF0000000000000ull);
include/fused_kernel/core/constexpr_libs/constexpr_cmath.h:875
log::BaseFunc::c_log(float)extracts the exponent assuming a normal float. For subnormal inputs (exponent field == 0 but mantissa != 0),e/mare computed incorrectly, so constexprcxp::log::f(x)will be wrong for subnormals. Add a normalization step similar toldexpf(scale by 2^24, adjust exponent) before computingeandm.
This issue also appears on line 904 of the same file.
uint bits = bit_cast<uint>(x);
int e = static_cast<int>((bits >> 23) & 0xFF) - 127;
float m = bit_cast<float>((bits & 0x007FFFFF) | 0x3F800000);
if (m > 1.41421356237f) {
Comment on lines
+1042
to
+1046
| FK_HOST_DEVICE_FUSE float exec(const float base, const float exponent) { | ||
| if (base::is_constant_evaluated()) { | ||
| return pow::BaseFunc::exec(base, exponent); | ||
| } else { | ||
| return std::pow(base, exponent); |
morousg
deleted the
348-add-pow-and-powf-along-with-dependencies-in-cxp
branch
August 19, 2026 21:19
Comment on lines
+783
to
+786
| if (x < -709.0) | ||
| return 0.0; | ||
| if (x > 709.0) | ||
| return base::numeric_limits<double>::infinity(); |
Comment on lines
+871
to
+874
| uint bits = bit_cast<uint>(x); | ||
| int e = static_cast<int>((bits >> 23) & 0xFF) - 127; | ||
|
|
||
| float m = bit_cast<float>((bits & 0x007FFFFF) | 0x3F800000); |
Comment on lines
+982
to
+991
| if (base == 0.0f) | ||
| return (exponent > 0.0f) ? 0.0f : base::numeric_limits<float>::infinity(); | ||
|
|
||
| if (base < 0.0f) { | ||
| int e_int = static_cast<int>(exponent); | ||
| if (static_cast<float>(e_int) == exponent) { | ||
| float res = exp::BaseFunc::exec(exponent * log::BaseFunc::exec(-base)); | ||
| return (e_int % 2 != 0) ? -res : res; | ||
| } | ||
| return base::numeric_limits<float>::quiet_NaN(); |
| #define CXP_F_FUNC \ | ||
| template <typename... Types> \ | ||
| FK_HOST_DEVICE_FUSE auto f(const Types... vals) { \ | ||
| FK_HOST_DEVICE_FUSE decltype(auto) f(const Types... vals) { \ |
Comment on lines
+730
to
+731
| template <std::floating_point ST> | ||
| FK_HOST_DEVICE_FUSE ST exec(const ST x) { |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.