[v5,03/16] ipa: libipa: fixedpoint: Fix unsigned usage
diff mbox series

Message ID 20260114173918.1744023-4-kieran.bingham@ideasonboard.com
State New
Headers show
Series
  • libipa: Introduce a Quantized type
Related show

Commit Message

Kieran Bingham Jan. 14, 2026, 5:39 p.m. UTC
The fixedToFloatingPoint does not support unsigned Q types, and
incorrectly sign-extends all values which have the top most bit set in
the quantized values.

Fix this by ensuring that only signed types perform sign extension, and
simplify the calculation for unsigned types.

Convert the storage of the test cases to signed types to correctly
represent their intended purpose, to prevent test failures.

Reviewed-by: Isaac Scott <isaac.scott@ideasonboard.com>
Signed-off-by: Kieran Bingham <kieran.bingham@ideasonboard.com>
---
 src/ipa/libipa/fixedpoint.h    | 3 +++
 test/ipa/libipa/fixedpoint.cpp | 6 +++---
 2 files changed, 6 insertions(+), 3 deletions(-)

Patch
diff mbox series

diff --git a/src/ipa/libipa/fixedpoint.h b/src/ipa/libipa/fixedpoint.h
index 709cf50f0fcd..b4a7fa5e0ecd 100644
--- a/src/ipa/libipa/fixedpoint.h
+++ b/src/ipa/libipa/fixedpoint.h
@@ -49,6 +49,9 @@  constexpr R fixedToFloatingPoint(T number)
 	static_assert(sizeof(int) >= sizeof(T));
 	static_assert(I + F <= sizeof(T) * 8);
 
+	if constexpr (std::is_unsigned_v<T>)
+		return static_cast<R>(number) / static_cast<R>(1 << F);
+
 	/*
 	 * Recreate the upper bits in case of a negative number by shifting the sign
 	 * bit from the fixed point to the first bit of the unsigned and then right shifting
diff --git a/test/ipa/libipa/fixedpoint.cpp b/test/ipa/libipa/fixedpoint.cpp
index 99eb662ddf4e..4b017e86a74f 100644
--- a/test/ipa/libipa/fixedpoint.cpp
+++ b/test/ipa/libipa/fixedpoint.cpp
@@ -68,7 +68,7 @@  protected:
 		 * The second 7.992 test is to test that unused bits don't
 		 * affect the result.
 		 */
-		std::map<double, uint16_t> testCases = {
+		std::map<double, int16_t> testCases = {
 			{ 7.992, 0x3ff },
 			{   0.2, 0x01a },
 			{  -0.2, 0x7e6 },
@@ -81,14 +81,14 @@  protected:
 
 		int ret;
 		for (const auto &testCase : testCases) {
-			ret = testSingleFixedPoint<4, 7, uint16_t>(testCase.first,
+			ret = testSingleFixedPoint<4, 7, int16_t>(testCase.first,
 								   testCase.second);
 			if (ret != TestPass)
 				return ret;
 		}
 
 		/* Special case with a superfluous one in the unused bits */
-		ret = testFixedToFloat<4, 7, uint16_t, double>(0xbff, 7.992);
+		ret = testFixedToFloat<4, 7, int16_t, double>(0xbff, 7.992);
 		if (ret != TestPass)
 			return ret;