ogl_beamforming

Commit: f3a4a25c5d11bc9317e4b7ca26aa2dd2eb8121a4
Parent: 3771478afe8241300735eb5597f814a15f3c51e7
Author: Randy Palamar
Date:   Sat,  3 Oct 2026 20:17:37 -0700

das: don't use explicit Catmull-Rom for Float16Complex

the last commit caused a performance regression for TPW/VLS which
are loading Float16 data in DAS. Some instructions which were
previously hidden by the load latency got pushed after the load
completion which hurt performance.

Diffstat:
M
shaders/das.glsl
 | 
2
+-

1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/shaders/das.glsl b/shaders/das.glsl
@@ -111,7 +111,7 @@ SAMPLE_TYPE cubic(const u64 rf_pointer, const f32 t)
 	#endif
 
 	SAMPLE_TYPE result;
-	if (C_SPLINE == 0.5f) {
+	if (C_SPLINE == 0.5f && InputDataKind != DataKind_Float16Complex) {
 		// NOTE(rnp): expanded Catmull-Rom spline
 		// (mathematically equivalent but easier on the compiler)
 		SAMPLE_TYPE P0 = samples[0];