Revision 11fbc531fd50ffd3130885f267e19a5fa6d8e491 authored by Andrew Adams on 29 April 2024, 17:56:07 UTC, committed by Andrew Adams on 29 April 2024, 17:56:07 UTC
The old pattern was min(u16(x8 + y8), any u16 constant at all), which is wrong in several ways. I replaced that with all the variants of saturating_add I could think of. It would be nice to be able to call find_intrinsics here to normalize, but that would require making find_intrinsics trigger on scalars, and I wanted to make the minimal change here to fix things.
1 parent d55d82b
| File | Mode | Size |
|---|---|---|
| CMakeLists.txt | -rw-r--r-- | 2.6 KB |
| GenGen.cpp | -rw-r--r-- | 120 bytes |
| RunGen.h | -rw-r--r-- | 57.2 KB |
| RunGenMain.cpp | -rw-r--r-- | 22.9 KB |
| binary2cpp.cpp | -rw-r--r-- | 1.7 KB |
| build_halide_h.cpp | -rw-r--r-- | 2.6 KB |
| find_inverse.cpp | -rw-r--r-- | 11.8 KB |
| halide_benchmark.h | -rw-r--r-- | 9.4 KB |
| halide_config.make.tpl | -rw-r--r-- | 157 bytes |
| halide_image.h | -rw-r--r-- | 743 bytes |
| halide_image_info.h | -rw-r--r-- | 10.4 KB |
| halide_image_io.h | -rw-r--r-- | 89.0 KB |
| halide_malloc_trace.h | -rw-r--r-- | 3.3 KB |
| halide_thread_pool.h | -rw-r--r-- | 4.8 KB |
| halide_trace_config.h | -rw-r--r-- | 19.1 KB |
| launch_wasm_test.js | -rw-r--r-- | 936 bytes |
| makelib.sh | -rwxr-xr-x | 2.2 KB |
| regexp_replace.cpp | -rw-r--r-- | 855 bytes |
Computing file changes ...
