diff options
Diffstat (limited to 'src/compiler/nir')
| -rw-r--r-- | src/compiler/nir/nir_opt_load_store_vectorize.c | 2 | ||||
| -rw-r--r-- | src/compiler/nir/tests/load_store_vectorizer_tests.cpp | 8 |
2 files changed, 6 insertions, 4 deletions
diff --git a/src/compiler/nir/nir_opt_load_store_vectorize.c b/src/compiler/nir/nir_opt_load_store_vectorize.c index 6587251c7e7..414a97fec45 100644 --- a/src/compiler/nir/nir_opt_load_store_vectorize.c +++ b/src/compiler/nir/nir_opt_load_store_vectorize.c @@ -38,6 +38,8 @@ * - It won't turn four consecutive vec3 loads into 3 vec4 loads. * - It doesn't do global vectorization. * Handling these cases probably wouldn't provide much benefit though. + * + * This probably doesn't handle big-endian GPUs correctly. */ #include "nir.h" diff --git a/src/compiler/nir/tests/load_store_vectorizer_tests.cpp b/src/compiler/nir/tests/load_store_vectorizer_tests.cpp index cd623704fc3..0d749debd4c 100644 --- a/src/compiler/nir/tests/load_store_vectorizer_tests.cpp +++ b/src/compiler/nir/tests/load_store_vectorizer_tests.cpp @@ -961,10 +961,10 @@ TEST_F(nir_load_store_vectorize_test, ssbo_store_adjacent_8_8_16) ASSERT_EQ(val->bit_size, 8); ASSERT_EQ(val->num_components, 4); nir_const_value *cv = nir_instr_as_load_const(val->parent_instr)->value; - ASSERT_EQ(nir_const_value_as_uint(cv[0], 32), 0x10); - ASSERT_EQ(nir_const_value_as_uint(cv[1], 32), 0x20); - ASSERT_EQ(nir_const_value_as_uint(cv[2], 32), 0x30); - ASSERT_EQ(nir_const_value_as_uint(cv[3], 32), 0x0); + ASSERT_EQ(nir_const_value_as_uint(cv[0], 8), 0x10); + ASSERT_EQ(nir_const_value_as_uint(cv[1], 8), 0x20); + ASSERT_EQ(nir_const_value_as_uint(cv[2], 8), 0x30); + ASSERT_EQ(nir_const_value_as_uint(cv[3], 8), 0x0); } TEST_F(nir_load_store_vectorize_test, ssbo_store_adjacent_32_32_64) |