AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

nir/builder: Emit x != 0 for nir_i2b

Browse Source 
There are a lot of optimizations in opt_algebraic that match ('ine', a,
0), but there are almost none that match i2b.  Instead of adding a huge
pile of additional patterns (including variation that include both ine
and i2b), just emit a != 0 instead of i2b(a).

I think that the changes to the unit tests weaken them slightly, but
perhaps that's okay?

No shader-db changes on any Intel platform.  The GLSL paths use other
means to generate i2b operations, but the SPIR-V paths use nir_i2b.
Presumably since 4676b3d3dd (nir: Use nir_test_mask instead of
i2b(iand)), no fossil-db changes either.

v2: Use nir_ine_imm.  Suggested by Jesse.

Acked-by: Jesse Natalie <jenatali@microsoft.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
Tested-by: Daniel Schürmann <daniel@schuermann.dev>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15121>
This commit is contained in:
Ian Romanick
2022-02-15 15:29:42 -08:00
committed by Marge Bot
parent 7a5e9df39d
commit 9342c14eeb
2 changed files with 6 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/compiler/nir/nir_builder.h
View File

@@ -413,7 +413,7 @@ nir_f2b(nir_builder *b, nir_ssa_def *src)
static inline nir_ssa_def *
nir_i2b(nir_builder *b, nir_ssa_def *src)
{
return nir_type_convert(b, src, nir_type_int, nir_type_bool1);
return nir_ine(b, src, nir_imm_intN_t(b, 0, src->bit_size));
}
static inline nir_ssa_def *

8
src/compiler/nir/tests/load_store_vectorizer_tests.cpp
View File

@@ -1461,8 +1461,9 @@ TEST_F(nir_load_store_vectorize_test, shared_load_bool)
ASSERT_EQ(deref->deref_type, nir_deref_type_var);
ASSERT_EQ(deref->var, var);
ASSERT_TRUE(test_alu(loads[0x1]->src.ssa->parent_instr, nir_op_i2b1));
ASSERT_TRUE(test_alu(loads[0x2]->src.ssa->parent_instr, nir_op_i2b1));
/* The loaded value is converted to Boolean by (loaded != 0). */
ASSERT_TRUE(test_alu(loads[0x1]->src.ssa->parent_instr, nir_op_ine));
ASSERT_TRUE(test_alu(loads[0x2]->src.ssa->parent_instr, nir_op_ine));
ASSERT_TRUE(test_alu_def(loads[0x1]->src.ssa->parent_instr, 0, &load->dest.ssa, 0));
ASSERT_TRUE(test_alu_def(loads[0x2]->src.ssa->parent_instr, 0, &load->dest.ssa, 1));
}
@@ -1500,7 +1501,8 @@ TEST_F(nir_load_store_vectorize_test, shared_load_bool_mixed)
ASSERT_EQ(deref->deref_type, nir_deref_type_var);
ASSERT_EQ(deref->var, var);
ASSERT_TRUE(test_alu(loads[0x1]->src.ssa->parent_instr, nir_op_i2b1));
/* The loaded value is converted to Boolean by (loaded != 0). */
ASSERT_TRUE(test_alu(loads[0x1]->src.ssa->parent_instr, nir_op_ine));
ASSERT_TRUE(test_alu_def(loads[0x1]->src.ssa->parent_instr, 0, &load->dest.ssa, 0));
EXPECT_INSTR_SWIZZLES(movs[0x2], load, "y");