aco: Fix the meaning of is_atomic.

Previously, is_atomic really meant "is not atomic", contrary to its name.
This commit fixes it to mean what one would think it means.

Fixes: 69bed1c9186c3e24ad54089218d58c5f7b83befe
Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Daniel Schürmann <daniel@schuermann.dev>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3618>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3618>

diff --git a/src/amd/compiler/aco_dead_code_analysis.cpp b/src/amd/compiler/aco_dead_code_analysis.cpp
index 443ba188c186cae399cb94d52f76fbe00031489f..87b3112dc971083b1588e6af0a6abe6057655fa7 100644 (file)
--- a/src/amd/compiler/aco_dead_code_analysis.cpp
+++ b/src/amd/compiler/aco_dead_code_analysis.cpp
@@ -84,7 +84,7 @@ bool is_dead(const std::vector<uint16_t>& uses, Instruction *instr)
    if (std::any_of(instr->definitions.begin(), instr->definitions.end(),
           [&uses] (const Definition& def) { return uses[def.tempId()];}))
       return false;
-   return instr_info.is_atomic[(int)instr->opcode];
+   return !instr_info.is_atomic[(int)instr->opcode];
 }
 
 std::vector<uint16_t> dead_code_analysis(Program *program) {

diff --git a/src/amd/compiler/aco_opcodes.py b/src/amd/compiler/aco_opcodes.py
index d3b7252414ff526644fda38350b941d211386779..a0bc601192c14765593c1a75a7a05d78c3527acf 100644 (file)
--- a/src/amd/compiler/aco_opcodes.py
+++ b/src/amd/compiler/aco_opcodes.py
@@ -186,7 +186,7 @@ class Opcode(object):
 # global dictionary of opcodes
 opcodes = {}
 
-def opcode(name, opcode_gfx7 = -1, opcode_gfx9 = -1, opcode_gfx10 = -1, format = Format.PSEUDO, input_mod = False, output_mod = False, is_atomic = True):
+def opcode(name, opcode_gfx7 = -1, opcode_gfx9 = -1, opcode_gfx10 = -1, format = Format.PSEUDO, input_mod = False, output_mod = False, is_atomic = False):
    assert name not in opcodes
    opcodes[name] = Opcode(name, opcode_gfx7, opcode_gfx9, opcode_gfx10, format, input_mod, output_mod, is_atomic)
 
@@ -586,7 +586,7 @@ SMEM = {
    (  -1,   -1,   -1, 0xac, 0xac, "s_atomic_dec_x2"),
 }
 for (gfx6, gfx7, gfx8, gfx9, gfx10, name) in SMEM:
-   opcode(name, gfx7, gfx9, gfx10, Format.SMEM, is_atomic = "atomic" not in name)
+   opcode(name, gfx7, gfx9, gfx10, Format.SMEM, is_atomic = "atomic" in name)
 
 
 # VOP2 instructions: 2 inputs, 1 output (+ optional vcc)
@@ -1269,7 +1269,7 @@ MUBUF = {
    (  -1,   -1,   -1,   -1, 0x72, "buffer_gl1_inv"),
 }
 for (gfx6, gfx7, gfx8, gfx9, gfx10, name) in MUBUF:
-    opcode(name, gfx7, gfx9, gfx10, Format.MUBUF, is_atomic = "atomic" not in name)
+    opcode(name, gfx7, gfx9, gfx10, Format.MUBUF, is_atomic = "atomic" in name)
 
 MTBUF = {
    (0x00, 0x00, 0x00, 0x00, 0x00, "tbuffer_load_format_x"),
@@ -1333,7 +1333,7 @@ IMAGE_ATOMIC = {
 # (gfx6, gfx7, gfx8, gfx9, gfx10, name) = (gfx6, gfx7, gfx89, gfx89, ???, name)
 # gfx7 and gfx10 opcodes are the same here
 for (gfx6, gfx7, gfx89, name) in IMAGE_ATOMIC:
-   opcode(name, gfx7, gfx89, gfx7, Format.MIMG, is_atomic = False)
+   opcode(name, gfx7, gfx89, gfx7, Format.MIMG, is_atomic = True)
 
 IMAGE_SAMPLE = {
    (0x20, "image_sample"),
@@ -1473,7 +1473,7 @@ FLAT = {
    (0x60,   -1, 0x60, "flat_atomic_fmax_x2"),
 }
 for (gfx7, gfx8, gfx10, name) in FLAT:
-    opcode(name, gfx7, gfx8, gfx10, Format.FLAT, is_atomic = "atomic" not in name)
+    opcode(name, gfx7, gfx8, gfx10, Format.FLAT, is_atomic = "atomic" in name)
 
 GLOBAL = {
    #GFX8_9, GFX10
@@ -1533,7 +1533,7 @@ GLOBAL = {
    (  -1, 0x60, "global_atomic_fmax_x2"),
 }
 for (gfx8, gfx10, name) in GLOBAL:
-    opcode(name, -1, gfx8, gfx10, Format.GLOBAL, is_atomic = "atomic" not in name)
+    opcode(name, -1, gfx8, gfx10, Format.GLOBAL, is_atomic = "atomic" in name)
 
 SCRATCH = {
    #GFX8_9, GFX10