shifathosn
/
KytyPS5
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

tests: run the saveexec mask cases in a populated subgroup (#876) 

tests: run saveexec masks in a populated subgroup

Co-authored-by: Matthew Hurley <171001439+theantipopau@users.noreply.github.com>
This commit is contained in:
Matthew Hurley 2026-09-29 07:42:11 +10:00 committed by GitHub
parent 49db3ea54c
commit 21d7e8e1f0
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 28 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

43
tests/ShaderRecompilerComputeTests.cpp
View File

@ -17910,15 +17910,22 @@ TestCase ScalarSaveExecOps() {
AppendStoreSgpr(&code, 253, 12);
AppendEnd(&code);
return {"ScalarSaveExecOps",
code,
{},
{1, 0, 1, 0, 0xffffffffu, 0xffffffffu, 0xffffffffu,
3, 0xffffffffu, 3, 2, 0xffffffffu, 1},
{O::S_MOV_B32, O::S_AND_SAVEEXEC_B64, O::S_ORN2_SAVEEXEC_B64,
O::S_ANDN1_SAVEEXEC_B64, O::S_AND_SAVEEXEC_B32,
O::S_ANDN1_SAVEEXEC_B32, O::S_MOV_B64, O::V_MOV_B32,
O::BUFFER_STORE_DWORD, O::S_ENDPGM}};
TestCase test;
test.name = "ScalarSaveExecOps";
test.code = code;
// A full wave enters with EXEC set for all 64 lanes, so the two saved EXEC
// pairs are 0xffffffff/0xffffffff rather than the 1/0 a single-invocation
// dispatch would report.
test.expected = {0xffffffffu, 0xffffffffu, 0xffffffffu, 0xffffffffu,
0xffffffffu, 0xffffffffu, 0xffffffffu, 3, 0xffffffffu,
3, 2, 0xffffffffu, 1};
test.opcodes = {O::S_MOV_B32, O::S_AND_SAVEEXEC_B64, O::S_ORN2_SAVEEXEC_B64,
O::S_ANDN1_SAVEEXEC_B64, O::S_AND_SAVEEXEC_B32,
O::S_ANDN1_SAVEEXEC_B32, O::S_MOV_B64, O::V_MOV_B32,
O::BUFFER_STORE_DWORD, O::S_ENDPGM};
test.compute_info.threads_num[0] = 64;
test.has_compute_info = true;
return test;
}
TestCase ScalarOrn2SaveexecUsesSourceOrNotExec() {
@ -17938,12 +17945,18 @@ TestCase ScalarOrn2SaveexecUsesSourceOrNotExec() {
AppendStoreSgpr(&code, 253, 4);
AppendEnd(&code);
return {"ScalarOrn2SaveexecUsesSourceOrNotExec",
code,
{},
{0x0000000cu, 0x80000000u, 0xfffffff3u, 0x7fffffffu, 1},
{O::S_MOV_B32, O::S_ORN2_SAVEEXEC_B64, O::S_MOV_B64, O::V_MOV_B32,
O::BUFFER_STORE_DWORD, O::S_ENDPGM}};
TestCase test;
test.name = "ScalarOrn2SaveexecUsesSourceOrNotExec";
test.code = code;
test.expected = {0x0000000cu, 0x80000000u, 0xfffffff3u, 0x7fffffffu, 1};
test.opcodes = {O::S_MOV_B32, O::S_ORN2_SAVEEXEC_B64, O::S_MOV_B64, O::V_MOV_B32,
O::BUFFER_STORE_DWORD, O::S_ENDPGM};
// The case reads a whole 64-bit EXEC mask, so it needs a populated subgroup.
// A 1x1x1 dispatch leaves OpGroupNonUniformBallot with a single invocation
// and every lane above zero reads back as zero.
test.compute_info.threads_num[0] = 64;
test.has_compute_info = true;
return test;
}
TestCase ScalarOrn2SaveexecB32(u32 wave_size, u32 threads) {