[llvm-branch-commits] [llvm] 3a042dc - [AMDGPU] Fix default value of glc for mubuf rtn atomics
Petar Avramovic via llvm-branch-commits
llvm-branch-commits at lists.llvm.org
Mon Dec 7 05:00:54 PST 2020
Author: Petar Avramovic
Date: 2020-12-07T14:00:08+01:00
New Revision: 3a042dcd2e1ab53244d504cd87a81b577a62a6ea
URL: https://github.com/llvm/llvm-project/commit/3a042dcd2e1ab53244d504cd87a81b577a62a6ea
DIFF: https://github.com/llvm/llvm-project/commit/3a042dcd2e1ab53244d504cd87a81b577a62a6ea.diff
LOG: [AMDGPU] Fix default value of glc for mubuf rtn atomics
Mubuf rtn atomics use GLC_1 thus default value for glc operand
should be -1, see https://reviews.llvm.org/D90730.
This allows us to report error when rtn atomic requires glc=1
but does not have glc operand in input.
Differential Revision: https://reviews.llvm.org/D92654
Added:
Modified:
llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
llvm/test/MC/AMDGPU/gfx1030_err.s
llvm/test/MC/AMDGPU/gfx1030_new.s
Removed:
################################################################################
diff --git a/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp b/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
index 45774935287b..56d97588df6e 100644
--- a/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
+++ b/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
@@ -6691,7 +6691,8 @@ void AMDGPUAsmParser::cvtMubufImpl(MCInst &Inst,
addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyOffset);
if (!IsAtomic || IsAtomicReturn) {
- addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyGLC);
+ addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTyGLC,
+ IsAtomicReturn ? -1 : 0);
}
addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySLC);
diff --git a/llvm/test/MC/AMDGPU/gfx1030_err.s b/llvm/test/MC/AMDGPU/gfx1030_err.s
index 5a57d3b3048a..dbee18bd2d91 100644
--- a/llvm/test/MC/AMDGPU/gfx1030_err.s
+++ b/llvm/test/MC/AMDGPU/gfx1030_err.s
@@ -140,3 +140,9 @@ ds_write_src2_b32 v1 offset:65535
ds_write_src2_b64 v1 offset:65535
// GFX10: :[[@LINE-1]]:{{[0-9]+}}: error: instruction not supported on this GPU
+
+buffer_atomic_csub v5, off, s[8:11], s3 offset:4095
+// GFX10: :[[@LINE-1]]:{{[0-9]+}}: error: instruction must use glc
+
+global_atomic_csub v2, v[0:1], v2, off offset:100 slc
+// GFX10: :[[@LINE-1]]:{{[0-9]+}}: error: instruction must use glc
diff --git a/llvm/test/MC/AMDGPU/gfx1030_new.s b/llvm/test/MC/AMDGPU/gfx1030_new.s
index 8dc977498cd5..94f4ff3a237a 100644
--- a/llvm/test/MC/AMDGPU/gfx1030_new.s
+++ b/llvm/test/MC/AMDGPU/gfx1030_new.s
@@ -27,7 +27,7 @@ global_atomic_csub v2, v0, v2, s[2:3] glc
global_atomic_csub v2, v0, v2, s[2:3] offset:100 glc slc
// GFX10: encoding: [0x64,0x80,0xd3,0xdc,0x00,0x02,0x02,0x02]
-buffer_atomic_csub v5, off, s[8:11], s3
+buffer_atomic_csub v5, off, s[8:11], s3 glc
// GFX10: encoding: [0x00,0x40,0xd0,0xe0,0x00,0x05,0x02,0x03]
buffer_atomic_csub v5, off, s[8:11], s3 offset:4095 glc
More information about the llvm-branch-commits
mailing list