diff options
author | Matt Arsenault <Matthew.Arsenault@amd.com> | 2015-11-03 22:39:50 +0000 |
---|---|---|
committer | Matt Arsenault <Matthew.Arsenault@amd.com> | 2015-11-03 22:39:50 +0000 |
commit | 192b282bf3dd285da991a6aeb49831d3fd93b85c (patch) | |
tree | 26796093b51d9e72792cd0a1f9e41b6a28645801 /llvm/lib/Target/AMDGPU | |
parent | d77f0d2526e28611b09121324540e8faaa5cab1a (diff) | |
download | bcm5719-llvm-192b282bf3dd285da991a6aeb49831d3fd93b85c.tar.gz bcm5719-llvm-192b282bf3dd285da991a6aeb49831d3fd93b85c.zip |
AMDGPU: Define correct number of SGPRs
There are actually 104 so 2 were missing.
More assembler tests with high register number tuples
will be included in later patches.
llvm-svn: 251999
Diffstat (limited to 'llvm/lib/Target/AMDGPU')
-rw-r--r-- | llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp | 4 | ||||
-rw-r--r-- | llvm/lib/Target/AMDGPU/SIRegisterInfo.td | 12 |
2 files changed, 10 insertions, 6 deletions
diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp index 30920a0bb93..1315b6a7b3a 100644 --- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp +++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp @@ -41,6 +41,10 @@ BitVector SIRegisterInfo::getReservedRegs(const MachineFunction &MF) const { reserveRegisterTuples(Reserved, AMDGPU::EXEC); reserveRegisterTuples(Reserved, AMDGPU::FLAT_SCR); + // Reserve the last 2 registers so we will always have at least 2 more that + // will physically contain VCC. + reserveRegisterTuples(Reserved, AMDGPU::SGPR102_SGPR103); + // Tonga and Iceland can only allocate a fixed number of SGPRs due // to a hw bug. if (MF.getSubtarget<AMDGPUSubtarget>().hasSGPRInitBug()) { diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.td b/llvm/lib/Target/AMDGPU/SIRegisterInfo.td index 299c6a69cbd..f7abe53d828 100644 --- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.td +++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.td @@ -56,7 +56,7 @@ def FLAT_SCR : RegisterWithSubRegs <"flat_scr", [FLAT_SCR_LO, FLAT_SCR_HI]>, } // SGPR registers -foreach Index = 0-101 in { +foreach Index = 0-103 in { def SGPR#Index : SIReg <"SGPR"#Index, Index>; } @@ -75,23 +75,23 @@ foreach Index = 0-255 in { // SGPR 32-bit registers def SGPR_32 : RegisterClass<"AMDGPU", [i32, f32], 32, - (add (sequence "SGPR%u", 0, 101))>; + (add (sequence "SGPR%u", 0, 103))>; // SGPR 64-bit registers def SGPR_64Regs : RegisterTuples<[sub0, sub1], - [(add (decimate (trunc SGPR_32, 101), 2)), + [(add (decimate SGPR_32, 2)), (add (decimate (shl SGPR_32, 1), 2))]>; // SGPR 128-bit registers def SGPR_128 : RegisterTuples<[sub0, sub1, sub2, sub3], - [(add (decimate (trunc SGPR_32, 99), 4)), + [(add (decimate SGPR_32, 4)), (add (decimate (shl SGPR_32, 1), 4)), (add (decimate (shl SGPR_32, 2), 4)), (add (decimate (shl SGPR_32, 3), 4))]>; // SGPR 256-bit registers def SGPR_256 : RegisterTuples<[sub0, sub1, sub2, sub3, sub4, sub5, sub6, sub7], - [(add (decimate (trunc SGPR_32, 95), 4)), + [(add (decimate SGPR_32, 4)), (add (decimate (shl SGPR_32, 1), 4)), (add (decimate (shl SGPR_32, 2), 4)), (add (decimate (shl SGPR_32, 3), 4)), @@ -103,7 +103,7 @@ def SGPR_256 : RegisterTuples<[sub0, sub1, sub2, sub3, sub4, sub5, sub6, sub7], // SGPR 512-bit registers def SGPR_512 : RegisterTuples<[sub0, sub1, sub2, sub3, sub4, sub5, sub6, sub7, sub8, sub9, sub10, sub11, sub12, sub13, sub14, sub15], - [(add (decimate (trunc SGPR_32, 87), 4)), + [(add (decimate SGPR_32, 4)), (add (decimate (shl SGPR_32, 1), 4)), (add (decimate (shl SGPR_32, 2), 4)), (add (decimate (shl SGPR_32, 3), 4)), |