replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

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
This commit is contained in:
Matt Arsenault 2015-11-03 22:39:50 +00:00
parent d77f0d2526
commit 192b282bf3
3 changed files with 13 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
View File

@ -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()) {

12
llvm/lib/Target/AMDGPU/SIRegisterInfo.td
View File

@ -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)),

3
llvm/test/MC/AMDGPU/sop2.s
View File

@ -129,3 +129,6 @@ s_cbranch_g_fork s[4:5], s[6:7]
// CHECK: s_absdiff_i32 s2, s4, s6 ; encoding: [0x04,0x06,0x02,0x96]
s_absdiff_i32 s2, s4, s6
// CHECK: s_add_u32 s101, s102, s103 ; encoding: [0x66,0x67,0x65,0x80]
s_add_u32 s101, s102, s103