class R600Reg_128<string n, list<Register> subregs, bits<16> encoding> :
RegisterWithSubRegs<n, subregs> {
let Namespace = "AMDGPU";
- let SubRegIndices = [sel_x, sel_y, sel_z, sel_w];
+ let SubRegIndices = [sub0, sub1, sub2, sub3];
let HWEncoding = encoding;
}
foreach Chan = [ "X", "Y", "Z", "W" ] in {
// 32-bit Temporary Registers
def T#Index#_#Chan : R600RegWithChan <"T"#Index#"."#Chan, Index, Chan>;
+
+ // Indirect addressing offset registers
+ def Addr#Index#_#Chan : R600RegWithChan <"T("#Index#" + AR.x)."#Chan,
+ Index, Chan>;
+ def TRegMem#Index#_#Chan : R600RegWithChan <"T"#Index#"."#Chan, Index,
+ Chan>;
}
// 128-bit Temporary Registers
def T#Index#_XYZW : R600Reg_128 <"T"#Index#".XYZW",
Index>;
}
+// KCACHE_BANK0
+foreach Index = 159-128 in {
+ foreach Chan = [ "X", "Y", "Z", "W" ] in {
+ // 32-bit Temporary Registers
+ def KC0_#Index#_#Chan : R600RegWithChan <"KC0["#Index#"-128]."#Chan, Index, Chan>;
+ }
+ // 128-bit Temporary Registers
+ def KC0_#Index#_XYZW : R600Reg_128 <"KC0["#Index#"-128].XYZW",
+ [!cast<Register>("KC0_"#Index#"_X"),
+ !cast<Register>("KC0_"#Index#"_Y"),
+ !cast<Register>("KC0_"#Index#"_Z"),
+ !cast<Register>("KC0_"#Index#"_W")],
+ Index>;
+}
+
+// KCACHE_BANK1
+foreach Index = 191-160 in {
+ foreach Chan = [ "X", "Y", "Z", "W" ] in {
+ // 32-bit Temporary Registers
+ def KC1_#Index#_#Chan : R600RegWithChan <"KC1["#Index#"-160]."#Chan, Index, Chan>;
+ }
+ // 128-bit Temporary Registers
+ def KC1_#Index#_XYZW : R600Reg_128 <"KC1["#Index#"-160].XYZW",
+ [!cast<Register>("KC1_"#Index#"_X"),
+ !cast<Register>("KC1_"#Index#"_Y"),
+ !cast<Register>("KC1_"#Index#"_Z"),
+ !cast<Register>("KC1_"#Index#"_W")],
+ Index>;
+}
+
+
// Array Base Register holding input in FS
-foreach Index = 448-464 in {
+foreach Index = 448-480 in {
def ArrayBase#Index : R600Reg<"ARRAY_BASE", Index>;
}
def ONE_INT : R600Reg<"1", 250>;
def HALF : R600Reg<"0.5", 252>;
def NEG_HALF : R600Reg<"-0.5", 252>;
-def ALU_LITERAL_X : R600Reg<"literal.x", 253>;
-def PV_X : R600Reg<"pv.x", 254>;
+def ALU_LITERAL_X : R600RegWithChan<"literal.x", 253, "X">;
+def ALU_LITERAL_Y : R600RegWithChan<"literal.y", 253, "Y">;
+def ALU_LITERAL_Z : R600RegWithChan<"literal.z", 253, "Z">;
+def ALU_LITERAL_W : R600RegWithChan<"literal.w", 253, "W">;
+def PV_X : R600RegWithChan<"PV.x", 254, "X">;
+def PV_Y : R600RegWithChan<"PV.y", 254, "Y">;
+def PV_Z : R600RegWithChan<"PV.z", 254, "Z">;
+def PV_W : R600RegWithChan<"PV.w", 254, "W">;
def PREDICATE_BIT : R600Reg<"PredicateBit", 0>;
def PRED_SEL_OFF: R600Reg<"Pred_sel_off", 0>;
def PRED_SEL_ZERO : R600Reg<"Pred_sel_zero", 2>;
def PRED_SEL_ONE : R600Reg<"Pred_sel_one", 3>;
+def AR_X : R600Reg<"AR.x", 0>;
def R600_ArrayBase : RegisterClass <"AMDGPU", [f32, i32], 32,
- (add (sequence "ArrayBase%u", 448, 464))>;
+ (add (sequence "ArrayBase%u", 448, 480))>;
// special registers for ALU src operands
// const buffer reference, SRCx_SEL contains index
def ALU_CONST : R600Reg<"CBuf", 0>;
// interpolation param reference, SRCx_SEL contains index
def ALU_PARAM : R600Reg<"Param", 0>;
+let isAllocatable = 0 in {
+
+// XXX: Only use the X channel, until we support wider stack widths
+def R600_Addr : RegisterClass <"AMDGPU", [i32], 127, (add (sequence "Addr%u_X", 0, 127))>;
+
+} // End isAllocatable = 0
+
+def R600_KC0_X : RegisterClass <"AMDGPU", [f32, i32], 32,
+ (add (sequence "KC0_%u_X", 128, 159))>;
+
+def R600_KC0_Y : RegisterClass <"AMDGPU", [f32, i32], 32,
+ (add (sequence "KC0_%u_Y", 128, 159))>;
+
+def R600_KC0_Z : RegisterClass <"AMDGPU", [f32, i32], 32,
+ (add (sequence "KC0_%u_Z", 128, 159))>;
+
+def R600_KC0_W : RegisterClass <"AMDGPU", [f32, i32], 32,
+ (add (sequence "KC0_%u_W", 128, 159))>;
+
+def R600_KC0 : RegisterClass <"AMDGPU", [f32, i32], 32,
+ (interleave R600_KC0_X, R600_KC0_Y,
+ R600_KC0_Z, R600_KC0_W)>;
+
+def R600_KC1_X : RegisterClass <"AMDGPU", [f32, i32], 32,
+ (add (sequence "KC1_%u_X", 160, 191))>;
+
+def R600_KC1_Y : RegisterClass <"AMDGPU", [f32, i32], 32,
+ (add (sequence "KC1_%u_Y", 160, 191))>;
+
+def R600_KC1_Z : RegisterClass <"AMDGPU", [f32, i32], 32,
+ (add (sequence "KC1_%u_Z", 160, 191))>;
+
+def R600_KC1_W : RegisterClass <"AMDGPU", [f32, i32], 32,
+ (add (sequence "KC1_%u_W", 160, 191))>;
+
+def R600_KC1 : RegisterClass <"AMDGPU", [f32, i32], 32,
+ (interleave R600_KC1_X, R600_KC1_Y,
+ R600_KC1_Z, R600_KC1_W)>;
+
def R600_TReg32_X : RegisterClass <"AMDGPU", [f32, i32], 32,
- (add (sequence "T%u_X", 0, 127))>;
+ (add (sequence "T%u_X", 0, 127), AR_X)>;
def R600_TReg32_Y : RegisterClass <"AMDGPU", [f32, i32], 32,
(add (sequence "T%u_Y", 0, 127))>;
def R600_Reg32 : RegisterClass <"AMDGPU", [f32, i32], 32, (add
R600_TReg32,
R600_ArrayBase,
+ R600_Addr,
ZERO, HALF, ONE, ONE_INT, PV_X, ALU_LITERAL_X, NEG_ONE, NEG_HALF,
ALU_CONST, ALU_PARAM
)>;
(add (sequence "T%u_XYZW", 0, 127))> {
let CopyCost = -1;
}
+
+//===----------------------------------------------------------------------===//
+// Register classes for indirect addressing
+//===----------------------------------------------------------------------===//
+
+// Super register for all the Indirect Registers. This register class is used
+// by the REG_SEQUENCE instruction to specify the registers to use for direct
+// reads / writes which may be written / read by an indirect address.
+class IndirectSuper<string n, list<Register> subregs> :
+ RegisterWithSubRegs<n, subregs> {
+ let Namespace = "AMDGPU";
+ let SubRegIndices =
+ [sub0, sub1, sub2, sub3, sub4, sub5, sub6, sub7,
+ sub8, sub9, sub10, sub11, sub12, sub13, sub14, sub15];
+}
+
+def IndirectSuperReg : IndirectSuper<"Indirect",
+ [TRegMem0_X, TRegMem1_X, TRegMem2_X, TRegMem3_X, TRegMem4_X, TRegMem5_X,
+ TRegMem6_X, TRegMem7_X, TRegMem8_X, TRegMem9_X, TRegMem10_X, TRegMem11_X,
+ TRegMem12_X, TRegMem13_X, TRegMem14_X, TRegMem15_X]
+>;
+
+def IndirectReg : RegisterClass<"AMDGPU", [f32, i32], 32, (add IndirectSuperReg)>;
+
+// This register class defines the registers that are the storage units for
+// the "Indirect Addressing" pseudo memory space.
+// XXX: Only use the X channel, until we support wider stack widths
+def TRegMem : RegisterClass<"AMDGPU", [f32, i32], 32,
+ (add (sequence "TRegMem%u_X", 0, 16))
+>;
projects / oota-llvm.git / blobdiff