Index: lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp =================================================================== --- lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp +++ lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp @@ -1042,6 +1042,8 @@ bool validateIntClampSupported(const MCInst &Inst); bool validateMIMGAtomicDMask(const MCInst &Inst); bool validateMIMGDataSize(const MCInst &Inst); + bool validateMIMGR128(const MCInst &Inst); + bool validateMIMGD16(const MCInst &Inst); bool usesConstantBus(const MCInst &Inst, unsigned OpIdx); bool isInlineConstant(const MCInst &Inst, unsigned OpIdx) const; unsigned findImplicitSGPRReadInVOP(const MCInst &Inst) const; @@ -2326,6 +2328,35 @@ return DMask == 0x1 || DMask == 0x3 || DMask == 0xf; } +bool AMDGPUAsmParser::validateMIMGR128(const MCInst &Inst) { + + const unsigned Opc = Inst.getOpcode(); + const MCInstrDesc &Desc = MII.get(Opc); + + if ((Desc.TSFlags & SIInstrFlags::MIMG) == 0) + return true; + + int Idx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::r128); + assert(Idx != -1); + + bool R128 = (Inst.getOperand(Idx).getImm() != 0); + + return !R128 || isCI() || isSI() || isVI(); +} + +bool AMDGPUAsmParser::validateMIMGD16(const MCInst &Inst) { + + const unsigned Opc = Inst.getOpcode(); + const MCInstrDesc &Desc = MII.get(Opc); + + if ((Desc.TSFlags & SIInstrFlags::MIMG) == 0) + return true; + if ((Desc.TSFlags & SIInstrFlags::D16) == 0) + return true; + + return !isCI() && !isSI(); +} + bool AMDGPUAsmParser::validateInstruction(const MCInst &Inst, const SMLoc &IDLoc) { if (!validateConstantBusLimitations(Inst)) { @@ -2353,6 +2384,17 @@ "invalid atomic image dmask"); return false; } + if (!validateMIMGR128(Inst)) { + Error(IDLoc, + "r128 modifier is not supported on this GPU"); + return false; + } + // For MUBUF/MTBUF d16 is a part of opcode, so there is nothing to validate. + if (!validateMIMGD16(Inst)) { + Error(IDLoc, + "d16 modifier is not supported on this GPU"); + return false; + } return true; } Index: test/MC/AMDGPU/mimg.s =================================================================== --- test/MC/AMDGPU/mimg.s +++ test/MC/AMDGPU/mimg.s @@ -1,115 +1,138 @@ -// RUN: llvm-mc -arch=amdgcn -show-encoding %s | FileCheck %s --check-prefix=SICI -// RUN: llvm-mc -arch=amdgcn -mcpu=tahiti -show-encoding %s | FileCheck %s --check-prefix=SICI -// RUN: llvm-mc -arch=amdgcn -mcpu=fiji -show-encoding %s | FileCheck %s --check-prefix=VI +// RUN: not llvm-mc -arch=amdgcn -show-encoding %s | FileCheck %s --check-prefix=GCN --check-prefix=SICI --check-prefix=SICIVI +// RUN: not llvm-mc -arch=amdgcn -mcpu=tahiti -show-encoding %s | FileCheck %s --check-prefix=GCN --check-prefix=SICI --check-prefix=SICIVI +// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx900 -show-encoding %s | FileCheck %s --check-prefix=GCN --check-prefix=GFX9 --check-prefix=GFX89 + +// RUN: llvm-mc -arch=amdgcn -mcpu=fiji -show-encoding %s | FileCheck %s --check-prefix=GCN --check-prefix=SICIVI --check-prefix=VI --check-prefix=GFX89 + +// RUN: not llvm-mc -arch=amdgcn -show-encoding %s 2>&1 | FileCheck %s --check-prefix=NOSICI +// RUN: not llvm-mc -arch=amdgcn -mcpu=tahiti -show-encoding %s 2>&1 | FileCheck %s --check-prefix=NOSICI +// RUN: not llvm-mc -arch=amdgcn -mcpu=gfx900 -show-encoding %s 2>&1 | FileCheck %s --check-prefix=NOGFX9 //===----------------------------------------------------------------------===// // Image Load/Store //===----------------------------------------------------------------------===// image_load v[4:6], v[237:240], s[28:35] dmask:0x7 unorm -// SICI: image_load v[4:6], v[237:240], s[28:35] dmask:0x7 unorm ; encoding: [0x00,0x17,0x00,0xf0,0xed,0x04,0x07,0x00] -// VI: image_load v[4:6], v[237:240], s[28:35] dmask:0x7 unorm ; encoding: [0x00,0x17,0x00,0xf0,0xed,0x04,0x07,0x00] +// GCN: image_load v[4:6], v[237:240], s[28:35] dmask:0x7 unorm ; encoding: [0x00,0x17,0x00,0xf0,0xed,0x04,0x07,0x00] image_load v4, v[237:240], s[28:35] -// SICI: image_load v4, v[237:240], s[28:35] ; encoding: [0x00,0x00,0x00,0xf0,0xed,0x04,0x07,0x00] -// VI: image_load v4, v[237:240], s[28:35] ; encoding: [0x00,0x00,0x00,0xf0,0xed,0x04,0x07,0x00] +// GCN: image_load v4, v[237:240], s[28:35] ; encoding: [0x00,0x00,0x00,0xf0,0xed,0x04,0x07,0x00] image_load v[4:7], v[237:240], s[28:35] dmask:0x7 tfe -// SICI: image_load v[4:7], v[237:240], s[28:35] dmask:0x7 tfe ; encoding: [0x00,0x07,0x01,0xf0,0xed,0x04,0x07,0x00] -// VI: image_load v[4:7], v[237:240], s[28:35] dmask:0x7 tfe ; encoding: [0x00,0x07,0x01,0xf0,0xed,0x04,0x07,0x00] +// GCN: image_load v[4:7], v[237:240], s[28:35] dmask:0x7 tfe ; encoding: [0x00,0x07,0x01,0xf0,0xed,0x04,0x07,0x00] // Verify support of all possible modifiers. // FIXME: This test is incorrect because r128 assumes a 128-bit SRSRC. -// FIXME: Check that d16 is not supported before VI image_load v[5:6], v[1:4], s[8:15] dmask:0x1 unorm glc slc r128 tfe lwe da d16 -// VI: image_load v[5:6], v[1:4], s[8:15] dmask:0x1 unorm glc slc r128 tfe lwe da d16 ; encoding: [0x00,0xf1,0x03,0xf2,0x01,0x05,0x02,0x80] +// NOSICI: error: d16 modifier is not supported on this GPU +// VI: image_load v[5:6], v[1:4], s[8:15] dmask:0x1 unorm glc slc r128 tfe lwe da d16 ; encoding: [0x00,0xf1,0x03,0xf2,0x01,0x05,0x02,0x80] +// NOGFX9: error: r128 modifier is not supported on this GPU -// FIXME: Check that d16 is not supported before VI image_load v5, v[1:4], s[8:15] d16 -// VI: image_load v5, v[1:4], s[8:15] d16 ; encoding: [0x00,0x00,0x00,0xf0,0x01,0x05,0x02,0x80] +// NOSICI: error: d16 modifier is not supported on this GPU +// GFX89: image_load v5, v[1:4], s[8:15] d16 ; encoding: [0x00,0x00,0x00,0xf0,0x01,0x05,0x02,0x80] + +image_load v5, v[1:4], s[8:15] r128 +// SICIVI: image_load v5, v[1:4], s[8:15] r128 ; encoding: [0x00,0x80,0x00,0xf0,0x01,0x05,0x02,0x00] +// NOGFX9: error: r128 modifier is not supported on this GPU image_store v[193:195], v[237:240], s[28:35] dmask:0x7 unorm -// SICI: image_store v[193:195], v[237:240], s[28:35] dmask:0x7 unorm ; encoding: [0x00,0x17,0x20,0xf0,0xed,0xc1,0x07,0x00] -// VI: image_store v[193:195], v[237:240], s[28:35] dmask:0x7 unorm ; encoding: [0x00,0x17,0x20,0xf0,0xed,0xc1,0x07,0x00] +// GCN: image_store v[193:195], v[237:240], s[28:35] dmask:0x7 unorm ; encoding: [0x00,0x17,0x20,0xf0,0xed,0xc1,0x07,0x00] image_store v193, v[237:240], s[28:35] -// SICI: image_store v193, v[237:240], s[28:35] ; encoding: [0x00,0x00,0x20,0xf0,0xed,0xc1,0x07,0x00] -// VI: image_store v193, v[237:240], s[28:35] ; encoding: [0x00,0x00,0x20,0xf0,0xed,0xc1,0x07,0x00] +// GCN: image_store v193, v[237:240], s[28:35] ; encoding: [0x00,0x00,0x20,0xf0,0xed,0xc1,0x07,0x00] image_store v[193:194], v[237:240], s[28:35] tfe -// SICI: image_store v[193:194], v[237:240], s[28:35] tfe ; encoding: [0x00,0x00,0x21,0xf0,0xed,0xc1,0x07,0x00] -// VI: image_store v[193:194], v[237:240], s[28:35] tfe ; encoding: [0x00,0x00,0x21,0xf0,0xed,0xc1,0x07,0x00] +// GCN: image_store v[193:194], v[237:240], s[28:35] tfe ; encoding: [0x00,0x00,0x21,0xf0,0xed,0xc1,0x07,0x00] // Verify support of all possible modifiers. // FIXME: This test is incorrect because r128 assumes a 128-bit SRSRC. -// FIXME: Check that d16 is not supported before VI image_store v5, v[1:4], s[8:15] dmask:0x1 unorm glc slc r128 lwe da d16 -// VI: image_store v5, v[1:4], s[8:15] dmask:0x1 unorm glc slc r128 lwe da d16 ; encoding: [0x00,0xf1,0x22,0xf2,0x01,0x05,0x02,0x80] +// NOSICI: error: d16 modifier is not supported on this GPU +// VI: image_store v5, v[1:4], s[8:15] dmask:0x1 unorm glc slc r128 lwe da d16 ; encoding: [0x00,0xf1,0x22,0xf2,0x01,0x05,0x02,0x80] +// NOGFX9: error: r128 modifier is not supported on this GPU -// FIXME: Check that d16 is not supported before VI image_store v5, v[1:4], s[8:15] d16 -// VI: image_store v5, v[1:4], s[8:15] d16 ; encoding: [0x00,0x00,0x20,0xf0,0x01,0x05,0x02,0x80] +// NOSICI: error: d16 modifier is not supported on this GPU +// GFX89: image_store v5, v[1:4], s[8:15] d16 ; encoding: [0x00,0x00,0x20,0xf0,0x01,0x05,0x02,0x80] + +// FIXME: This test is incorrect because r128 assumes a 128-bit SRSRC. +image_store v5, v[1:4], s[8:15] r128 +// SICIVI: image_store v5, v[1:4], s[8:15] r128 ; encoding: [0x00,0x80,0x20,0xf0,0x01,0x05,0x02,0x00] +// NOGFX9: error: r128 modifier is not supported on this GPU //===----------------------------------------------------------------------===// // Image Sample //===----------------------------------------------------------------------===// image_sample v[193:195], v[237:240], s[28:35], s[4:7] dmask:0x7 unorm -// SICI: image_sample v[193:195], v[237:240], s[28:35], s[4:7] dmask:0x7 unorm ; encoding: [0x00,0x17,0x80,0xf0,0xed,0xc1,0x27,0x00] -// VI: image_sample v[193:195], v[237:240], s[28:35], s[4:7] dmask:0x7 unorm ; encoding: [0x00,0x17,0x80,0xf0,0xed,0xc1,0x27,0x00] +// GCN: image_sample v[193:195], v[237:240], s[28:35], s[4:7] dmask:0x7 unorm ; encoding: [0x00,0x17,0x80,0xf0,0xed,0xc1,0x27,0x00] image_sample v193, v[237:240], s[28:35], s[4:7] -// SICI: image_sample v193, v[237:240], s[28:35], s[4:7] ; encoding: [0x00,0x00,0x80,0xf0,0xed,0xc1,0x27,0x00] -// VI: image_sample v193, v[237:240], s[28:35], s[4:7] ; encoding: [0x00,0x00,0x80,0xf0,0xed,0xc1,0x27,0x00] +// GCN: image_sample v193, v[237:240], s[28:35], s[4:7] ; encoding: [0x00,0x00,0x80,0xf0,0xed,0xc1,0x27,0x00] image_sample v[193:194], v[237:240], s[28:35], s[4:7] tfe -// SICI: image_sample v[193:194], v[237:240], s[28:35], s[4:7] tfe ; encoding: [0x00,0x00,0x81,0xf0,0xed,0xc1,0x27,0x00] -// VI: image_sample v[193:194], v[237:240], s[28:35], s[4:7] tfe ; encoding: [0x00,0x00,0x81,0xf0,0xed,0xc1,0x27,0x00] +// GCN: image_sample v[193:194], v[237:240], s[28:35], s[4:7] tfe ; encoding: [0x00,0x00,0x81,0xf0,0xed,0xc1,0x27,0x00] + +// FIXME: This test is incorrect because r128 assumes a 128-bit SRSRC. +image_sample v193, v[237:240], s[28:35], s[4:7] r128 +// SICIVI: image_sample v193, v[237:240], s[28:35], s[4:7] r128 ; encoding: [0x00,0x80,0x80,0xf0,0xed,0xc1,0x27,0x00] +// NOGFX9: error: r128 modifier is not supported on this GPU + +image_sample v193, v[237:240], s[28:35], s[4:7] d16 +// NOSICI: error: d16 modifier is not supported on this GPU +// GFX89: image_sample v193, v[237:240], s[28:35], s[4:7] d16 ; encoding: [0x00,0x00,0x80,0xf0,0xed,0xc1,0x27,0x80] //===----------------------------------------------------------------------===// // Image Atomics //===----------------------------------------------------------------------===// image_atomic_add v4, v[192:195], s[28:35] dmask:0x1 unorm glc -// SICI: image_atomic_add v4, v[192:195], s[28:35] dmask:0x1 unorm glc ; encoding: [0x00,0x31,0x44,0xf0,0xc0,0x04,0x07,0x00] -// VI: image_atomic_add v4, v[192:195], s[28:35] dmask:0x1 unorm glc ; encoding: [0x00,0x31,0x48,0xf0,0xc0,0x04,0x07,0x00] +// SICI: image_atomic_add v4, v[192:195], s[28:35] dmask:0x1 unorm glc ; encoding: [0x00,0x31,0x44,0xf0,0xc0,0x04,0x07,0x00] +// GFX89: image_atomic_add v4, v[192:195], s[28:35] dmask:0x1 unorm glc ; encoding: [0x00,0x31,0x48,0xf0,0xc0,0x04,0x07,0x00] image_atomic_add v252, v2, s[8:15] dmask:0x1 unorm -// SICI: image_atomic_add v252, v2, s[8:15] dmask:0x1 unorm ; encoding: [0x00,0x11,0x44,0xf0,0x02,0xfc,0x02,0x00] -// VI: image_atomic_add v252, v2, s[8:15] dmask:0x1 unorm ; encoding: [0x00,0x11,0x48,0xf0,0x02,0xfc,0x02,0x00] +// SICI: image_atomic_add v252, v2, s[8:15] dmask:0x1 unorm ; encoding: [0x00,0x11,0x44,0xf0,0x02,0xfc,0x02,0x00] +// GFX89: image_atomic_add v252, v2, s[8:15] dmask:0x1 unorm ; encoding: [0x00,0x11,0x48,0xf0,0x02,0xfc,0x02,0x00] image_atomic_add v[6:7], v255, s[8:15] dmask:0x3 -// SICI: image_atomic_add v[6:7], v255, s[8:15] dmask:0x3 ; encoding: [0x00,0x03,0x44,0xf0,0xff,0x06,0x02,0x00] -// VI: image_atomic_add v[6:7], v255, s[8:15] dmask:0x3 ; encoding: [0x00,0x03,0x48,0xf0,0xff,0x06,0x02,0x00] +// SICI: image_atomic_add v[6:7], v255, s[8:15] dmask:0x3 ; encoding: [0x00,0x03,0x44,0xf0,0xff,0x06,0x02,0x00] +// GFX89: image_atomic_add v[6:7], v255, s[8:15] dmask:0x3 ; encoding: [0x00,0x03,0x48,0xf0,0xff,0x06,0x02,0x00] image_atomic_add v7, v3, s[0:7] dmask:0x1 glc -// SICI: image_atomic_add v7, v3, s[0:7] dmask:0x1 glc ; encoding: [0x00,0x21,0x44,0xf0,0x03,0x07,0x00,0x00] -// VI: image_atomic_add v7, v3, s[0:7] dmask:0x1 glc ; encoding: [0x00,0x21,0x48,0xf0,0x03,0x07,0x00,0x00] +// SICI: image_atomic_add v7, v3, s[0:7] dmask:0x1 glc ; encoding: [0x00,0x21,0x44,0xf0,0x03,0x07,0x00,0x00] +// GFX89: image_atomic_add v7, v3, s[0:7] dmask:0x1 glc ; encoding: [0x00,0x21,0x48,0xf0,0x03,0x07,0x00,0x00] image_atomic_add v8, v4, s[8:15] dmask:0x1 slc -// SICI: image_atomic_add v8, v4, s[8:15] dmask:0x1 slc ; encoding: [0x00,0x01,0x44,0xf2,0x04,0x08,0x02,0x00] -// VI: image_atomic_add v8, v4, s[8:15] dmask:0x1 slc ; encoding: [0x00,0x01,0x48,0xf2,0x04,0x08,0x02,0x00] +// SICI: image_atomic_add v8, v4, s[8:15] dmask:0x1 slc ; encoding: [0x00,0x01,0x44,0xf2,0x04,0x08,0x02,0x00] +// GFX89: image_atomic_add v8, v4, s[8:15] dmask:0x1 slc ; encoding: [0x00,0x01,0x48,0xf2,0x04,0x08,0x02,0x00] image_atomic_add v9, v5, s[8:15] dmask:0x1 unorm glc slc lwe da -// SICI: image_atomic_add v9, v5, s[8:15] dmask:0x1 unorm glc slc lwe da ; encoding: [0x00,0x71,0x46,0xf2,0x05,0x09,0x02,0x00] -// VI: image_atomic_add v9, v5, s[8:15] dmask:0x1 unorm glc slc lwe da ; encoding: [0x00,0x71,0x4a,0xf2,0x05,0x09,0x02,0x00] +// SICI: image_atomic_add v9, v5, s[8:15] dmask:0x1 unorm glc slc lwe da ; encoding: [0x00,0x71,0x46,0xf2,0x05,0x09,0x02,0x00] +// GFX89: image_atomic_add v9, v5, s[8:15] dmask:0x1 unorm glc slc lwe da ; encoding: [0x00,0x71,0x4a,0xf2,0x05,0x09,0x02,0x00] image_atomic_add v10, v6, s[8:15] dmask:0x1 lwe -// SICI: image_atomic_add v10, v6, s[8:15] dmask:0x1 lwe ; encoding: [0x00,0x01,0x46,0xf0,0x06,0x0a,0x02,0x00] -// VI: image_atomic_add v10, v6, s[8:15] dmask:0x1 lwe ; encoding: [0x00,0x01,0x4a,0xf0,0x06,0x0a,0x02,0x00] +// SICI: image_atomic_add v10, v6, s[8:15] dmask:0x1 lwe ; encoding: [0x00,0x01,0x46,0xf0,0x06,0x0a,0x02,0x00] +// GFX89: image_atomic_add v10, v6, s[8:15] dmask:0x1 lwe ; encoding: [0x00,0x01,0x4a,0xf0,0x06,0x0a,0x02,0x00] image_atomic_add v11, v7, s[8:15] dmask:0x1 da -// SICI: image_atomic_add v11, v7, s[8:15] dmask:0x1 da ; encoding: [0x00,0x41,0x44,0xf0,0x07,0x0b,0x02,0x00] -// VI: image_atomic_add v11, v7, s[8:15] dmask:0x1 da ; encoding: [0x00,0x41,0x48,0xf0,0x07,0x0b,0x02,0x00] +// SICI: image_atomic_add v11, v7, s[8:15] dmask:0x1 da ; encoding: [0x00,0x41,0x44,0xf0,0x07,0x0b,0x02,0x00] +// GFX89: image_atomic_add v11, v7, s[8:15] dmask:0x1 da ; encoding: [0x00,0x41,0x48,0xf0,0x07,0x0b,0x02,0x00] image_atomic_swap v4, v[192:195], s[28:35] dmask:0x1 unorm glc -// SICI: image_atomic_swap v4, v[192:195], s[28:35] dmask:0x1 unorm glc ; encoding: [0x00,0x31,0x3c,0xf0,0xc0,0x04,0x07,0x00] -// VI: image_atomic_swap v4, v[192:195], s[28:35] dmask:0x1 unorm glc ; encoding: [0x00,0x31,0x40,0xf0,0xc0,0x04,0x07,0x00] +// SICI: image_atomic_swap v4, v[192:195], s[28:35] dmask:0x1 unorm glc ; encoding: [0x00,0x31,0x3c,0xf0,0xc0,0x04,0x07,0x00] +// GFX89: image_atomic_swap v4, v[192:195], s[28:35] dmask:0x1 unorm glc ; encoding: [0x00,0x31,0x40,0xf0,0xc0,0x04,0x07,0x00] image_atomic_cmpswap v[4:5], v[192:195], s[28:35] dmask:0x3 unorm glc -// SICI: image_atomic_cmpswap v[4:5], v[192:195], s[28:35] dmask:0x3 unorm glc ; encoding: [0x00,0x33,0x40,0xf0,0xc0,0x04,0x07,0x00] -// VI: image_atomic_cmpswap v[4:5], v[192:195], s[28:35] dmask:0x3 unorm glc ; encoding: [0x00,0x33,0x44,0xf0,0xc0,0x04,0x07,0x00] +// SICI: image_atomic_cmpswap v[4:5], v[192:195], s[28:35] dmask:0x3 unorm glc ; encoding: [0x00,0x33,0x40,0xf0,0xc0,0x04,0x07,0x00] +// GFX89: image_atomic_cmpswap v[4:5], v[192:195], s[28:35] dmask:0x3 unorm glc ; encoding: [0x00,0x33,0x44,0xf0,0xc0,0x04,0x07,0x00] image_atomic_cmpswap v[4:7], v[192:195], s[28:35] dmask:0xf unorm glc -// SICI: image_atomic_cmpswap v[4:7], v[192:195], s[28:35] dmask:0xf unorm glc ; encoding: [0x00,0x3f,0x40,0xf0,0xc0,0x04,0x07,0x00] -// VI: image_atomic_cmpswap v[4:7], v[192:195], s[28:35] dmask:0xf unorm glc ; encoding: [0x00,0x3f,0x44,0xf0,0xc0,0x04,0x07,0x00] +// SICI: image_atomic_cmpswap v[4:7], v[192:195], s[28:35] dmask:0xf unorm glc ; encoding: [0x00,0x3f,0x40,0xf0,0xc0,0x04,0x07,0x00] +// GFX89: image_atomic_cmpswap v[4:7], v[192:195], s[28:35] dmask:0xf unorm glc ; encoding: [0x00,0x3f,0x44,0xf0,0xc0,0x04,0x07,0x00] + +// FIXME: This test is incorrect because r128 assumes a 128-bit SRSRC. +image_atomic_add v10, v6, s[8:15] dmask:0x1 r128 +// SICI: image_atomic_add v10, v6, s[8:15] dmask:0x1 r128 ; encoding: [0x00,0x81,0x44,0xf0,0x06,0x0a,0x02,0x00] +// VI: image_atomic_add v10, v6, s[8:15] dmask:0x1 r128 ; encoding: [0x00,0x81,0x48,0xf0,0x06,0x0a,0x02,0x00] +// NOGFX9: error: r128 modifier is not supported on this GPU