[PATCH] D101619: [AMDGPU][AsmParser] Adjust img instruction address field if a16 present
David Stuttard via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Fri Apr 30 05:29:05 PDT 2021
dstuttard created this revision.
Herald added subscribers: kerbowa, hiraditya, t-tye, tpr, yaxunl, nhaehnle, jvesely, kzhuravl, arsenm.
dstuttard requested review of this revision.
Herald added subscribers: llvm-commits, wdng.
Herald added a project: LLVM.
A16 support for image instructions assembly (gfx10) was missing
Repository:
rG LLVM Github Monorepo
https://reviews.llvm.org/D101619
Files:
llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
llvm/test/MC/AMDGPU/gfx10_asm_mimg.s
Index: llvm/test/MC/AMDGPU/gfx10_asm_mimg.s
===================================================================
--- llvm/test/MC/AMDGPU/gfx10_asm_mimg.s
+++ llvm/test/MC/AMDGPU/gfx10_asm_mimg.s
@@ -379,11 +379,35 @@
image_sample_c_cd_cl_o v[64:66], [v32, v16, v0, v2, v1, v4, v5, v6, v7, v8, v9, v10], s[4:11], s[100:103] dmask:0x7 dim:SQ_RSRC_IMG_3D
; GFX10: image_sample_c_cd_cl_o v[64:66], [v32, v16, v0, v2, v1, v4, v5, v6, v7, v8, v9, v10], s[4:11], s[100:103] dmask:0x7 dim:SQ_RSRC_IMG_3D ; encoding: [0x16,0x07,0xbc,0xf1,0x20,0x40,0x21,0x03,0x10,0x00,0x02,0x01,0x04,0x05,0x06,0x07,0x08,0x09,0x0a,0x00]
-image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16
-; GFX10: image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16 ; encoding: [0x08,0x1f,0x00,0xf0,0x00,0x00,0x00,0x40]
+image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16
+; GFX10: image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16 ; encoding: [0x08,0x1f,0x00,0xf0,0x00,0x00,0x00,0x40]
image_load v[0:4], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm tfe
; GFX10: image_load v[0:4], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm tfe ; encoding: [0x08,0x1f,0x01,0xf0,0x00,0x00,0x00,0x00]
-image_load v[0:4], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16 tfe
-; GFX10: image_load v[0:4], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16 tfe ; encoding: [0x08,0x1f,0x01,0xf0,0x00,0x00,0x00,0x40]
+image_load v[0:4], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16 tfe
+; GFX10: image_load v[0:4], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16 tfe ; encoding: [0x08,0x1f,0x01,0xf0,0x00,0x00,0x00,0x40]
+
+image_load v1, v1, s[16:23] dmask:0x1 dim:SQ_RSRC_IMG_2D a16
+; GFX10: image_load v1, v1, s[16:23] dmask:0x1 dim:SQ_RSRC_IMG_2D a16 ; encoding: [0x08,0x01,0x00,0xf0,0x01,0x01,0x04,0x40]
+
+image_load v[1:2], v1, s[16:23] dmask:0x1 dim:SQ_RSRC_IMG_2D a16 tfe
+; GFX10: image_load v[1:2], v1, s[16:23] dmask:0x1 dim:SQ_RSRC_IMG_2D a16 tfe ; encoding: [0x08,0x01,0x01,0xf0,0x01,0x01,0x04,0x40]
+
+image_load v1, v1, s[16:23] dmask:0x1 dim:SQ_RSRC_IMG_2D a16 lwe
+; GFX10: image_load v1, v1, s[16:23] dmask:0x1 dim:SQ_RSRC_IMG_2D a16 lwe ; encoding: [0x08,0x01,0x02,0xf0,0x01,0x01,0x04,0x40]
+
+image_load v[1:2], v1, s[16:23] dmask:0x1 dim:SQ_RSRC_IMG_2D a16 tfe lwe
+; GFX10: image_load v[1:2], v1, s[16:23] dmask:0x1 dim:SQ_RSRC_IMG_2D a16 tfe lwe ; encoding: [0x08,0x01,0x03,0xf0,0x01,0x01,0x04,0x40]
+
+image_load v[1:2], v1, s[16:23] dmask:0x3 dim:SQ_RSRC_IMG_2D a16
+; GFX10: image_load v[1:2], v1, s[16:23] dmask:0x3 dim:SQ_RSRC_IMG_2D a16 ; encoding: [0x08,0x03,0x00,0xf0,0x01,0x01,0x04,0x40]
+
+image_load v[1:4], v1, s[16:23] dmask:0x7 dim:SQ_RSRC_IMG_2D a16 tfe
+; GFX10: image_load v[1:4], v1, s[16:23] dmask:0x7 dim:SQ_RSRC_IMG_2D a16 tfe ; encoding: [0x08,0x07,0x01,0xf0,0x01,0x01,0x04,0x40]
+
+image_load v[1:4], v1, s[16:23] dmask:0xf dim:SQ_RSRC_IMG_2D a16 lwe
+; GFX10: image_load v[1:4], v1, s[16:23] dmask:0xf dim:SQ_RSRC_IMG_2D a16 lwe ; encoding: [0x08,0x0f,0x02,0xf0,0x01,0x01,0x04,0x40]
+
+image_load v[1:3], v1, s[16:23] dmask:0x5 dim:SQ_RSRC_IMG_2D a16 tfe lwe
+; GFX10: image_load v[1:3], v1, s[16:23] dmask:0x5 dim:SQ_RSRC_IMG_2D a16 tfe lwe ; encoding: [0x08,0x05,0x03,0xf0,0x01,0x01,0x04,0x40]
Index: llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
===================================================================
--- llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
+++ llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
@@ -3414,6 +3414,7 @@
int VAddr0Idx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::vaddr0);
int SrsrcIdx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::srsrc);
int DimIdx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::dim);
+ int A16Idx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::a16);
assert(VAddr0Idx != -1);
assert(SrsrcIdx != -1);
@@ -3428,11 +3429,24 @@
unsigned VAddrSize =
IsNSA ? SrsrcIdx - VAddr0Idx
: AMDGPU::getRegOperandSize(getMRI(), Desc, VAddr0Idx) / 4;
+ bool IsA16 = (A16Idx != -1 && Inst.getOperand(A16Idx).getImm()) ? 1 : 0;
+
+ // This mimics the calculation of AddrSize in SIInstrInfo::verifyInstruction.
+ unsigned AddrSize = BaseOpcode->NumExtraArgs;
+ unsigned AddrComponents = (BaseOpcode->Coordinates ? DimInfo->NumCoords : 0) +
+ (BaseOpcode->LodOrClampOrMip ? 1 : 0);
+ if (IsA16)
+ AddrComponents = divideCeil(AddrComponents, 2);
+
+ AddrSize += AddrComponents;
+
+ if (BaseOpcode->Gradients) {
+ if (IsA16 || BaseOpcode->G16)
+ AddrSize += alignTo<2>(DimInfo->NumGradients / 2);
+ else
+ AddrSize += DimInfo->NumGradients;
+ }
- unsigned AddrSize = BaseOpcode->NumExtraArgs +
- (BaseOpcode->Gradients ? DimInfo->NumGradients : 0) +
- (BaseOpcode->Coordinates ? DimInfo->NumCoords : 0) +
- (BaseOpcode->LodOrClampOrMip ? 1 : 0);
if (!IsNSA) {
if (AddrSize > 8)
AddrSize = 16;
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D101619.341857.patch
Type: text/x-patch
Size: 5035 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210430/7448ede5/attachment.bin>
More information about the llvm-commits
mailing list