[AMDGPU] Add v5f32/VReg_160 support for MIMG instructions

[critson]

Summary

Avoid having to round up to v8f32/VReg_256 when only 5 VGPRs are
required for a MIMG address operand.

Maintain _V8 instruction variants of pseudo instructions allowing
assembly prior to GFX10 to work as-is. Currently the validator
can tell for GFX10 what the correct size is, so will disallow
oversize address registers.

[foad]

My only slight concern is if people have GFX10 disassemblies stored somewhere, they might find they can no longer reassemble them after your change. I don't know how serious that is.

Will there be a follow up patch to support v6f32/VReg_192? I think the register classes already exist for that.

[critson]

In D103672#2798535, @foad wrote:

My only slight concern is if people have GFX10 disassemblies stored somewhere, they might find they can no longer reassemble them after your change. I don't know how serious that is.

I believe we could allow these by changing the validator.

Will there be a follow up patch to support v6f32/VReg_192? I think the register classes already exist for that.

v6f32 is not actually defined in LLVM (MVT class), so I am not sure how possible this is. I realized this as I initially tried to include v6f32 support in this change.

[rampitec]

I believe it was long needed.

llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp

bool AMDGPUAsmParser::validateMIMGAddrSize(const MCInst &Inst) {
   bool IsA16 = (A16Idx != -1 && Inst.getOperand(A16Idx).getImm());
 
   unsigned AddrSize =
       AMDGPU::getAddrSizeMIMGOp(BaseOpcode, DimInfo, IsA16, hasG16());
 
   if (!IsNSA) {
     if (AddrSize > 8)
       AddrSize = 16;
-    else if (AddrSize > 4)
+    else if (AddrSize > 5)
       AddrSize = 8;
   }
 
   return VAddrSize == AddrSize;
 }
 
 bool AMDGPUAsmParser::validateMIMGAtomicDMask(const MCInst &Inst) {
 

llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp

if (STI.getFeatureBits()[AMDGPU::FeatureGFX10]) {
 
     AddrSize =
         AMDGPU::getAddrSizeMIMGOp(BaseOpcode, Dim, IsA16, AMDGPU::hasG16(STI));
 
     IsNSA = Info->MIMGEncoding == AMDGPU::MIMGEncGfx10NSA;
     if (!IsNSA) {
       if (AddrSize > 8)
         AddrSize = 16;
-      else if (AddrSize > 4)
+      else if (AddrSize > 5)
         AddrSize = 8;
     } else {
       if (AddrSize > Info->VAddrDwords) {
         // The NSA encoding does not contain enough operands for the combination
         // of base opcode / dimension. Should this be an error?
         return MCDisassembler::Success;
       }
     }

llvm/lib/Target/AMDGPU/MIMGInstructions.td

 class MIMGAddrSize<int dw, bit enable_disasm> {
   int NumWords = dw;
 
   RegisterClass RegClass = !if(!le(NumWords, 0), ?,
                            !if(!eq(NumWords, 1), VGPR_32,
                            !if(!eq(NumWords, 2), VReg_64,
                            !if(!eq(NumWords, 3), VReg_96,
                            !if(!eq(NumWords, 4), VReg_128,
+                           !if(!eq(NumWords, 5), VReg_160,
                            !if(!le(NumWords, 8), VReg_256,
-                           !if(!le(NumWords, 16), VReg_512, ?)))))));
+                           !if(!le(NumWords, 16), VReg_512, ?))))))));
 
   // Whether the instruction variant with this vaddr size should be enabled for
   // the auto-generated disassembler.
   bit Disassemble = enable_disasm;
 }
 
 // Return whether x is in lst.
 class isIntInList<int x, list<int> lst> {
!foreach(dw, !if(sample.Gradients,
                          [1, 2, 3],
                          [1, 2, 3, 4])),
              !add(dw, !size(sample.ExtraAddrArgs)));
 
   // Generate machine instructions based on possible register classes for the
   // required numbers of address words. The disassembler defaults to the
   // smallest register class.
   list<MIMGAddrSize> MachineInstrs =
-    !foldl(MIMGAddrSizes_tmp<[], 0>, [1, 2, 3, 4, 8, 16], lhs, dw,
+    !foldl(MIMGAddrSizes_tmp<[], 0>, [1, 2, 3, 4, 5, 8, 16], lhs, dw,
            !if(isRangeInList<lhs.Min, dw, AllNumAddrWords>.ret,
                MIMGAddrSizes_tmp<
                   !listconcat(lhs.List, [MIMGAddrSize<dw, !empty(lhs.List)>]),
-                  !if(!eq(dw, 3), 3, !add(dw, 1))>, // we still need _V4 for codegen w/ 3 dwords
+                  !if(!or(!eq(dw, 3), !eq(dw, 5)), dw, !add(dw, 1))>,
+                  // we still want _V4/_V8 for codegen with 3/5 dwords
                lhs)).List;
 
   // For NSA, generate machine instructions for all possible numbers of words
   // except 1 (which is already covered by the non-NSA case).
   // The disassembler defaults to the largest number of arguments among the
   // variants with the same number of NSA words, and custom code then derives
   // the exact variant based on the sample variant and the image dimension.
   list<MIMGAddrSize> NSAInstrs =

llvm/lib/Target/AMDGPU/SIISelLowering.cpp

if (Elts.size() == 1) {
     Type = MVT::v2f32;
     NumElts = 2;
   } else if (Elts.size() == 3) {
     Type = MVT::v3f32;
     NumElts = 3;
   } else if (Elts.size() <= 4) {
     Type = MVT::v4f32;
     NumElts = 4;
+  } else if (Elts.size() <= 5) {
+    Type = MVT::v5f32;
+    NumElts = 5;
   } else if (Elts.size() <= 8) {
     Type = MVT::v8f32;
     NumElts = 8;
   } else {
     assert(Elts.size() <= 16);
     Type = MVT::v16f32;
     NumElts = 16;
   }

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp

if (DimOp) {
       unsigned VAddrWords;
       if (IsNSA) {
         VAddrWords = SRsrcIdx - VAddr0Idx;
       } else {
         const TargetRegisterClass *RC = getOpRegClass(MI, VAddr0Idx);
         VAddrWords = MRI.getTargetRegisterInfo()->getRegSizeInBits(*RC) / 32;
         if (AddrWords > 8)
           AddrWords = 16;
-        else if (AddrWords > 4)
+        else if (AddrWords > 5)
           AddrWords = 8;
-        else if (AddrWords == 4)
-          AddrWords = 4;
-        else if (AddrWords == 3)
-          AddrWords = 3;
       }
 
       if (VAddrWords != AddrWords) {
         LLVM_DEBUG(dbgs() << "bad vaddr size, expected " << AddrWords
                           << " but got " << VAddrWords << "\n");
         ErrInfo = "bad vaddr size";
         return false;
       }

llvm/lib/Target/AMDGPU/SIShrinkInstructions.cpp

void SIShrinkInstructions::shrinkMIMG(MachineInstr &MI) {
   const TargetRegisterClass *RC;
 
   if (Info->VAddrDwords == 2) {
     RC = &AMDGPU::VReg_64RegClass;
   } else if (Info->VAddrDwords == 3) {
     RC = &AMDGPU::VReg_96RegClass;
   } else if (Info->VAddrDwords == 4) {
     RC = &AMDGPU::VReg_128RegClass;
+  } else if (Info->VAddrDwords == 5) {
+    RC = &AMDGPU::VReg_160RegClass;
   } else if (Info->VAddrDwords <= 8) {
     RC = &AMDGPU::VReg_256RegClass;
     NewAddrDwords = 8;
   } else {
     RC = &AMDGPU::VReg_512RegClass;
     NewAddrDwords = 16;
   }
 

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.gather4.dim.ll

 ; GFX10NSA-NEXT:    s_mov_b32 s4, s6
 ; GFX10NSA-NEXT:    s_mov_b32 s5, s7
 ; GFX10NSA-NEXT:    s_mov_b32 s6, s8
 ; GFX10NSA-NEXT:    s_mov_b32 s7, s9
 ; GFX10NSA-NEXT:    s_mov_b32 s8, s10
 ; GFX10NSA-NEXT:    s_mov_b32 s9, s11
 ; GFX10NSA-NEXT:    s_mov_b32 s10, s12
 ; GFX10NSA-NEXT:    s_mov_b32 s11, s13
-; GFX10NSA-NEXT:    image_gather4_c_b_cl v[0:3], v[0:7], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D
+; GFX10NSA-NEXT:    image_gather4_c_b_cl v[0:3], v[0:4], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D
 ; GFX10NSA-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10NSA-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.b.cl.2d.v4f32.f32.f32(i32 1, float %bias, float %zcompare, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @gather4_l_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %s, float %t, float %lod) {

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.gather4.o.dim.ll

 ; GFX10-NEXT:    s_mov_b32 s4, s6
 ; GFX10-NEXT:    s_mov_b32 s5, s7
 ; GFX10-NEXT:    s_mov_b32 s6, s8
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    s_mov_b32 s8, s10
 ; GFX10-NEXT:    s_mov_b32 s9, s11
 ; GFX10-NEXT:    s_mov_b32 s10, s12
 ; GFX10-NEXT:    s_mov_b32 s11, s13
-; GFX10-NEXT:    image_gather4_c_cl_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D
+; GFX10-NEXT:    image_gather4_c_cl_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.cl.o.2d.v4f32.f32(i32 1, i32 %offset, float %zcompare, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @gather4_b_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %bias, float %s, float %t) {

 ; GFX10-NEXT:    s_mov_b32 s4, s6
 ; GFX10-NEXT:    s_mov_b32 s5, s7
 ; GFX10-NEXT:    s_mov_b32 s6, s8
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    s_mov_b32 s8, s10
 ; GFX10-NEXT:    s_mov_b32 s9, s11
 ; GFX10-NEXT:    s_mov_b32 s10, s12
 ; GFX10-NEXT:    s_mov_b32 s11, s13
-; GFX10-NEXT:    image_gather4_c_b_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D
+; GFX10-NEXT:    image_gather4_c_b_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.b.o.2d.v4f32.f32.f32(i32 1, i32 %offset, float %bias, float %zcompare, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @gather4_b_cl_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %bias, float %s, float %t, float %clamp) {

 ; GFX10-NEXT:    s_mov_b32 s4, s6
 ; GFX10-NEXT:    s_mov_b32 s5, s7
 ; GFX10-NEXT:    s_mov_b32 s6, s8
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    s_mov_b32 s8, s10
 ; GFX10-NEXT:    s_mov_b32 s9, s11
 ; GFX10-NEXT:    s_mov_b32 s10, s12
 ; GFX10-NEXT:    s_mov_b32 s11, s13
-; GFX10-NEXT:    image_gather4_b_cl_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D
+; GFX10-NEXT:    image_gather4_b_cl_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.b.cl.o.2d.v4f32.f32.f32(i32 1, i32 %offset, float %bias, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @gather4_c_b_cl_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %bias, float %zcompare, float %s, float %t, float %clamp) {

 ; GFX10-NEXT:    s_mov_b32 s4, s6
 ; GFX10-NEXT:    s_mov_b32 s5, s7
 ; GFX10-NEXT:    s_mov_b32 s6, s8
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    s_mov_b32 s8, s10
 ; GFX10-NEXT:    s_mov_b32 s9, s11
 ; GFX10-NEXT:    s_mov_b32 s10, s12
 ; GFX10-NEXT:    s_mov_b32 s11, s13
-; GFX10-NEXT:    image_gather4_c_l_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D
+; GFX10-NEXT:    image_gather4_c_l_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.l.o.2d.v4f32.f32(i32 1, i32 %offset, float %zcompare, float %s, float %t, float %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 false, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @gather4_lz_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %s, float %t) {

llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.sample.g16.ll

 
 define amdgpu_ps <4 x float> @sample_c_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
 ; GFX10-LABEL: sample_c_d_cl_1d:
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_mov_b32_e32 v5, 0xffff
 ; GFX10-NEXT:    s_lshl_b32 s12, s0, 16
 ; GFX10-NEXT:    v_and_or_b32 v1, v1, v5, s12
 ; GFX10-NEXT:    v_and_or_b32 v2, v2, v5, s12
-; GFX10-NEXT:    image_sample_c_d_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
+; GFX10-NEXT:    image_sample_c_d_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {

 
 define amdgpu_ps <4 x float> @sample_c_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
 ; GFX10-LABEL: sample_c_cd_cl_1d:
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_mov_b32_e32 v5, 0xffff
 ; GFX10-NEXT:    s_lshl_b32 s12, s0, 16
 ; GFX10-NEXT:    v_and_or_b32 v1, v1, v5, s12
 ; GFX10-NEXT:    v_and_or_b32 v2, v2, v5, s12
-; GFX10-NEXT:    image_sample_c_cd_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
+; GFX10-NEXT:    image_sample_c_cd_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.gather4.dim.ll

 ; GFX10: image_gather4_b_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_b_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %bias, float %s, float %t, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.b.cl.2d.v4f32.f32.f32(i32 1, float %bias, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_c_b_cl_2d:
-; GFX6789: image_gather4_c_b_cl v[0:3], v[0:7], s[0:7], s[8:11] dmask:0x1{{$}}
-; GFX10: image_gather4_c_b_cl v[0:3], v[0:7], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ;
+; GFX6789: image_gather4_c_b_cl v[0:3], v[0:4], s[0:7], s[8:11] dmask:0x1{{$}}
+; GFX10: image_gather4_c_b_cl v[0:3], v[0:4], s[0:7], s[8:11] dmask:0x1 dim:SQ_RSRC_IMG_2D ;
 define amdgpu_ps <4 x float> @gather4_c_b_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %bias, float %zcompare, float %s, float %t, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.b.cl.2d.v4f32.f32.f32(i32 1, float %bias, float %zcompare, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_l_2d:
 ; GFX6789: image_gather4_l v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1{{$}}

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.gather4.o.dim.ll

 ; GCN: image_gather4_cl_o v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
 define amdgpu_ps <4 x float> @gather4_cl_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %s, float %t, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.cl.o.2d.v4f32.f32(i32 1, i32 %offset, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_c_cl_o_2d:
-; GCN: image_gather4_c_cl_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0x1{{$}}
+; GCN: image_gather4_c_cl_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0x1{{$}}
 define amdgpu_ps <4 x float> @gather4_c_cl_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, float %s, float %t, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.cl.o.2d.v4f32.f32(i32 1, i32 %offset, float %zcompare, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_b_o_2d:
 ; GCN: image_gather4_b_o v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
 define amdgpu_ps <4 x float> @gather4_b_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %bias, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.b.o.2d.v4f32.f32.f32(i32 1, i32 %offset, float %bias, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_c_b_o_2d:
-; GCN: image_gather4_c_b_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0x1{{$}}
+; GCN: image_gather4_c_b_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0x1{{$}}
 define amdgpu_ps <4 x float> @gather4_c_b_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %bias, float %zcompare, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.b.o.2d.v4f32.f32.f32(i32 1, i32 %offset, float %bias, float %zcompare, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_b_cl_o_2d:
-; GCN: image_gather4_b_cl_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0x1{{$}}
+; GCN: image_gather4_b_cl_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0x1{{$}}
 define amdgpu_ps <4 x float> @gather4_b_cl_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %bias, float %s, float %t, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.b.cl.o.2d.v4f32.f32.f32(i32 1, i32 %offset, float %bias, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_c_b_cl_o_2d:
 ; GCN: image_gather4_c_b_cl_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0x1{{$}}
 define amdgpu_ps <4 x float> @gather4_c_b_cl_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %bias, float %zcompare, float %s, float %t, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.b.cl.o.2d.v4f32.f32.f32(i32 1, i32 %offset, float %bias, float %zcompare, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_l_o_2d:
 ; GCN: image_gather4_l_o v[0:3], v[0:3], s[0:7], s[8:11] dmask:0x1{{$}}
 define amdgpu_ps <4 x float> @gather4_l_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %s, float %t, float %lod) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.l.o.2d.v4f32.f32(i32 1, i32 %offset, float %s, float %t, float %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_c_l_o_2d:
-; GCN: image_gather4_c_l_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0x1{{$}}
+; GCN: image_gather4_c_l_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0x1{{$}}
 define amdgpu_ps <4 x float> @gather4_c_l_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, float %s, float %t, float %lod) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.gather4.c.l.o.2d.v4f32.f32(i32 1, i32 %offset, float %zcompare, float %s, float %t, float %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}gather4_lz_o_2d:
 ; GCN: image_gather4_lz_o v[0:3], v[0:2], s[0:7], s[8:11] dmask:0x1{{$}}

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.nsa.ll

 ; NSA: image_sample_d v[0:3], [v3, v8, v7, v5, v4, v6, v0, v2, v1],
 define amdgpu_ps <4 x float> @sample_d_3d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %s, float %r, float %t, float %dsdh, float %dtdv, float %dsdv, float %drdv, float %drdh, float %dtdh) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.d.3d.v4f32.f32(i32 15, float %dsdh, float %dtdh, float %drdh, float %dsdv, float %dtdv, float %drdv, float %s, float %t, float %r, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_contig_nsa:
-; NONSA: image_sample_c_l v5, v[0:7],
-; NSA: image_sample_c_l v8, v[0:7],
+; NONSA: image_sample_c_l v5, v[0:4],
+; NSA: image_sample_c_l v8, v[0:4],
 ; NSA: image_sample v9, [v6, v7, v5],
 define amdgpu_ps <2 x float> @sample_contig_nsa(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %s1, float %t1, float %r1, float %lod, float %r2, float %s2, float %t2) {
 main_body:
   %v1 = call float @llvm.amdgcn.image.sample.c.l.3d.f32.f32(i32 1, float %zcompare, float %s1, float %t1, float %r1, float %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   %v2 = call float @llvm.amdgcn.image.sample.3d.f32.f32(i32 1, float %s2, float %t2, float %r2, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   %r.0 = insertelement <2 x float> undef, float %v1, i32 0
   %r = insertelement <2 x float> %r.0, float %v2, i32 1
   ret <2 x float> %r
main_body:
   %v1 = call float @llvm.amdgcn.image.sample.c.l.3d.f32.f32(i32 1, float %zcompare, float %s1, float %t1, float %r1, float %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   %v2 = call float @llvm.amdgcn.image.sample.3d.f32.f32(i32 1, float %s2, float %t2, float %r2, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   %r.0 = insertelement <2 x float> undef, float %v1, i32 0
   %r = insertelement <2 x float> %r.0, float %v2, i32 1
   ret <2 x float> %r
 }
 
 ; GCN-LABEL: {{^}}sample_contig_contig:
-; NSA: image_sample_c_l v8, v[0:7],
+; NSA: image_sample_c_l v8, v[0:4],
 ; NSA: image_sample v9, v[5:7],
-; NONSA: image_sample_c_l v8, v[0:7],
+; NONSA: image_sample_c_l v8, v[0:4],
 ; NONSA: image_sample v9, v[5:7],
 define amdgpu_ps <2 x float> @sample_contig_contig(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %s1, float %t1, float %r1, float %lod, float %s2, float %t2, float %r2) {
 main_body:
   %v1 = call float @llvm.amdgcn.image.sample.c.l.3d.f32.f32(i32 1, float %zcompare, float %s1, float %t1, float %r1, float %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   %v2 = call float @llvm.amdgcn.image.sample.3d.f32.f32(i32 1, float %s2, float %t2, float %r2, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   %r.0 = insertelement <2 x float> undef, float %v1, i32 0
   %r = insertelement <2 x float> %r.0, float %v2, i32 1
   ret <2 x float> %r

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.a16.dim.ll

 ; GFX9-NEXT:    v_mov_b32_e32 v7, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 0xffff
 ; GFX9-NEXT:    v_and_b32_e32 v5, v0, v5
 ; GFX9-NEXT:    v_and_b32_e32 v3, v0, v3
 ; GFX9-NEXT:    v_and_b32_e32 v0, v0, v1
 ; GFX9-NEXT:    v_lshl_or_b32 v10, v6, 16, v5
 ; GFX9-NEXT:    v_lshl_or_b32 v9, v4, 16, v3
 ; GFX9-NEXT:    v_lshl_or_b32 v8, v2, 16, v0
-; GFX9-NEXT:    image_sample_c_d_cl v[0:3], v[7:14], s[0:7], s[8:11] dmask:0xf a16
+; GFX9-NEXT:    image_sample_c_d_cl v[0:3], v[7:11], s[0:7], s[8:11] dmask:0xf a16
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: sample_c_d_cl_2d:
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_mov_b32_e32 v8, 0xffff
 ; GFX10-NEXT:    v_and_b32_e32 v5, v8, v5
 ; GFX10-NEXT:    v_and_b32_e32 v3, v8, v3

 ; GFX9-NEXT:    v_mov_b32_e32 v7, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v0, 0xffff
 ; GFX9-NEXT:    v_and_b32_e32 v5, v0, v5
 ; GFX9-NEXT:    v_and_b32_e32 v3, v0, v3
 ; GFX9-NEXT:    v_and_b32_e32 v0, v0, v1
 ; GFX9-NEXT:    v_lshl_or_b32 v10, v6, 16, v5
 ; GFX9-NEXT:    v_lshl_or_b32 v9, v4, 16, v3
 ; GFX9-NEXT:    v_lshl_or_b32 v8, v2, 16, v0
-; GFX9-NEXT:    image_sample_c_cd_cl v[0:3], v[7:14], s[0:7], s[8:11] dmask:0xf a16
+; GFX9-NEXT:    image_sample_c_cd_cl v[0:3], v[7:11], s[0:7], s[8:11] dmask:0xf a16
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: sample_c_cd_cl_2d:
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_mov_b32_e32 v8, 0xffff
 ; GFX10-NEXT:    v_and_b32_e32 v5, v8, v5
 ; GFX10-NEXT:    v_and_b32_e32 v3, v8, v3

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.dim.ll

 }
 
 define amdgpu_ps <4 x float> @sample_c_b_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %bias, float %zcompare, float %s, float %t, float %clamp) {
 ; VERDE-LABEL: sample_c_b_cl_2d:
 ; VERDE:       ; %bb.0: ; %main_body
 ; VERDE-NEXT:    s_mov_b64 s[12:13], exec
 ; VERDE-NEXT:    s_wqm_b64 exec, exec
 ; VERDE-NEXT:    s_and_b64 exec, exec, s[12:13]
-; VERDE-NEXT:    image_sample_c_b_cl v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf
+; VERDE-NEXT:    image_sample_c_b_cl v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf
 ; VERDE-NEXT:    s_waitcnt vmcnt(0)
 ; VERDE-NEXT:    ; return to shader part epilog
 ;
 ; GFX6789-LABEL: sample_c_b_cl_2d:
 ; GFX6789:       ; %bb.0: ; %main_body
 ; GFX6789-NEXT:    s_mov_b64 s[12:13], exec
 ; GFX6789-NEXT:    s_wqm_b64 exec, exec
 ; GFX6789-NEXT:    s_and_b64 exec, exec, s[12:13]
-; GFX6789-NEXT:    image_sample_c_b_cl v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf
+; GFX6789-NEXT:    image_sample_c_b_cl v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf
 ; GFX6789-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6789-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: sample_c_b_cl_2d:
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    s_mov_b32 s12, exec_lo ; encoding: [0x7e,0x03,0x8c,0xbe]
 ; GFX10-NEXT:    s_wqm_b32 exec_lo, exec_lo ; encoding: [0x7e,0x09,0xfe,0xbe]
 ; GFX10-NEXT:    s_and_b32 exec_lo, exec_lo, s12 ; encoding: [0x7e,0x0c,0x7e,0x87]
-; GFX10-NEXT:    image_sample_c_b_cl v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x08,0x0f,0xb8,0xf0,0x00,0x00,0x40,0x00]
+; GFX10-NEXT:    image_sample_c_b_cl v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x08,0x0f,0xb8,0xf0,0x00,0x00,0x40,0x00]
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
 ; GFX10-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.b.cl.2d.v4f32.f32.f32(i32 15, float %bias, float %zcompare, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_d_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %dsdh, float %dsdv, float %s) {

 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f32.f32(i32 15, float %dsdh, float %dtdh, float %dsdv, float %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %dsdh, float %dsdv, float %s, float %clamp) {
 ; VERDE-LABEL: sample_c_d_cl_1d:
 ; VERDE:       ; %bb.0: ; %main_body
-; VERDE-NEXT:    image_sample_c_d_cl v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf
+; VERDE-NEXT:    image_sample_c_d_cl v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf
 ; VERDE-NEXT:    s_waitcnt vmcnt(0)
 ; VERDE-NEXT:    ; return to shader part epilog
 ;
 ; GFX6789-LABEL: sample_c_d_cl_1d:
 ; GFX6789:       ; %bb.0: ; %main_body
-; GFX6789-NEXT:    image_sample_c_d_cl v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf
+; GFX6789-NEXT:    image_sample_c_d_cl v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf
 ; GFX6789-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6789-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: sample_c_d_cl_1d:
 ; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_sample_c_d_cl v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x00,0x0f,0xac,0xf0,0x00,0x00,0x40,0x00]
+; GFX10-NEXT:    image_sample_c_d_cl v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x00,0x0f,0xac,0xf0,0x00,0x00,0x40,0x00]
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
 ; GFX10-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.1d.v4f32.f32.f32(i32 15, float %zcompare, float %dsdh, float %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %dsdh, float %dtdh, float %dsdv, float %dtdv, float %s, float %t, float %clamp) {

 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f32.f32(i32 15, float %dsdh, float %dtdh, float %dsdv, float %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %dsdh, float %dsdv, float %s, float %clamp) {
 ; VERDE-LABEL: sample_c_cd_cl_1d:
 ; VERDE:       ; %bb.0: ; %main_body
-; VERDE-NEXT:    image_sample_c_cd_cl v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf
+; VERDE-NEXT:    image_sample_c_cd_cl v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf
 ; VERDE-NEXT:    s_waitcnt vmcnt(0)
 ; VERDE-NEXT:    ; return to shader part epilog
 ;
 ; GFX6789-LABEL: sample_c_cd_cl_1d:
 ; GFX6789:       ; %bb.0: ; %main_body
-; GFX6789-NEXT:    image_sample_c_cd_cl v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf
+; GFX6789-NEXT:    image_sample_c_cd_cl v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf
 ; GFX6789-NEXT:    s_waitcnt vmcnt(0)
 ; GFX6789-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: sample_c_cd_cl_1d:
 ; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_sample_c_cd_cl v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x00,0x0f,0xac,0xf1,0x00,0x00,0x40,0x00]
+; GFX10-NEXT:    image_sample_c_cd_cl v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x00,0x0f,0xac,0xf1,0x00,0x00,0x40,0x00]
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
 ; GFX10-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.1d.v4f32.f32.f32(i32 15, float %zcompare, float %dsdh, float %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %dsdh, float %dtdh, float %dsdv, float %dtdv, float %s, float %t, float %clamp) {

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.g16.a16.dim.ll

main_body:
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_d_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t) {
 ; GFX10-LABEL: sample_d_2d:
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_and_b32_e32 v4, 0xffff, v4
 ; GFX10-NEXT:    v_lshl_or_b32 v4, v5, 16, v4
-; GFX10-NEXT:    image_sample_d v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
+; GFX10-NEXT:    image_sample_d v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_d_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_lshlrev_b32_e32 v5, 16, v5
 ; GFX10GISEL-NEXT:    v_and_or_b32 v4, 0xffff, v4, v5
-; GFX10GISEL-NEXT:    image_sample_d v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
+; GFX10GISEL-NEXT:    image_sample_d v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f32.f16(i32 15, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_d_3d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %dsdh, float %dtdh, float %drdh, float %dsdv, float %dtdv, float %drdv, half %s, half %t, half %r) {
main_body:
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_cd_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t) {
 ; GFX10-LABEL: sample_cd_2d:
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_and_b32_e32 v4, 0xffff, v4
 ; GFX10-NEXT:    v_lshl_or_b32 v4, v5, 16, v4
-; GFX10-NEXT:    image_sample_cd v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
+; GFX10-NEXT:    image_sample_cd v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_cd_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_lshlrev_b32_e32 v5, 16, v5
 ; GFX10GISEL-NEXT:    v_and_or_b32 v4, 0xffff, v4, v5
-; GFX10GISEL-NEXT:    image_sample_cd v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
+; GFX10GISEL-NEXT:    image_sample_cd v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.cd.2d.v4f32.f32.f16(i32 15, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_cd_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %dsdh, float %dsdv, half %s) {

 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_g16_noa16_c_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
 ; GFX10-LABEL: sample_g16_noa16_c_d_cl_1d:
 ; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_sample_c_d_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
+; GFX10-NEXT:    image_sample_c_d_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_g16_noa16_c_d_cl_1d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v5, 0xffff
 ; GFX10GISEL-NEXT:    s_lshl_b32 s12, s0, 16
 ; GFX10GISEL-NEXT:    v_and_or_b32 v1, v1, v5, s12
 ; GFX10GISEL-NEXT:    v_and_or_b32 v2, v2, v5, s12
-; GFX10GISEL-NEXT:    image_sample_c_d_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
+; GFX10GISEL-NEXT:    image_sample_c_d_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_g16_noa16_c_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {

 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_g16_noa16_c_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
 ; GFX10-LABEL: sample_g16_noa16_c_cd_cl_1d:
 ; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_sample_c_cd_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
+; GFX10-NEXT:    image_sample_c_cd_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_g16_noa16_c_cd_cl_1d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v5, 0xffff
 ; GFX10GISEL-NEXT:    s_lshl_b32 s12, s0, 16
 ; GFX10GISEL-NEXT:    v_and_or_b32 v1, v1, v5, s12
 ; GFX10GISEL-NEXT:    v_and_or_b32 v2, v2, v5, s12
-; GFX10GISEL-NEXT:    image_sample_c_cd_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
+; GFX10GISEL-NEXT:    image_sample_c_cd_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_g16_noa16_c_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.g16.encode.ll

 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
 ; GFX10-LABEL: sample_c_d_cl_1d:
 ; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_sample_c_d_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xac,0xf0,0x00,0x00,0x40,0x00]
+; GFX10-NEXT:    image_sample_c_d_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xac,0xf0,0x00,0x00,0x40,0x00]
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
 ; GFX10-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {

 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
 ; GFX10-LABEL: sample_c_cd_cl_1d:
 ; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_sample_c_cd_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xac,0xf1,0x00,0x00,0x40,0x00]
+; GFX10-NEXT:    image_sample_c_cd_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xac,0xf1,0x00,0x00,0x40,0x00]
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) ; encoding: [0x70,0x3f,0x8c,0xbf]
 ; GFX10-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.g16.ll

 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
 ; GFX10-LABEL: sample_c_d_cl_1d:
 ; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_sample_c_d_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
+; GFX10-NEXT:    image_sample_c_d_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {

 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp) {
 ; GFX10-LABEL: sample_c_cd_cl_1d:
 ; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_sample_c_cd_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
+; GFX10-NEXT:    image_sample_c_cd_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.1d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp) {

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.o.dim.ll

 ; GCN: image_sample_c_cl_o v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_c_cl_o_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, float %s, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.cl.o.1d.v4f32.f32(i32 15, i32 %offset, float %zcompare, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_c_cl_o_2d:
-; GCN: image_sample_c_cl_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}
+; GCN: image_sample_c_cl_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_c_cl_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, float %s, float %t, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.cl.o.2d.v4f32.f32(i32 15, i32 %offset, float %zcompare, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_b_o_1d:
 ; GCN: image_sample_b_o v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf{{$}}

 ; GCN: image_sample_c_b_o v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_c_b_o_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %bias, float %zcompare, float %s) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.b.o.1d.v4f32.f32.f32(i32 15, i32 %offset, float %bias, float %zcompare, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_c_b_o_2d:
-; GCN: image_sample_c_b_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}
+; GCN: image_sample_c_b_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_c_b_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %bias, float %zcompare, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.b.o.2d.v4f32.f32.f32(i32 15, i32 %offset, float %bias, float %zcompare, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_b_cl_o_1d:
 ; GCN: image_sample_b_cl_o v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_b_cl_o_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %bias, float %s, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.b.cl.o.1d.v4f32.f32.f32(i32 15, i32 %offset, float %bias, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_b_cl_o_2d:
-; GCN: image_sample_b_cl_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}
+; GCN: image_sample_b_cl_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_b_cl_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %bias, float %s, float %t, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.b.cl.o.2d.v4f32.f32.f32(i32 15, i32 %offset, float %bias, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_c_b_cl_o_1d:
-; GCN: image_sample_c_b_cl_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}
+; GCN: image_sample_c_b_cl_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_c_b_cl_o_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %bias, float %zcompare, float %s, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.b.cl.o.1d.v4f32.f32.f32(i32 15, i32 %offset, float %bias, float %zcompare, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_c_b_cl_o_2d:
 ; GCN: image_sample_c_b_cl_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}

 ; GCN: image_sample_d_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_d_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %dsdh, float %dtdh, float %dsdv, float %dtdv, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.d.o.2d.v4f32.f32.f32(i32 15, i32 %offset, float %dsdh, float %dtdh, float %dsdv, float %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_c_d_o_1d:
-; GCN: image_sample_c_d_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}
+; GCN: image_sample_c_d_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_c_d_o_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, float %dsdh, float %dsdv, float %s) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.o.1d.v4f32.f32.f32(i32 15, i32 %offset, float %zcompare, float %dsdh, float %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_c_d_o_2d:
 ; GCN: image_sample_c_d_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_c_d_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, float %dsdh, float %dtdh, float %dsdv, float %dtdv, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.o.2d.v4f32.f32.f32(i32 15, i32 %offset, float %zcompare, float %dsdh, float %dtdh, float %dsdv, float %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_d_cl_o_1d:
-; GCN: image_sample_d_cl_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}
+; GCN: image_sample_d_cl_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_d_cl_o_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %dsdh, float %dsdv, float %s, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.o.1d.v4f32.f32.f32(i32 15, i32 %offset, float %dsdh, float %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_d_cl_o_2d:
 ; GCN: image_sample_d_cl_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}

 ; GCN: image_sample_cd_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_cd_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %dsdh, float %dtdh, float %dsdv, float %dtdv, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.cd.o.2d.v4f32.f32.f32(i32 15, i32 %offset, float %dsdh, float %dtdh, float %dsdv, float %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_c_cd_o_1d:
-; GCN: image_sample_c_cd_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}
+; GCN: image_sample_c_cd_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_c_cd_o_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, float %dsdh, float %dsdv, float %s) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.o.1d.v4f32.f32.f32(i32 15, i32 %offset, float %zcompare, float %dsdh, float %dsdv, float %s, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_c_cd_o_2d:
 ; GCN: image_sample_c_cd_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_c_cd_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, float %dsdh, float %dtdh, float %dsdv, float %dtdv, float %s, float %t) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.o.2d.v4f32.f32.f32(i32 15, i32 %offset, float %zcompare, float %dsdh, float %dtdh, float %dsdv, float %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_cd_cl_o_1d:
-; GCN: image_sample_cd_cl_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}
+; GCN: image_sample_cd_cl_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_cd_cl_o_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %dsdh, float %dsdv, float %s, float %clamp) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.o.1d.v4f32.f32.f32(i32 15, i32 %offset, float %dsdh, float %dsdv, float %s, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_cd_cl_o_2d:
 ; GCN: image_sample_cd_cl_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}

 ; GCN: image_sample_c_l_o v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_c_l_o_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, float %s, float %lod) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.l.o.1d.v4f32.f32(i32 15, i32 %offset, float %zcompare, float %s, float %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_c_l_o_2d:
-; GCN: image_sample_c_l_o v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf{{$}}
+; GCN: image_sample_c_l_o v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf{{$}}
 define amdgpu_ps <4 x float> @sample_c_l_o_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, i32 %offset, float %zcompare, float %s, float %t, float %lod) {
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.l.o.2d.v4f32.f32(i32 15, i32 %offset, float %zcompare, float %s, float %t, float %lod, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 ; GCN-LABEL: {{^}}sample_lz_o_1d:
 ; GCN: image_sample_lz_o v[0:3], v[0:1], s[0:7], s[8:11] dmask:0xf{{$}}

llvm/test/CodeGen/AMDGPU/nsa-reassign.mir

 # RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs -run-pass greedy,amdgpu-nsa-reassign,virtregrewriter,si-shrink-instructions -o - %s | FileCheck -check-prefix=GCN %s
 
 --- |
   define amdgpu_kernel void @nsa_reassign() #0 { ret void }
   define amdgpu_kernel void @do_not_reassign_spill() #0 { ret void }
 
   attributes #0 = { "amdgpu-num-vgpr"="8" }
 ...
 
 # GCN-LABEL: name: nsa_reassign
-# GCN: IMAGE_SAMPLE_C_L_V1_V8_gfx10
+# GCN: IMAGE_SAMPLE_C_L_V1_V5_gfx10
 ---
 name:            nsa_reassign
 tracksRegLiveness: true
 machineFunctionInfo:
   stackPtrOffsetReg:  $sgpr32
 stack:
   - { id: 0, type: default, offset: 0, size: 4, alignment: 4 }
 registers:

llvm/test/CodeGen/AMDGPU/vgpr-tuple-allocation.ll

 
 ; GFX9: v_mov_b32_e32 v36, v16
 ; GFX9-NEXT: v_mov_b32_e32 v35, v15
 ; GFX9-NEXT: v_mov_b32_e32 v34, v14
 ; GFX9-NEXT: v_mov_b32_e32 v33, v13
 ; GFX9-NEXT: v_mov_b32_e32 v32, v12
 ; GFX9: ;;#ASMSTART
 ; GFX9-NEXT: ;;#ASMEND
-; GFX9: image_gather4_c_b_cl v[40:43], v[32:39], s[4:11], s[4:7] dmask:0x1
+; GFX9: image_gather4_c_b_cl v[40:43], v[32:36], s[4:11], s[4:7] dmask:0x1
 ; GFX9-NEXT: s_getpc_b64 s[4:5]
 ; GFX9-NEXT: s_add_u32 s4, s4, extern_func@gotpcrel32@lo+4
 ; GFX9-NEXT: s_addc_u32 s5, s5, extern_func@gotpcrel32@hi+12
 ; GFX9-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x0
 ; GFX9-NEXT: v_writelane_b32 v44, s30, 0
 ; GFX9: s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5]
 

 ; GFX10-NEXT: v_mov_b32_e32 v35, v15
 ; GFX10-NEXT: v_mov_b32_e32 v34, v14
 ; GFX10-NEXT: v_mov_b32_e32 v33, v13
 ; GFX10-NEXT: v_mov_b32_e32 v32, v12
 
 ; GFX10: ;;#ASMSTART
 ; GFX10-NEXT: ;;#ASMEND
 
-; GFX10: image_gather4_c_b_cl v[40:43], v[32:39], s[4:11], s[4:7] dmask:0x1 dim:SQ_RSRC_IMG_2D
+; GFX10: image_gather4_c_b_cl v[40:43], v[32:36], s[4:11], s[4:7] dmask:0x1 dim:SQ_RSRC_IMG_2D
 ; GFX10-NEXT: s_waitcnt_depctr 0xffe3
 ; GFX10-NEXT: s_getpc_b64 s[4:5]
 ; GFX10-NEXT: s_add_u32 s4, s4, extern_func@gotpcrel32@lo+4
 ; GFX10-NEXT: s_addc_u32 s5, s5, extern_func@gotpcrel32@hi+12
 ; GFX10: s_load_dwordx2 s[4:5], s[4:5], 0x0
 ; GFX10: s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT: s_swappc_b64 s[30:31], s[4:5]
 

 
 define <4 x float> @call_preserved_vgpr_tuple8(<8 x i32> %rsrc, <4 x i32> %samp, float %bias, float %zcompare, float %s, float %t, float %clamp) {
 ; The vgpr tuple8 operand in image_gather4_c_b_cl instruction needs to be preserved
 ; across the call and should get allcoated to 8 CSRs.
 ; Only the lower 5 sub-registers of the tuple are preserved.
 ; The upper 3 sub-registers are unused.
 
 ; GFX9-LABEL: call_preserved_vgpr_tuple8:
-; GFX9: buffer_store_dword v56, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
+; GFX9: buffer_store_dword v45, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
 ; GFX9: buffer_store_dword v40, off, s[0:3], s33 offset:16 ; 4-byte Folded Spill
 ; GFX9-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:12 ; 4-byte Folded Spill
 ; GFX9-NEXT: buffer_store_dword v42, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
 ; GFX9-NEXT: buffer_store_dword v43, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
 ; GFX9-NEXT: buffer_store_dword v44, off, s[0:3], s33 ; 4-byte Folded Spill
 
 ; GFX9: v_mov_b32_e32 v44, v16
 ; GFX9-NEXT: v_mov_b32_e32 v43, v15
 ; GFX9-NEXT: v_mov_b32_e32 v42, v14
 ; GFX9-NEXT: v_mov_b32_e32 v41, v13
 ; GFX9-NEXT: v_mov_b32_e32 v40, v12
 
-; GFX9: image_gather4_c_b_cl v[0:3], v[40:47], s[36:43], s[4:7] dmask:0x1
+; GFX9: image_gather4_c_b_cl v[0:3], v[40:44], s[36:43], s[4:7] dmask:0x1
 ; GFX9-NEXT: s_getpc_b64 s[4:5]
 ; GFX9-NEXT: s_add_u32 s4, s4, extern_func@gotpcrel32@lo+4
 ; GFX9-NEXT: s_addc_u32 s5, s5, extern_func@gotpcrel32@hi+12
 ; GFX9-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x0
 ; GFX9: s_waitcnt vmcnt(0)
 ; GFX9-NEXT: global_store_dwordx4 v[0:1], v[0:3], off
 ; GFX9-NEXT: s_waitcnt lgkmcnt(0)
 ; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5]
-; GFX9-NEXT: image_gather4_c_b_cl v[0:3], v[40:47], s[36:43], s[4:7] dmask:0x1
+; GFX9-NEXT: image_gather4_c_b_cl v[0:3], v[40:44], s[36:43], s[4:7] dmask:0x1
 
 ; GFX9: buffer_load_dword v44, off, s[0:3], s33 ; 4-byte Folded Reload
 ; GFX9-NEXT: buffer_load_dword v43, off, s[0:3], s33 offset:4 ; 4-byte Folded Reload
 ; GFX9-NEXT: buffer_load_dword v42, off, s[0:3], s33 offset:8 ; 4-byte Folded Reload
 ; GFX9-NEXT: buffer_load_dword v41, off, s[0:3], s33 offset:12 ; 4-byte Folded Reload
 ; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 offset:16 ; 4-byte Folded Reload
 
-; GFX9: buffer_load_dword v56, off, s[0:3], s32 offset:20 ; 4-byte Folded Reload
+; GFX9: buffer_load_dword v45, off, s[0:3], s32 offset:20 ; 4-byte Folded Reload
 ; GFX9: s_setpc_b64 s[4:5]
 ;
 ; GFX10-LABEL: call_preserved_vgpr_tuple8:
 ; GFX10: buffer_store_dword v45, off, s[0:3], s32 offset:20 ; 4-byte Folded Spill
 ; GFX10: buffer_store_dword v40, off, s[0:3], s33 offset:16 ; 4-byte Folded Spill
 ; GFX10-NEXT: buffer_store_dword v41, off, s[0:3], s33 offset:12 ; 4-byte Folded Spill
 ; GFX10-NEXT: buffer_store_dword v42, off, s[0:3], s33 offset:8 ; 4-byte Folded Spill
 ; GFX10-NEXT: buffer_store_dword v43, off, s[0:3], s33 offset:4 ; 4-byte Folded Spill
 ; GFX10-NEXT: buffer_store_dword v44, off, s[0:3], s33 ; 4-byte Folded Spill
 
 
-; GFX10:      image_gather4_c_b_cl v[0:3], v[12:19], s[36:43], s[4:7] dmask:0x1 dim:SQ_RSRC_IMG_2D
+; GFX10:      image_gather4_c_b_cl v[0:3], v[12:16], s[36:43], s[4:7] dmask:0x1 dim:SQ_RSRC_IMG_2D
 ; GFX10-NEXT: s_waitcnt_depctr 0xffe3
 ; GFX10-NEXT: s_getpc_b64 s[4:5]
 ; GFX10-NEXT: s_add_u32 s4, s4, extern_func@gotpcrel32@lo+4
 ; GFX10-NEXT: s_addc_u32 s5, s5, extern_func@gotpcrel32@hi+12
 ; GFX10-NEXT: v_mov_b32_e32 v40, v16
 ; GFX10-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x0
 ; GFX10-NEXT: v_mov_b32_e32 v41, v15
 ; GFX10-NEXT: v_mov_b32_e32 v42, v14

llvm/test/MC/AMDGPU/gfx10_asm_mimg.s

 ; 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]
 
 image_sample_d v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16
 ; GFX10: image_sample_d v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16 ; encoding: [0x00,0x0f,0x88,0xf0,0x00,0x00,0x40,0x40]
 
-image_sample_d v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
-; GFX10: image_sample_d v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; encoding: [0x08,0x0f,0x88,0xf0,0x00,0x00,0x40,0x40]
+image_sample_d v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
+; GFX10: image_sample_d v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; encoding: [0x08,0x0f,0x88,0xf0,0x00,0x00,0x40,0x40]
 
 image_sample_d v[0:3], [v0, v1, v2, v3, v4, v5, v6, v8], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_3D a16
 ; GFX10: image_sample_d v[0:3], [v0, v1, v2, v3, v4, v5, v6, v8], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_3D a16 ; encoding: [0x14,0x0f,0x88,0xf0,0x00,0x00,0x40,0x40,0x01,0x02,0x03,0x04,0x05,0x06,0x08,0x00]
 
 image_sample_c_d v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16
 ; GFX10: image_sample_c_d v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16 ; encoding: [0x00,0x0f,0xa8,0xf0,0x00,0x00,0x40,0x40]
 
 image_sample_c_d v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16

 ; GFX10: image_sample_c_d_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16 ; encoding: [0x00,0x0f,0xac,0xf0,0x00,0x00,0x40,0x40]
 
 image_sample_c_d_cl v[0:3], [v0, v1, v2, v3, v4, v5, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10: image_sample_c_d_cl v[0:3], [v0, v1, v2, v3, v4, v5, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; encoding: [0x0c,0x0f,0xac,0xf0,0x00,0x00,0x40,0x40,0x01,0x02,0x03,0x04,0x05,0x07,0x00,0x00]
 
 image_sample_cd v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16
 ; GFX10: image_sample_cd v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16 ; encoding: [0x00,0x0f,0xa0,0xf1,0x00,0x00,0x40,0x40]
 
-image_sample_cd v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
-; GFX10: image_sample_cd v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; encoding: [0x08,0x0f,0xa0,0xf1,0x00,0x00,0x40,0x40]
+image_sample_cd v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
+; GFX10: image_sample_cd v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; encoding: [0x08,0x0f,0xa0,0xf1,0x00,0x00,0x40,0x40]
 
 image_sample_c_cd v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16
 ; GFX10: image_sample_c_cd v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16 ; encoding: [0x00,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x40]
 
 image_sample_c_cd v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10: image_sample_c_cd v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; encoding: [0x08,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x40]
 
 image_sample_cd_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16

 ; GFX10: image_sample_c_d_g16 v[0:3], [v0, v1, v3, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0xa8,0xf0,0x00,0x00,0x40,0x00,0x01,0x03,0x05,0x06]
 
 image_sample_d_cl_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
 ; GFX10: image_sample_d_cl_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0x8c,0xf0,0x00,0x00,0x40,0x00]
 
 image_sample_d_cl_g16 v[0:3], [v0, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10: image_sample_d_cl_g16 v[0:3], [v0, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0x8c,0xf0,0x00,0x00,0x40,0x00,0x02,0x04,0x05,0x06]
 
-image_sample_c_d_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
-; GFX10: image_sample_c_d_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xac,0xf0,0x00,0x00,0x40,0x00]
+image_sample_c_d_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
+; GFX10: image_sample_c_d_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xac,0xf0,0x00,0x00,0x40,0x00]
 
 image_sample_c_d_cl_g16 v[0:3], [v0, v1, v3, v5, v6, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10: image_sample_c_d_cl_g16 v[0:3], [v0, v1, v3, v5, v6, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0d,0x0f,0xac,0xf0,0x00,0x00,0x40,0x00,0x01,0x03,0x05,0x06,0x07,0x00,0x00,0x00]
 
 image_sample_cd_g16 v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
 ; GFX10: image_sample_cd_g16 v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xa0,0xf1,0x00,0x00,0x40,0x00]
 
 image_sample_cd_g16 v[0:3], [v0, v2, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10: image_sample_cd_g16 v[0:3], [v0, v2, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0xa0,0xf1,0x00,0x00,0x40,0x00,0x02,0x04,0x05,0x00]
 
 image_sample_c_cd_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
 ; GFX10: image_sample_c_cd_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x00]
 
 image_sample_c_cd_g16 v[0:3], [v0, v1, v3, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10: image_sample_c_cd_g16 v[0:3], [v0, v1, v3, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x00,0x01,0x03,0x05,0x06]
 
 image_sample_cd_cl_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
 ; GFX10: image_sample_cd_cl_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xa4,0xf1,0x00,0x00,0x40,0x00]
 
 image_sample_cd_cl_g16 v[0:3], [v0, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10: image_sample_cd_cl_g16 v[0:3], [v0, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0xa4,0xf1,0x00,0x00,0x40,0x00,0x02,0x04,0x05,0x06]
 
-image_sample_c_cd_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
-; GFX10: image_sample_c_cd_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xac,0xf1,0x00,0x00,0x40,0x00]
+image_sample_c_cd_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D
+; GFX10: image_sample_c_cd_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xac,0xf1,0x00,0x00,0x40,0x00]
 
 image_sample_c_cd_cl_g16 v[0:3], [v0, v1, v3, v5, v6, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10: image_sample_c_cd_cl_g16 v[0:3], [v0, v1, v3, v5, v6, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0d,0x0f,0xac,0xf1,0x00,0x00,0x40,0x00,0x01,0x03,0x05,0x06,0x07,0x00,0x00,0x00]
 
 image_sample_c_d_o_g16 v0, [v0, v1, v2, v4, v6, v7, v8], s[0:7], s[8:11] dmask:0x4 dim:SQ_RSRC_IMG_2D_ARRAY
 ; GFX10: image_sample_c_d_o_g16 v0, [v0, v1, v2, v4, v6, v7, v8], s[0:7], s[8:11] dmask:0x4 dim:SQ_RSRC_IMG_2D_ARRAY ; encoding: [0x2d,0x04,0xe8,0xf0,0x00,0x00,0x40,0x00,0x01,0x02,0x04,0x06,0x07,0x08,0x00,0x00]
 
 image_sample_c_d_o_g16 v[0:1], [v0, v1, v2, v4, v6, v7, v8], s[0:7], s[8:11] dmask:0x6 dim:SQ_RSRC_IMG_2D_ARRAY
 ; GFX10: image_sample_c_d_o_g16 v[0:1], [v0, v1, v2, v4, v6, v7, v8], s[0:7], s[8:11] dmask:0x6 dim:SQ_RSRC_IMG_2D_ARRAY ; encoding: [0x2d,0x06,0xe8,0xf0,0x00,0x00,0x40,0x00,0x01,0x02,0x04,0x06,0x07,0x08,0x00,0x00]

llvm/test/MC/Disassembler/AMDGPU/gfx10_mimg.txt

 0x16,0x0f,0xa8,0xf0,0x08,0x10,0x25,0x03,0x09,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0x10,0x11,0x12,0x13,0x14
 
 # GFX10: image_sample_c_d_cl v[16:19], [v8, v9, v10, v11, v12, v13, v14, v15, v16], s[20:27], s[100:103] dmask:0xf dim:SQ_RSRC_IMG_CUBE ; encoding: [0x1c,0x0f,0xac,0xf0,0x08,0x10,0x25,0x03,0x09,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0x10]
 0x1c,0x0f,0xac,0xf0,0x08,0x10,0x25,0x03,0x09,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0x10
 
 # GFX10: image_sample_c_l v[16:19], [v8, v9, v10, v11, v12], s[20:27], s[100:103] dmask:0xf dim:SQ_RSRC_IMG_CUBE ; encoding: [0x1a,0x0f,0xb0,0xf0,0x08,0x10,0x25,0x03,0x09,0x0a,0x0b,0x0c]
 0x1a,0x0f,0xb0,0xf0,0x08,0x10,0x25,0x03,0x09,0x0a,0x0b,0x0c
 
-# GFX10: image_sample_c_b v[16:19], v[8:15], s[20:27], s[100:103] dmask:0xf dim:SQ_RSRC_IMG_CUBE ; encoding: [0x18,0x0f,0xb4,0xf0,0x08,0x10,0x25,0x03]
+# GFX10: image_sample_c_b v[16:19], v[8:12], s[20:27], s[100:103] dmask:0xf dim:SQ_RSRC_IMG_CUBE ; encoding: [0x18,0x0f,0xb4,0xf0,0x08,0x10,0x25,0x03]
 0x18,0x0f,0xb4,0xf0,0x08,0x10,0x25,0x03
 
 # TODO: address of this instruction is v[250:255], but this register class does
 #       not exist, and the next-larger size goes beyond the last register, so
 #       the disassembly is not adjusted properly
 # GFX10: image_sample_c_b_cl v16, v[250:252], s[20:27], s[100:103] dmask:0xf dim:SQ_RSRC_IMG_CUBE ; encoding: [0x18,0x0f,0xb8,0xf0,0xfa,0x10,0x25,0x03]
 0x18,0x0f,0xb8,0xf0,0xfa,0x10,0x25,0x03
 

llvm/test/MC/Disassembler/AMDGPU/mimg_gfx10.txt

 0x08,0x0f,0x02,0xf0,0x01,0x01,0x04,0x40
 
 # 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]
 0x08,0x05,0x03,0xf0,0x01,0x01,0x04,0x40
 
 # GFX10: image_sample_d v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16 ; encoding: [0x00,0x0f,0x88,0xf0,0x00,0x00,0x40,0x40]
 0x00,0x0f,0x88,0xf0,0x00,0x00,0x40,0x40
 
-# GFX10: image_sample_d v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; encoding: [0x08,0x0f,0x88,0xf0,0x00,0x00,0x40,0x40]
+# GFX10: image_sample_d v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; encoding: [0x08,0x0f,0x88,0xf0,0x00,0x00,0x40,0x40]
 0x08,0x0f,0x88,0xf0,0x00,0x00,0x40,0x40
 
 # GFX10: image_sample_d v[0:3], [v0, v1, v2, v3, v4, v5, v6, v8], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_3D a16 ; encoding: [0x14,0x0f,0x88,0xf0,0x00,0x00,0x40,0x40,0x01,0x02,0x03,0x04,0x05,0x06,0x08,0x00]
 0x14,0x0f,0x88,0xf0,0x00,0x00,0x40,0x40,0x01,0x02,0x03,0x04,0x05,0x06,0x08,0x00
 
 # GFX10: image_sample_c_d v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16 ; encoding: [0x00,0x0f,0xa8,0xf0,0x00,0x00,0x40,0x40]
 0x00,0x0f,0xa8,0xf0,0x00,0x00,0x40,0x40
 

 0x00,0x0f,0xac,0xf0,0x00,0x00,0x40,0x40
 
 # GFX10: image_sample_c_d_cl v[0:3], [v0, v1, v2, v3, v4, v5, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; encoding: [0x0c,0x0f,0xac,0xf0,0x00,0x00,0x40,0x40,0x01,0x02,0x03,0x04,0x05,0x07,0x00,0x00]
 0x0c,0x0f,0xac,0xf0,0x00,0x00,0x40,0x40,0x01,0x02,0x03,0x04,0x05,0x07,0x00,0x00
 
 # GFX10: image_sample_cd v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16 ; encoding: [0x00,0x0f,0xa0,0xf1,0x00,0x00,0x40,0x40]
 0x00,0x0f,0xa0,0xf1,0x00,0x00,0x40,0x40
 
-# GFX10: image_sample_cd v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; encoding: [0x08,0x0f,0xa0,0xf1,0x00,0x00,0x40,0x40]
+# GFX10: image_sample_cd v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; encoding: [0x08,0x0f,0xa0,0xf1,0x00,0x00,0x40,0x40]
 0x08,0x0f,0xa0,0xf1,0x00,0x00,0x40,0x40
 
 # GFX10: image_sample_c_cd v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16 ; encoding: [0x00,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x40]
 0x00,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x40
 
 # GFX10: image_sample_c_cd v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16 ; encoding: [0x08,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x40]
 0x08,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x40
 

 0x0b,0x0f,0xa8,0xf0,0x00,0x00,0x40,0x00,0x01,0x03,0x05,0x06
 
 # GFX10: image_sample_d_cl_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0x8c,0xf0,0x00,0x00,0x40,0x00]
 0x01,0x0f,0x8c,0xf0,0x00,0x00,0x40,0x00
 
 # GFX10: image_sample_d_cl_g16 v[0:3], [v0, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0x8c,0xf0,0x00,0x00,0x40,0x00,0x02,0x04,0x05,0x06]
 0x0b,0x0f,0x8c,0xf0,0x00,0x00,0x40,0x00,0x02,0x04,0x05,0x06
 
-# GFX10: image_sample_c_d_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xac,0xf0,0x00,0x00,0x40,0x00]
+# GFX10: image_sample_c_d_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xac,0xf0,0x00,0x00,0x40,0x00]
 0x01,0x0f,0xac,0xf0,0x00,0x00,0x40,0x00
 
 # GFX10: image_sample_c_d_cl_g16 v[0:3], [v0, v1, v3, v5, v6, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0d,0x0f,0xac,0xf0,0x00,0x00,0x40,0x00,0x01,0x03,0x05,0x06,0x07,0x00,0x00,0x00]
 0x0d,0x0f,0xac,0xf0,0x00,0x00,0x40,0x00,0x01,0x03,0x05,0x06,0x07,0x00,0x00,0x00
 
 # GFX10: image_sample_cd_g16 v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xa0,0xf1,0x00,0x00,0x40,0x00]
 0x01,0x0f,0xa0,0xf1,0x00,0x00,0x40,0x00
 
 # GFX10: image_sample_cd_g16 v[0:3], [v0, v2, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0xa0,0xf1,0x00,0x00,0x40,0x00,0x02,0x04,0x05,0x00]
 0x0b,0x0f,0xa0,0xf1,0x00,0x00,0x40,0x00,0x02,0x04,0x05,0x00
 
 # GFX10: image_sample_c_cd_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x00]
 0x01,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x00
 
 # GFX10: image_sample_c_cd_g16 v[0:3], [v0, v1, v3, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x00,0x01,0x03,0x05,0x06]
 0x0b,0x0f,0xa8,0xf1,0x00,0x00,0x40,0x00,0x01,0x03,0x05,0x06
 
 # GFX10: image_sample_cd_cl_g16 v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xa4,0xf1,0x00,0x00,0x40,0x00]
 0x01,0x0f,0xa4,0xf1,0x00,0x00,0x40,0x00
 
 # GFX10: image_sample_cd_cl_g16 v[0:3], [v0, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0b,0x0f,0xa4,0xf1,0x00,0x00,0x40,0x00,0x02,0x04,0x05,0x06]
 0x0b,0x0f,0xa4,0xf1,0x00,0x00,0x40,0x00,0x02,0x04,0x05,0x06
 
-# GFX10: image_sample_c_cd_cl_g16 v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xac,0xf1,0x00,0x00,0x40,0x00]
+# GFX10: image_sample_c_cd_cl_g16 v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D ; encoding: [0x01,0x0f,0xac,0xf1,0x00,0x00,0x40,0x00]
 0x01,0x0f,0xac,0xf1,0x00,0x00,0x40,0x00
 
 # GFX10: image_sample_c_cd_cl_g16 v[0:3], [v0, v1, v3, v5, v6, v7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D ; encoding: [0x0d,0x0f,0xac,0xf1,0x00,0x00,0x40,0x00,0x01,0x03,0x05,0x06,0x07,0x00,0x00,0x00]
 0x0d,0x0f,0xac,0xf1,0x00,0x00,0x40,0x00,0x01,0x03,0x05,0x06,0x07,0x00,0x00,0x00
 
 # GFX10: image_sample_c_d_o_g16 v0, [v0, v1, v2, v4, v6, v7, v8], s[0:7], s[8:11] dmask:0x4 dim:SQ_RSRC_IMG_2D_ARRAY ; encoding: [0x2d,0x04,0xe8,0xf0,0x00,0x00,0x40,0x00,0x01,0x02,0x04,0x06,0x07,0x08,0x00,0x00]
 0x2d,0x04,0xe8,0xf0,0x00,0x00,0x40,0x00,0x01,0x02,0x04,0x06,0x07,0x08,0x00,0x00
 
 # GFX10: image_sample_c_d_o_g16 v[0:1], [v0, v1, v2, v4, v6, v7, v8], s[0:7], s[8:11] dmask:0x6 dim:SQ_RSRC_IMG_2D_ARRAY ; encoding: [0x2d,0x06,0xe8,0xf0,0x00,0x00,0x40,0x00,0x01,0x02,0x04,0x06,0x07,0x08,0x00,0x00]
 0x2d,0x06,0xe8,0xf0,0x00,0x00,0x40,0x00,0x01,0x02,0x04,0x06,0x07,0x08,0x00,0x00