[llvm] r229605 - R600/SI: Add missing offset operand to buffer bothen
Matt Arsenault
Matthew.Arsenault at amd.com
Tue Feb 17 18:04:39 PST 2015
Author: arsenm
Date: Tue Feb 17 20:04:38 2015
New Revision: 229605
URL: http://llvm.org/viewvc/llvm-project?rev=229605&view=rev
Log:
R600/SI: Add missing offset operand to buffer bothen
Modified:
llvm/trunk/lib/Target/R600/SIInstrInfo.td
llvm/trunk/lib/Target/R600/SIInstructions.td
llvm/trunk/test/CodeGen/R600/llvm.SI.load.dword.ll
Modified: llvm/trunk/lib/Target/R600/SIInstrInfo.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/R600/SIInstrInfo.td?rev=229605&r1=229604&r2=229605&view=diff
==============================================================================
--- llvm/trunk/lib/Target/R600/SIInstrInfo.td (original)
+++ llvm/trunk/lib/Target/R600/SIInstrInfo.td Tue Feb 17 20:04:38 2015
@@ -1856,8 +1856,8 @@ multiclass MUBUF_Load_Helper <mubuf op,
let offen = 1, idxen = 1 in {
defm _BOTHEN : MUBUF_m <op, name#"_bothen", (outs regClass:$vdata),
(ins SReg_128:$srsrc, VReg_64:$vaddr,
- SCSrc_32:$soffset, glc:$glc, slc:$slc, tfe:$tfe),
- name#" $vdata, $vaddr, $srsrc, $soffset, idxen offen"#"$glc"#"$slc"#"$tfe", []>;
+ SCSrc_32:$soffset, mbuf_offset:$offset, glc:$glc, slc:$slc, tfe:$tfe),
+ name#" $vdata, $vaddr, $srsrc, $soffset idxen offen"#"$offset"#"$glc"#"$slc"#"$tfe", []>;
}
let offen = 0, idxen = 0, glc = 0, slc = 0, tfe = 0 in {
Modified: llvm/trunk/lib/Target/R600/SIInstructions.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/R600/SIInstructions.td?rev=229605&r1=229604&r2=229605&view=diff
==============================================================================
--- llvm/trunk/lib/Target/R600/SIInstructions.td (original)
+++ llvm/trunk/lib/Target/R600/SIInstructions.td Tue Feb 17 20:04:38 2015
@@ -2948,9 +2948,9 @@ multiclass MUBUF_Load_Dword <ValueType v
def : Pat <
(vt (int_SI_buffer_load_dword v4i32:$rsrc, v2i32:$vaddr, i32:$soffset,
- imm, 1, 1, imm:$glc, imm:$slc,
+ imm:$offset, 1, 1, imm:$glc, imm:$slc,
imm:$tfe)),
- (bothen $rsrc, $vaddr, $soffset, (as_i1imm $glc), (as_i1imm $slc),
+ (bothen $rsrc, $vaddr, $soffset, (as_i16imm $offset), (as_i1imm $glc), (as_i1imm $slc),
(as_i1imm $tfe))
>;
}
Modified: llvm/trunk/test/CodeGen/R600/llvm.SI.load.dword.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/R600/llvm.SI.load.dword.ll?rev=229605&r1=229604&r2=229605&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/R600/llvm.SI.load.dword.ll (original)
+++ llvm/trunk/test/CodeGen/R600/llvm.SI.load.dword.ll Tue Feb 17 20:04:38 2015
@@ -1,29 +1,41 @@
-;RUN: llc < %s -march=amdgcn -mcpu=verde -verify-machineinstrs | FileCheck %s
-;RUN: llc < %s -march=amdgcn -mcpu=tonga -verify-machineinstrs | FileCheck %s
+; RUN: llc -march=amdgcn -mcpu=verde -show-mc-encoding -verify-machineinstrs < %s | FileCheck %s
+; RUN: llc -march=amdgcn -mcpu=tonga -show-mc-encoding -verify-machineinstrs < %s | FileCheck %s
; Example of a simple geometry shader loading vertex attributes from the
; ESGS ring buffer
+; FIXME: Out of bounds immediate offset crashes
+
; CHECK-LABEL: {{^}}main:
-; CHECK: buffer_load_dword
-; CHECK: buffer_load_dword
-; CHECK: buffer_load_dword
-; CHECK: buffer_load_dword
+; CHECK: buffer_load_dword {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 glc slc
+; CHECK: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 offen glc slc
+; CHECK: buffer_load_dword {{v[0-9]+}}, {{v[0-9]+}}, {{s\[[0-9]+:[0-9]+\]}}, 0 idxen glc slc
+; CHECK: buffer_load_dword {{v[0-9]+}}, {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, 0 idxen offen glc slc
+; CHECK: s_movk_i32 [[K:s[0-9]+]], 0x4d2 ; encoding
+; CHECK: buffer_load_dword {{v[0-9]+}}, {{v\[[0-9]+:[0-9]+\]}}, {{s\[[0-9]+:[0-9]+\]}}, [[K]] idxen offen offset:65535 glc slc
-define void @main([17 x <16 x i8>] addrspace(2)* byval, [32 x <16 x i8>] addrspace(2)* byval, [16 x <32 x i8>] addrspace(2)* byval, [2 x <16 x i8>] addrspace(2)* byval, [17 x <16 x i8>] addrspace(2)* inreg, [17 x <16 x i8>] addrspace(2)* inreg, i32, i32, i32, i32) #0 {
+define void @main([17 x <16 x i8>] addrspace(2)* byval %arg, [32 x <16 x i8>] addrspace(2)* byval %arg1, [16 x <32 x i8>] addrspace(2)* byval %arg2, [2 x <16 x i8>] addrspace(2)* byval %arg3, [17 x <16 x i8>] addrspace(2)* inreg %arg4, [17 x <16 x i8>] addrspace(2)* inreg %arg5, i32 %arg6, i32 %arg7, i32 %arg8, i32 %arg9) #0 {
main_body:
- %10 = getelementptr [2 x <16 x i8>] addrspace(2)* %3, i64 0, i32 1
- %11 = load <16 x i8> addrspace(2)* %10, !tbaa !0
- %12 = shl i32 %6, 2
- %13 = call i32 @llvm.SI.buffer.load.dword.i32.i32(<16 x i8> %11, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 1, i32 0)
- %14 = bitcast i32 %13 to float
- %15 = call i32 @llvm.SI.buffer.load.dword.i32.i32(<16 x i8> %11, i32 %12, i32 0, i32 0, i32 1, i32 0, i32 1, i32 1, i32 0)
- %16 = bitcast i32 %15 to float
- %17 = call i32 @llvm.SI.buffer.load.dword.i32.i32(<16 x i8> %11, i32 %12, i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 0)
- %18 = bitcast i32 %17 to float
- %19 = call i32 @llvm.SI.buffer.load.dword.i32.v2i32(<16 x i8> %11, <2 x i32> <i32 0, i32 0>, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1, i32 0)
- %20 = bitcast i32 %19 to float
- call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %14, float %16, float %18, float %20)
+ %tmp = getelementptr [2 x <16 x i8>] addrspace(2)* %arg3, i64 0, i32 1
+ %tmp10 = load <16 x i8> addrspace(2)* %tmp, !tbaa !0
+ %tmp11 = shl i32 %arg6, 2
+ %tmp12 = call i32 @llvm.SI.buffer.load.dword.i32.i32(<16 x i8> %tmp10, i32 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 1, i32 0)
+ %tmp13 = bitcast i32 %tmp12 to float
+ %tmp14 = call i32 @llvm.SI.buffer.load.dword.i32.i32(<16 x i8> %tmp10, i32 %tmp11, i32 0, i32 0, i32 1, i32 0, i32 1, i32 1, i32 0)
+ %tmp15 = bitcast i32 %tmp14 to float
+ %tmp16 = call i32 @llvm.SI.buffer.load.dword.i32.i32(<16 x i8> %tmp10, i32 %tmp11, i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 0)
+ %tmp17 = bitcast i32 %tmp16 to float
+ %tmp18 = call i32 @llvm.SI.buffer.load.dword.i32.v2i32(<16 x i8> %tmp10, <2 x i32> zeroinitializer, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1, i32 0)
+ %tmp19 = bitcast i32 %tmp18 to float
+
+ %tmp20 = call i32 @llvm.SI.buffer.load.dword.i32.v2i32(<16 x i8> %tmp10, <2 x i32> zeroinitializer, i32 0, i32 123, i32 1, i32 1, i32 1, i32 1, i32 0)
+ %tmp21 = bitcast i32 %tmp20 to float
+
+ %tmp22 = call i32 @llvm.SI.buffer.load.dword.i32.v2i32(<16 x i8> %tmp10, <2 x i32> zeroinitializer, i32 1234, i32 65535, i32 1, i32 1, i32 1, i32 1, i32 0)
+ %tmp23 = bitcast i32 %tmp22 to float
+
+ call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %tmp13, float %tmp15, float %tmp17, float %tmp19)
+ call void @llvm.SI.export(i32 15, i32 0, i32 1, i32 12, i32 0, float %tmp21, float %tmp23, float %tmp23, float %tmp23)
ret void
}
More information about the llvm-commits
mailing list