maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/AMDGPU/flat-address-space.ll

228 lines
9.0 KiB
LLVM
Raw Normal View History

AMDGPU: Start selecting flat instruction offsets llvm-svn: 305201
2017-06-13 00:53:51 +08:00
; RUN: llc -O0 -mtriple=amdgcn-mesa-mesa3d -mcpu=bonaire < %s | FileCheck -check-prefixes=CHECK,CIVI %s
; RUN: llc -O0 -mtriple=amdgcn-mesa-mesa3d -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefixes=CHECK,CIVI %s
Enable FeatureFlatForGlobal on Volcanic Islands This switches to the workaround that HSA defaults to for the mesa path. This should be applied to the 4.0 branch. Patch by Vedran Miletić <vedran@miletic.net> llvm-svn: 292982
2017-01-25 06:02:15 +08:00
; RUN: llc -O0 -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -mattr=-flat-for-global < %s | FileCheck -check-prefixes=CHECK,HSA %s
AMDGPU: Start selecting flat instruction offsets llvm-svn: 305201
2017-06-13 00:53:51 +08:00
; RUN: llc -O0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -mattr=-flat-for-global < %s | FileCheck -check-prefixes=CHECK,HSA,GFX9 %s
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
; Disable optimizations in case there are optimizations added that
; specialize away generic pointer accesses.
; These testcases might become useless when there are optimizations to
; remove generic pointers.
R600: Call EmitFunctionHeader() in the AsmPrinter to populate the ELF symbol table llvm-svn: 218776
2014-10-02 01:15:17 +08:00
; CHECK-LABEL: {{^}}store_flat_i32:
AMDGPU: fix overlapping copies in copyPhysReg Summary: When copying aggregate registers within the same register class, there may be an overlap between source and destination that forces us to do the copy backwards. Do the simplest possible thing that guarantees the correct order of moves when there are overlaps, and does whatever when there is no overlap. (The last part forces some trivial adjustments to test cases.) Together with r255906, this fixes a VM fault in Unreal Elemental Demo. While at it, change the generation of kill and def flags to something that looks more reasonable. This method is used very late during compilation, so it probably doesn't matter in practice, and to be honest, I don't know if this change is actually correct because the semantics in connection with aggregate registers vs. sub-registers are not clear to me. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=93264 Reviewers: arsenm, tstellarAMD Subscribers: arsenm, llvm-commits Differential Revision: http://reviews.llvm.org/D15622 llvm-svn: 256072
2015-12-19 09:16:06 +08:00
; CHECK-DAG: s_load_dwordx2 s{{\[}}[[LO_SREG:[0-9]+]]:[[HI_SREG:[0-9]+]]],
; CHECK-DAG: s_load_dword s[[SDATA:[0-9]+]],
; CHECK: s_waitcnt lgkmcnt(0)
; CHECK-DAG: v_mov_b32_e32 v[[DATA:[0-9]+]], s[[SDATA]]
; CHECK-DAG: v_mov_b32_e32 v[[LO_VREG:[0-9]+]], s[[LO_SREG]]
; CHECK-DAG: v_mov_b32_e32 v[[HI_VREG:[0-9]+]], s[[HI_SREG]]
[AMDGPU] Assembler: Swap operands of flat_store instructions to match AMD assembler Historically, AMD internal sp3 assembler has flat_store* addr, data format. To match existing code and to enable reuse, change LLVM definitions to match. Also update MC and CodeGen tests. Differential Revision: http://reviews.llvm.org/D16927 Patch by: Nikolay Haustov llvm-svn: 260694
2016-02-13 01:57:54 +08:00
; CHECK: flat_store_dword v{{\[}}[[LO_VREG]]:[[HI_VREG]]{{\]}}, v[[DATA]]
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @store_flat_i32(i32 addrspace(1)* %gptr, i32 %x) #0 {
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%fptr = addrspacecast i32 addrspace(1)* %gptr to i32 addrspace(4)*
AMDGPU: Enable InferAddressSpaces llvm-svn: 294408
2017-02-08 14:16:04 +08:00
store volatile i32 %x, i32 addrspace(4)* %fptr, align 4
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
ret void
}
R600: Call EmitFunctionHeader() in the AsmPrinter to populate the ELF symbol table llvm-svn: 218776
2014-10-02 01:15:17 +08:00
; CHECK-LABEL: {{^}}store_flat_i64:
R600/SI: Change all instruction assembly names to lowercase. This matches the format produced by the AMD proprietary driver. //==================================================================// // Shell script for converting .ll test cases: (Pass the .ll files you want to convert to this script as arguments). //==================================================================// ; This was necessary on my system so that A-Z in sed would match only ; upper case. I'm not sure why. export LC_ALL='C' TEST_FILES="$*" MATCHES=`grep -v Patterns SIInstructions.td | grep -o '"[A-Z0-9_]\+["e]' | grep -o '[A-Z0-9_]\+' | sort -r` for f in $TEST_FILES; do # Check that there are SI tests: grep -q -e 'verde' -e 'bonaire' -e 'SI' -e 'tahiti' $f if [ $? -eq 0 ]; then for match in $MATCHES; do sed -i -e "s/\([ :]$match\)/\L\1/" $f done # Try to get check lines with partial instruction names sed -i 's/\(;[ ]*SI[A-Z\\-]*: \)\([A-Z_0-9]\+\)/\1\L\2/' $f fi done sed -i -e 's/bb0_1/BB0_1/g' ../../../test/CodeGen/R600/infinite-loop.ll sed -i -e 's/SI-NOT: bfe/SI-NOT: {{[^@]}}bfe/g'../../../test/CodeGen/R600/llvm.AMDGPU.bfe.*32.ll ../../../test/CodeGen/R600/sext-in-reg.ll sed -i -e 's/exp_IEEE/EXP_IEEE/g' ../../../test/CodeGen/R600/llvm.exp2.ll sed -i -e 's/numVgprs/NumVgprs/g' ../../../test/CodeGen/R600/register-count-comments.ll sed -i 's/\(; CHECK[-NOT]*: \)\([A-Z_0-9]\+\)/\1\L\2/' ../../../test/CodeGen/R600/select64.ll ../../../test/CodeGen/R600/sgpr-copy.ll //==================================================================// // Shell script for converting .td files (run this last) //==================================================================// export LC_ALL='C' sed -i -e '/Patterns/!s/\("[A-Z0-9_]\+[ "e]\)/\L\1/g' SIInstructions.td sed -i -e 's/"EXP/"exp/g' SIInstrInfo.td llvm-svn: 221350
2014-11-05 22:50:53 +08:00
; CHECK: flat_store_dwordx2
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @store_flat_i64(i64 addrspace(1)* %gptr, i64 %x) #0 {
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%fptr = addrspacecast i64 addrspace(1)* %gptr to i64 addrspace(4)*
AMDGPU: Enable InferAddressSpaces llvm-svn: 294408
2017-02-08 14:16:04 +08:00
store volatile i64 %x, i64 addrspace(4)* %fptr, align 8
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
ret void
}
R600: Call EmitFunctionHeader() in the AsmPrinter to populate the ELF symbol table llvm-svn: 218776
2014-10-02 01:15:17 +08:00
; CHECK-LABEL: {{^}}store_flat_v4i32:
R600/SI: Change all instruction assembly names to lowercase. This matches the format produced by the AMD proprietary driver. //==================================================================// // Shell script for converting .ll test cases: (Pass the .ll files you want to convert to this script as arguments). //==================================================================// ; This was necessary on my system so that A-Z in sed would match only ; upper case. I'm not sure why. export LC_ALL='C' TEST_FILES="$*" MATCHES=`grep -v Patterns SIInstructions.td | grep -o '"[A-Z0-9_]\+["e]' | grep -o '[A-Z0-9_]\+' | sort -r` for f in $TEST_FILES; do # Check that there are SI tests: grep -q -e 'verde' -e 'bonaire' -e 'SI' -e 'tahiti' $f if [ $? -eq 0 ]; then for match in $MATCHES; do sed -i -e "s/\([ :]$match\)/\L\1/" $f done # Try to get check lines with partial instruction names sed -i 's/\(;[ ]*SI[A-Z\\-]*: \)\([A-Z_0-9]\+\)/\1\L\2/' $f fi done sed -i -e 's/bb0_1/BB0_1/g' ../../../test/CodeGen/R600/infinite-loop.ll sed -i -e 's/SI-NOT: bfe/SI-NOT: {{[^@]}}bfe/g'../../../test/CodeGen/R600/llvm.AMDGPU.bfe.*32.ll ../../../test/CodeGen/R600/sext-in-reg.ll sed -i -e 's/exp_IEEE/EXP_IEEE/g' ../../../test/CodeGen/R600/llvm.exp2.ll sed -i -e 's/numVgprs/NumVgprs/g' ../../../test/CodeGen/R600/register-count-comments.ll sed -i 's/\(; CHECK[-NOT]*: \)\([A-Z_0-9]\+\)/\1\L\2/' ../../../test/CodeGen/R600/select64.ll ../../../test/CodeGen/R600/sgpr-copy.ll //==================================================================// // Shell script for converting .td files (run this last) //==================================================================// export LC_ALL='C' sed -i -e '/Patterns/!s/\("[A-Z0-9_]\+[ "e]\)/\L\1/g' SIInstructions.td sed -i -e 's/"EXP/"exp/g' SIInstrInfo.td llvm-svn: 221350
2014-11-05 22:50:53 +08:00
; CHECK: flat_store_dwordx4
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @store_flat_v4i32(<4 x i32> addrspace(1)* %gptr, <4 x i32> %x) #0 {
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%fptr = addrspacecast <4 x i32> addrspace(1)* %gptr to <4 x i32> addrspace(4)*
AMDGPU: Enable InferAddressSpaces llvm-svn: 294408
2017-02-08 14:16:04 +08:00
store volatile <4 x i32> %x, <4 x i32> addrspace(4)* %fptr, align 16
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
ret void
}
R600: Call EmitFunctionHeader() in the AsmPrinter to populate the ELF symbol table llvm-svn: 218776
2014-10-02 01:15:17 +08:00
; CHECK-LABEL: {{^}}store_flat_trunc_i16:
R600/SI: Change all instruction assembly names to lowercase. This matches the format produced by the AMD proprietary driver. //==================================================================// // Shell script for converting .ll test cases: (Pass the .ll files you want to convert to this script as arguments). //==================================================================// ; This was necessary on my system so that A-Z in sed would match only ; upper case. I'm not sure why. export LC_ALL='C' TEST_FILES="$*" MATCHES=`grep -v Patterns SIInstructions.td | grep -o '"[A-Z0-9_]\+["e]' | grep -o '[A-Z0-9_]\+' | sort -r` for f in $TEST_FILES; do # Check that there are SI tests: grep -q -e 'verde' -e 'bonaire' -e 'SI' -e 'tahiti' $f if [ $? -eq 0 ]; then for match in $MATCHES; do sed -i -e "s/\([ :]$match\)/\L\1/" $f done # Try to get check lines with partial instruction names sed -i 's/\(;[ ]*SI[A-Z\\-]*: \)\([A-Z_0-9]\+\)/\1\L\2/' $f fi done sed -i -e 's/bb0_1/BB0_1/g' ../../../test/CodeGen/R600/infinite-loop.ll sed -i -e 's/SI-NOT: bfe/SI-NOT: {{[^@]}}bfe/g'../../../test/CodeGen/R600/llvm.AMDGPU.bfe.*32.ll ../../../test/CodeGen/R600/sext-in-reg.ll sed -i -e 's/exp_IEEE/EXP_IEEE/g' ../../../test/CodeGen/R600/llvm.exp2.ll sed -i -e 's/numVgprs/NumVgprs/g' ../../../test/CodeGen/R600/register-count-comments.ll sed -i 's/\(; CHECK[-NOT]*: \)\([A-Z_0-9]\+\)/\1\L\2/' ../../../test/CodeGen/R600/select64.ll ../../../test/CodeGen/R600/sgpr-copy.ll //==================================================================// // Shell script for converting .td files (run this last) //==================================================================// export LC_ALL='C' sed -i -e '/Patterns/!s/\("[A-Z0-9_]\+[ "e]\)/\L\1/g' SIInstructions.td sed -i -e 's/"EXP/"exp/g' SIInstrInfo.td llvm-svn: 221350
2014-11-05 22:50:53 +08:00
; CHECK: flat_store_short
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @store_flat_trunc_i16(i16 addrspace(1)* %gptr, i32 %x) #0 {
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%fptr = addrspacecast i16 addrspace(1)* %gptr to i16 addrspace(4)*
%y = trunc i32 %x to i16
AMDGPU: Enable InferAddressSpaces llvm-svn: 294408
2017-02-08 14:16:04 +08:00
store volatile i16 %y, i16 addrspace(4)* %fptr, align 2
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
ret void
}
R600: Call EmitFunctionHeader() in the AsmPrinter to populate the ELF symbol table llvm-svn: 218776
2014-10-02 01:15:17 +08:00
; CHECK-LABEL: {{^}}store_flat_trunc_i8:
R600/SI: Change all instruction assembly names to lowercase. This matches the format produced by the AMD proprietary driver. //==================================================================// // Shell script for converting .ll test cases: (Pass the .ll files you want to convert to this script as arguments). //==================================================================// ; This was necessary on my system so that A-Z in sed would match only ; upper case. I'm not sure why. export LC_ALL='C' TEST_FILES="$*" MATCHES=`grep -v Patterns SIInstructions.td | grep -o '"[A-Z0-9_]\+["e]' | grep -o '[A-Z0-9_]\+' | sort -r` for f in $TEST_FILES; do # Check that there are SI tests: grep -q -e 'verde' -e 'bonaire' -e 'SI' -e 'tahiti' $f if [ $? -eq 0 ]; then for match in $MATCHES; do sed -i -e "s/\([ :]$match\)/\L\1/" $f done # Try to get check lines with partial instruction names sed -i 's/\(;[ ]*SI[A-Z\\-]*: \)\([A-Z_0-9]\+\)/\1\L\2/' $f fi done sed -i -e 's/bb0_1/BB0_1/g' ../../../test/CodeGen/R600/infinite-loop.ll sed -i -e 's/SI-NOT: bfe/SI-NOT: {{[^@]}}bfe/g'../../../test/CodeGen/R600/llvm.AMDGPU.bfe.*32.ll ../../../test/CodeGen/R600/sext-in-reg.ll sed -i -e 's/exp_IEEE/EXP_IEEE/g' ../../../test/CodeGen/R600/llvm.exp2.ll sed -i -e 's/numVgprs/NumVgprs/g' ../../../test/CodeGen/R600/register-count-comments.ll sed -i 's/\(; CHECK[-NOT]*: \)\([A-Z_0-9]\+\)/\1\L\2/' ../../../test/CodeGen/R600/select64.ll ../../../test/CodeGen/R600/sgpr-copy.ll //==================================================================// // Shell script for converting .td files (run this last) //==================================================================// export LC_ALL='C' sed -i -e '/Patterns/!s/\("[A-Z0-9_]\+[ "e]\)/\L\1/g' SIInstructions.td sed -i -e 's/"EXP/"exp/g' SIInstrInfo.td llvm-svn: 221350
2014-11-05 22:50:53 +08:00
; CHECK: flat_store_byte
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @store_flat_trunc_i8(i8 addrspace(1)* %gptr, i32 %x) #0 {
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%fptr = addrspacecast i8 addrspace(1)* %gptr to i8 addrspace(4)*
%y = trunc i32 %x to i8
AMDGPU: Enable InferAddressSpaces llvm-svn: 294408
2017-02-08 14:16:04 +08:00
store volatile i8 %y, i8 addrspace(4)* %fptr, align 2
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
ret void
}
Fix CHECK directives that weren't checking. llvm-svn: 246485
2015-09-01 05:10:35 +08:00
; CHECK-LABEL: load_flat_i32:
R600/SI: Change all instruction assembly names to lowercase. This matches the format produced by the AMD proprietary driver. //==================================================================// // Shell script for converting .ll test cases: (Pass the .ll files you want to convert to this script as arguments). //==================================================================// ; This was necessary on my system so that A-Z in sed would match only ; upper case. I'm not sure why. export LC_ALL='C' TEST_FILES="$*" MATCHES=`grep -v Patterns SIInstructions.td | grep -o '"[A-Z0-9_]\+["e]' | grep -o '[A-Z0-9_]\+' | sort -r` for f in $TEST_FILES; do # Check that there are SI tests: grep -q -e 'verde' -e 'bonaire' -e 'SI' -e 'tahiti' $f if [ $? -eq 0 ]; then for match in $MATCHES; do sed -i -e "s/\([ :]$match\)/\L\1/" $f done # Try to get check lines with partial instruction names sed -i 's/\(;[ ]*SI[A-Z\\-]*: \)\([A-Z_0-9]\+\)/\1\L\2/' $f fi done sed -i -e 's/bb0_1/BB0_1/g' ../../../test/CodeGen/R600/infinite-loop.ll sed -i -e 's/SI-NOT: bfe/SI-NOT: {{[^@]}}bfe/g'../../../test/CodeGen/R600/llvm.AMDGPU.bfe.*32.ll ../../../test/CodeGen/R600/sext-in-reg.ll sed -i -e 's/exp_IEEE/EXP_IEEE/g' ../../../test/CodeGen/R600/llvm.exp2.ll sed -i -e 's/numVgprs/NumVgprs/g' ../../../test/CodeGen/R600/register-count-comments.ll sed -i 's/\(; CHECK[-NOT]*: \)\([A-Z_0-9]\+\)/\1\L\2/' ../../../test/CodeGen/R600/select64.ll ../../../test/CodeGen/R600/sgpr-copy.ll //==================================================================// // Shell script for converting .td files (run this last) //==================================================================// export LC_ALL='C' sed -i -e '/Patterns/!s/\("[A-Z0-9_]\+[ "e]\)/\L\1/g' SIInstructions.td sed -i -e 's/"EXP/"exp/g' SIInstrInfo.td llvm-svn: 221350
2014-11-05 22:50:53 +08:00
; CHECK: flat_load_dword
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @load_flat_i32(i32 addrspace(1)* noalias %out, i32 addrspace(1)* noalias %gptr) #0 {
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%fptr = addrspacecast i32 addrspace(1)* %gptr to i32 addrspace(4)*
AMDGPU: Enable InferAddressSpaces llvm-svn: 294408
2017-02-08 14:16:04 +08:00
%fload = load volatile i32, i32 addrspace(4)* %fptr, align 4
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
store i32 %fload, i32 addrspace(1)* %out, align 4
ret void
}
Fix CHECK directives that weren't checking. llvm-svn: 246485
2015-09-01 05:10:35 +08:00
; CHECK-LABEL: load_flat_i64:
R600/SI: Change all instruction assembly names to lowercase. This matches the format produced by the AMD proprietary driver. //==================================================================// // Shell script for converting .ll test cases: (Pass the .ll files you want to convert to this script as arguments). //==================================================================// ; This was necessary on my system so that A-Z in sed would match only ; upper case. I'm not sure why. export LC_ALL='C' TEST_FILES="$*" MATCHES=`grep -v Patterns SIInstructions.td | grep -o '"[A-Z0-9_]\+["e]' | grep -o '[A-Z0-9_]\+' | sort -r` for f in $TEST_FILES; do # Check that there are SI tests: grep -q -e 'verde' -e 'bonaire' -e 'SI' -e 'tahiti' $f if [ $? -eq 0 ]; then for match in $MATCHES; do sed -i -e "s/\([ :]$match\)/\L\1/" $f done # Try to get check lines with partial instruction names sed -i 's/\(;[ ]*SI[A-Z\\-]*: \)\([A-Z_0-9]\+\)/\1\L\2/' $f fi done sed -i -e 's/bb0_1/BB0_1/g' ../../../test/CodeGen/R600/infinite-loop.ll sed -i -e 's/SI-NOT: bfe/SI-NOT: {{[^@]}}bfe/g'../../../test/CodeGen/R600/llvm.AMDGPU.bfe.*32.ll ../../../test/CodeGen/R600/sext-in-reg.ll sed -i -e 's/exp_IEEE/EXP_IEEE/g' ../../../test/CodeGen/R600/llvm.exp2.ll sed -i -e 's/numVgprs/NumVgprs/g' ../../../test/CodeGen/R600/register-count-comments.ll sed -i 's/\(; CHECK[-NOT]*: \)\([A-Z_0-9]\+\)/\1\L\2/' ../../../test/CodeGen/R600/select64.ll ../../../test/CodeGen/R600/sgpr-copy.ll //==================================================================// // Shell script for converting .td files (run this last) //==================================================================// export LC_ALL='C' sed -i -e '/Patterns/!s/\("[A-Z0-9_]\+[ "e]\)/\L\1/g' SIInstructions.td sed -i -e 's/"EXP/"exp/g' SIInstrInfo.td llvm-svn: 221350
2014-11-05 22:50:53 +08:00
; CHECK: flat_load_dwordx2
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @load_flat_i64(i64 addrspace(1)* noalias %out, i64 addrspace(1)* noalias %gptr) #0 {
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%fptr = addrspacecast i64 addrspace(1)* %gptr to i64 addrspace(4)*
AMDGPU: Enable InferAddressSpaces llvm-svn: 294408
2017-02-08 14:16:04 +08:00
%fload = load volatile i64, i64 addrspace(4)* %fptr, align 8
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
store i64 %fload, i64 addrspace(1)* %out, align 8
ret void
}
Fix CHECK directives that weren't checking. llvm-svn: 246485
2015-09-01 05:10:35 +08:00
; CHECK-LABEL: load_flat_v4i32:
R600/SI: Change all instruction assembly names to lowercase. This matches the format produced by the AMD proprietary driver. //==================================================================// // Shell script for converting .ll test cases: (Pass the .ll files you want to convert to this script as arguments). //==================================================================// ; This was necessary on my system so that A-Z in sed would match only ; upper case. I'm not sure why. export LC_ALL='C' TEST_FILES="$*" MATCHES=`grep -v Patterns SIInstructions.td | grep -o '"[A-Z0-9_]\+["e]' | grep -o '[A-Z0-9_]\+' | sort -r` for f in $TEST_FILES; do # Check that there are SI tests: grep -q -e 'verde' -e 'bonaire' -e 'SI' -e 'tahiti' $f if [ $? -eq 0 ]; then for match in $MATCHES; do sed -i -e "s/\([ :]$match\)/\L\1/" $f done # Try to get check lines with partial instruction names sed -i 's/\(;[ ]*SI[A-Z\\-]*: \)\([A-Z_0-9]\+\)/\1\L\2/' $f fi done sed -i -e 's/bb0_1/BB0_1/g' ../../../test/CodeGen/R600/infinite-loop.ll sed -i -e 's/SI-NOT: bfe/SI-NOT: {{[^@]}}bfe/g'../../../test/CodeGen/R600/llvm.AMDGPU.bfe.*32.ll ../../../test/CodeGen/R600/sext-in-reg.ll sed -i -e 's/exp_IEEE/EXP_IEEE/g' ../../../test/CodeGen/R600/llvm.exp2.ll sed -i -e 's/numVgprs/NumVgprs/g' ../../../test/CodeGen/R600/register-count-comments.ll sed -i 's/\(; CHECK[-NOT]*: \)\([A-Z_0-9]\+\)/\1\L\2/' ../../../test/CodeGen/R600/select64.ll ../../../test/CodeGen/R600/sgpr-copy.ll //==================================================================// // Shell script for converting .td files (run this last) //==================================================================// export LC_ALL='C' sed -i -e '/Patterns/!s/\("[A-Z0-9_]\+[ "e]\)/\L\1/g' SIInstructions.td sed -i -e 's/"EXP/"exp/g' SIInstrInfo.td llvm-svn: 221350
2014-11-05 22:50:53 +08:00
; CHECK: flat_load_dwordx4
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @load_flat_v4i32(<4 x i32> addrspace(1)* noalias %out, <4 x i32> addrspace(1)* noalias %gptr) #0 {
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%fptr = addrspacecast <4 x i32> addrspace(1)* %gptr to <4 x i32> addrspace(4)*
AMDGPU: Enable InferAddressSpaces llvm-svn: 294408
2017-02-08 14:16:04 +08:00
%fload = load volatile <4 x i32>, <4 x i32> addrspace(4)* %fptr, align 32
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
store <4 x i32> %fload, <4 x i32> addrspace(1)* %out, align 8
ret void
}
Fix CHECK directives that weren't checking. llvm-svn: 246485
2015-09-01 05:10:35 +08:00
; CHECK-LABEL: sextload_flat_i8:
R600/SI: Change all instruction assembly names to lowercase. This matches the format produced by the AMD proprietary driver. //==================================================================// // Shell script for converting .ll test cases: (Pass the .ll files you want to convert to this script as arguments). //==================================================================// ; This was necessary on my system so that A-Z in sed would match only ; upper case. I'm not sure why. export LC_ALL='C' TEST_FILES="$*" MATCHES=`grep -v Patterns SIInstructions.td | grep -o '"[A-Z0-9_]\+["e]' | grep -o '[A-Z0-9_]\+' | sort -r` for f in $TEST_FILES; do # Check that there are SI tests: grep -q -e 'verde' -e 'bonaire' -e 'SI' -e 'tahiti' $f if [ $? -eq 0 ]; then for match in $MATCHES; do sed -i -e "s/\([ :]$match\)/\L\1/" $f done # Try to get check lines with partial instruction names sed -i 's/\(;[ ]*SI[A-Z\\-]*: \)\([A-Z_0-9]\+\)/\1\L\2/' $f fi done sed -i -e 's/bb0_1/BB0_1/g' ../../../test/CodeGen/R600/infinite-loop.ll sed -i -e 's/SI-NOT: bfe/SI-NOT: {{[^@]}}bfe/g'../../../test/CodeGen/R600/llvm.AMDGPU.bfe.*32.ll ../../../test/CodeGen/R600/sext-in-reg.ll sed -i -e 's/exp_IEEE/EXP_IEEE/g' ../../../test/CodeGen/R600/llvm.exp2.ll sed -i -e 's/numVgprs/NumVgprs/g' ../../../test/CodeGen/R600/register-count-comments.ll sed -i 's/\(; CHECK[-NOT]*: \)\([A-Z_0-9]\+\)/\1\L\2/' ../../../test/CodeGen/R600/select64.ll ../../../test/CodeGen/R600/sgpr-copy.ll //==================================================================// // Shell script for converting .td files (run this last) //==================================================================// export LC_ALL='C' sed -i -e '/Patterns/!s/\("[A-Z0-9_]\+[ "e]\)/\L\1/g' SIInstructions.td sed -i -e 's/"EXP/"exp/g' SIInstrInfo.td llvm-svn: 221350
2014-11-05 22:50:53 +08:00
; CHECK: flat_load_sbyte
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @sextload_flat_i8(i32 addrspace(1)* noalias %out, i8 addrspace(1)* noalias %gptr) #0 {
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%fptr = addrspacecast i8 addrspace(1)* %gptr to i8 addrspace(4)*
AMDGPU: Enable InferAddressSpaces llvm-svn: 294408
2017-02-08 14:16:04 +08:00
%fload = load volatile i8, i8 addrspace(4)* %fptr, align 4
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%ext = sext i8 %fload to i32
store i32 %ext, i32 addrspace(1)* %out, align 4
ret void
}
Fix CHECK directives that weren't checking. llvm-svn: 246485
2015-09-01 05:10:35 +08:00
; CHECK-LABEL: zextload_flat_i8:
R600/SI: Change all instruction assembly names to lowercase. This matches the format produced by the AMD proprietary driver. //==================================================================// // Shell script for converting .ll test cases: (Pass the .ll files you want to convert to this script as arguments). //==================================================================// ; This was necessary on my system so that A-Z in sed would match only ; upper case. I'm not sure why. export LC_ALL='C' TEST_FILES="$*" MATCHES=`grep -v Patterns SIInstructions.td | grep -o '"[A-Z0-9_]\+["e]' | grep -o '[A-Z0-9_]\+' | sort -r` for f in $TEST_FILES; do # Check that there are SI tests: grep -q -e 'verde' -e 'bonaire' -e 'SI' -e 'tahiti' $f if [ $? -eq 0 ]; then for match in $MATCHES; do sed -i -e "s/\([ :]$match\)/\L\1/" $f done # Try to get check lines with partial instruction names sed -i 's/\(;[ ]*SI[A-Z\\-]*: \)\([A-Z_0-9]\+\)/\1\L\2/' $f fi done sed -i -e 's/bb0_1/BB0_1/g' ../../../test/CodeGen/R600/infinite-loop.ll sed -i -e 's/SI-NOT: bfe/SI-NOT: {{[^@]}}bfe/g'../../../test/CodeGen/R600/llvm.AMDGPU.bfe.*32.ll ../../../test/CodeGen/R600/sext-in-reg.ll sed -i -e 's/exp_IEEE/EXP_IEEE/g' ../../../test/CodeGen/R600/llvm.exp2.ll sed -i -e 's/numVgprs/NumVgprs/g' ../../../test/CodeGen/R600/register-count-comments.ll sed -i 's/\(; CHECK[-NOT]*: \)\([A-Z_0-9]\+\)/\1\L\2/' ../../../test/CodeGen/R600/select64.ll ../../../test/CodeGen/R600/sgpr-copy.ll //==================================================================// // Shell script for converting .td files (run this last) //==================================================================// export LC_ALL='C' sed -i -e '/Patterns/!s/\("[A-Z0-9_]\+[ "e]\)/\L\1/g' SIInstructions.td sed -i -e 's/"EXP/"exp/g' SIInstrInfo.td llvm-svn: 221350
2014-11-05 22:50:53 +08:00
; CHECK: flat_load_ubyte
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @zextload_flat_i8(i32 addrspace(1)* noalias %out, i8 addrspace(1)* noalias %gptr) #0 {
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%fptr = addrspacecast i8 addrspace(1)* %gptr to i8 addrspace(4)*
AMDGPU: Enable InferAddressSpaces llvm-svn: 294408
2017-02-08 14:16:04 +08:00
%fload = load volatile i8, i8 addrspace(4)* %fptr, align 4
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%ext = zext i8 %fload to i32
store i32 %ext, i32 addrspace(1)* %out, align 4
ret void
}
Fix CHECK directives that weren't checking. llvm-svn: 246485
2015-09-01 05:10:35 +08:00
; CHECK-LABEL: sextload_flat_i16:
R600/SI: Change all instruction assembly names to lowercase. This matches the format produced by the AMD proprietary driver. //==================================================================// // Shell script for converting .ll test cases: (Pass the .ll files you want to convert to this script as arguments). //==================================================================// ; This was necessary on my system so that A-Z in sed would match only ; upper case. I'm not sure why. export LC_ALL='C' TEST_FILES="$*" MATCHES=`grep -v Patterns SIInstructions.td | grep -o '"[A-Z0-9_]\+["e]' | grep -o '[A-Z0-9_]\+' | sort -r` for f in $TEST_FILES; do # Check that there are SI tests: grep -q -e 'verde' -e 'bonaire' -e 'SI' -e 'tahiti' $f if [ $? -eq 0 ]; then for match in $MATCHES; do sed -i -e "s/\([ :]$match\)/\L\1/" $f done # Try to get check lines with partial instruction names sed -i 's/\(;[ ]*SI[A-Z\\-]*: \)\([A-Z_0-9]\+\)/\1\L\2/' $f fi done sed -i -e 's/bb0_1/BB0_1/g' ../../../test/CodeGen/R600/infinite-loop.ll sed -i -e 's/SI-NOT: bfe/SI-NOT: {{[^@]}}bfe/g'../../../test/CodeGen/R600/llvm.AMDGPU.bfe.*32.ll ../../../test/CodeGen/R600/sext-in-reg.ll sed -i -e 's/exp_IEEE/EXP_IEEE/g' ../../../test/CodeGen/R600/llvm.exp2.ll sed -i -e 's/numVgprs/NumVgprs/g' ../../../test/CodeGen/R600/register-count-comments.ll sed -i 's/\(; CHECK[-NOT]*: \)\([A-Z_0-9]\+\)/\1\L\2/' ../../../test/CodeGen/R600/select64.ll ../../../test/CodeGen/R600/sgpr-copy.ll //==================================================================// // Shell script for converting .td files (run this last) //==================================================================// export LC_ALL='C' sed -i -e '/Patterns/!s/\("[A-Z0-9_]\+[ "e]\)/\L\1/g' SIInstructions.td sed -i -e 's/"EXP/"exp/g' SIInstrInfo.td llvm-svn: 221350
2014-11-05 22:50:53 +08:00
; CHECK: flat_load_sshort
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @sextload_flat_i16(i32 addrspace(1)* noalias %out, i16 addrspace(1)* noalias %gptr) #0 {
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%fptr = addrspacecast i16 addrspace(1)* %gptr to i16 addrspace(4)*
AMDGPU: Enable InferAddressSpaces llvm-svn: 294408
2017-02-08 14:16:04 +08:00
%fload = load volatile i16, i16 addrspace(4)* %fptr, align 4
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%ext = sext i16 %fload to i32
store i32 %ext, i32 addrspace(1)* %out, align 4
ret void
}
Fix CHECK directives that weren't checking. llvm-svn: 246485
2015-09-01 05:10:35 +08:00
; CHECK-LABEL: zextload_flat_i16:
R600/SI: Change all instruction assembly names to lowercase. This matches the format produced by the AMD proprietary driver. //==================================================================// // Shell script for converting .ll test cases: (Pass the .ll files you want to convert to this script as arguments). //==================================================================// ; This was necessary on my system so that A-Z in sed would match only ; upper case. I'm not sure why. export LC_ALL='C' TEST_FILES="$*" MATCHES=`grep -v Patterns SIInstructions.td | grep -o '"[A-Z0-9_]\+["e]' | grep -o '[A-Z0-9_]\+' | sort -r` for f in $TEST_FILES; do # Check that there are SI tests: grep -q -e 'verde' -e 'bonaire' -e 'SI' -e 'tahiti' $f if [ $? -eq 0 ]; then for match in $MATCHES; do sed -i -e "s/\([ :]$match\)/\L\1/" $f done # Try to get check lines with partial instruction names sed -i 's/\(;[ ]*SI[A-Z\\-]*: \)\([A-Z_0-9]\+\)/\1\L\2/' $f fi done sed -i -e 's/bb0_1/BB0_1/g' ../../../test/CodeGen/R600/infinite-loop.ll sed -i -e 's/SI-NOT: bfe/SI-NOT: {{[^@]}}bfe/g'../../../test/CodeGen/R600/llvm.AMDGPU.bfe.*32.ll ../../../test/CodeGen/R600/sext-in-reg.ll sed -i -e 's/exp_IEEE/EXP_IEEE/g' ../../../test/CodeGen/R600/llvm.exp2.ll sed -i -e 's/numVgprs/NumVgprs/g' ../../../test/CodeGen/R600/register-count-comments.ll sed -i 's/\(; CHECK[-NOT]*: \)\([A-Z_0-9]\+\)/\1\L\2/' ../../../test/CodeGen/R600/select64.ll ../../../test/CodeGen/R600/sgpr-copy.ll //==================================================================// // Shell script for converting .td files (run this last) //==================================================================// export LC_ALL='C' sed -i -e '/Patterns/!s/\("[A-Z0-9_]\+[ "e]\)/\L\1/g' SIInstructions.td sed -i -e 's/"EXP/"exp/g' SIInstrInfo.td llvm-svn: 221350
2014-11-05 22:50:53 +08:00
; CHECK: flat_load_ushort
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @zextload_flat_i16(i32 addrspace(1)* noalias %out, i16 addrspace(1)* noalias %gptr) #0 {
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%fptr = addrspacecast i16 addrspace(1)* %gptr to i16 addrspace(4)*
AMDGPU: Enable InferAddressSpaces llvm-svn: 294408
2017-02-08 14:16:04 +08:00
%fload = load volatile i16, i16 addrspace(4)* %fptr, align 4
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
%ext = zext i16 %fload to i32
store i32 %ext, i32 addrspace(1)* %out, align 4
ret void
}
AMDGPU/SI: Don't allow unaligned scratch access Summary: The hardware doesn't support this. Reviewers: arsenm Subscribers: kzhuravl, wdng, nhaehnle, yaxunl, llvm-commits, tony-tye Differential Revision: https://reviews.llvm.org/D25523 llvm-svn: 284257
2016-10-15 02:10:39 +08:00
; CHECK-LABEL: flat_scratch_unaligned_load:
; CHECK: flat_load_ubyte
; CHECK: flat_load_ubyte
; CHECK: flat_load_ubyte
; CHECK: flat_load_ubyte
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @flat_scratch_unaligned_load() {
AMDGPU/SI: Don't allow unaligned scratch access Summary: The hardware doesn't support this. Reviewers: arsenm Subscribers: kzhuravl, wdng, nhaehnle, yaxunl, llvm-commits, tony-tye Differential Revision: https://reviews.llvm.org/D25523 llvm-svn: 284257
2016-10-15 02:10:39 +08:00
%scratch = alloca i32
%fptr = addrspacecast i32* %scratch to i32 addrspace(4)*
%ld = load volatile i32, i32 addrspace(4)* %fptr, align 1
ret void
}
; CHECK-LABEL: flat_scratch_unaligned_store:
; CHECK: flat_store_byte
; CHECK: flat_store_byte
; CHECK: flat_store_byte
; CHECK: flat_store_byte
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @flat_scratch_unaligned_store() {
AMDGPU/SI: Don't allow unaligned scratch access Summary: The hardware doesn't support this. Reviewers: arsenm Subscribers: kzhuravl, wdng, nhaehnle, yaxunl, llvm-commits, tony-tye Differential Revision: https://reviews.llvm.org/D25523 llvm-svn: 284257
2016-10-15 02:10:39 +08:00
%scratch = alloca i32
%fptr = addrspacecast i32* %scratch to i32 addrspace(4)*
store volatile i32 0, i32 addrspace(4)* %fptr, align 1
ret void
}
AMDGPU/SI: Don't emit multi-dword flat memory ops when they might access scratch Summary: A single flat memory operations that might access the scratch buffer can only access MaxPrivateElementSize bytes. Reviewers: arsenm Subscribers: kzhuravl, wdng, nhaehnle, yaxunl, tony-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D25788 llvm-svn: 285198
2016-10-26 22:38:47 +08:00
; CHECK-LABEL: flat_scratch_multidword_load:
; HSA: flat_load_dword
; HSA: flat_load_dword
; FIXME: These tests are broken for os = mesa3d, becasue it doesn't initialize flat_scr
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @flat_scratch_multidword_load() {
AMDGPU/SI: Don't emit multi-dword flat memory ops when they might access scratch Summary: A single flat memory operations that might access the scratch buffer can only access MaxPrivateElementSize bytes. Reviewers: arsenm Subscribers: kzhuravl, wdng, nhaehnle, yaxunl, tony-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D25788 llvm-svn: 285198
2016-10-26 22:38:47 +08:00
%scratch = alloca <2 x i32>
%fptr = addrspacecast <2 x i32>* %scratch to <2 x i32> addrspace(4)*
%ld = load volatile <2 x i32>, <2 x i32> addrspace(4)* %fptr
ret void
}
; CHECK-LABEL: flat_scratch_multidword_store:
; HSA: flat_store_dword
; HSA: flat_store_dword
; FIXME: These tests are broken for os = mesa3d, becasue it doesn't initialize flat_scr
AMDGPU: Mark all unspecified CC functions in tests as amdgpu_kernel Currently the default C calling convention functions are treated the same as compute kernels. Make this explicit so the default calling convention can be changed to a non-kernel. Converted with perl -pi -e 's/define void/define amdgpu_kernel void/' on the relevant test directories (and undoing in one place that actually wanted a non-kernel). llvm-svn: 298444
2017-03-22 05:39:51 +08:00
define amdgpu_kernel void @flat_scratch_multidword_store() {
AMDGPU/SI: Don't emit multi-dword flat memory ops when they might access scratch Summary: A single flat memory operations that might access the scratch buffer can only access MaxPrivateElementSize bytes. Reviewers: arsenm Subscribers: kzhuravl, wdng, nhaehnle, yaxunl, tony-tye, llvm-commits Differential Revision: https://reviews.llvm.org/D25788 llvm-svn: 285198
2016-10-26 22:38:47 +08:00
%scratch = alloca <2 x i32>
%fptr = addrspacecast <2 x i32>* %scratch to <2 x i32> addrspace(4)*
store volatile <2 x i32> zeroinitializer, <2 x i32> addrspace(4)* %fptr
ret void
}
AMDGPU: Start selecting flat instruction offsets llvm-svn: 305201
2017-06-13 00:53:51 +08:00
; CHECK-LABEL: {{^}}store_flat_i8_max_offset:
; CIVI: flat_store_byte v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}{{$}}
; GFX9: flat_store_byte v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}} offset:4095{{$}}
define amdgpu_kernel void @store_flat_i8_max_offset(i8 addrspace(4)* %fptr, i8 %x) #0 {
%fptr.offset = getelementptr inbounds i8, i8 addrspace(4)* %fptr, i64 4095
store volatile i8 %x, i8 addrspace(4)* %fptr.offset
ret void
}
; CHECK-LABEL: {{^}}store_flat_i8_max_offset_p1:
; CHECK: flat_store_byte v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}{{$}}
define amdgpu_kernel void @store_flat_i8_max_offset_p1(i8 addrspace(4)* %fptr, i8 %x) #0 {
%fptr.offset = getelementptr inbounds i8, i8 addrspace(4)* %fptr, i64 4096
store volatile i8 %x, i8 addrspace(4)* %fptr.offset
ret void
}
; CHECK-LABEL: {{^}}store_flat_i8_neg_offset:
; CHECK: flat_store_byte v{{\[[0-9]+:[0-9]+\]}}, v{{[0-9]+}}{{$}}
define amdgpu_kernel void @store_flat_i8_neg_offset(i8 addrspace(4)* %fptr, i8 %x) #0 {
%fptr.offset = getelementptr inbounds i8, i8 addrspace(4)* %fptr, i64 -2
store volatile i8 %x, i8 addrspace(4)* %fptr.offset
ret void
}
; CHECK-LABEL: {{^}}load_flat_i8_max_offset:
; CIVI: flat_load_ubyte v{{[0-9]+}}, v{{\[[0-9]+:[0-9]+\]}}{{$}}
; GFX9: flat_load_ubyte v{{[0-9]+}}, v{{\[[0-9]+:[0-9]+\]}} offset:4095{{$}}
define amdgpu_kernel void @load_flat_i8_max_offset(i8 addrspace(4)* %fptr) #0 {
%fptr.offset = getelementptr inbounds i8, i8 addrspace(4)* %fptr, i64 4095
%val = load volatile i8, i8 addrspace(4)* %fptr.offset
ret void
}
; CHECK-LABEL: {{^}}load_flat_i8_max_offset_p1:
; CHECK: flat_load_ubyte v{{[0-9]+}}, v{{\[[0-9]+:[0-9]+\]}}{{$}}
define amdgpu_kernel void @load_flat_i8_max_offset_p1(i8 addrspace(4)* %fptr) #0 {
%fptr.offset = getelementptr inbounds i8, i8 addrspace(4)* %fptr, i64 4096
%val = load volatile i8, i8 addrspace(4)* %fptr.offset
ret void
}
; CHECK-LABEL: {{^}}load_flat_i8_neg_offset:
; CHECK: flat_load_ubyte v{{[0-9]+}}, v{{\[[0-9]+:[0-9]+\]}}{{$}}
define amdgpu_kernel void @load_flat_i8_neg_offset(i8 addrspace(4)* %fptr) #0 {
%fptr.offset = getelementptr inbounds i8, i8 addrspace(4)* %fptr, i64 -2
%val = load volatile i8, i8 addrspace(4)* %fptr.offset
ret void
}
R600/SI: Add preliminary support for flat address space llvm-svn: 217777
2014-09-15 23:41:53 +08:00
attributes #0 = { nounwind }
AMDGPU: Switch barrier intrinsics to using convergent noduplicate prevents unrolling of small loops that happen to have barriers in them. If a loop has a barrier in it, it is OK to duplicate it for the unroll. llvm-svn: 256075
2015-12-19 09:46:41 +08:00
attributes #1 = { nounwind convergent }