Skip to content

Add option to change SVE vector length for current and children processes #101295

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 18 commits into from
Jun 7, 2024
Merged

Add option to change SVE vector length for current and children processes #101295

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
18 commits
Select commit Hold shift + click to select a range
1393c30
Add option to change SVE vector length for current and children proce…
SwapnilGaikwad Apr 16, 2024
2c040a7
Use maxVectorTBitWidth to get desired SVE length
SwapnilGaikwad Apr 24, 2024
19dce8d
Merge main
SwapnilGaikwad Apr 30, 2024
22aa47e
Use CNTB to determine current vector length
SwapnilGaikwad Apr 30, 2024
43ece1e
Use ACLE to detect vector length on Linux and hardcode on Windows
SwapnilGaikwad May 3, 2024
9140d82
Use rdvl instead of cntb to count vector length
SwapnilGaikwad May 7, 2024
aca2809
Add inline assembly suitable for MSVC
SwapnilGaikwad May 8, 2024
1d14227
Add MSCV compatible assembly helper instead of inline assembly
SwapnilGaikwad May 10, 2024
23dda80
Merge main
SwapnilGaikwad May 20, 2024
c080aad
Declare GetSveLengthFromOS on Arm64
SwapnilGaikwad May 20, 2024
a0c5247
Remove definition of GetSveLengthFromOS from .S file
SwapnilGaikwad May 21, 2024
4c0ffa7
Move non-windows definition of GetSveLengthFromOS to .S file
SwapnilGaikwad Jun 5, 2024
aaeb733
Merge main
SwapnilGaikwad Jun 5, 2024
8b07210
Apply suggestions from code review
jkotas Jun 5, 2024
c0a6910
Move declaration of GetSveLengthFromOS to .cpp file
SwapnilGaikwad Jun 5, 2024
528e157
Disable temporarily for Windows on Arm machines
SwapnilGaikwad Jun 5, 2024
63a2be8
Merge main
SwapnilGaikwad Jun 6, 2024
ec0c317
Hardcode GetSveLengthFromOS() to return SVE length as 128
SwapnilGaikwad Jun 6, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions src/coreclr/vm/arm64/asmhelpers.S
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,13 @@ LEAF_ENTRY GetDataCacheZeroIDReg, _TEXT
ret lr
LEAF_END GetDataCacheZeroIDReg, _TEXT

// uint64_t GetSveLengthFromOS(void);
.arch_extension sve
LEAF_ENTRY GetSveLengthFromOS, _TEXT
rdvl x0, 1
ret lr
LEAF_END GetSveLengthFromOS, _TEXT

//-----------------------------------------------------------------------------
// This routine captures the machine state. It is used by helper method frame
//-----------------------------------------------------------------------------
Expand Down
8 changes: 8 additions & 0 deletions src/coreclr/vm/arm64/asmhelpers.asm
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -78,6 +78,14 @@
ret lr
LEAF_END

;; uint64_t GetSveLengthFromOS(void);
LEAF_ENTRY GetSveLengthFromOS
;; TODO-SVE: Remove the hardcoded value 128 and uncomment once CI machines are updated to use MASM 14.4 or later
;; rdvl x0, 1
mov x0, #128
ret lr
LEAF_END

;;-----------------------------------------------------------------------------
;; This routine captures the machine state. It is used by helper method frame
;;-----------------------------------------------------------------------------
Expand Down
16 changes: 12 additions & 4 deletions src/coreclr/vm/codeman.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1243,6 +1243,7 @@ EEJitManager::EEJitManager()

#ifdef TARGET_ARM64
extern "C" DWORD64 __stdcall GetDataCacheZeroIDReg();
extern "C" uint64_t GetSveLengthFromOS();
#endif

void EEJitManager::SetCpuInfo()
Expand All @@ -1257,12 +1258,12 @@ void EEJitManager::SetCpuInfo()

int cpuFeatures = minipal_getcpufeatures();

#if defined(TARGET_X86) || defined(TARGET_AMD64)
CPUCompileFlags.Set(InstructionSet_VectorT128);

// Get the maximum bitwidth of Vector<T>, rounding down to the nearest multiple of 128-bits
uint32_t maxVectorTBitWidth = (CLRConfig::GetConfigValue(CLRConfig::EXTERNAL_MaxVectorTBitWidth) / 128) * 128;

#if defined(TARGET_X86) || defined(TARGET_AMD64)
CPUCompileFlags.Set(InstructionSet_VectorT128);

if (((cpuFeatures & XArchIntrinsicConstants_Avx2) != 0) && ((maxVectorTBitWidth == 0) || (maxVectorTBitWidth >= 256)))
{
// We allow 256-bit Vector<T> by default
Expand Down Expand Up @@ -1514,7 +1515,14 @@ void EEJitManager::SetCpuInfo()

if (((cpuFeatures & ARM64IntrinsicConstants_Sve) != 0) && CLRConfig::GetConfigValue(CLRConfig::EXTERNAL_EnableArm64Sve))
{
CPUCompileFlags.Set(InstructionSet_Sve);
uint32_t maxVectorTLength = (maxVectorTBitWidth / 8);
uint64_t sveLengthFromOS = GetSveLengthFromOS();

// Do not enable SVE when the user specified vector length is smaller than the one offered by underlying OS.
if ((maxVectorTLength >= sveLengthFromOS) || (maxVectorTBitWidth == 0))
{
CPUCompileFlags.Set(InstructionSet_Sve);
}
}

// DCZID_EL0<4> (DZP) indicates whether use of DC ZVA instructions is permitted (0) or prohibited (1).
Expand Down