6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2025-07-02 04:24:22 +00:00
Code Issues Projects Releases Wiki Activity

Revert "R600: Add work-around for the CF stack entry HW bug"

Browse Source 
This reverts commit 35b8331cad6eb512a2506adbc394201181da94ba.

The -debug-only flag for llc doesn't appear to be available in
all build configurations.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@199845 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Tom Stellard
2014-01-22 22:20:54 +00:00
parent ac2d2e0a3d
commit 8a4f11e3b6
6 changed files with 7 additions and 288 deletions
Download Patch File Download Diff File Expand all files Collapse all files

43
lib/Target/R600/R600ControlFlowFinalizer.cpp
View File

@ -73,44 +73,6 @@ bool CFStack::branchStackContains(CFStack::StackItem Item) {
return false;
}
bool CFStack::requiresWorkAroundForInst(unsigned Opcode) {
if (Opcode == AMDGPU::CF_ALU_PUSH_BEFORE && ST.hasCaymanISA() &&
getLoopDepth() > 1)
return true;
if (!ST.hasCFAluBug())
return false;
switch(Opcode) {
default: return false;
case AMDGPU::CF_ALU_PUSH_BEFORE:
case AMDGPU::CF_ALU_ELSE_AFTER:
case AMDGPU::CF_ALU_BREAK:
case AMDGPU::CF_ALU_CONTINUE:
if (CurrentSubEntries == 0)
return false;
if (ST.getWavefrontSize() == 64) {
// We are being conservative here. We only require this work-around if
// CurrentSubEntries > 3 &&
// (CurrentSubEntries % 4 == 3 || CurrentSubEntries % 4 == 0)
//
// We have to be conservative, because we don't know for certain that
// our stack allocation algorithm for Evergreen/NI is correct. Applying this
// work-around when CurrentSubEntries > 3 allows us to over-allocate stack
// resources without any problems.
return CurrentSubEntries > 3;
} else {
assert(ST.getWavefrontSize() == 32);
// We are being conservative here. We only require the work-around if
// CurrentSubEntries > 7 &&
// (CurrentSubEntries % 8 == 7 || CurrentSubEntries % 8 == 0)
// See the comment on the wavefront size == 64 case for why we are
// being conservative.
return CurrentSubEntries > 7;
}
}
}
unsigned CFStack::getSubEntrySize(CFStack::StackItem Item) {
switch(Item) {
default:
@ -510,12 +472,9 @@ public:
if (MI->getOpcode() == AMDGPU::CF_ALU)
LastAlu.back() = MI;
I++;
bool RequiresWorkAround =
CFStack.requiresWorkAroundForInst(MI->getOpcode());
switch (MI->getOpcode()) {
case AMDGPU::CF_ALU_PUSH_BEFORE:
if (RequiresWorkAround) {
DEBUG(dbgs() << "Applying bug work-around for ALU_PUSH_BEFORE\n");
if (ST.hasCaymanISA() && CFStack.getLoopDepth() > 1) {
BuildMI(MBB, MI, MBB.findDebugLoc(MI), TII->get(AMDGPU::CF_PUSH_EG))
.addImm(CfCount + 1)
.addImm(1);