AI/rocm-systems
2
0
Garpu 0
Kode Masalah Tarik Permintaan Actions Packages Projects Rilis Wiki Kegiatan

P4 to Git Change 1266028 by smekhano@stas-rampitec-hsa on 2016/05/06 13:22:26

Telusuri Sumber 
SWDEV-93545 - HSA HLC: produce v_fma_f32 instead of v_mad_f32/v_mac_f32 on GFX9 if denorms are supported

	1. Added means to know HW target for HSAIL BE.
	2. Simplified compiler lib logic in handling target capabilities.
	3. Used target info to fuse mul/add into fma_f32 on GFX9.
	Previously it was disabled because mad/mac always flush. v_fma_f32 does not flush,
	but only fast starting with GFX9.

	Testing: smoke, precheckin
	Reviewed by Nikolay Haustov and Evgeny Mankov

Affected files ...

... //depot/stg/opencl/drivers/opencl/compiler/lib/backends/common/codegen.cpp#69 edit
... //depot/stg/opencl/drivers/opencl/compiler/lib/backends/common/linker.cpp#142 edit
... //depot/stg/opencl/drivers/opencl/compiler/lib/utils/v0_8/libUtils.cpp#17 edit
... //depot/stg/opencl/drivers/opencl/compiler/lib/utils/v0_8/libUtils.h#26 edit
... //depot/stg/opencl/drivers/opencl/compiler/llvm/lib/Target/HSAIL/HSAIL.h#42 edit
... //depot/stg/opencl/drivers/opencl/compiler/llvm/lib/Target/HSAIL/HSAILFusion.td#29 edit
... //depot/stg/opencl/drivers/opencl/compiler/llvm/lib/Target/HSAIL/HSAILISelDAGToDAG.cpp#70 edit
... //depot/stg/opencl/drivers/opencl/compiler/llvm/lib/Target/HSAIL/HSAILISelLowering.cpp#115 edit
... //depot/stg/opencl/drivers/opencl/compiler/llvm/lib/Target/HSAIL/HSAILInstructions.td#21 edit
... //depot/stg/opencl/drivers/opencl/compiler/llvm/lib/Target/HSAIL/HSAILTargetMachine.cpp#56 edit
... //depot/stg/opencl/drivers/opencl/tests/hsa/tlst/llc_opt.tlst#95 edit
This commit is contained in:
foreman
2016-05-06 13:39:37 -04:00
orang tua feca1fb254
melakukan 4de15fc86f
4 mengubah file dengan 41 tambahan dan 21 penghapusan
Download Patch File Download Diff File Expand all files Collapse all files
rocclr/compiler/lib/backends/common/codegen.cpp
+5
Melihat File
@@ -42,6 +42,10 @@
using namespace amdcl;
using namespace llvm;
namespace llvm {
extern int HsailOptimizeFor;
}
//!--------------------------------------------------------------------------!//
// JIT Memory manager
//!--------------------------------------------------------------------------!//
@@ -614,6 +618,7 @@ llvmCodeGen(
#ifdef WITH_TARGET_HSAIL
if (isHSAILTarget(binary->target)) {
llvm::HsailOptimizeFor = getIsaType(aclutGetTargetInfo(binary));
if (Target.addPassesToEmitFile(Passes, *Out, TargetMachine::CGFT_ObjectFile, true)) {
delete Out;
return 1;