From: Krzysztof Raszkowski <krzysztof.raszkowski@intel.com>
Date: Fri, 6 Mar 2020 15:09:24 +0000 (+0100)
Subject: gallium/swr: Fix vcvtph2ps llvm intrinsic compile error
X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=ad66b25415745383aa9380975f16967bfb1022a8;p=mesa.git

gallium/swr: Fix vcvtph2ps llvm intrinsic compile error

Reviewed-by: Jan Zielinski <jan.zielinski@intel.com>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4090>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4090>
---

diff --git a/src/gallium/drivers/swr/rasterizer/codegen/gen_llvm_ir_macros.py b/src/gallium/drivers/swr/rasterizer/codegen/gen_llvm_ir_macros.py
index 656a7fa3cd3..b35a2a4ae24 100644
--- a/src/gallium/drivers/swr/rasterizer/codegen/gen_llvm_ir_macros.py
+++ b/src/gallium/drivers/swr/rasterizer/codegen/gen_llvm_ir_macros.py
@@ -53,7 +53,6 @@ intrinsics = [
     ['VPERMD',      ['a', 'idx'], 'a'],
     ['VPERMPS',     ['idx', 'a'], 'a'],
     ['VCVTPD2PS',   ['a'], 'VectorType::get(mFP32Ty, a->getType()->getVectorNumElements())'],
-    ['VCVTPH2PS',   ['a'], 'VectorType::get(mFP32Ty, a->getType()->getVectorNumElements())'],
     ['VCVTPS2PH',   ['a', 'round'], 'mSimdInt16Ty'],
     ['VHSUBPS',     ['a', 'b'], 'a'],
     ['VPTESTC',     ['a', 'b'], 'mInt32Ty'],
diff --git a/src/gallium/drivers/swr/rasterizer/jitter/builder_misc.cpp b/src/gallium/drivers/swr/rasterizer/jitter/builder_misc.cpp
index 4f6b0b46a60..96175692259 100644
--- a/src/gallium/drivers/swr/rasterizer/jitter/builder_misc.cpp
+++ b/src/gallium/drivers/swr/rasterizer/jitter/builder_misc.cpp
@@ -621,37 +621,11 @@ namespace SwrJit
     /// @param a - 128bit SIMD lane(8x16bit) of float16 in int16 format.
     Value* Builder::CVTPH2PS(Value* a, const llvm::Twine& name)
     {
-        if (JM()->mArch.F16C())
-        {
-            return VCVTPH2PS(a, name);
-        }
-        else
-        {
-            FunctionType* pFuncTy   = FunctionType::get(mFP32Ty, mInt16Ty);
-            Function*     pCvtPh2Ps = cast<Function>(
-#if LLVM_VERSION_MAJOR >= 9
-                JM()->mpCurrentModule->getOrInsertFunction("ConvertFloat16ToFloat32", pFuncTy).getCallee());
-#else
-                JM()->mpCurrentModule->getOrInsertFunction("ConvertFloat16ToFloat32", pFuncTy));
-#endif
+        // Bitcast Nxint16 to Nxhalf
+        uint32_t numElems = a->getType()->getVectorNumElements();
+        Value*   input    = BITCAST(a, VectorType::get(mFP16Ty, numElems));
 
-            if (sys::DynamicLibrary::SearchForAddressOfSymbol("ConvertFloat16ToFloat32") == nullptr)
-            {
-                sys::DynamicLibrary::AddSymbol("ConvertFloat16ToFloat32",
-                                               (void*)&ConvertFloat16ToFloat32);
-            }
-
-            Value* pResult = UndefValue::get(mSimdFP32Ty);
-            for (uint32_t i = 0; i < mVWidth; ++i)
-            {
-                Value* pSrc  = VEXTRACT(a, C(i));
-                Value* pConv = CALL(pCvtPh2Ps, std::initializer_list<Value*>{pSrc});
-                pResult      = VINSERT(pResult, pConv, C(i));
-            }
-
-            pResult->setName(name);
-            return pResult;
-        }
+        return FP_EXT(input, VectorType::get(mFP32Ty, numElems), name);
     }
 
     //////////////////////////////////////////////////////////////////////////
diff --git a/src/gallium/drivers/swr/rasterizer/jitter/functionpasses/lower_x86.cpp b/src/gallium/drivers/swr/rasterizer/jitter/functionpasses/lower_x86.cpp
index 4eb0162d5d9..3c728084afe 100644
--- a/src/gallium/drivers/swr/rasterizer/jitter/functionpasses/lower_x86.cpp
+++ b/src/gallium/drivers/swr/rasterizer/jitter/functionpasses/lower_x86.cpp
@@ -127,7 +127,6 @@ namespace SwrJit
             {"meta.intrinsic.VGATHERDD", {{Intrinsic::not_intrinsic,            Intrinsic::not_intrinsic},  VGATHER_EMU}},
             {"meta.intrinsic.VSCATTERPS", {{Intrinsic::not_intrinsic,           Intrinsic::not_intrinsic}, VSCATTER_EMU}},
             {"meta.intrinsic.VCVTPD2PS", {{Intrinsic::x86_avx_cvt_pd2_ps_256,   Intrinsic::not_intrinsic},  NO_EMU}},
-            {"meta.intrinsic.VCVTPH2PS", {{Intrinsic::x86_vcvtph2ps_256,        Intrinsic::not_intrinsic},  NO_EMU}},
             {"meta.intrinsic.VROUND",    {{Intrinsic::x86_avx_round_ps_256,     DOUBLE},                    NO_EMU}},
             {"meta.intrinsic.VHSUBPS",   {{Intrinsic::x86_avx_hsub_ps_256,      DOUBLE},                    NO_EMU}},
         },
@@ -141,7 +140,6 @@ namespace SwrJit
             {"meta.intrinsic.VGATHERDD",    {{Intrinsic::not_intrinsic,         Intrinsic::not_intrinsic},  VGATHER_EMU}},
             {"meta.intrinsic.VSCATTERPS", {{Intrinsic::not_intrinsic,           Intrinsic::not_intrinsic}, VSCATTER_EMU}},
             {"meta.intrinsic.VCVTPD2PS",    {{Intrinsic::x86_avx_cvt_pd2_ps_256, DOUBLE},                   NO_EMU}},
-            {"meta.intrinsic.VCVTPH2PS",    {{Intrinsic::x86_vcvtph2ps_256,     Intrinsic::not_intrinsic},  NO_EMU}},
             {"meta.intrinsic.VROUND",       {{Intrinsic::x86_avx_round_ps_256,  DOUBLE},                    NO_EMU}},
             {"meta.intrinsic.VHSUBPS",      {{Intrinsic::x86_avx_hsub_ps_256,   DOUBLE},                    NO_EMU}},
         },
@@ -164,7 +162,6 @@ namespace SwrJit
 #else
             {"meta.intrinsic.VCVTPD2PS", {{Intrinsic::not_intrinsic,            Intrinsic::not_intrinsic}, VCONVERT_EMU}},
 #endif
-            {"meta.intrinsic.VCVTPH2PS", {{Intrinsic::x86_avx512_mask_vcvtph2ps_256, Intrinsic::x86_avx512_mask_vcvtph2ps_512}, NO_EMU}},
             {"meta.intrinsic.VROUND", {{Intrinsic::not_intrinsic,               Intrinsic::not_intrinsic}, VROUND_EMU}},
             {"meta.intrinsic.VHSUBPS", {{Intrinsic::not_intrinsic,              Intrinsic::not_intrinsic}, VHSUB_EMU}},
         }};