From: Stephane Marchesin Date: Tue, 30 Sep 2008 22:00:58 +0000 (+0200) Subject: Gallivm: add slt. glxgears should be running, except it isn't. X-Git-Url: https://git.libre-soc.org/?a=commitdiff_plain;h=a6ff215777da2181d7099284f2da28eff78273a9;p=mesa.git Gallivm: add slt. glxgears should be running, except it isn't. --- diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index 5863f370954..a6580725512 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -181,6 +181,7 @@ void InstructionsSoa::createFunctionMap() m_functionsMap[TGSI_OPCODE_POWER] = "pow"; m_functionsMap[TGSI_OPCODE_LIT] = "lit"; m_functionsMap[TGSI_OPCODE_RSQ] = "rsq"; + m_functionsMap[TGSI_OPCODE_SLT] = "slt"; } void InstructionsSoa::createDependencies() @@ -280,6 +281,14 @@ std::vector InstructionsSoa::dp3(const std::vector i return callBuiltin(func, in1, in2); } + +std::vector InstructionsSoa::slt(const std::vector in1, + const std::vector in2) +{ + llvm::Function *func = function(TGSI_OPCODE_SLT); + return callBuiltin(func, in1, in2); +} + llvm::Value * InstructionsSoa::allocaTemp() { VectorType *vector = VectorType::get(Type::FloatTy, 4); diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.h b/src/gallium/auxiliary/gallivm/instructionssoa.h index 20cab3b3bbd..3817fdc904b 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.h +++ b/src/gallium/auxiliary/gallivm/instructionssoa.h @@ -69,6 +69,8 @@ public: std::vector pow(const std::vector in1, const std::vector in2); std::vector rsq(const std::vector in1); + std::vector slt(const std::vector in1, + const std::vector in2); std::vector sub(const std::vector in1, const std::vector in2); void end(); diff --git a/src/gallium/auxiliary/gallivm/soabuiltins.c b/src/gallium/auxiliary/gallivm/soabuiltins.c index 78f84510e29..cb85e1734ec 100644 --- a/src/gallium/auxiliary/gallivm/soabuiltins.c +++ b/src/gallium/auxiliary/gallivm/soabuiltins.c @@ -36,6 +36,8 @@ typedef __attribute__(( ext_vector_type(4) )) float float4; extern float fabsf(float val); +/* helpers */ + float4 absvec(float4 vec) { float4 res; @@ -47,6 +49,58 @@ float4 absvec(float4 vec) return res; } +float4 maxvec(float4 a, float4 b) +{ + return (float4){(a.x > b.x) ? a.x : b.x, + (a.y > b.y) ? a.y : b.y, + (a.z > b.z) ? a.z : b.z, + (a.w > b.w) ? a.w : b.w}; +} + +float4 minvec(float4 a, float4 b) +{ + return (float4){(a.x < b.x) ? a.x : b.x, + (a.y < b.y) ? a.y : b.y, + (a.z < b.z) ? a.z : b.z, + (a.w < b.w) ? a.w : b.w}; +} + +extern float powf(float num, float p); +extern float sqrtf(float x); + +float4 powvec(float4 vec, float4 q) +{ + float4 p; + p.x = powf(vec.x, q.x); + p.y = powf(vec.y, q.y); + p.z = powf(vec.z, q.z); + p.w = powf(vec.w, q.w); + return p; +} + +float4 sqrtvec(float4 vec) +{ + float4 p; + p.x = sqrtf(vec.x); + p.y = sqrtf(vec.y); + p.z = sqrtf(vec.z); + p.w = sqrtf(vec.w); + return p; +} + +float4 sltvec(float4 v1, float4 v2) +{ + float4 p; + p.x = (v1.x < v2.x) ? 1.0 : 0.0; + p.y = (v1.y < v2.y) ? 1.0 : 0.0; + p.z = (v1.z < v2.z) ? 1.0 : 0.0; + p.w = (v1.w < v2.w) ? 1.0 : 0.0; + return p; +} + + +/* instructions */ + void abs(float4 *res, float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w) { @@ -69,7 +123,6 @@ void dp3(float4 *res, res[3] = dot; } - void dp4(float4 *res, float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w, float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w) @@ -83,35 +136,25 @@ void dp4(float4 *res, res[3] = dot; } -extern float powf(float num, float p); -extern float sqrtf(float x); - -float4 powvec(float4 vec, float4 q) -{ - float4 p; - p.x = powf(vec.x, q.x); - p.y = powf(vec.y, q.y); - p.z = powf(vec.z, q.z); - p.w = powf(vec.w, q.w); - return p; -} - -void pow(float4 *res, - float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w, - float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w) +void lit(float4 *res, + float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w) { - res[0] = powvec(tmp0x, tmp1x); - res[1] = res[0]; - res[2] = res[0]; - res[3] = res[0]; -} + const float4 zerovec = (float4) {0.0, 0.0, 0.0, 0.0}; + const float4 min128 = (float4) {-128.f, -128.f, -128.f, -128.f}; + const float4 plus128 = (float4) {128.f, 128.f, 128.f, 128.f}; -float4 minvec(float4 a, float4 b) -{ - return (float4){(a.x < b.x) ? a.x : b.x, - (a.y < b.y) ? a.y : b.y, - (a.z < b.z) ? a.z : b.z, - (a.w < b.w) ? a.w : b.w}; + res[0] = (float4){1.0, 1.0, 1.0, 1.0}; + if (tmp0x.x > 0) { + float4 tmpy = maxvec(tmp0y, zerovec); + float4 tmpw = minvec(tmp0w, plus128); + tmpw = maxvec(tmpw, min128); + res[1] = tmp0x; + res[2] = powvec(tmpy, tmpw); + } else { + res[1] = zerovec; + res[2] = zerovec; + } + res[3] = (float4){1.0, 1.0, 1.0, 1.0}; } void min(float4 *res, @@ -125,14 +168,6 @@ void min(float4 *res, } -float4 maxvec(float4 a, float4 b) -{ - return (float4){(a.x > b.x) ? a.x : b.x, - (a.y > b.y) ? a.y : b.y, - (a.z > b.z) ? a.z : b.z, - (a.w > b.w) ? a.w : b.w}; -} - void max(float4 *res, float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w, float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w) @@ -143,37 +178,14 @@ void max(float4 *res, res[3] = maxvec(tmp0w, tmp1w); } - -void lit(float4 *res, - float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w) -{ - const float4 zerovec = (float4) {0.0, 0.0, 0.0, 0.0}; - const float4 min128 = (float4) {-128.f, -128.f, -128.f, -128.f}; - const float4 plus128 = (float4) {128.f, 128.f, 128.f, 128.f}; - - res[0] = (float4){1.0, 1.0, 1.0, 1.0}; - if (tmp0x.x > 0) { - float4 tmpy = maxvec(tmp0y, zerovec); - float4 tmpw = minvec(tmp0w, plus128); - tmpw = maxvec(tmpw, min128); - res[1] = tmp0x; - res[2] = powvec(tmpy, tmpw); - } else { - res[1] = zerovec; - res[2] = zerovec; - } - res[3] = (float4){1.0, 1.0, 1.0, 1.0}; -} - - -float4 sqrtvec(float4 vec) +void pow(float4 *res, + float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w, + float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w) { - float4 p; - p.x = sqrtf(vec.x); - p.y = sqrtf(vec.y); - p.z = sqrtf(vec.z); - p.w = sqrtf(vec.w); - return p; + res[0] = powvec(tmp0x, tmp1x); + res[1] = res[0]; + res[2] = res[0]; + res[3] = res[0]; } void rsq(float4 *res, @@ -185,3 +197,14 @@ void rsq(float4 *res, res[2] = onevec/sqrtvec(absvec(tmp0z)); res[3] = onevec/sqrtvec(absvec(tmp0w)); } + +void slt(float4 *res, + float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w, + float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w) +{ + res[0] = sltvec(tmp0x, tmp1x); + res[1] = sltvec(tmp0y, tmp1y); + res[2] = sltvec(tmp0z, tmp1z); + res[3] = sltvec(tmp0w, tmp1w); +} + diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index fdfbb76c16a..7292c0e366c 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -767,6 +767,7 @@ translate_instructionir(llvm::Module *module, } break; case TGSI_OPCODE_SLT: { + out = instr->slt(inputs[0], inputs[1]); } break; case TGSI_OPCODE_SGE: {