Merge remote-tracking branch 'mesa/13.0' into nougat-x86

6 files changed, 31 insertions, 16 deletions

diff --git a/src/compiler/Makefile.glsl.am b/src/compiler/Makefile.glsl.am
index 80dfb73..15bea6b 100644
--- a/src/compiler/Makefile.glsl.am
+++ b/src/compiler/Makefile.glsl.am
@@ -62,8 +62,11 @@ glsl_tests_blob_test_LDADD =				\
 
 glsl_tests_cache_test_SOURCES =				\
 	glsl/tests/cache_test.c
+glsl_tests_cache_test_CFLAGS =				\
+	$(PTHREAD_CFLAGS)
 glsl_tests_cache_test_LDADD =				\
-	glsl/libglsl.la
+	glsl/libglsl.la					\
+	$(PTHREAD_LIBS)
 
 glsl_tests_general_ir_test_SOURCES =			\
 	glsl/tests/builtin_variable_test.cpp		\
diff --git a/src/compiler/glsl/builtin_functions.cpp b/src/compiler/glsl/builtin_functions.cpp
index 3e4bcbb..3dead1a 100644
--- a/src/compiler/glsl/builtin_functions.cpp
+++ b/src/compiler/glsl/builtin_functions.cpp
@@ -3563,9 +3563,17 @@ builtin_builder::_tanh(const glsl_type *type)
    ir_variable *x = in_var(type, "x");
    MAKE_SIG(type, v130, 1, x);
 
+   /* Clamp x to [-10, +10] to avoid precision problems.
+    * When x > 10, e^(-x) is so small relative to e^x that it gets flushed to
+    * zero in the computation e^x + e^(-x). The same happens in the other
+    * direction when x < -10.
+    */
+   ir_variable *t = body.make_temp(type, "tmp");
+   body.emit(assign(t, min2(max2(x, imm(-10.0f)), imm(10.0f))));
+
    /* (e^x - e^(-x)) / (e^x + e^(-x)) */
-   body.emit(ret(div(sub(exp(x), exp(neg(x))),
-                     add(exp(x), exp(neg(x))))));
+   body.emit(ret(div(sub(exp(t), exp(neg(t))),
+                     add(exp(t), exp(neg(t))))));
 
    return sig;
 }
diff --git a/src/compiler/glsl/link_uniforms.cpp b/src/compiler/glsl/link_uniforms.cpp
index b3c3c5a..8529b74 100644
--- a/src/compiler/glsl/link_uniforms.cpp
+++ b/src/compiler/glsl/link_uniforms.cpp
@@ -633,6 +633,8 @@ private:
          uniform->opaque[shader_type].index = this->next_subroutine;
          uniform->opaque[shader_type].active = true;
 
+         prog->_LinkedShaders[shader_type]->NumSubroutineUniforms++;
+
          /* Increment the subroutine index by 1 for non-arrays and by the
           * number of array elements for arrays.
           */
diff --git a/src/compiler/glsl/linker.cpp b/src/compiler/glsl/linker.cpp
index f62a848..b71c51e 100644
--- a/src/compiler/glsl/linker.cpp
+++ b/src/compiler/glsl/linker.cpp
@@ -3118,7 +3118,6 @@ link_calculate_subroutine_compat(struct gl_shader_program *prog)
          if (!uni)
             continue;
 
-         sh->NumSubroutineUniforms++;
          count = 0;
          if (sh->NumSubroutineFunctions == 0) {
             linker_error(prog, "subroutine uniform %s defined but no valid functions found\n", uni->type->name);
diff --git a/src/compiler/nir/nir_opt_undef.c b/src/compiler/nir/nir_opt_undef.c
index 0f8ba31..c4777a8 100644
--- a/src/compiler/nir/nir_opt_undef.c
+++ b/src/compiler/nir/nir_opt_undef.c
@@ -79,9 +79,7 @@ opt_undef_vecN(nir_builder *b, nir_alu_instr *alu)
 {
    if (alu->op != nir_op_vec2 &&
        alu->op != nir_op_vec3 &&
-       alu->op != nir_op_vec4 &&
-       alu->op != nir_op_fmov &&
-       alu->op != nir_op_imov)
+       alu->op != nir_op_vec4)
       return false;
 
    assert(alu->dest.dest.is_ssa);
diff --git a/src/compiler/spirv/vtn_glsl450.c b/src/compiler/spirv/vtn_glsl450.c
index cb0570d..fbc7ce6 100644
--- a/src/compiler/spirv/vtn_glsl450.c
+++ b/src/compiler/spirv/vtn_glsl450.c
@@ -565,16 +565,21 @@ handle_glsl450_alu(struct vtn_builder *b, enum GLSLstd450 entrypoint,
                                    build_exp(nb, nir_fneg(nb, src[0]))));
       return;
 
-   case GLSLstd450Tanh:
-      /* (0.5 * (e^x - e^(-x))) / (0.5 * (e^x + e^(-x))) */
-      val->ssa->def =
-         nir_fdiv(nb, nir_fmul(nb, nir_imm_float(nb, 0.5f),
-                                   nir_fsub(nb, build_exp(nb, src[0]),
-                                                build_exp(nb, nir_fneg(nb, src[0])))),
-                      nir_fmul(nb, nir_imm_float(nb, 0.5f),
-                                   nir_fadd(nb, build_exp(nb, src[0]),
-                                                build_exp(nb, nir_fneg(nb, src[0])))));
+   case GLSLstd450Tanh: {
+      /* tanh(x) := (0.5 * (e^x - e^(-x))) / (0.5 * (e^x + e^(-x)))
+       *
+       * With a little algebra this reduces to (e^2x - 1) / (e^2x + 1)
+       *
+       * We clamp x to (-inf, +10] to avoid precision problems.  When x > 10,
+       * e^2x is so much larger than 1.0 that 1.0 gets flushed to zero in the
+       * computation e^2x +/- 1 so it can be ignored.
+       */
+      nir_ssa_def *x = nir_fmin(nb, src[0], nir_imm_float(nb, 10));
+      nir_ssa_def *exp2x = build_exp(nb, nir_fmul(nb, x, nir_imm_float(nb, 2)));
+      val->ssa->def = nir_fdiv(nb, nir_fsub(nb, exp2x, nir_imm_float(nb, 1)),
+                                   nir_fadd(nb, exp2x, nir_imm_float(nb, 1)));
       return;
+   }
 
    case GLSLstd450Asinh:
       val->ssa->def = nir_fmul(nb, nir_fsign(nb, src[0]),