gallivm: Introduce lp_format_intrinsic.

For adding .v4f32 like suffixes to intrinsics, taking special care for
scalar case, which was being often neglected.

This fixes invalid IR when doing mipmap filtering on SSE2 (the only
case where we'd use intrinsics with scalars.)

Reviewed-by: Roland Scheidegger <sroland@vmware.com>

diff --git a/src/gallium/auxiliary/gallivm/lp_bld_arit.c b/src/gallium/auxiliary/gallivm/lp_bld_arit.c
index 587c83acfa0d90a038aa83ad427213a3d1fd2c2a..0c43617d5315c6c356d1c6409fe53c55caf600e5 100644 (file)
--- a/src/gallium/auxiliary/gallivm/lp_bld_arit.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_arit.c
@@ -1493,7 +1493,7 @@ lp_build_abs(struct lp_build_context *bld,
 
    if(type.floating) {
       char intrinsic[32];
-      util_snprintf(intrinsic, sizeof intrinsic, "llvm.fabs.v%uf%u", type.length, type.width);
+      lp_format_intrinsic(intrinsic, sizeof intrinsic, "llvm.fabs", vec_type);
       return lp_build_intrinsic_unary(builder, intrinsic, vec_type, a);
    }
 
@@ -1880,9 +1880,7 @@ lp_build_round_arch(struct lp_build_context *bld,
          break;
       }
 
-      util_snprintf(intrinsic, sizeof intrinsic, "%s.v%uf%u",
-                    intrinsic_root, type.length, type.width);
-
+      lp_format_intrinsic(intrinsic, sizeof intrinsic, intrinsic_root, bld->vec_type);
       return lp_build_intrinsic_unary(builder, intrinsic, bld->vec_type, a);
    }
    else /* (util_cpu_caps.has_altivec) */
@@ -2026,7 +2024,7 @@ lp_build_floor(struct lp_build_context *bld,
 
       if (type.width != 32) {
          char intrinsic[32];
-         util_snprintf(intrinsic, sizeof intrinsic, "llvm.floor.v%uf%u", type.length, type.width);
+         lp_format_intrinsic(intrinsic, sizeof intrinsic, "llvm.floor", vec_type);
          return lp_build_intrinsic_unary(builder, intrinsic, vec_type, a);
       }
 
@@ -2101,7 +2099,7 @@ lp_build_ceil(struct lp_build_context *bld,
 
       if (type.width != 32) {
          char intrinsic[32];
-         util_snprintf(intrinsic, sizeof intrinsic, "llvm.ceil.v%uf%u", type.length, type.width);
+         lp_format_intrinsic(intrinsic, sizeof intrinsic, "llvm.ceil", vec_type);
          return lp_build_intrinsic_unary(builder, intrinsic, vec_type, a);
       }
 
@@ -2438,15 +2436,8 @@ lp_build_sqrt(struct lp_build_context *bld,
 
    assert(lp_check_value(type, a));
 
-   /* TODO: optimize the constant case */
-
    assert(type.floating);
-   if (type.length == 1) {
-      util_snprintf(intrinsic, sizeof intrinsic, "llvm.sqrt.f%u", type.width);
-   }
-   else {
-      util_snprintf(intrinsic, sizeof intrinsic, "llvm.sqrt.v%uf%u", type.length, type.width);
-   }
+   lp_format_intrinsic(intrinsic, sizeof intrinsic, "llvm.sqrt", vec_type);
 
    return lp_build_intrinsic_unary(builder, intrinsic, vec_type, a);
 }

diff --git a/src/gallium/auxiliary/gallivm/lp_bld_intr.c b/src/gallium/auxiliary/gallivm/lp_bld_intr.c
index 30f4863ec443f8cc9c650de299008b1a7dd87ca2..e153389e6a9bd0b7041fc0b74a9fbb80b564b908 100644 (file)
--- a/src/gallium/auxiliary/gallivm/lp_bld_intr.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_intr.c
@@ -45,6 +45,7 @@
 
 
 #include "util/u_debug.h"
+#include "util/u_string.h"
 
 #include "lp_bld_const.h"
 #include "lp_bld_intr.h"
@@ -52,6 +53,48 @@
 #include "lp_bld_pack.h"
 
 
+void
+lp_format_intrinsic(char *name,
+                    size_t size,
+                    const char *name_root,
+                    LLVMTypeRef type)
+{
+   unsigned length = 0;
+   unsigned width;
+   char c;
+
+   LLVMTypeKind kind = LLVMGetTypeKind(type);
+   if (kind == LLVMVectorTypeKind) {
+      length = LLVMGetVectorSize(type);
+      type = LLVMGetElementType(type);
+      kind = LLVMGetTypeKind(type);
+   }
+
+   switch (kind) {
+   case LLVMIntegerTypeKind:
+      c = 'i';
+      width = LLVMGetIntTypeWidth(type);
+      break;
+   case LLVMFloatTypeKind:
+      c = 'f';
+      width = 32;
+      break;
+   case LLVMDoubleTypeKind:
+      c = 'f';
+      width = 64;
+      break;
+   default:
+      assert(0);
+   }
+
+   if (length) {
+      util_snprintf(name, size, "%s.v%u%c%u", name_root, length, c, width);
+   } else {
+      util_snprintf(name, size, "%s.%c%u", name_root, c, width);
+   }
+}
+
+
 LLVMValueRef
 lp_declare_intrinsic(LLVMModuleRef module,
                      const char *name,

diff --git a/src/gallium/auxiliary/gallivm/lp_bld_intr.h b/src/gallium/auxiliary/gallivm/lp_bld_intr.h
index a54b367961afecaad8d91a2c05e9cc5ee155e642..7d80ac28f708d0ceb8ad270c58ae8a30c17fb4aa 100644 (file)
--- a/src/gallium/auxiliary/gallivm/lp_bld_intr.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_intr.h
@@ -47,6 +47,12 @@
 #define LP_MAX_FUNC_ARGS 32
 
 
+void
+lp_format_intrinsic(char *name,
+                    size_t size,
+                    const char *name_root,
+                    LLVMTypeRef type);
+
 LLVMValueRef
 lp_declare_intrinsic(LLVMModuleRef module,
                      const char *name,