nir/intrinsics: Allow deref sources to consume anything

This commit adds a new num_components value for intrinsic sources of -1
which means that it consumes everything and the number of components
effectively isn't validated.  This is useful for deref sources which
just take the result of the deref and we leave it up to the driver to
decide what that size should be.

Reviewed-by: Alejandro Piñeiro <apinheiro@igalia.com>
Reviewed-by: Caio Marcelo de Oliveira Filho <caio.oliveira@intel.com>

diff --git a/src/compiler/nir/nir.h b/src/compiler/nir/nir.h
index 971d8c4979f92c6d0c4633fc258052a2af384819..0042f803eb2d89b9a5f2cbcf8e28424ff1c60211 100644 (file)
--- a/src/compiler/nir/nir.h
+++ b/src/compiler/nir/nir.h
@@ -1292,9 +1292,11 @@ typedef struct {
    /** number of components of each input register
     *
     * If this value is 0, the number of components is given by the
-    * num_components field of nir_intrinsic_instr.
+    * num_components field of nir_intrinsic_instr.  If this value is -1, the
+    * intrinsic consumes however many components are provided and it is not
+    * validated at all.
     */
-   unsigned src_components[NIR_INTRINSIC_MAX_INPUTS];
+   int src_components[NIR_INTRINSIC_MAX_INPUTS];
 
    bool has_dest;
 
@@ -1322,10 +1324,12 @@ nir_intrinsic_src_components(nir_intrinsic_instr *intr, unsigned srcn)
 {
    const nir_intrinsic_info *info = &nir_intrinsic_infos[intr->intrinsic];
    assert(srcn < info->num_srcs);
-   if (info->src_components[srcn])
+   if (info->src_components[srcn] > 0)
       return info->src_components[srcn];
-   else
+   else if (info->src_components[srcn] == 0)
       return intr->num_components;
+   else
+      return nir_src_num_components(intr->src[srcn]);
 }
 
 static inline unsigned

diff --git a/src/compiler/nir/nir_intrinsics.py b/src/compiler/nir/nir_intrinsics.py
index 6ea6ad1198fe78a5945258eb6ae9921b1ef93e47..24bcd6bb169ecab3878e7fec9d4c03d96d2d55f9 100644 (file)
--- a/src/compiler/nir/nir_intrinsics.py
+++ b/src/compiler/nir/nir_intrinsics.py
@@ -132,9 +132,9 @@ intrinsic("nop", flags=[CAN_ELIMINATE])
 
 intrinsic("load_param", dest_comp=0, indices=[PARAM_IDX], flags=[CAN_ELIMINATE])
 
-intrinsic("load_deref", dest_comp=0, src_comp=[1], flags=[CAN_ELIMINATE])
-intrinsic("store_deref", src_comp=[1, 0], indices=[WRMASK])
-intrinsic("copy_deref", src_comp=[1, 1])
+intrinsic("load_deref", dest_comp=0, src_comp=[-1], flags=[CAN_ELIMINATE])
+intrinsic("store_deref", src_comp=[-1, 0], indices=[WRMASK])
+intrinsic("copy_deref", src_comp=[-1, -1])
 
 # Interpolation of input.  The interp_deref_at* intrinsics are similar to the
 # load_var intrinsic acting on a shader input except that they interpolate the
@@ -269,15 +269,15 @@ intrinsic("set_vertex_count", src_comp=[1])
 # lowered, variants take a constant buffer index and register offset.
 
 def atomic(name, flags=[]):
-    intrinsic(name + "_deref", src_comp=[1], dest_comp=1, flags=flags)
+    intrinsic(name + "_deref", src_comp=[-1], dest_comp=1, flags=flags)
     intrinsic(name, src_comp=[1], dest_comp=1, indices=[BASE], flags=flags)
 
 def atomic2(name):
-    intrinsic(name + "_deref", src_comp=[1, 1], dest_comp=1)
+    intrinsic(name + "_deref", src_comp=[-1, 1], dest_comp=1)
     intrinsic(name, src_comp=[1, 1], dest_comp=1, indices=[BASE])
 
 def atomic3(name):
-    intrinsic(name + "_deref", src_comp=[1, 1, 1], dest_comp=1)
+    intrinsic(name + "_deref", src_comp=[-1, 1, 1], dest_comp=1)
     intrinsic(name, src_comp=[1, 1, 1], dest_comp=1, indices=[BASE])
 
 atomic("atomic_counter_inc")
@@ -374,20 +374,20 @@ intrinsic("vulkan_resource_reindex", src_comp=[1, 1], dest_comp=1,
 # 1: The data parameter to the atomic function (i.e. the value to add
 #    in shared_atomic_add, etc).
 # 2: For CompSwap only: the second data parameter.
-intrinsic("deref_atomic_add",  src_comp=[1, 1], dest_comp=1)
-intrinsic("deref_atomic_imin", src_comp=[1, 1], dest_comp=1)
-intrinsic("deref_atomic_umin", src_comp=[1, 1], dest_comp=1)
-intrinsic("deref_atomic_imax", src_comp=[1, 1], dest_comp=1)
-intrinsic("deref_atomic_umax", src_comp=[1, 1], dest_comp=1)
-intrinsic("deref_atomic_and",  src_comp=[1, 1], dest_comp=1)
-intrinsic("deref_atomic_or",   src_comp=[1, 1], dest_comp=1)
-intrinsic("deref_atomic_xor",  src_comp=[1, 1], dest_comp=1)
-intrinsic("deref_atomic_exchange", src_comp=[1, 1], dest_comp=1)
-intrinsic("deref_atomic_comp_swap", src_comp=[1, 1, 1], dest_comp=1)
-intrinsic("deref_atomic_fadd",  src_comp=[1, 1], dest_comp=1)
-intrinsic("deref_atomic_fmin",  src_comp=[1, 1], dest_comp=1)
-intrinsic("deref_atomic_fmax",  src_comp=[1, 1], dest_comp=1)
-intrinsic("deref_atomic_fcomp_swap", src_comp=[1, 1, 1], dest_comp=1)
+intrinsic("deref_atomic_add",  src_comp=[-1, 1], dest_comp=1)
+intrinsic("deref_atomic_imin", src_comp=[-1, 1], dest_comp=1)
+intrinsic("deref_atomic_umin", src_comp=[-1, 1], dest_comp=1)
+intrinsic("deref_atomic_imax", src_comp=[-1, 1], dest_comp=1)
+intrinsic("deref_atomic_umax", src_comp=[-1, 1], dest_comp=1)
+intrinsic("deref_atomic_and",  src_comp=[-1, 1], dest_comp=1)
+intrinsic("deref_atomic_or",   src_comp=[-1, 1], dest_comp=1)
+intrinsic("deref_atomic_xor",  src_comp=[-1, 1], dest_comp=1)
+intrinsic("deref_atomic_exchange", src_comp=[-1, 1], dest_comp=1)
+intrinsic("deref_atomic_comp_swap", src_comp=[-1, 1, 1], dest_comp=1)
+intrinsic("deref_atomic_fadd",  src_comp=[-1, 1], dest_comp=1)
+intrinsic("deref_atomic_fmin",  src_comp=[-1, 1], dest_comp=1)
+intrinsic("deref_atomic_fmax",  src_comp=[-1, 1], dest_comp=1)
+intrinsic("deref_atomic_fcomp_swap", src_comp=[-1, 1, 1], dest_comp=1)
 
 # SSBO atomic intrinsics
 #