ac/llvm: Fix ac_build_reduce in wave32 mode.

Previously, when cluster_size was set to 0, it always worked as if
the cluster size was 64. This commit fixes it in wave32 mode by
changing to work as if the cluster size was set to 32.

Signed-off-by: Timur Kristóf <timur.kristof@gmail.com>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>

diff --git a/src/amd/llvm/ac_llvm_build.c b/src/amd/llvm/ac_llvm_build.c
index 79831260066a83e5bdd8789bc6d0da23b1c5a35a..e97eb919c6aa37fc8289be82ab34a02bd6c0869d 100644 (file)
--- a/src/amd/llvm/ac_llvm_build.c
+++ b/src/amd/llvm/ac_llvm_build.c
@@ -4344,12 +4344,15 @@ ac_build_reduce(struct ac_llvm_context *ctx, LLVMValueRef src, nir_op op, unsign
        if (cluster_size == 32) return ac_build_wwm(ctx, result);
 
        if (ctx->chip_class >= GFX8) {
-               if (ctx->chip_class >= GFX10)
-                       swap = ac_build_readlane(ctx, result, LLVMConstInt(ctx->i32, 31, false));
-               else
-                       swap = ac_build_dpp(ctx, identity, result, dpp_row_bcast31, 0xc, 0xf, false);
-               result = ac_build_alu_op(ctx, result, swap, op);
-               result = ac_build_readlane(ctx, result, LLVMConstInt(ctx->i32, 63, 0));
+               if (ctx->wave_size == 64) {
+                       if (ctx->chip_class >= GFX10)
+                               swap = ac_build_readlane(ctx, result, LLVMConstInt(ctx->i32, 31, false));
+                       else
+                               swap = ac_build_dpp(ctx, identity, result, dpp_row_bcast31, 0xc, 0xf, false);
+                       result = ac_build_alu_op(ctx, result, swap, op);
+                       result = ac_build_readlane(ctx, result, LLVMConstInt(ctx->i32, 63, 0));
+               }
+
                return ac_build_wwm(ctx, result);
        } else {
                swap = ac_build_readlane(ctx, result, ctx->i32_0);