Remove ixsimpl git dep and update lit tests for new simplification forms

ixsimpl is now vendored as a subfolder, so the pip git dependency is
no longer needed. Updated affine map CHECK lines in gather_to_shared
and gemm lit tests to match the new canonical form where `mod` is
expanded to `x - (x floordiv N) * N`.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Signed-off-by: Ivan Butygin <ivan.butygin@gmail.com>

Author: Hardcode84

lit_tests/kernel/wave/gather_to_shared.py

@@ -329,9 +329,9 @@ def repeat(acc: tkl.Register[M, N, tkl.f32]) -> tkl.Register[M, N, tkl.f32]:
     print(scaled_gemm.asm)
 
     # CHECK-LABEL:    test_gather_to_shared_scaled_dims
-    # CHECK-DAG:      #[[MAP_COL:.*]] = affine_map<()[s0] -> ((s0 floordiv 8) mod 8)>
+    # CHECK-DAG:      #[[MAP_COL:.*]] = affine_map<()[s0] -> (s0 floordiv 8 - (s0 floordiv 64) * 8)>
     # CHECK-DAG:      #[[MAP_ROW:.*]] = affine_map<()[s0] -> (s0 mod 8)>
-    # CHECK-DAG:      #[[MAP_COL_SCALE:.*]] = affine_map<()[s0] -> ((s0 floordiv 2) mod 2)>
+    # CHECK-DAG:      #[[MAP_COL_SCALE:.*]] = affine_map<()[s0] -> (s0 floordiv 2 - (s0 floordiv 4) * 2)>
     # CHECK-DAG:      #[[MAP_ROW_SCALE:.*]] = affine_map<()[s0] -> (s0 mod 2)>
     # CHECK-DAG:      #[[MAP_ROW_SWIZZLED:.*]] = affine_map<()[s0] -> ((s0 mod 64) floordiv 16)>
     # CHECK-DAG:      #[[MAP_ROW_SWIZZLED_2:.*]] = affine_map<()[s0] -> ((s0 mod 64) floordiv 16 + 4)>

lit_tests/kernel/wave/gemm.py

@@ -280,7 +280,7 @@ def repeat(acc: tkl.Register[M, N, tkl.f32]) -> tkl.Register[M, N, tkl.f32]:
 
     # CHECK-LABEL:    test_reordered_gemm
     # CHECK-DAG:        #[[MAP_LIN_A:.+]] = affine_map<()[s0, s1, s2, s3, s4] -> (s0 * 65536 + s2 * 16384 + s3 * 8 + s4 * 32 + (s1 floordiv 4) * 32768 + (s3 floordiv 4) * 480 - ((s2 * 32 + s3 floordiv 4) floordiv 64) * 32768 - ((s0 * 8 + s1) floordiv 32) * 262144)>
-    # CHECK-DAG:        #[[MAP_LIN_B:.+]] = affine_map<()[s0, s1, s2, s3, s4] -> (s0 * 32768 + s2 * 16384 + s3 * 8 + s4 * 32 + ((s0 + s1 * 8) floordiv 32) * 131072 + (s3 floordiv 4) * 480 - ((s2 * 32 + s3 floordiv 4) floordiv 64) * 32768 - (s0 floordiv 4) * 131072)>
+    # CHECK-DAG:        #[[MAP_LIN_B:.+]] = affine_map<()[s0, s1, s2, s3, s4] -> (s0 * 32768 + s2 * 16384 + s3 * 8 + s4 * 32 + ((s0 + s1 * 8) floordiv 32) * 131072 - (s0 floordiv 4) * 131072 + (s3 floordiv 4) * 480 - ((s2 * 32 + s3 floordiv 4) floordiv 64) * 32768)>
     # CHECK-DAG:        affine.apply #[[MAP_LIN_A]]()[%block_id_y, %block_id_x, %thread_id_y, %thread_id_x, {{.*}}]
     # CHECK-DAG:        affine.apply #[[MAP_LIN_B]]()[%block_id_x, %block_id_y, %thread_id_y, %thread_id_x, {{.*}}]
     # CHECK-DAG:        vector.load {{.*}} : memref<{{.*}}xf16, strided<[1]>>, vector<8xf16>

pyproject.toml

@@ -37,7 +37,6 @@ dependencies = [
     "sympy",
     "torch>=2.6,<2.10",
     "typing_extensions",
-    "ixsimpl @ git+https://github.com/hardcode84/4.git@d41a6255b62eb5b9192a2fecee0ef0330ce59f69",
 ]
 
 dynamic = ["version"]