vulkan: for scalar FA, select between 1 and 8 rows

Author: jeffbolznv

ggml/src/ggml-vulkan/ggml-vulkan.cpp

@@ -1590,7 +1590,8 @@ static void ggml_vk_wait_events(vk_context& ctx, std::vector<vk::Event>&& events
 
 // number of rows/cols for flash attention shader
 static constexpr uint32_t flash_attention_num_small_rows = 32;
-static constexpr uint32_t scalar_flash_attention_num_small_rows = 8;
+static constexpr uint32_t scalar_flash_attention_num_small_rows = 1;
+static constexpr uint32_t scalar_flash_attention_num_large_rows = 8;
 
 static uint32_t get_fa_num_small_rows(bool scalar) {
     return scalar ? scalar_flash_attention_num_small_rows : flash_attention_num_small_rows;
@@ -1599,8 +1600,16 @@ static uint32_t get_fa_num_small_rows(bool scalar) {
 static std::array<uint32_t, 2> fa_rows_cols(bool scalar, uint32_t D, uint32_t clamp, ggml_type type, bool small_rows) {
     GGML_UNUSED(clamp);
 
+    if (scalar) {
+        if (small_rows) {
+            return {scalar_flash_attention_num_small_rows, 64};
+        } else {
+            return {scalar_flash_attention_num_large_rows, 32};
+        }
+    }
+
     // small rows, large cols
-    if (small_rows || scalar) {
+    if (small_rows) {
         return {get_fa_num_small_rows(scalar), 32};
     }
 
@@ -5729,8 +5738,29 @@ static void ggml_vk_flash_attn(ggml_backend_vk_context * ctx, vk_context& subctx
     assert(q->type == GGML_TYPE_F32);
     assert(k->type == v->type);
 
-    vk_pipeline *pipelines;
     bool scalar = !ctx->device->coopmat2;
+
+    uint32_t gqa_ratio = 1;
+    uint32_t qk_ratio = neq2 / nek2;
+    uint32_t workgroups_x = (uint32_t)neq1;
+    uint32_t workgroups_y = (uint32_t)neq2;
+    uint32_t workgroups_z = (uint32_t)neq3;
+
+    // For scalar FA, we can use the "large" size to accommodate qga.
+    // For coopmat FA, we always use the small size (which is still pretty large for gqa).
+    const uint32_t max_gqa = scalar ? scalar_flash_attention_num_large_rows : get_fa_num_small_rows(false);
+
+    if (N == 1 && qk_ratio > 1 && qk_ratio <= max_gqa &&
+        qk_ratio * nek2 == neq2 && nek2 == nev2 && neq3 == 1 && nek3 == 1 && nev3 == 1) {
+        // grouped query attention - make the N dimension equal to gqa_ratio, reduce
+        // workgroups proportionally in y dimension. The shader will detect gqa_ratio > 1
+        // and change addressing calculations to index Q's dimension 2.
+        gqa_ratio = qk_ratio;
+        N = gqa_ratio;
+        workgroups_y /= N;
+    }
+
+    vk_pipeline *pipelines;
     // XXX TODO other backends may be changing accumulator precision to default to f32 soon
     bool f32acc = scalar || dst->op_params[3] == GGML_PREC_F32;
     bool small_rows = N <= get_fa_num_small_rows(scalar);
@@ -5776,24 +5806,6 @@ static void ggml_vk_flash_attn(ggml_backend_vk_context * ctx, vk_context& subctx
     vk_pipeline pipeline = pipelines[aligned];
     assert(pipeline);
 
-    uint32_t gqa_ratio = 1;
-    uint32_t qk_ratio = neq2 / nek2;
-    uint32_t workgroups_x = (uint32_t)neq1;
-    uint32_t workgroups_y = (uint32_t)neq2;
-    uint32_t workgroups_z = (uint32_t)neq3;
-
-    const uint32_t max_gqa = get_fa_num_small_rows(scalar);
-
-    if (N == 1 && qk_ratio > 1 && qk_ratio <= max_gqa &&
-        qk_ratio * nek2 == neq2 && nek2 == nev2 && neq3 == 1 && nek3 == 1 && nev3 == 1) {
-        // grouped query attention - make the N dimension equal to gqa_ratio, reduce
-        // workgroups proportionally in y dimension. The shader will detect gqa_ratio > 1
-        // and change addressing calculations to index Q's dimension 2.
-        gqa_ratio = qk_ratio;
-        N = gqa_ratio;
-        workgroups_y /= N;
-    }
-
     uint32_t split_kv = KV;
     uint32_t split_k = 1;
 

ggml/src/ggml-vulkan/vulkan-shaders/flash_attn.comp

@@ -295,7 +295,9 @@ void main() {
             [[unroll]] for (uint32_t idx = 0; idx < Bc * Br; idx += gl_WorkGroupSize.x) {
                 uint32_t c = (idx + tid) % Bc;
                 uint32_t r = (idx + tid) / Bc;
-                masksh[c][r] = data_m[(i * Br + r) * m_stride + (j * Bc + c)];
+                if (idx + tid < Bc * Br) {
+                    masksh[c][r] = data_m[(i * Br + r) * m_stride + (j * Bc + c)];
+                }
             }
             barrier();