/* * Copyright © 2022 Collabora Ltd. * SPDX-License-Identifier: MIT */ #include "mme_runner.h" #include "mme_tu104_sim.h" #include "clcd97.h" #include "nv_push_clc597.h" class mme_tu104_sim_test : public ::testing::Test, public mme_hw_runner { public: mme_tu104_sim_test(); ~mme_tu104_sim_test(); void SetUp(); void test_macro(const mme_builder *b, const std::vector& macro, const std::vector& params); }; mme_tu104_sim_test::mme_tu104_sim_test() : ::testing::Test(), mme_hw_runner() { } mme_tu104_sim_test::~mme_tu104_sim_test() { } void mme_tu104_sim_test::SetUp() { ASSERT_TRUE(set_up_hw(TURING_A, UINT16_MAX)); } void mme_tu104_sim_test::test_macro(const mme_builder *b, const std::vector& macro, const std::vector& params) { const uint32_t data_dwords = DATA_BO_SIZE / sizeof(uint32_t); std::vector insts(macro.size() / 3); mme_tu104_decode(&insts[0], ¯o[0], macro.size() / 3); /* First, make a copy of the data and simulate the macro */ std::vector sim_data(data, data + (DATA_BO_SIZE / 4)); mme_tu104_sim_mem sim_mem = { .addr = data_addr, .data = &sim_data[0], .size = DATA_BO_SIZE, }; mme_tu104_sim(insts.size(), &insts[0], params.size(), params.size() ? ¶ms[0] : NULL, 1, &sim_mem); /* Now run the macro on the GPU */ push_macro(0, macro); P_1INC(p, NVC597, CALL_MME_MACRO(0)); if (params.empty()) { P_NVC597_CALL_MME_MACRO(p, 0, 0); } else { P_INLINE_ARRAY(p, ¶ms[0], params.size()); } submit_push(); /* Check the results */ for (uint32_t i = 0; i < data_dwords; i++) ASSERT_EQ(data[i], sim_data[i]); } static mme_tu104_reg mme_value_as_reg(mme_value val) { assert(val.type == MME_VALUE_TYPE_REG); return (mme_tu104_reg)(MME_TU104_REG_R0 + val.reg); } TEST_F(mme_tu104_sim_test, sanity) { const uint32_t canary = 0xc0ffee01; mme_builder b; mme_builder_init(&b, devinfo); mme_store_imm_addr(&b, data_addr, mme_imm(canary)); auto macro = mme_builder_finish_vec(&b); std::vector params; test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, multi_param) { mme_builder b; mme_builder_init(&b, devinfo); mme_value v0 = mme_alloc_reg(&b); mme_value v1 = mme_alloc_reg(&b); mme_tu104_asm(&b, i) { i.alu[0].dst = mme_value_as_reg(v0); i.alu[0].src[0] = MME_TU104_REG_LOAD1; i.alu[1].dst = mme_value_as_reg(v1); i.alu[1].src[0] = MME_TU104_REG_LOAD0; i.imm[0] = (1<<12) | (NVC597_SET_MME_SHADOW_SCRATCH(12) >> 2); i.out[0].mthd = MME_TU104_OUT_OP_IMM0; i.out[0].emit = MME_TU104_OUT_OP_LOAD0; i.imm[1] = (1<<12) | (NVC597_SET_MME_SHADOW_SCRATCH(35) >> 2); i.out[1].mthd = MME_TU104_OUT_OP_IMM1; i.out[1].emit = MME_TU104_OUT_OP_LOAD1; } mme_value v2 = mme_state(&b, NVC597_SET_MME_SHADOW_SCRATCH(12)); mme_value v3 = mme_state(&b, NVC597_SET_MME_SHADOW_SCRATCH(35)); mme_store_imm_addr(&b, data_addr + 0, v0); mme_store_imm_addr(&b, data_addr + 4, v1); mme_store_imm_addr(&b, data_addr + 8, v2); mme_store_imm_addr(&b, data_addr + 12, v3); auto macro = mme_builder_finish_vec(&b); std::vector params; params.push_back(2581); params.push_back(3048); test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, pred_param) { mme_builder b; mme_builder_init(&b, devinfo); mme_value v0 = mme_load(&b); mme_value v1 = mme_mov(&b, mme_imm(240)); mme_tu104_asm(&b, i) { i.pred_mode = MME_TU104_PRED_TTTT; i.alu[0].dst = mme_value_as_reg(v1); i.alu[0].src[0] = MME_TU104_REG_LOAD0; } mme_value v2 = mme_load(&b); mme_store_imm_addr(&b, data_addr + 0, v0); mme_store_imm_addr(&b, data_addr + 4, v1); mme_store_imm_addr(&b, data_addr + 8, v2); auto macro = mme_builder_finish_vec(&b); for (uint32_t j = 0; j < 4; j++) { reset_push(); std::vector params; params.push_back((j & 1) * 2043); params.push_back((j & 2) * 523); params.push_back(2581); params.push_back(3048); test_macro(&b, macro, params); } } TEST_F(mme_tu104_sim_test, out_imm0) { mme_builder b; mme_builder_init(&b, devinfo); mme_mthd(&b, NVC597_SET_REPORT_SEMAPHORE_A); mme_emit(&b, mme_imm(high32(data_addr + 0))); mme_emit(&b, mme_imm(low32(data_addr + 0))); mme_tu104_asm(&b, i) { i.imm[0] = 0x1234; i.out[0].emit = MME_TU104_OUT_OP_IMM0; } mme_emit(&b, mme_imm(0x10000000)); mme_mthd(&b, NVC597_SET_REPORT_SEMAPHORE_A); mme_emit(&b, mme_imm(high32(data_addr + 4))); mme_emit(&b, mme_imm(low32(data_addr + 4))); mme_tu104_asm(&b, i) { i.imm[0] = 0x8765; i.out[0].emit = MME_TU104_OUT_OP_IMM0; } mme_emit(&b, mme_imm(0x10000000)); auto macro = mme_builder_finish_vec(&b); std::vector params; test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, out_imm1) { mme_builder b; mme_builder_init(&b, devinfo); mme_mthd(&b, NVC597_SET_REPORT_SEMAPHORE_A); mme_emit(&b, mme_imm(high32(data_addr + 0))); mme_emit(&b, mme_imm(low32(data_addr + 0))); mme_tu104_asm(&b, i) { i.imm[1] = 0x1234; i.out[0].emit = MME_TU104_OUT_OP_IMM1; } mme_emit(&b, mme_imm(0x10000000)); mme_mthd(&b, NVC597_SET_REPORT_SEMAPHORE_A); mme_emit(&b, mme_imm(high32(data_addr + 4))); mme_emit(&b, mme_imm(low32(data_addr + 4))); mme_tu104_asm(&b, i) { i.imm[1] = 0x8765; i.out[0].emit = MME_TU104_OUT_OP_IMM1; } mme_emit(&b, mme_imm(0x10000000)); auto macro = mme_builder_finish_vec(&b); std::vector params; test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, out_immhigh0) { mme_builder b; mme_builder_init(&b, devinfo); mme_mthd(&b, NVC597_SET_REPORT_SEMAPHORE_A); mme_emit(&b, mme_imm(high32(data_addr + 0))); mme_emit(&b, mme_imm(low32(data_addr + 0))); mme_tu104_asm(&b, i) { i.imm[0] = 0x1234; i.out[0].emit = MME_TU104_OUT_OP_IMMHIGH0; } mme_emit(&b, mme_imm(0x10000000)); mme_mthd(&b, NVC597_SET_REPORT_SEMAPHORE_A); mme_emit(&b, mme_imm(high32(data_addr + 4))); mme_emit(&b, mme_imm(low32(data_addr + 4))); mme_tu104_asm(&b, i) { i.imm[0] = 0x8765; i.out[1].emit = MME_TU104_OUT_OP_IMMHIGH0; } mme_emit(&b, mme_imm(0x10000000)); auto macro = mme_builder_finish_vec(&b); std::vector params; test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, out_immhigh1) { mme_builder b; mme_builder_init(&b, devinfo); mme_mthd(&b, NVC597_SET_REPORT_SEMAPHORE_A); mme_emit(&b, mme_imm(high32(data_addr + 0))); mme_emit(&b, mme_imm(low32(data_addr + 0))); mme_tu104_asm(&b, i) { i.imm[1] = 0x1234; i.out[0].emit = MME_TU104_OUT_OP_IMMHIGH1; } mme_emit(&b, mme_imm(0x10000000)); mme_mthd(&b, NVC597_SET_REPORT_SEMAPHORE_A); mme_emit(&b, mme_imm(high32(data_addr + 4))); mme_emit(&b, mme_imm(low32(data_addr + 4))); mme_tu104_asm(&b, i) { i.imm[1] = 0x8765; i.out[1].emit = MME_TU104_OUT_OP_IMMHIGH1; } mme_emit(&b, mme_imm(0x10000000)); auto macro = mme_builder_finish_vec(&b); std::vector params; test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, out_imm32) { mme_builder b; mme_builder_init(&b, devinfo); mme_mthd(&b, NVC597_SET_REPORT_SEMAPHORE_A); mme_emit(&b, mme_imm(high32(data_addr + 0))); mme_emit(&b, mme_imm(low32(data_addr + 0))); mme_tu104_asm(&b, i) { i.imm[0] = 0x1234; i.imm[1] = 0x7654; i.out[0].emit = MME_TU104_OUT_OP_IMM32; } mme_emit(&b, mme_imm(0x10000000)); mme_mthd(&b, NVC597_SET_REPORT_SEMAPHORE_A); mme_emit(&b, mme_imm(high32(data_addr + 4))); mme_emit(&b, mme_imm(low32(data_addr + 4))); mme_tu104_asm(&b, i) { i.imm[0] = 0x1234; i.imm[1] = 0x7654; i.out[1].emit = MME_TU104_OUT_OP_IMM32; } mme_emit(&b, mme_imm(0x10000000)); auto macro = mme_builder_finish_vec(&b); std::vector params; test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, reg_imm32) { const uint32_t canary = 0xc0ffee01; mme_builder b; mme_builder_init(&b, devinfo); mme_value v = mme_alloc_reg(&b); mme_tu104_asm(&b, i) { i.alu[0].dst = mme_value_as_reg(v); i.alu[0].op = MME_TU104_ALU_OP_ADD; i.alu[0].src[0] = MME_TU104_REG_IMM32, i.imm[0] = (uint16_t)canary; i.imm[1] = (uint16_t)(canary >> 16); } mme_store_imm_addr(&b, data_addr, v); auto macro = mme_builder_finish_vec(&b); std::vector params; test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, pred_alu) { static const enum mme_tu104_pred preds[] = { MME_TU104_PRED_UUUU, MME_TU104_PRED_TTTT, MME_TU104_PRED_FFFF, MME_TU104_PRED_TTUU, MME_TU104_PRED_FFUU, MME_TU104_PRED_TFUU, MME_TU104_PRED_TUUU, MME_TU104_PRED_FUUU, MME_TU104_PRED_UUTT, MME_TU104_PRED_UUTF, MME_TU104_PRED_UUTU, MME_TU104_PRED_UUFT, MME_TU104_PRED_UUFF, MME_TU104_PRED_UUFU, MME_TU104_PRED_UUUT, MME_TU104_PRED_UUUF, }; for (uint32_t i = 0; i < ARRAY_SIZE(preds); i++) { mme_builder b; mme_builder_init(&b, devinfo); mme_value pred = mme_load(&b); mme_value v0 = mme_mov(&b, mme_imm(i * 100 + 13)); mme_value v1 = mme_mov(&b, mme_imm(i * 100 + 62)); mme_tu104_asm(&b, inst) { inst.pred = mme_value_as_reg(pred); inst.pred_mode = preds[i]; inst.alu[0].dst = mme_value_as_reg(v0); inst.alu[0].src[0] = MME_TU104_REG_IMM; inst.imm[0] = i * 100 + 25; inst.alu[1].dst = mme_value_as_reg(v1); inst.alu[1].src[0] = MME_TU104_REG_IMM; inst.imm[1] = i * 100 + 73; } mme_store_imm_addr(&b, data_addr + i * 8 + 0, v0); mme_store_imm_addr(&b, data_addr + i * 8 + 4, v1); auto macro = mme_builder_finish_vec(&b); for (uint32_t j = 0; j < 2; j++) { reset_push(); std::vector params; params.push_back(j * 25894); test_macro(&b, macro, params); } } } TEST_F(mme_tu104_sim_test, pred_out) { static const enum mme_tu104_pred preds[] = { MME_TU104_PRED_UUUU, MME_TU104_PRED_TTTT, MME_TU104_PRED_FFFF, MME_TU104_PRED_TTUU, MME_TU104_PRED_FFUU, MME_TU104_PRED_TFUU, MME_TU104_PRED_TUUU, MME_TU104_PRED_FUUU, MME_TU104_PRED_UUTT, MME_TU104_PRED_UUTF, MME_TU104_PRED_UUTU, MME_TU104_PRED_UUFT, MME_TU104_PRED_UUFF, MME_TU104_PRED_UUFU, MME_TU104_PRED_UUUT, MME_TU104_PRED_UUUF, }; for (uint32_t i = 0; i < ARRAY_SIZE(preds); i++) { mme_builder b; mme_builder_init(&b, devinfo); mme_value pred = mme_load(&b); mme_tu104_asm(&b, inst) { inst.imm[0] = (1<<12) | (NVC597_SET_MME_SHADOW_SCRATCH(i*2 + 0) >> 2); inst.imm[1] = i * 100 + 25; inst.out[0].mthd = MME_TU104_OUT_OP_IMM0; inst.out[0].emit = MME_TU104_OUT_OP_IMM1; } mme_tu104_asm(&b, inst) { inst.imm[0] = (1<<12) | (NVC597_SET_MME_SHADOW_SCRATCH(i*2 + 1) >> 2); inst.imm[1] = i * 100 + 75; inst.out[0].mthd = MME_TU104_OUT_OP_IMM0; inst.out[0].emit = MME_TU104_OUT_OP_IMM1; } mme_tu104_asm(&b, inst) { inst.pred = mme_value_as_reg(pred); inst.pred_mode = preds[i]; inst.imm[0] = (1<<12) | (NVC597_SET_MME_SHADOW_SCRATCH(i*2 + 0) >> 2); inst.imm[1] = (1<<12) | (NVC597_SET_MME_SHADOW_SCRATCH(i*2 + 1) >> 2); inst.out[0].mthd = MME_TU104_OUT_OP_IMM0; inst.out[0].emit = MME_TU104_OUT_OP_IMM1; inst.out[1].mthd = MME_TU104_OUT_OP_IMM1; inst.out[1].emit = MME_TU104_OUT_OP_IMM0; } mme_value v0 = mme_state(&b, NVC597_SET_MME_SHADOW_SCRATCH(i*2 + 0)); mme_value v1 = mme_state(&b, NVC597_SET_MME_SHADOW_SCRATCH(i*2 + 1)); mme_store_imm_addr(&b, data_addr + i * 8 + 0, v0); mme_store_imm_addr(&b, data_addr + i * 8 + 4, v1); auto macro = mme_builder_finish_vec(&b); for (uint32_t j = 0; j < 2; j++) { reset_push(); std::vector params; params.push_back(j * 25894); test_macro(&b, macro, params); } } } TEST_F(mme_tu104_sim_test, add) { mme_builder b; mme_builder_init(&b, devinfo); mme_value x = mme_load(&b); mme_value y = mme_load(&b); mme_value sum = mme_add(&b, x, y); mme_store_imm_addr(&b, data_addr, sum); auto macro = mme_builder_finish_vec(&b); std::vector params; params.push_back(25); params.push_back(138); test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, add_imm) { mme_builder b; mme_builder_init(&b, devinfo); mme_value x = mme_load(&b); mme_value v0 = mme_add(&b, x, mme_imm(0x00000001)); mme_value v1 = mme_add(&b, x, mme_imm(0xffffffff)); mme_value v2 = mme_add(&b, x, mme_imm(0xffff8000)); mme_value v3 = mme_add(&b, mme_imm(0x00000001), x); mme_value v4 = mme_add(&b, mme_imm(0xffffffff), x); mme_value v5 = mme_add(&b, mme_imm(0xffff8000), x); mme_value v6 = mme_add(&b, mme_zero(), mme_imm(0x00000001)); mme_value v7 = mme_add(&b, mme_zero(), mme_imm(0xffffffff)); mme_value v8 = mme_add(&b, mme_zero(), mme_imm(0xffff8000)); mme_store_imm_addr(&b, data_addr + 0, v0); mme_store_imm_addr(&b, data_addr + 4, v1); mme_store_imm_addr(&b, data_addr + 8, v2); mme_store_imm_addr(&b, data_addr + 12, v3); mme_store_imm_addr(&b, data_addr + 16, v4); mme_store_imm_addr(&b, data_addr + 20, v5); mme_store_imm_addr(&b, data_addr + 24, v6); mme_store_imm_addr(&b, data_addr + 28, v7); mme_store_imm_addr(&b, data_addr + 32, v8); auto macro = mme_builder_finish_vec(&b); uint32_t vals[] = { 0x0000ffff, 0x00008000, 0x0001ffff, 0xffffffff, }; for (uint32_t i = 0; i < ARRAY_SIZE(vals); i++) { reset_push(); std::vector params; params.push_back(vals[i]); test_macro(&b, macro, params); } } TEST_F(mme_tu104_sim_test, addc) { mme_builder b; mme_builder_init(&b, devinfo); struct mme_value64 x = { mme_load(&b), mme_load(&b) }; struct mme_value64 y = { mme_load(&b), mme_load(&b) }; struct mme_value64 sum = mme_add64(&b, x, y); mme_store_imm_addr(&b, data_addr + 0, sum.lo); mme_store_imm_addr(&b, data_addr + 4, sum.hi); auto macro = mme_builder_finish_vec(&b); std::vector params; params.push_back(0x80008650); params.push_back(0x596); params.push_back(0x8000a8f6); params.push_back(0x836); test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, addc_imm) { mme_builder b; mme_builder_init(&b, devinfo); mme_value x_lo = mme_load(&b); mme_value x_hi = mme_load(&b); mme_value v1_lo = mme_alloc_reg(&b); mme_value v1_hi = mme_alloc_reg(&b); mme_tu104_asm(&b, i) { i.alu[0].dst = mme_value_as_reg(v1_lo); i.alu[0].op = MME_TU104_ALU_OP_ADD; i.alu[0].src[0] = mme_value_as_reg(x_lo); i.alu[0].src[1] = MME_TU104_REG_IMM; i.imm[0] = 0x0001; i.alu[1].dst = mme_value_as_reg(v1_hi); i.alu[1].op = MME_TU104_ALU_OP_ADDC; i.alu[1].src[0] = mme_value_as_reg(x_hi); i.alu[1].src[1] = MME_TU104_REG_IMM; i.imm[1] = 0x0000; } mme_value v2_lo = mme_alloc_reg(&b); mme_value v2_hi = mme_alloc_reg(&b); mme_tu104_asm(&b, i) { i.alu[0].dst = mme_value_as_reg(v2_lo); i.alu[0].op = MME_TU104_ALU_OP_ADD; i.alu[0].src[0] = mme_value_as_reg(x_lo); i.alu[0].src[1] = MME_TU104_REG_IMM; i.imm[0] = 0x0000; i.alu[1].dst = mme_value_as_reg(v2_hi); i.alu[1].op = MME_TU104_ALU_OP_ADDC; i.alu[1].src[0] = mme_value_as_reg(x_hi); i.alu[1].src[1] = MME_TU104_REG_IMM; i.imm[1] = 0x0001; } mme_value v3_lo = mme_alloc_reg(&b); mme_value v3_hi = mme_alloc_reg(&b); mme_tu104_asm(&b, i) { i.alu[0].dst = mme_value_as_reg(v3_lo); i.alu[0].op = MME_TU104_ALU_OP_ADD; i.alu[0].src[0] = mme_value_as_reg(x_lo); i.alu[0].src[1] = MME_TU104_REG_IMM; i.imm[0] = 0x0000; i.alu[1].dst = mme_value_as_reg(v3_hi); i.alu[1].op = MME_TU104_ALU_OP_ADDC; i.alu[1].src[0] = mme_value_as_reg(x_hi); i.alu[1].src[1] = MME_TU104_REG_IMM; i.imm[1] = 0xffff; } mme_value v4_lo = mme_alloc_reg(&b); mme_value v4_hi = mme_alloc_reg(&b); mme_tu104_asm(&b, i) { i.alu[0].dst = mme_value_as_reg(v4_lo); i.alu[0].op = MME_TU104_ALU_OP_ADD; i.alu[0].src[0] = mme_value_as_reg(x_lo); i.alu[0].src[1] = MME_TU104_REG_IMM; i.imm[0] = 0x0000; i.alu[1].dst = mme_value_as_reg(v4_hi); i.alu[1].op = MME_TU104_ALU_OP_ADDC; i.alu[1].src[0] = mme_value_as_reg(x_hi); i.alu[1].src[1] = MME_TU104_REG_IMM; i.imm[1] = 0x8000; } mme_store_imm_addr(&b, data_addr + 0, v1_lo); mme_store_imm_addr(&b, data_addr + 4, v1_hi); mme_store_imm_addr(&b, data_addr + 8, v2_lo); mme_store_imm_addr(&b, data_addr + 12, v2_hi); mme_store_imm_addr(&b, data_addr + 16, v3_lo); mme_store_imm_addr(&b, data_addr + 20, v3_hi); mme_store_imm_addr(&b, data_addr + 24, v4_lo); mme_store_imm_addr(&b, data_addr + 28, v4_hi); auto macro = mme_builder_finish_vec(&b); uint64_t vals[] = { 0x0000ffffffffffffull, 0x0000ffffffff8000ull, 0x0000ffff00000000ull, 0x0000800000000000ull, 0x00008000ffffffffull, 0x0001ffff00000000ull, 0xffffffff00000000ull, 0xffffffffffffffffull, }; for (uint32_t i = 0; i < ARRAY_SIZE(vals); i++) { reset_push(); std::vector params; params.push_back(low32(vals[i])); params.push_back(high32(vals[i])); test_macro(&b, macro, params); } } TEST_F(mme_tu104_sim_test, sub) { mme_builder b; mme_builder_init(&b, devinfo); mme_value x = mme_load(&b); mme_value y = mme_load(&b); mme_value diff = mme_sub(&b, x, y); mme_store_imm_addr(&b, data_addr, diff); auto macro = mme_builder_finish_vec(&b); std::vector params; params.push_back(25); params.push_back(138); test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, subb) { mme_builder b; mme_builder_init(&b, devinfo); struct mme_value64 x = { mme_load(&b), mme_load(&b) }; struct mme_value64 y = { mme_load(&b), mme_load(&b) }; struct mme_value64 diff = mme_sub64(&b, x, y); mme_store_imm_addr(&b, data_addr + 0, diff.lo); mme_store_imm_addr(&b, data_addr + 4, diff.hi); auto macro = mme_builder_finish_vec(&b); std::vector params; params.push_back(0x80008650); params.push_back(0x596); params.push_back(0x8000a8f6); params.push_back(0x836); test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, mul) { mme_builder b; mme_builder_init(&b, devinfo); mme_value x = mme_load(&b); mme_value y = mme_load(&b); mme_value sum = mme_mul(&b, x, y); mme_store_imm_addr(&b, data_addr, sum); auto macro = mme_builder_finish_vec(&b); std::vector params; params.push_back(25); params.push_back(138); test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, mul_imm) { mme_builder b; mme_builder_init(&b, devinfo); mme_value x = mme_load(&b); mme_value v0 = mme_mul(&b, x, mme_imm(0x00000001)); mme_value v1 = mme_mul(&b, x, mme_imm(0xffffffff)); mme_value v2 = mme_mul(&b, x, mme_imm(0xffff8000)); mme_value v3 = mme_mul(&b, mme_imm(0x00000001), x); mme_value v4 = mme_mul(&b, mme_imm(0xffffffff), x); mme_value v5 = mme_mul(&b, mme_imm(0xffff8000), x); mme_store_imm_addr(&b, data_addr + 0, v0); mme_store_imm_addr(&b, data_addr + 4, v1); mme_store_imm_addr(&b, data_addr + 8, v2); mme_store_imm_addr(&b, data_addr + 12, v3); mme_store_imm_addr(&b, data_addr + 16, v4); mme_store_imm_addr(&b, data_addr + 20, v5); auto macro = mme_builder_finish_vec(&b); int32_t vals[] = { 1, -5, -1, 5 }; for (uint32_t i = 0; i < ARRAY_SIZE(vals); i++) { reset_push(); std::vector params; params.push_back(vals[i]); test_macro(&b, macro, params); } } TEST_F(mme_tu104_sim_test, mul_mulh) { mme_builder b; mme_builder_init(&b, devinfo); mme_value x = mme_load(&b); mme_value y = mme_load(&b); struct mme_value64 prod = mme_imul_32x32_64(&b, x, y); mme_store_imm_addr(&b, data_addr + 0, prod.lo); mme_store_imm_addr(&b, data_addr + 4, prod.hi); auto macro = mme_builder_finish_vec(&b); std::vector params; params.push_back(0x80008650); params.push_back(0x596); test_macro(&b, macro, params); } static inline struct mme_value mme_mulu(struct mme_builder *b, struct mme_value x, struct mme_value y) { return mme_alu(b, MME_ALU_OP_MULU, x, y); } TEST_F(mme_tu104_sim_test, mulu_imm) { mme_builder b; mme_builder_init(&b, devinfo); mme_value x = mme_load(&b); mme_value v0 = mme_mulu(&b, x, mme_imm(0x00000001)); mme_value v1 = mme_mulu(&b, x, mme_imm(0xffffffff)); mme_value v2 = mme_mulu(&b, x, mme_imm(0xffff8000)); mme_value v3 = mme_mulu(&b, mme_imm(0x00000001), x); mme_value v4 = mme_mulu(&b, mme_imm(0xffffffff), x); mme_value v5 = mme_mulu(&b, mme_imm(0xffff8000), x); mme_store_imm_addr(&b, data_addr + 0, v0); mme_store_imm_addr(&b, data_addr + 4, v1); mme_store_imm_addr(&b, data_addr + 8, v2); mme_store_imm_addr(&b, data_addr + 12, v3); mme_store_imm_addr(&b, data_addr + 16, v4); mme_store_imm_addr(&b, data_addr + 20, v5); auto macro = mme_builder_finish_vec(&b); int32_t vals[] = { 1, -5, -1, 5 }; for (uint32_t i = 0; i < ARRAY_SIZE(vals); i++) { reset_push(); std::vector params; params.push_back(vals[i]); test_macro(&b, macro, params); } } TEST_F(mme_tu104_sim_test, mulu_mulh) { mme_builder b; mme_builder_init(&b, devinfo); mme_value x = mme_load(&b); mme_value y = mme_load(&b); struct mme_value64 prod = mme_umul_32x32_64(&b, x, y); mme_store_imm_addr(&b, data_addr + 0, prod.lo); mme_store_imm_addr(&b, data_addr + 4, prod.hi); auto macro = mme_builder_finish_vec(&b); std::vector params; params.push_back(0x80008650); params.push_back(0x596); test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, clz) { mme_builder b; mme_builder_init(&b, devinfo); mme_value bits = mme_clz(&b, mme_load(&b)); mme_store_imm_addr(&b, data_addr, bits); auto macro = mme_builder_finish_vec(&b); std::vector params; params.push_back(0x00406fe0); test_macro(&b, macro, params); } #define SHIFT_TEST(op) \ TEST_F(mme_tu104_sim_test, op) \ { \ mme_builder b; \ mme_builder_init(&b, devinfo); \ \ mme_value val = mme_load(&b); \ mme_value shift1 = mme_load(&b); \ mme_value shift2 = mme_load(&b); \ mme_store_imm_addr(&b, data_addr + 0, mme_##op(&b, val, shift1)); \ mme_store_imm_addr(&b, data_addr + 4, mme_##op(&b, val, shift2)); \ \ auto macro = mme_builder_finish_vec(&b); \ \ std::vector params; \ params.push_back(0x0c406fe0); \ params.push_back(5); \ params.push_back(51); \ \ test_macro(&b, macro, params); \ } SHIFT_TEST(sll) SHIFT_TEST(srl) SHIFT_TEST(sra) #undef SHIFT_TEST TEST_F(mme_tu104_sim_test, bfe) { const uint32_t canary = 0xc0ffee01; mme_builder b; mme_builder_init(&b, devinfo); mme_value val = mme_load(&b); mme_value pos = mme_load(&b); mme_store_imm_addr(&b, data_addr + 0, mme_bfe(&b, val, pos, 1), true); mme_store_imm_addr(&b, data_addr + 4, mme_bfe(&b, val, pos, 2), true); mme_store_imm_addr(&b, data_addr + 8, mme_bfe(&b, val, pos, 5), true); auto macro = mme_builder_finish_vec(&b); for (unsigned i = 0; i < 31; i++) { std::vector params; params.push_back(canary); params.push_back(i); test_macro(&b, macro, params); ASSERT_EQ(data[0], (canary >> i) & 0x1); ASSERT_EQ(data[1], (canary >> i) & 0x3); ASSERT_EQ(data[2], (canary >> i) & 0x1f); } } TEST_F(mme_tu104_sim_test, not) { mme_builder b; mme_builder_init(&b, devinfo); mme_value x = mme_load(&b); mme_value v1 = mme_not(&b, x); mme_store_imm_addr(&b, data_addr + 0, v1); auto macro = mme_builder_finish_vec(&b); std::vector params; params.push_back(0x0c406fe0); test_macro(&b, macro, params); } #define BITOP_TEST(op) \ TEST_F(mme_tu104_sim_test, op) \ { \ mme_builder b; \ mme_builder_init(&b, devinfo); \ \ mme_value x = mme_load(&b); \ mme_value y = mme_load(&b); \ mme_value v1 = mme_##op(&b, x, y); \ mme_value v2 = mme_##op(&b, x, mme_imm(0xffff8000)); \ mme_value v3 = mme_##op(&b, x, mme_imm(0xffffffff)); \ mme_store_imm_addr(&b, data_addr + 0, v1); \ mme_store_imm_addr(&b, data_addr + 4, v2); \ mme_store_imm_addr(&b, data_addr + 8, v3); \ \ auto macro = mme_builder_finish_vec(&b); \ \ std::vector params; \ params.push_back(0x0c406fe0); \ params.push_back(0x00fff0c0); \ \ test_macro(&b, macro, params); \ } BITOP_TEST(and) BITOP_TEST(and_not) BITOP_TEST(nand) BITOP_TEST(or) BITOP_TEST(xor) #undef BITOP_TEST TEST_F(mme_tu104_sim_test, merge) { mme_builder b; mme_builder_init(&b, devinfo); mme_value x = mme_load(&b); mme_value y = mme_load(&b); mme_value m1 = mme_merge(&b, x, y, 12, 12, 20); mme_value m2 = mme_merge(&b, x, y, 12, 8, 20); mme_value m3 = mme_merge(&b, x, y, 8, 12, 20); mme_value m4 = mme_merge(&b, x, y, 12, 16, 8); mme_value m5 = mme_merge(&b, x, y, 24, 12, 8); mme_store_imm_addr(&b, data_addr + 0, m1); mme_store_imm_addr(&b, data_addr + 4, m2); mme_store_imm_addr(&b, data_addr + 8, m3); mme_store_imm_addr(&b, data_addr + 12, m4); mme_store_imm_addr(&b, data_addr + 16, m5); auto macro = mme_builder_finish_vec(&b); std::vector params; params.push_back(0x0c406fe0); params.push_back(0x76543210u); test_macro(&b, macro, params); } #define COMPARISON_TEST(op) \ TEST_F(mme_tu104_sim_test, op) \ { \ mme_builder b; \ mme_builder_init(&b, devinfo); \ \ mme_value x = mme_load(&b); \ mme_value y = mme_load(&b); \ mme_value z = mme_load(&b); \ mme_value w = mme_load(&b); \ \ mme_value v1 = mme_##op(&b, x, y); \ mme_value v2 = mme_##op(&b, y, x); \ mme_value v3 = mme_##op(&b, y, z); \ mme_value v4 = mme_##op(&b, z, y); \ mme_value v5 = mme_##op(&b, w, z); \ mme_value v6 = mme_##op(&b, z, w); \ mme_value v7 = mme_##op(&b, w, w); \ \ mme_store_imm_addr(&b, data_addr + 0, v1); \ mme_store_imm_addr(&b, data_addr + 4, v2); \ mme_store_imm_addr(&b, data_addr + 8, v3); \ mme_store_imm_addr(&b, data_addr + 12, v4); \ mme_store_imm_addr(&b, data_addr + 16, v5); \ mme_store_imm_addr(&b, data_addr + 20, v6); \ mme_store_imm_addr(&b, data_addr + 24, v7); \ \ auto macro = mme_builder_finish_vec(&b); \ \ std::vector params; \ params.push_back(-5); \ params.push_back(-10); \ params.push_back(5); \ params.push_back(10); \ \ test_macro(&b, macro, params); \ } COMPARISON_TEST(slt) COMPARISON_TEST(sltu) COMPARISON_TEST(sle) COMPARISON_TEST(sleu) COMPARISON_TEST(seq) #undef COMPARISON_TEST static inline void mme_inc_whole_inst(mme_builder *b, mme_value val) { mme_tu104_asm(b, i) { i.alu[0].dst = mme_value_as_reg(val); i.alu[0].op = MME_TU104_ALU_OP_ADD; i.alu[0].src[0] = mme_value_as_reg(val); i.alu[0].src[1] = MME_TU104_REG_IMM; i.imm[0] = 1; } } TEST_F(mme_tu104_sim_test, loop) { mme_builder b; mme_builder_init(&b, devinfo); mme_value count = mme_load(&b); mme_value x = mme_mov(&b, mme_zero()); mme_value y = mme_mov(&b, mme_zero()); mme_loop(&b, count) { mme_tu104_asm(&b, i) { } /* noop */ mme_add_to(&b, x, x, count); } mme_add_to(&b, y, y, mme_imm(1)); mme_tu104_asm(&b, i) { } /* noop */ mme_tu104_asm(&b, i) { } /* noop */ mme_tu104_asm(&b, i) { } /* noop */ mme_store_imm_addr(&b, data_addr + 0, count); mme_store_imm_addr(&b, data_addr + 4, x); mme_store_imm_addr(&b, data_addr + 8, y); auto macro = mme_builder_finish_vec(&b); uint32_t counts[] = {0, 1, 5, 9}; for (uint32_t i = 0; i < ARRAY_SIZE(counts); i++) { reset_push(); std::vector params; params.push_back(counts[i]); test_macro(&b, macro, params); ASSERT_EQ(data[0], counts[i]); ASSERT_EQ(data[1], counts[i] * counts[i]); ASSERT_EQ(data[2], 1); } } TEST_F(mme_tu104_sim_test, jal) { mme_builder b; mme_builder_init(&b, devinfo); mme_value x = mme_mov(&b, mme_zero()); mme_value y = mme_mov(&b, mme_zero()); mme_tu104_asm(&b, i) { i.alu[0].op = MME_TU104_ALU_OP_JAL; i.imm[0] = (1 << 15) | 6; } for (uint32_t j = 0; j < 10; j++) mme_inc_whole_inst(&b, x); // mme_tu104_asm(&b, i) { // i.alu[0].op = MME_TU104_ALU_OP_JAL; // i.imm[0] = 6; // } // // for (uint32_t j = 0; j < 10; j++) // mme_inc_whole_inst(&b, y); mme_store_imm_addr(&b, data_addr + 0, x); mme_store_imm_addr(&b, data_addr + 4, y); auto macro = mme_builder_finish_vec(&b); std::vector params; test_macro(&b, macro, params); ASSERT_EQ(data[0], 5); } TEST_F(mme_tu104_sim_test, bxx_fwd) { mme_builder b; mme_builder_init(&b, devinfo); mme_value vals[10]; for (uint32_t i = 0; i < 10; i++) vals[i] = mme_mov(&b, mme_zero()); mme_tu104_asm(&b, i) { i.alu[0].op = MME_TU104_ALU_OP_BEQ; i.imm[0] = (1 << 15) | 6; } for (uint32_t j = 0; j < 10; j++) mme_inc_whole_inst(&b, vals[j]); for (uint32_t j = 0; j < 10; j++) mme_store_imm_addr(&b, data_addr + j * 4, vals[j]); auto macro = mme_builder_finish_vec(&b); std::vector params; test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, bxx_bwd) { mme_builder b; mme_builder_init(&b, devinfo); mme_value vals[15]; for (uint32_t i = 0; i < 15; i++) vals[i] = mme_mov(&b, mme_zero()); mme_tu104_asm(&b, i) { i.alu[0].op = MME_TU104_ALU_OP_JAL; i.imm[0] = (1 << 15) | 12; } for (uint32_t j = 0; j < 10; j++) mme_inc_whole_inst(&b, vals[j]); mme_tu104_asm(&b, i) { i.alu[0].op = MME_TU104_ALU_OP_JAL; i.imm[0] = (1 << 15) | 2; } mme_tu104_asm(&b, i) { i.alu[0].op = MME_TU104_ALU_OP_BEQ; i.imm[0] = (1 << 15) | ((-8) & 0x1fff); } for (uint32_t j = 10; j < 15; j++) mme_inc_whole_inst(&b, vals[j]); for (uint32_t j = 0; j < 15; j++) mme_store_imm_addr(&b, data_addr + j * 4, vals[j]); auto macro = mme_builder_finish_vec(&b); std::vector params; test_macro(&b, macro, params); for (uint32_t j = 0; j < 3; j++) ASSERT_EQ(data[j], 0); for (uint32_t j = 3; j < 15; j++) ASSERT_EQ(data[j], 1); } TEST_F(mme_tu104_sim_test, bxx_exit) { mme_builder b; mme_builder_init(&b, devinfo); mme_value vals[10]; for (uint32_t i = 0; i < 10; i++) vals[i] = mme_mov(&b, mme_zero()); for (uint32_t i = 0; i < 10; i++) mme_store_imm_addr(&b, data_addr + i * 4, mme_imm(0)); mme_tu104_asm(&b, i) { i.alu[0].op = MME_TU104_ALU_OP_BEQ; i.imm[0] = (1 << 15) | 0x1000; } /* those writes won't be visible */ for (uint32_t j = 0; j < 10; j++) mme_inc_whole_inst(&b, vals[j]); for (uint32_t i = 0; i < 10; i++) mme_store_imm_addr(&b, data_addr + i * 4, vals[i]); std::vector params; auto macro = mme_builder_finish_vec(&b); test_macro(&b, macro, params); uint32_t i; for (i = 0; i < 10; i++) ASSERT_EQ(data[i], 0); } TEST_F(mme_tu104_sim_test, mme_exit) { mme_builder b; mme_builder_init(&b, devinfo); mme_value vals[10]; for (uint32_t i = 0; i < 10; i++) vals[i] = mme_mov(&b, mme_zero()); for (uint32_t i = 0; i < 10; i++) mme_store_imm_addr(&b, data_addr + i * 4, mme_imm(0)); /* abort */ mme_exit(&b); /* those writes won't be visible */ for (uint32_t i = 0; i < 10; i++) vals[i] = mme_mov(&b, mme_imm(i)); for (uint32_t i = 0; i < 10; i++) { mme_store_imm_addr(&b, data_addr + i * 4, vals[i]); } std::vector params; auto macro = mme_builder_finish_vec(&b); test_macro(&b, macro, params); uint32_t i; for (i = 0; i < 10; i++) ASSERT_EQ(data[i], 0); } TEST_F(mme_tu104_sim_test, mme_exit_if) { mme_builder b; mme_builder_init(&b, devinfo); mme_value vals[10]; for (uint32_t i = 0; i < 10; i++) vals[i] = mme_mov(&b, mme_zero()); for (uint32_t i = 0; i < 10; i++) mme_store_imm_addr(&b, data_addr + i * 4, mme_imm(0)); /* shouldn't do anything */ mme_exit_if(&b, ieq, mme_zero(), mme_imm(1)); for (uint32_t i = 0; i < 10; i++) vals[i] = mme_mov(&b, mme_imm(i)); for (uint32_t i = 0; i < 10; i++) { /* abort on reaching 5 */ mme_exit_if(&b, ile, mme_imm(5), vals[i]); mme_store_imm_addr(&b, data_addr + i * 4, vals[i]); } std::vector params; auto macro = mme_builder_finish_vec(&b); test_macro(&b, macro, params); uint32_t i; for (i = 0; i < 10; i++) ASSERT_EQ(data[i], i < 5 ? i : 0); } static bool c_ilt(int32_t x, int32_t y) { return x < y; }; static bool c_ult(uint32_t x, uint32_t y) { return x < y; }; static bool c_ile(int32_t x, int32_t y) { return x <= y; }; static bool c_ule(uint32_t x, uint32_t y) { return x <= y; }; static bool c_ieq(int32_t x, int32_t y) { return x == y; }; static bool c_ige(int32_t x, int32_t y) { return x >= y; }; static bool c_uge(uint32_t x, uint32_t y) { return x >= y; }; static bool c_igt(int32_t x, int32_t y) { return x > y; }; static bool c_ugt(uint32_t x, uint32_t y) { return x > y; }; static bool c_ine(int32_t x, int32_t y) { return x != y; }; #define IF_TEST(op) \ TEST_F(mme_tu104_sim_test, if_##op) \ { \ mme_builder b; \ mme_builder_init(&b, devinfo); \ \ mme_value x = mme_load(&b); \ mme_value y = mme_load(&b); \ mme_value i = mme_mov(&b, mme_zero()); \ \ mme_start_if_##op(&b, x, y); \ { \ mme_add_to(&b, i, i, mme_imm(1)); \ mme_add_to(&b, i, i, mme_imm(1)); \ } \ mme_end_if(&b); \ mme_start_if_##op(&b, x, mme_imm(56)); \ { \ mme_add_to(&b, i, i, mme_imm(1)); \ } \ mme_end_if(&b); \ mme_add_to(&b, i, i, mme_imm(1)); \ mme_add_to(&b, i, i, mme_imm(1)); \ mme_add_to(&b, i, i, mme_imm(1)); \ \ mme_store_imm_addr(&b, data_addr + 0, i); \ \ auto macro = mme_builder_finish_vec(&b); \ \ uint32_t vals[] = {23, 56, (uint32_t)-5, (uint32_t)-10, 56, 14}; \ \ for (uint32_t i = 0; i < ARRAY_SIZE(vals) - 1; i++) { \ reset_push(); \ \ std::vector params; \ params.push_back(vals[i + 0]); \ params.push_back(vals[i + 1]); \ \ test_macro(&b, macro, params); \ \ uint32_t expected = 3; \ if (c_##op(params[0], params[1])) \ expected += 2; \ if (c_##op(params[0], 56)) \ expected += 1; \ \ ASSERT_EQ(data[0], expected); \ } \ } IF_TEST(ilt) IF_TEST(ult) IF_TEST(ile) IF_TEST(ule) IF_TEST(ieq) IF_TEST(ige) IF_TEST(uge) IF_TEST(igt) IF_TEST(ugt) IF_TEST(ine) #undef IF_TEST #define WHILE_TEST(op, start, step, bound) \ TEST_F(mme_tu104_sim_test, while_##op) \ { \ mme_builder b; \ mme_builder_init(&b, devinfo); \ \ mme_value x = mme_mov(&b, mme_zero()); \ mme_value y = mme_mov(&b, mme_zero()); \ mme_value z = mme_mov(&b, mme_imm(start)); \ mme_value w = mme_mov(&b, mme_zero()); \ mme_value v = mme_mov(&b, mme_zero()); \ \ for (uint32_t j = 0; j < 5; j++) \ mme_inc_whole_inst(&b, x); \ \ mme_while(&b, op, z, mme_imm(bound)) { \ for (uint32_t j = 0; j < 5; j++) \ mme_inc_whole_inst(&b, y); \ \ mme_add_to(&b, z, z, mme_imm(step)); \ \ for (uint32_t j = 0; j < 5; j++) \ mme_inc_whole_inst(&b, w); \ } \ \ for (uint32_t j = 0; j < 5; j++) \ mme_inc_whole_inst(&b, v); \ \ mme_store_imm_addr(&b, data_addr + 0, x); \ mme_store_imm_addr(&b, data_addr + 4, y); \ mme_store_imm_addr(&b, data_addr + 8, z); \ mme_store_imm_addr(&b, data_addr + 12, w); \ mme_store_imm_addr(&b, data_addr + 16, v); \ \ auto macro = mme_builder_finish_vec(&b); \ \ uint32_t end = (uint32_t)(start), count = 0; \ while (c_##op(end, (bound))) { \ end += (uint32_t)(step); \ count++; \ } \ \ std::vector params; \ test_macro(&b, macro, params); \ ASSERT_EQ(data[0], 5); \ ASSERT_EQ(data[1], 5 * count); \ ASSERT_EQ(data[2], end); \ ASSERT_EQ(data[3], 5 * count); \ ASSERT_EQ(data[4], 5); \ } WHILE_TEST(ilt, 0, 1, 7) WHILE_TEST(ult, 0, 1, 7) WHILE_TEST(ile, -10, 2, 0) WHILE_TEST(ule, 0, 1, 7) WHILE_TEST(ieq, 0, 5, 0) WHILE_TEST(ige, 5, -1, -5) WHILE_TEST(uge, 15, -2, 2) WHILE_TEST(igt, 7, -3, -10) WHILE_TEST(ugt, 1604, -30, 1000) WHILE_TEST(ine, 0, 1, 7) #undef WHILE_TEST TEST_F(mme_tu104_sim_test, nested_while) { mme_builder b; mme_builder_init(&b, devinfo); mme_value n = mme_load(&b); mme_value m = mme_load(&b); mme_value count = mme_mov(&b, mme_zero()); mme_value i = mme_mov(&b, mme_zero()); mme_value j = mme_mov(&b, mme_imm(0xffff)); mme_while(&b, ine, i, n) { mme_mov_to(&b, j, mme_zero()); mme_while(&b, ine, j, m) { mme_add_to(&b, count, count, mme_imm(1)); mme_add_to(&b, j, j, mme_imm(1)); } mme_add_to(&b, i, i, mme_imm(1)); } mme_store_imm_addr(&b, data_addr + 0, i); mme_store_imm_addr(&b, data_addr + 4, j); mme_store_imm_addr(&b, data_addr + 8, count); auto macro = mme_builder_finish_vec(&b); std::vector params; params.push_back(3); params.push_back(5); test_macro(&b, macro, params); ASSERT_EQ(data[0], 3); ASSERT_EQ(data[1], 5); ASSERT_EQ(data[2], 15); } #if 0 TEST_F(mme_tu104_sim_test, do_ble) { mme_builder b; mme_builder_init(&b, devinfo); mme_alu(&b, R5, ADD, LOAD0, ZERO); mme_alu(&b, R6, ADD, ZERO, ZERO); mme_alu(&b, R7, ADD, ZERO, ZERO); mme_alu_imm(&b, R7, ADD, R7, IMM, 1); mme_alu_imm(&b, R7, ADD, R7, IMM, 1); mme_alu_imm(&b, R7, ADD, R7, IMM, 1); mme_alu_imm(&b, R7, ADD, R7, IMM, 1); mme_alu_imm(&b, R6, ADD, R6, IMM, 1); mme_branch(&b, BLE, R6, R5, -3, 2); mme_alu_imm(&b, R7, ADD, R7, IMM, 1); mme_alu_imm(&b, R7, ADD, R7, IMM, 1); mme_store_imm_addr(&b, data_addr + 0, MME_TU104_REG_R7); mme_end(&b); uint32_t counts[] = {0, 1, 5, 9}; for (uint32_t i = 0; i < ARRAY_SIZE(counts); i++) { reset_push(); std::vector params; params.push_back(counts[i]); test_macro(&b, params); } } #endif TEST_F(mme_tu104_sim_test, dread_dwrite) { mme_builder b; mme_builder_init(&b, devinfo); mme_value x = mme_load(&b); mme_value y = mme_load(&b); mme_dwrite(&b, mme_imm(5), x); mme_dwrite(&b, mme_imm(8), y); mme_value y2 = mme_dread(&b, mme_imm(8)); mme_value x2 = mme_dread(&b, mme_imm(5)); mme_store_imm_addr(&b, data_addr + 0, y2); mme_store_imm_addr(&b, data_addr + 4, x2); auto macro = mme_builder_finish_vec(&b); std::vector params; params.push_back(-10); params.push_back(5); test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, dwrite_dma) { const uint32_t canary5 = 0xc0ffee01; const uint32_t canary8 = canary5 & 0x00ffff00; mme_builder b; mme_builder_init(&b, devinfo); mme_value x = mme_load(&b); mme_value y = mme_load(&b); mme_dwrite(&b, mme_imm(5), x); mme_dwrite(&b, mme_imm(8), y); auto macro = mme_builder_finish_vec(&b); push_macro(0, macro); P_1INC(p, NVC597, CALL_MME_MACRO(0)); P_INLINE_DATA(p, canary5); P_INLINE_DATA(p, canary8); P_MTHD(p, NVC597, SET_MME_MEM_ADDRESS_A); P_NVC597_SET_MME_MEM_ADDRESS_A(p, high32(data_addr)); P_NVC597_SET_MME_MEM_ADDRESS_B(p, low32(data_addr)); /* Start 3 dwords into MME RAM */ P_NVC597_SET_MME_DATA_RAM_ADDRESS(p, 3); P_IMMD(p, NVC597, MME_DMA_WRITE, 20); submit_push(); for (uint32_t i = 0; i < 20; i++) { if (i == 5 - 3) { ASSERT_EQ(data[i], canary5); } else if (i == 8 - 3) { ASSERT_EQ(data[i], canary8); } else { ASSERT_EQ(data[i], 0); } } } TEST_F(mme_tu104_sim_test, dram_limit) { static const uint32_t chunk_size = 32; mme_builder b; mme_builder_init(&b, devinfo); mme_value start = mme_load(&b); mme_value count = mme_load(&b); mme_value i = mme_mov(&b, start); mme_loop(&b, count) { mme_dwrite(&b, i, i); mme_add_to(&b, i, i, mme_imm(1)); } mme_value j = mme_mov(&b, start); struct mme_value64 addr = mme_mov64(&b, mme_imm64(data_addr)); mme_loop(&b, count) { mme_value x = mme_dread(&b, j); mme_store(&b, addr, x); mme_add_to(&b, j, j, mme_imm(1)); mme_add64_to(&b, addr, addr, mme_imm64(4)); } auto macro = mme_builder_finish_vec(&b); for (uint32_t i = 0; i < MME_TU104_DRAM_COUNT; i += chunk_size) { reset_push(); push_macro(0, macro); P_1INC(p, NVC597, CALL_MME_MACRO(0)); P_INLINE_DATA(p, i); P_INLINE_DATA(p, chunk_size); submit_push(); for (uint32_t j = 0; j < chunk_size; j++) ASSERT_EQ(data[j], i + j); } } TEST_F(mme_tu104_sim_test, dma_read_fifoed) { mme_builder b; mme_builder_init(&b, devinfo); mme_mthd(&b, NVC597_SET_MME_DATA_RAM_ADDRESS); mme_emit(&b, mme_zero()); mme_mthd(&b, NVC597_SET_MME_MEM_ADDRESS_A); mme_emit(&b, mme_imm(high32(data_addr))); mme_emit(&b, mme_imm(low32(data_addr))); mme_mthd(&b, NVC597_MME_DMA_READ_FIFOED); mme_emit(&b, mme_imm(2)); mme_tu104_load_barrier(&b); mme_value x = mme_load(&b); mme_value y = mme_load(&b); mme_store_imm_addr(&b, data_addr + 256 + 0, x); mme_store_imm_addr(&b, data_addr + 256 + 4, y); auto macro = mme_builder_finish_vec(&b); if (devinfo->cls_eng3d < BLACKWELL_A) P_IMMD(p, NVC597, SET_MME_DATA_FIFO_CONFIG, FIFO_SIZE_SIZE_4KB); for (uint32_t i = 0; i < 64; i++) data[i] = 1000 + i; std::vector params; params.push_back(7); test_macro(&b, macro, params); } TEST_F(mme_tu104_sim_test, scratch_limit) { static const uint32_t chunk_size = 32; mme_builder b; mme_builder_init(&b, devinfo); mme_value start = mme_load(&b); mme_value count = mme_load(&b); mme_value i = mme_mov(&b, start); mme_loop(&b, count) { mme_mthd_arr(&b, NVC597_SET_MME_SHADOW_SCRATCH(0), i); mme_emit(&b, i); mme_add_to(&b, i, i, mme_imm(1)); } mme_value j = mme_mov(&b, start); struct mme_value64 addr = mme_mov64(&b, mme_imm64(data_addr)); mme_loop(&b, count) { mme_value x = mme_state_arr(&b, NVC597_SET_MME_SHADOW_SCRATCH(0), j); mme_store(&b, addr, x); mme_add_to(&b, j, j, mme_imm(1)); mme_add64_to(&b, addr, addr, mme_imm64(4)); } auto macro = mme_builder_finish_vec(&b); for (uint32_t i = 0; i < MME_TU104_SCRATCH_COUNT; i += chunk_size) { reset_push(); push_macro(0, macro); P_1INC(p, NVC597, CALL_MME_MACRO(0)); P_INLINE_DATA(p, i); P_INLINE_DATA(p, chunk_size); submit_push(); for (uint32_t j = 0; j < chunk_size; j++) ASSERT_EQ(data[j], i + j); } } TEST_F(mme_tu104_sim_test, sanity_compute) { const uint32_t canary = 0xc0ffee01; mme_builder b; mme_builder_init(&b, devinfo); mme_store_compute_imm_addr(&b, data_addr, mme_imm(canary)); auto macro = mme_builder_finish_vec(&b); reset_push(); push_macro(0, macro); P_1INC(p, NVC7C0, CALL_MME_MACRO(0)); P_NVC7C0_CALL_MME_MACRO(p, 0, 0); submit_push(); ASSERT_EQ(data[0], canary); } TEST_F(mme_tu104_sim_test, scratch_limit_compute) { static const uint32_t chunk_size = 4; mme_builder b; mme_builder_init(&b, devinfo); mme_value start = mme_load(&b); mme_value count = mme_load(&b); mme_value i = mme_mov(&b, start); mme_loop(&b, count) { mme_mthd_arr(&b, NVC7C0_SET_MME_SHADOW_SCRATCH(0), i); mme_emit(&b, i); mme_add_to(&b, i, i, mme_imm(1)); } mme_value j = mme_mov(&b, start); struct mme_value64 addr = mme_mov64(&b, mme_imm64(data_addr)); mme_loop(&b, count) { mme_value x = mme_state_arr(&b, NVC7C0_SET_MME_SHADOW_SCRATCH(0), j); mme_store_compute(&b, addr, x); mme_add_to(&b, j, j, mme_imm(1)); mme_add64_to(&b, addr, addr, mme_imm64(4)); } auto macro = mme_builder_finish_vec(&b); for (uint32_t i = 0; i < 8; i += chunk_size) { reset_push(); push_macro(0, macro); P_1INC(p, NVC7C0, CALL_MME_MACRO(1)); P_INLINE_DATA(p, i); P_INLINE_DATA(p, chunk_size); submit_push(); for (uint32_t j = 0; j < chunk_size; j++) ASSERT_EQ(data[j], i + j); } } TEST_F(mme_tu104_sim_test, scratch_share_3d_to_compute) { static const uint32_t chunk_size = 4; mme_builder b; mme_builder_init(&b, devinfo); mme_value start = mme_load(&b); mme_value count = mme_load(&b); mme_value channel = mme_load(&b); mme_if(&b, ieq, channel, mme_zero()) { mme_value i = mme_mov(&b, start); mme_loop(&b, count) { mme_mthd_arr(&b, NVC597_SET_MME_SHADOW_SCRATCH(0), i); mme_emit(&b, i); mme_add_to(&b, i, i, mme_imm(1)); } } mme_if(&b, ieq, channel, mme_imm(1)) { mme_value i = mme_mov(&b, start); struct mme_value64 addr = mme_mov64(&b, mme_imm64(data_addr)); mme_loop(&b, count) { mme_value val = mme_state_arr(&b, NVC7C0_SET_MME_SHADOW_SCRATCH(0), i); mme_store_compute(&b, addr, val); mme_add_to(&b, i, i, mme_imm(1)); mme_add64_to(&b, addr, addr, mme_imm64(4)); } } auto macro = mme_builder_finish_vec(&b); for (uint32_t i = 0; i < 8; i += chunk_size) { reset_push(); push_macro(0, macro); P_1INC(p, NVC597, CALL_MME_MACRO(0)); P_INLINE_DATA(p, i); P_INLINE_DATA(p, chunk_size); P_INLINE_DATA(p, 0); P_1INC(p, NVC7C0, CALL_MME_MACRO(0)); P_INLINE_DATA(p, i); P_INLINE_DATA(p, chunk_size); P_INLINE_DATA(p, 1); submit_push(); for (uint32_t j = 0; j < chunk_size; j++) ASSERT_EQ(data[j], i + j); } }